Skip to content

Instantly share code, notes, and snippets.

@45deg

45deg/learn.ipynb

Created July 2, 2026 15:03
Show Gist options

  • Select an option

    Learn more about clone URLs
  • Save 45deg/293e7a1f41b7ec00aebba70e7041b9da to your computer and use it in GitHub Desktop.

Select an option

Learn more about clone URLs
Save 45deg/293e7a1f41b7ec00aebba70e7041b9da to your computer and use it in GitHub Desktop.
Download ZIP
Display the source blob
Display the rendered blob
Raw
learn.ipynb
{
"cells": [
{
"cell_type": "markdown",
"id": "d5d61642",
"metadata": {
"id": "d5d61642"
},
"source": [
"# Japanese punctuation restoration\n",
"\n",
"Colab notebook version.\n",
"\n",
"Data source:\n",
"- LLM-jp Corpus v4 GitLab repository API\n",
"- Lists `.jsonl.gz` files through GitLab API\n",
"- Downloads selected gzip files on demand into `/content/llmjp_gitlab_cache`\n",
"- Reads JSONL lines from local gzip cache during training\n",
"\n",
"Each JSONL row is expected to contain:\n",
"```json\n",
"{\"text\": \"...\", \"meta\": {...}}\n",
"```\n",
"\n",
"Punctuation labels:\n",
"- `NONE`\n",
"- `、`\n",
"- `。`\n",
"\n",
"Newline labels:\n",
"- `0`\n",
"- `1`\n",
"\n",
"The model predicts punctuation and newline insertion immediately after each input character.\n",
"\n",
"## 1. Install"
]
},
{
"cell_type": "code",
"execution_count": 1,
"id": "589f82d1",
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"executionInfo": {
"elapsed": 8435,
"status": "ok",
"timestamp": 1777267844166,
"user": {
"displayName": "kz",
"userId": "04068570917018885044"
},
"user_tz": -540
},
"id": "589f82d1",
"outputId": "182856cc-99f1-472b-cc68-93540209fb7b"
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"\u001b[?25l \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m0.0/64.9 kB\u001b[0m \u001b[31m?\u001b[0m eta \u001b[36m-:--:--\u001b[0m\r\u001b[2K \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m64.9/64.9 kB\u001b[0m \u001b[31m6.7 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
"\u001b[?25h\u001b[?25l \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m0.0/8.9 MB\u001b[0m \u001b[31m?\u001b[0m eta \u001b[36m-:--:--\u001b[0m\r\u001b[2K \u001b[91m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[91m╸\u001b[0m \u001b[32m8.9/8.9 MB\u001b[0m \u001b[31m280.1 MB/s\u001b[0m eta \u001b[36m0:00:01\u001b[0m\r\u001b[2K \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m8.9/8.9 MB\u001b[0m \u001b[31m147.2 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
"\u001b[?25h\u001b[?25l \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m0.0/155.6 kB\u001b[0m \u001b[31m?\u001b[0m eta \u001b[36m-:--:--\u001b[0m\r\u001b[2K \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m155.6/155.6 kB\u001b[0m \u001b[31m13.8 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
"\u001b[?25h\u001b[31mERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.\n",
"google-colab 1.0.0 requires requests==2.32.4, but you have requests 2.33.1 which is incompatible.\u001b[0m\u001b[31m\n",
"\u001b[0m"
]
}
],
"source": [
"\n",
"!pip -q install -U requests scikit-learn psutil tqdm"
]
},
{
"cell_type": "markdown",
"id": "2430964a",
"metadata": {
"id": "2430964a"
},
"source": [
"## 2. Imports / Config"
]
},
{
"cell_type": "code",
"execution_count": 11,
"id": "d8da75ee",
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"executionInfo": {
"elapsed": 1028,
"status": "ok",
"timestamp": 1777267933242,
"user": {
"displayName": "kz",
"userId": "04068570917018885044"
},
"user_tz": -540
},
"id": "d8da75ee",
"outputId": "38cbae56-0feb-40f4-f2dc-862f7c0c389d"
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"DEVICE: cuda\n",
"NVIDIA L4\n",
"VRAM GB: 22.03\n",
"Drive already mounted at /content/drive; to attempt to forcibly remount, call drive.mount(\"/content/drive\", force_remount=True).\n",
"checkpoint dir: /content/drive/MyDrive/punct-to-text-2/checkpoints\n",
"TrainConfig(gitlab_base_url='https://gitlab.llm-jp.nii.ac.jp', gitlab_project_path='datasets/llm-jp-corpus-v4', gitlab_ref='main', gitlab_token_env='GITLAB_TOKEN', gitlab_timeout_sec=180, api_per_page=100, api_max_pages_per_corpus=80, cache_dir='/content/llmjp_gitlab_cache', force_redownload=False, max_files_per_corpus=32, download_workers=4, vocab_scan_docs_per_corpus=20000, vocab_min_freq=2, vocab_max_size=12000, experiment_name='vocab12000_swiglu', max_len=192, hidden=384, heads=6, layers=4, ffn_type='swiglu', swiglu_hidden=1024, norm_first=True, num_labels=3, dropout=0.12, epochs=5, steps_per_epoch=2000, batch_size=256, val_chunks=4096, shuffle_buffer_chunks=2048, num_workers=2, use_train_chunk_cache=True, use_val_chunk_cache=True, train_cache_chunks=1000000, train_chunk_cache_path=None, val_chunk_cache_path=None, min_chars_per_chunk=48, stride=144, max_chunks_per_document=6, val_doc_ratio=0.04, min_line_chars=8, max_line_chars=2000, min_text_chars=64, max_text_chars=12000, min_japanese_ratio=0.35, max_ascii_ratio=0.35, max_digit_ratio=0.25, min_punct_ratio=0.004, max_punct_ratio=0.22, drop_duplicate_lines=True, lr=0.0005, min_lr_ratio=0.06, warmup_ratio=0.06, weight_decay=0.01, label_smoothing=0.0, newline_loss_weight=0.5, newline_pos_weight=4.0, newline_threshold=0.5, punct_margin=0.0, newline_threshold_candidates=(0.3, 0.35, 0.4, 0.45, 0.5, 0.55, 0.6), punct_margin_candidates=(0.0, 0.25, 0.5, 0.75, 1.0), inference_mode='logits_center_weighted', auto_tune=True, auto_tune_results_path='stage_a_auto_tune_results.json', auto_tune_select_path='stage_a_selected_config.json', auto_tune_max_files_per_corpus=32, auto_tune_steps_per_epoch=600, auto_tune_epochs=2, auto_tune_val_chunks=4096, auto_tune_train_cache_chunks=1000000, auto_tune_candidates=({'name': 'A1_vocab16000_gelu', 'vocab_max_size': 16000, 'ffn_type': 'gelu'}, {'name': 'A2_vocab12000_gelu', 'vocab_max_size': 12000, 'ffn_type': 'gelu'}, {'name': 'A3_vocab8000_gelu', 'vocab_max_size': 8000, 'ffn_type': 'gelu'}, {'name': 'A4_vocab12000_swiglu', 'vocab_max_size': 12000, 'ffn_type': 'swiglu', 'swiglu_hidden': 1024}), grad_clip=1.0, use_torch_compile=False, class_weights=(0.25, 0.8, 1.0), work_dir='punct_train_llmjp_gitlab_work', ckpt_path='/content/drive/MyDrive/punct-to-text-2/checkpoints/best_punct_llmjp.pt', bundle_path='/content/drive/MyDrive/punct-to-text-2/checkpoints/punct_llmjp_bundle.json', save_checkpoint_to_drive=True, drive_mount_point='/content/drive', drive_dir='/content/drive/MyDrive/punct-to-text-2/checkpoints', resume_from_checkpoint=False, resume_ckpt_path=None, resume_ckpt_candidates=('/content/drive/MyDrive/punct-to-text-2/checkpoints/best_punct_llmjp.pt', 'best_punct_llmjp.pt'))\n"
]
}
],
"source": [
"\n",
"import os\n",
"import re\n",
"import gc\n",
"import json\n",
"import math\n",
"import time\n",
"import gzip\n",
"import glob\n",
"import random\n",
"import hashlib\n",
"import shutil\n",
"import uuid\n",
"from concurrent.futures import ThreadPoolExecutor, as_completed\n",
"from dataclasses import dataclass, asdict\n",
"from typing import Dict, List, Tuple, Optional, Iterator\n",
"from collections import defaultdict, Counter\n",
"from urllib.parse import quote\n",
"\n",
"import numpy as np\n",
"import requests\n",
"import psutil\n",
"from tqdm.auto import tqdm\n",
"from sklearn.metrics import precision_recall_fscore_support, classification_report\n",
"\n",
"import torch\n",
"import torch.nn as nn\n",
"from torch.utils.data import Dataset, IterableDataset, DataLoader\n",
"\n",
"SEED = 42\n",
"random.seed(SEED)\n",
"np.random.seed(SEED)\n",
"torch.manual_seed(SEED)\n",
"torch.cuda.manual_seed_all(SEED)\n",
"\n",
"DEVICE = \"cuda\" if torch.cuda.is_available() else \"cpu\"\n",
"print(\"DEVICE:\", DEVICE)\n",
"if DEVICE == \"cuda\":\n",
" print(torch.cuda.get_device_name(0))\n",
" print(\"VRAM GB:\", round(torch.cuda.get_device_properties(0).total_memory / 1024**3, 2))\n",
" torch.backends.cuda.matmul.allow_tf32 = True\n",
" torch.backends.cudnn.allow_tf32 = True\n",
" torch.set_float32_matmul_precision(\"high\")\n",
"\n",
"LABEL2ID = {\n",
" \"NONE\": 0,\n",
" \"、\": 1,\n",
" \"。\": 2,\n",
"}\n",
"ID2LABEL = {v: k for k, v in LABEL2ID.items()}\n",
"PAD_ID = 0\n",
"UNK_ID = 1\n",
"PUNCT_BOUNDARY_CHARS = {\"、\", \"。\"}\n",
"BOUNDARY_CHARS = PUNCT_BOUNDARY_CHARS | {\"\\n\"}\n",
"REMOVE_CHARS = {\"\\u3000\", \"\\t\", \"\\r\"}\n",
"\n",
"@dataclass\n",
"class TrainConfig:\n",
" # GitLab repository API\n",
" gitlab_base_url: str = \"https://gitlab.llm-jp.nii.ac.jp\"\n",
" gitlab_project_path: str = \"datasets/llm-jp-corpus-v4\"\n",
" gitlab_ref: str = \"main\"\n",
" gitlab_token_env: str = \"GITLAB_TOKEN\"\n",
" gitlab_timeout_sec: int = 180\n",
" api_per_page: int = 100\n",
" api_max_pages_per_corpus: int = 80\n",
"\n",
" # Download/cache\n",
" cache_dir: str = \"/content/llmjp_gitlab_cache\"\n",
" force_redownload: bool = False\n",
"\n",
" # Keep this small for GitLab API mode. Increase only after the first run is cached.\n",
" max_files_per_corpus: Optional[int] = 32\n",
" download_workers: int = 4\n",
"\n",
" # Compact character vocabulary. This keeps the web ONNX model much smaller\n",
" # than the old 65k hash embedding table.\n",
" vocab_scan_docs_per_corpus: int = 20_000\n",
" vocab_min_freq: int = 2\n",
" vocab_max_size: int = 12_000\n",
" experiment_name: Optional[str] = None\n",
"\n",
" # Model\n",
" max_len: int = 192\n",
" hidden: int = 384\n",
" heads: int = 6\n",
" layers: int = 4\n",
" ffn_type: str = \"swiglu\"\n",
" swiglu_hidden: int = 1_024\n",
" norm_first: bool = True\n",
" num_labels: int = len(LABEL2ID)\n",
" dropout: float = 0.12\n",
"\n",
" # Streaming training size.\n",
" # Effective train chunks = epochs * steps_per_epoch * batch_size.\n",
" epochs: int = 5\n",
" steps_per_epoch: int = 2_000\n",
" batch_size: int = 256\n",
"\n",
" # Small memory buffers.\n",
" val_chunks: int = 4_096\n",
" shuffle_buffer_chunks: int = 2_048\n",
" num_workers: int = 2\n",
" use_train_chunk_cache: bool = True\n",
" use_val_chunk_cache: bool = True\n",
" train_cache_chunks: Optional[int] = 1_000_000\n",
" train_chunk_cache_path: Optional[str] = None\n",
" val_chunk_cache_path: Optional[str] = None\n",
"\n",
" # Chunking\n",
" min_chars_per_chunk: int = 48\n",
" stride: int = 144\n",
" max_chunks_per_document: int = 6\n",
" val_doc_ratio: float = 0.04\n",
"\n",
" # Line/document quality filters.\n",
" # Each accepted line must contain 、 or 。.\n",
" min_line_chars: int = 8\n",
" max_line_chars: int = 2_000\n",
" min_text_chars: int = 64\n",
" max_text_chars: int = 12_000\n",
" min_japanese_ratio: float = 0.35\n",
" max_ascii_ratio: float = 0.35\n",
" max_digit_ratio: float = 0.25\n",
" min_punct_ratio: float = 0.004\n",
" max_punct_ratio: float = 0.22\n",
" drop_duplicate_lines: bool = True\n",
"\n",
" # Optimization\n",
" lr: float = 5e-4 # 少し学習率を上げて収束を早める\n",
" min_lr_ratio: float = 0.06\n",
" warmup_ratio: float = 0.06\n",
" weight_decay: float = 0.01\n",
"\n",
" # 修正ポイント: label_smoothing を 0.0 に変更\n",
" # PyTorchの ignore_index(-100) と label_smoothing が同時に使われると、\n",
" # パディング部分に確率が漏れて学習が歪むバグを回避します。\n",
" label_smoothing: float = 0.0\n",
" newline_loss_weight: float = 0.5\n",
" newline_pos_weight: float = 4.0\n",
"\n",
" # Inference threshold search.\n",
" newline_threshold: float = 0.5\n",
" punct_margin: float = 0.0\n",
" newline_threshold_candidates: Tuple[float, ...] = (0.30, 0.35, 0.40, 0.45, 0.50, 0.55, 0.60)\n",
" punct_margin_candidates: Tuple[float, ...] = (0.00, 0.25, 0.50, 0.75, 1.00)\n",
" inference_mode: str = \"logits_center_weighted\"\n",
"\n",
" # Short automatic Stage A sweep. Enable this before build_vocab runs.\n",
" auto_tune: bool = False\n",
" auto_tune_results_path: str = \"stage_a_auto_tune_results.json\"\n",
" auto_tune_select_path: str = \"stage_a_selected_config.json\"\n",
" auto_tune_max_files_per_corpus: int = 32\n",
" auto_tune_steps_per_epoch: int = 600\n",
" auto_tune_epochs: int = 2\n",
" auto_tune_val_chunks: int = 4_096\n",
" auto_tune_train_cache_chunks: int = 1_000_000\n",
" auto_tune_candidates: Tuple[Dict[str, object], ...] = (\n",
" {\"name\": \"A1_vocab16000_gelu\", \"vocab_max_size\": 16_000, \"ffn_type\": \"gelu\"},\n",
" {\"name\": \"A2_vocab12000_gelu\", \"vocab_max_size\": 12_000, \"ffn_type\": \"gelu\"},\n",
" {\"name\": \"A3_vocab8000_gelu\", \"vocab_max_size\": 8_000, \"ffn_type\": \"gelu\"},\n",
" {\"name\": \"A4_vocab12000_swiglu\", \"vocab_max_size\": 12_000, \"ffn_type\": \"swiglu\", \"swiglu_hidden\": 1_024},\n",
" )\n",
"\n",
" grad_clip: float = 1.0\n",
" use_torch_compile: bool = False\n",
"\n",
" # Manual class weights because we do not run a label-count pre-scan.\n",
" # Order: NONE, 、, 。\n",
" class_weights: Tuple[float, ...] = (\n",
" 0.25,\n",
" 0.80,\n",
" 1.00,\n",
" )\n",
"\n",
" # Output / resume\n",
" work_dir: str = \"punct_train_llmjp_gitlab_work\"\n",
" ckpt_path: str = \"best_punct_llmjp.pt\"\n",
" bundle_path: str = \"punct_llmjp_bundle.json\"\n",
" save_checkpoint_to_drive: bool = True\n",
" drive_mount_point: str = \"/content/drive\"\n",
" drive_dir: str = \"/content/drive/MyDrive/punct-to-text-2/checkpoints\"\n",
" resume_from_checkpoint: bool = False\n",
" resume_ckpt_path: Optional[str] = None\n",
" # Also checks the user-mentioned/older-looking file name if present.\n",
" resume_ckpt_candidates: Tuple[str, ...] = (\n",
" \"best_punct_llmjp.pt\",\n",
" )\n",
"\n",
"CFG = TrainConfig()\n",
"# Set True for a short Stage A sweep that estimates vocab/FFN/thresholds.\n",
"os.makedirs(CFG.work_dir, exist_ok=True)\n",
"os.makedirs(CFG.cache_dir, exist_ok=True)\n",
"\n",
"def configure_checkpoint_storage() -> None:\n",
" if not CFG.save_checkpoint_to_drive:\n",
" return\n",
"\n",
" try:\n",
" from google.colab import drive\n",
" except Exception as e:\n",
" print(\"Google Drive mount skipped:\", repr(e))\n",
" return\n",
"\n",
" drive.mount(CFG.drive_mount_point)\n",
" os.makedirs(CFG.drive_dir, exist_ok=True)\n",
"\n",
" CFG.ckpt_path = os.path.join(CFG.drive_dir, os.path.basename(CFG.ckpt_path))\n",
" CFG.bundle_path = os.path.join(CFG.drive_dir, os.path.basename(CFG.bundle_path))\n",
" CFG.resume_ckpt_candidates = (CFG.ckpt_path, *CFG.resume_ckpt_candidates)\n",
" print(\"checkpoint dir:\", CFG.drive_dir)\n",
"\n",
"configure_checkpoint_storage()\n",
"if CFG.experiment_name is None:\n",
" CFG.experiment_name = f\"vocab{CFG.vocab_max_size}_{CFG.ffn_type}\"\n",
"print(CFG)\n",
"\n",
"# Optional, if the repository requires auth:\n",
"# os.environ[\"GITLAB_TOKEN\"] = \"paste_your_token_here\""
]
},
{
"cell_type": "markdown",
"id": "79b81eb8",
"metadata": {
"id": "79b81eb8"
},
"source": [
"## 3. Corpus mix\n",
"\n",
"Wikiと議事録を中心にし、Web系を補助にする。"
]
},
{
"cell_type": "code",
"execution_count": 12,
"id": "fd0729b1",
"metadata": {
"executionInfo": {
"elapsed": 2,
"status": "ok",
"timestamp": 1777267933246,
"user": {
"displayName": "kz",
"userId": "04068570917018885044"
},
"user_tz": -540
},
"id": "fd0729b1"
},
"outputs": [],
"source": [
"\n",
"LLMJP_CORPORA = [\n",
" # Main: stable general prose and formal spoken text.\n",
" {\"key\": \"ja_wiki\", \"path\": \"ja/ja_wiki\", \"ratio\": 0.30},\n",
" {\"key\": \"ja_kokkai_giji\", \"path\": \"ja/ja_kokkai_giji\", \"ratio\": 0.20},\n",
"\n",
" # Secondary: web coverage for generalization.\n",
" {\"key\": \"ja_fineweb_2\", \"path\": \"ja/ja_fineweb-2\", \"ratio\": 0.40},\n",
"\n",
" # ダウンロードの重さを軽減するため、以下のコーパスは一旦無効化（必要に応じて復活させてください）\n",
" # {\"key\": \"ja_cc_level0\", \"path\": \"ja/ja_cc/level0\", \"ratio\": 0.08},\n",
" # {\"key\": \"ja_nwc2010\", \"path\": \"ja/ja_nwc2010\", \"ratio\": 0.05},\n",
" # {\"key\": \"ja_warp_html\", \"path\": \"ja/ja_warp_html\", \"ratio\": 0.03},\n",
" {\"key\": \"ja_aozorabunko\", \"path\": \"ja/ja_aozorabunko\", \"ratio\": 0.1},\n",
" # {\"key\": \"ja_kaken\", \"path\": \"ja/ja_kaken\", \"ratio\": 0.04},\n",
" # {\"key\": \"ja_e_gov\", \"path\": \"ja/ja_e-gov\", \"ratio\": 0.02},\n",
"]"
]
},
{
"cell_type": "markdown",
"id": "a95f7444",
"metadata": {
"id": "a95f7444"
},
"source": [
"## 4. GitLab API helpers"
]
},
{
"cell_type": "code",
"execution_count": 13,
"id": "091ce114",
"metadata": {
"executionInfo": {
"elapsed": 10,
"status": "ok",
"timestamp": 1777267933258,
"user": {
"displayName": "kz",
"userId": "04068570917018885044"
},
"user_tz": -540
},
"id": "091ce114",
"lines_to_next_cell": 1
},
"outputs": [],
"source": [
"\n",
"SESSION = requests.Session()\n",
"\n",
"def gitlab_headers() -> Dict[str, str]:\n",
" token = os.environ.get(CFG.gitlab_token_env, \"\").strip()\n",
" if token:\n",
" return {\"PRIVATE-TOKEN\": token}\n",
" return {}\n",
"\n",
"\n",
"def gitlab_project_id() -> str:\n",
" return quote(CFG.gitlab_project_path, safe=\"\")\n",
"\n",
"\n",
"def gitlab_api_url(path: str) -> str:\n",
" return f\"{CFG.gitlab_base_url}/api/v4/{path.lstrip('/')}\"\n",
"\n",
"\n",
"def request_gitlab_json(url: str, params: Dict) -> requests.Response:\n",
" r = SESSION.get(\n",
" url,\n",
" headers=gitlab_headers(),\n",
" params=params,\n",
" timeout=CFG.gitlab_timeout_sec,\n",
" )\n",
" if r.status_code in {401, 403}:\n",
" raise RuntimeError(\n",
" f\"GitLab auth failed: HTTP {r.status_code}. \"\n",
" f\"Set os.environ['{CFG.gitlab_token_env}'] to a valid token if required.\"\n",
" )\n",
" if r.status_code == 404:\n",
" raise FileNotFoundError(f\"GitLab API path not found: {r.url}\")\n",
" r.raise_for_status()\n",
" return r\n",
"\n",
"\n",
"def list_gitlab_jsonl_gz_paths(rel_path: str) -> List[str]:\n",
" \"\"\"List .jsonl.gz blobs under one repository path through GitLab API.\"\"\"\n",
" url = gitlab_api_url(f\"projects/{gitlab_project_id()}/repository/tree\")\n",
" files: List[str] = []\n",
"\n",
" page = 1\n",
" while True:\n",
" params = {\n",
" \"path\": rel_path,\n",
" \"ref\": CFG.gitlab_ref,\n",
" \"recursive\": \"true\",\n",
" \"per_page\": CFG.api_per_page,\n",
" \"page\": page,\n",
" }\n",
" try:\n",
" r = request_gitlab_json(url, params)\n",
" except FileNotFoundError:\n",
" return []\n",
"\n",
" rows = r.json()\n",
" for row in rows:\n",
" if row.get(\"type\") == \"blob\":\n",
" p = row.get(\"path\", \"\")\n",
" if p.endswith(\".jsonl.gz\"):\n",
" files.append(p)\n",
"\n",
" next_page = r.headers.get(\"X-Next-Page\")\n",
" if not next_page:\n",
" break\n",
" page = int(next_page)\n",
" if page > CFG.api_max_pages_per_corpus:\n",
" print(f\"API page cap reached for {rel_path}; files found so far: {len(files)}\")\n",
" break\n",
"\n",
" return sorted(set(files))\n",
"\n",
"\n",
"def cache_path_for_gitlab_file(file_path: str) -> str:\n",
" digest = hashlib.md5(file_path.encode(\"utf-8\")).hexdigest()[:16]\n",
" base = os.path.basename(file_path)\n",
" return os.path.join(CFG.cache_dir, f\"{digest}_{base}\")\n",
"\n",
"\n",
"def is_valid_gzip_file(path: str) -> bool:\n",
" if not os.path.exists(path) or os.path.getsize(path) < 3:\n",
" return False\n",
" try:\n",
" with open(path, \"rb\") as f:\n",
" return f.read(2) == b\"\\x1f\\x8b\"\n",
" except Exception:\n",
" return False\n",
"\n",
"\n",
"def download_gitlab_file_to_cache(file_path: str) -> str:\n",
" \"\"\"Download one repository file through GitLab API raw endpoint into cache.\"\"\"\n",
" out_path = cache_path_for_gitlab_file(file_path)\n",
" if not CFG.force_redownload and is_valid_gzip_file(out_path):\n",
" return out_path\n",
"\n",
" encoded_path = quote(file_path, safe=\"\")\n",
" url = gitlab_api_url(f\"projects/{gitlab_project_id()}/repository/files/{encoded_path}/raw\")\n",
" params = {\n",
" \"ref\": CFG.gitlab_ref,\n",
" \"lfs\": \"true\",\n",
" }\n",
"\n",
" # DataLoaderのマルチプロセスで同時にダウンロードしても壊れないように一時ファイル名にUUIDを付与\n",
" tmp_path = f\"{out_path}.{uuid.uuid4().hex}.part\"\n",
" if os.path.exists(tmp_path):\n",
" os.remove(tmp_path)\n",
"\n",
" print(\"download:\", file_path)\n",
" with SESSION.get(\n",
" url,\n",
" headers=gitlab_headers(),\n",
" params=params,\n",
" stream=True,\n",
" timeout=CFG.gitlab_timeout_sec,\n",
" ) as r:\n",
" if r.status_code in {401, 403}:\n",
" raise RuntimeError(\n",
" f\"GitLab auth failed: HTTP {r.status_code}. \"\n",
" f\"Set os.environ['{CFG.gitlab_token_env}'] to a valid token if required.\"\n",
" )\n",
" if r.status_code == 404:\n",
" raise FileNotFoundError(f\"GitLab raw file not found: {file_path}\")\n",
" r.raise_for_status()\n",
"\n",
" total = int(r.headers.get(\"Content-Length\", \"0\") or 0)\n",
" with open(tmp_path, \"wb\") as f:\n",
" if total > 0:\n",
" pbar = tqdm(total=total, unit=\"B\", unit_scale=True, desc=os.path.basename(file_path))\n",
" else:\n",
" pbar = None\n",
" for chunk in r.iter_content(chunk_size=1024 * 1024):\n",
" if not chunk:\n",
" continue\n",
" f.write(chunk)\n",
" if pbar:\n",
" pbar.update(len(chunk))\n",
" if pbar:\n",
" pbar.close()\n",
"\n",
" if not is_valid_gzip_file(tmp_path):\n",
" with open(tmp_path, \"rb\") as f:\n",
" head = f.read(300)\n",
" os.remove(tmp_path)\n",
" raise RuntimeError(\n",
" \"Downloaded file is not gzip. This can mean the repository returned a Git LFS pointer, \"\n",
" \"an HTML error page, or a permission error. First bytes: \" + repr(head[:200])\n",
" )\n",
"\n",
" os.replace(tmp_path, out_path)\n",
" return out_path"
]
},
{
"cell_type": "markdown",
"id": "79199192",
"metadata": {
"id": "79199192"
},
"source": [
"## 5. Discover remote corpus files"
]
},
{
"cell_type": "code",
"execution_count": 14,
"id": "6d379c31",
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"executionInfo": {
"elapsed": 6803,
"status": "ok",
"timestamp": 1777267940063,
"user": {
"displayName": "kz",
"userId": "04068570917018885044"
},
"user_tz": -540
},
"id": "6d379c31",
"lines_to_next_cell": 1,
"outputId": "e104a5b3-6f3c-4eb7-b7b7-e04bdcd541c0"
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"list GitLab: ja_wiki ja/ja_wiki\n",
"list GitLab: ja_kokkai_giji ja/ja_kokkai_giji\n",
"list GitLab: ja_fineweb_2 ja/ja_fineweb-2\n",
"list GitLab: ja_aozorabunko ja/ja_aozorabunko\n",
"available corpora:\n",
"ja_wiki files 6 ratio 0.3 sample ja/ja_wiki/0004.jsonl.gz\n",
"ja_kokkai_giji files 12 ratio 0.2 sample ja/ja_kokkai_giji/0002.jsonl.gz\n",
"ja_fineweb_2 files 32 ratio 0.4 sample ja/ja_fineweb-2/1558.jsonl.gz\n",
"ja_aozorabunko files 1 ratio 0.1 sample ja/ja_aozorabunko/0000.jsonl.gz\n"
]
}
],
"source": [
"\n",
"AVAILABLE_CORPORA = []\n",
"missing_corpora = []\n",
"\n",
"for spec in LLMJP_CORPORA:\n",
" print(\"list GitLab:\", spec[\"key\"], spec[\"path\"])\n",
" paths = list_gitlab_jsonl_gz_paths(spec[\"path\"])\n",
" if not paths:\n",
" missing_corpora.append(spec[\"key\"])\n",
" continue\n",
"\n",
" rng = random.Random(SEED + len(spec[\"key\"]))\n",
" rng.shuffle(paths)\n",
" if CFG.max_files_per_corpus is not None:\n",
" paths = paths[: CFG.max_files_per_corpus]\n",
"\n",
" item = dict(spec)\n",
" item[\"gitlab_paths\"] = paths\n",
" AVAILABLE_CORPORA.append(item)\n",
"\n",
"if not AVAILABLE_CORPORA:\n",
" raise FileNotFoundError(\n",
" \"No .jsonl.gz files were found through GitLab API. \"\n",
" \"Check gitlab_project_path/ref, repository access, or token.\"\n",
" )\n",
"\n",
"if missing_corpora:\n",
" print(\"missing corpora skipped:\")\n",
" for key in missing_corpora:\n",
" print(\" -\", key)\n",
"\n",
"ratio_sum = sum(c[\"ratio\"] for c in AVAILABLE_CORPORA)\n",
"for c in AVAILABLE_CORPORA:\n",
" c[\"ratio\"] = c[\"ratio\"] / ratio_sum\n",
"\n",
"print(\"available corpora:\")\n",
"for c in AVAILABLE_CORPORA:\n",
" print(\n",
" c[\"key\"],\n",
" \"files\", len(c[\"gitlab_paths\"]),\n",
" \"ratio\", round(c[\"ratio\"], 4),\n",
" \"sample\", c[\"gitlab_paths\"][0],\n",
" )"
]
},
{
"cell_type": "markdown",
"id": "d0efcf51",
"metadata": {
"id": "d0efcf51"
},
"source": [
"## 6. JSONL.GZ reader and text cleaning"
]
},
{
"cell_type": "code",
"execution_count": 15,
"id": "31ab495b",
"metadata": {
"executionInfo": {
"elapsed": 2,
"status": "ok",
"timestamp": 1777267940077,
"user": {
"displayName": "kz",
"userId": "04068570917018885044"
},
"user_tz": -540
},
"id": "31ab495b",
"lines_to_next_cell": 1
},
"outputs": [],
"source": [
"\n",
"def iter_jsonl_gz_texts_from_gitlab_path(file_path: str) -> Iterator[str]:\n",
" # In the fast path, file_path is already a local cached .jsonl.gz path.\n",
" # Fallback supports old calls with a GitLab repository path.\n",
" local_path = file_path if os.path.exists(file_path) else download_gitlab_file_to_cache(file_path)\n",
" with gzip.open(local_path, \"rt\", encoding=\"utf-8\", errors=\"ignore\") as f:\n",
" for line in f:\n",
" line = line.strip()\n",
" if not line:\n",
" continue\n",
" try:\n",
" row = json.loads(line)\n",
" except Exception:\n",
" continue\n",
" text = row.get(\"text\")\n",
" if isinstance(text, str) and text.strip():\n",
" yield text\n",
"\n",
"TRANSLATION = str.maketrans({\n",
" \"｡\": \"。\",\n",
" \"．\": \"。\",\n",
" \"！\": \"。\",\n",
" \"!\": \"。\",\n",
" \"？\": \"。\",\n",
" \"?\": \"。\",\n",
" \"，\": \"、\",\n",
" \"､\": \"、\",\n",
"})\n",
"\n",
"def japanese_ratio(text: str) -> float:\n",
" if not text:\n",
" return 0.0\n",
" jp = 0\n",
" for ch in text:\n",
" if (\"\\u3040\" <= ch <= \"\\u30ff\") or (\"\\u4e00\" <= ch <= \"\\u9fff\"):\n",
" jp += 1\n",
" return jp / max(1, len(text))\n",
"\n",
"def ascii_ratio(text: str) -> float:\n",
" return sum(1 for ch in text if ord(ch) < 128) / max(1, len(text))\n",
"\n",
"def digit_ratio(text: str) -> float:\n",
" return sum(1 for ch in text if ch.isdigit()) / max(1, len(text))\n",
"\n",
"def punct_ratio(text: str) -> float:\n",
" return sum(1 for ch in text if ch in {\"、\", \"。\"}) / max(1, len(text))\n",
"\n",
"def normalize_line(line: str) -> str:\n",
" line = line.translate(TRANSLATION)\n",
" line = line.replace(\"\\u3000\", \" \").replace(\"\\t\", \" \")\n",
" line = re.sub(r\"\\s+\", \" \", line)\n",
" return line.strip()\n",
"\n",
"def split_long_line(line: str) -> List[str]:\n",
" if len(line) <= CFG.max_line_chars:\n",
" return [line]\n",
"\n",
" parts = re.split(r\"(?<=。)\", line)\n",
" out = []\n",
" buf = \"\"\n",
" for part in parts:\n",
" if not part:\n",
" continue\n",
" if len(buf) + len(part) <= CFG.max_line_chars:\n",
" buf += part\n",
" else:\n",
" if buf:\n",
" out.append(buf.strip())\n",
" buf = part\n",
" if buf:\n",
" out.append(buf.strip())\n",
" return [x for x in out if x]\n",
"\n",
"def looks_like_noise_line(line: str) -> bool:\n",
" if len(line) < CFG.min_line_chars:\n",
" return True\n",
"\n",
" # Remove non-sentence-like lines.\n",
" if \"、\" not in line and \"。\" not in line:\n",
" return True\n",
"\n",
" if re.search(r\"https?://|www\\.|<[^>]+>|&[a-z]+;\", line, flags=re.IGNORECASE):\n",
" return True\n",
" if re.search(r\"^[\\-_=*#]{4,}$\", line):\n",
" return True\n",
" if line.count(\"|\") >= 3 or line.count(\",\") >= 8:\n",
" return True\n",
" if ascii_ratio(line) > CFG.max_ascii_ratio:\n",
" return True\n",
" if digit_ratio(line) > CFG.max_digit_ratio:\n",
" return True\n",
" if japanese_ratio(line) < CFG.min_japanese_ratio:\n",
" return True\n",
"\n",
" pr = punct_ratio(line)\n",
" if pr < CFG.min_punct_ratio or pr > CFG.max_punct_ratio:\n",
" return True\n",
"\n",
" if len(line) < 32 and re.match(r\"^[\\-・*●○◎◇◆■□▶▷※注\\d０-９一二三四五六七八九十]+[\\s、。.)）]\", line):\n",
" return True\n",
"\n",
" return False\n",
"\n",
"def clean_text(raw_text: str) -> str:\n",
" raw_text = str(raw_text).replace(\"\\r\\n\", \"\\n\").replace(\"\\r\", \"\\n\")\n",
" accepted: List[str] = []\n",
" seen = set()\n",
"\n",
" for raw_line in raw_text.split(\"\\n\"):\n",
" line = normalize_line(raw_line)\n",
" if not line:\n",
" continue\n",
" for segment in split_long_line(line):\n",
" segment = normalize_line(segment)\n",
" if looks_like_noise_line(segment):\n",
" continue\n",
" if CFG.drop_duplicate_lines:\n",
" h = hashlib.md5(segment.encode(\"utf-8\", errors=\"ignore\")).hexdigest()\n",
" if h in seen:\n",
" continue\n",
" seen.add(h)\n",
" accepted.append(segment)\n",
"\n",
" text = \"\\n\".join(accepted)\n",
" if len(text) < CFG.min_text_chars or len(text) > CFG.max_text_chars:\n",
" return \"\"\n",
" return text"
]
},
{
"cell_type": "markdown",
"id": "2eb3481e",
"metadata": {
"id": "2eb3481e"
},
"source": [
"## 7. Labels, vocab, chunking"
]
},
{
"cell_type": "code",
"execution_count": 16,
"id": "e2fd3252",
"metadata": {
"executionInfo": {
"elapsed": 3,
"status": "ok",
"timestamp": 1777267940096,
"user": {
"displayName": "kz",
"userId": "04068570917018885044"
},
"user_tz": -540
},
"id": "e2fd3252",
"lines_to_next_cell": 1
},
"outputs": [],
"source": [
"\n",
"def add_punct_label(current_label_id: int, boundary: str) -> int:\n",
" current = ID2LABEL[int(current_label_id)]\n",
" punct = \"\"\n",
"\n",
" if \"、\" in current:\n",
" punct = \"、\"\n",
" elif \"。\" in current:\n",
" punct = \"。\"\n",
"\n",
" if boundary in PUNCT_BOUNDARY_CHARS:\n",
" punct = boundary\n",
"\n",
" return LABEL2ID.get(punct, LABEL2ID[\"NONE\"])\n",
"\n",
"def text_to_pairs(text: str) -> Tuple[List[str], List[int], List[int]]:\n",
" chars: List[str] = []\n",
" punct_labels: List[int] = []\n",
" newline_labels: List[int] = []\n",
" last_idx: Optional[int] = None\n",
"\n",
" for ch in text:\n",
" if ch in BOUNDARY_CHARS:\n",
" if last_idx is not None:\n",
" if ch in PUNCT_BOUNDARY_CHARS:\n",
" punct_labels[last_idx] = add_punct_label(punct_labels[last_idx], ch)\n",
" elif ch == \"\\n\":\n",
" newline_labels[last_idx] = 1\n",
" continue\n",
" if ch in REMOVE_CHARS or ch == \" \":\n",
" continue\n",
" chars.append(ch)\n",
" punct_labels.append(LABEL2ID[\"NONE\"])\n",
" newline_labels.append(0)\n",
" last_idx = len(punct_labels) - 1\n",
"\n",
" return chars, punct_labels, newline_labels\n",
"\n",
"def make_chunks(chars: List[str], punct_labels: List[int], newline_labels: List[int]) -> List[Tuple[List[str], List[int], List[int]]]:\n",
" chunks = []\n",
" n = len(chars)\n",
" if n < CFG.min_chars_per_chunk:\n",
" return chunks\n",
" for start in range(0, n, CFG.stride):\n",
" end = min(start + CFG.max_len, n)\n",
" c = chars[start:end]\n",
" y_punct = punct_labels[start:end]\n",
" y_newline = newline_labels[start:end]\n",
" if len(c) >= CFG.min_chars_per_chunk and (any(v != 0 for v in y_punct) or any(v != 0 for v in y_newline)):\n",
" chunks.append((c, y_punct, y_newline))\n",
" if end >= n or len(chunks) >= CFG.max_chunks_per_document:\n",
" break\n",
" return chunks\n",
"\n",
"def build_vocab() -> Tuple[Dict[str, int], List[str]]:\n",
" counter = Counter()\n",
" per_corpus_docs = {}\n",
"\n",
" for spec in AVAILABLE_CORPORA:\n",
" seen = 0\n",
" kept = 0\n",
" print(\"vocab scan:\", spec[\"key\"])\n",
"\n",
" for gitlab_path in spec[\"gitlab_paths\"]:\n",
" try:\n",
" for raw_text in iter_jsonl_gz_texts_from_gitlab_path(gitlab_path):\n",
" text = clean_text(raw_text)\n",
" if not text:\n",
" continue\n",
"\n",
" seen += 1\n",
" chars, punct_labels, newline_labels = text_to_pairs(text)\n",
" if len(chars) >= CFG.min_chars_per_chunk and make_chunks(chars, punct_labels, newline_labels):\n",
" counter.update(chars)\n",
" kept += 1\n",
"\n",
" if seen >= CFG.vocab_scan_docs_per_corpus:\n",
" break\n",
" except Exception as e:\n",
" print(\"SKIP vocab file\", gitlab_path, repr(e))\n",
"\n",
" if seen >= CFG.vocab_scan_docs_per_corpus:\n",
" break\n",
"\n",
" per_corpus_docs[spec[\"key\"]] = {\"seen\": seen, \"kept\": kept}\n",
"\n",
" itos = [\"<pad>\", \"<unk>\"]\n",
" for ch, freq in counter.most_common(CFG.vocab_max_size - len(itos)):\n",
" if freq >= CFG.vocab_min_freq:\n",
" itos.append(ch)\n",
"\n",
" stoi = {ch: i for i, ch in enumerate(itos)}\n",
" if len(itos) >= 65_000:\n",
" raise ValueError(\"vocab too large for compact assumptions; lower vocab_max_size\")\n",
"\n",
" print(\"vocab size:\", len(itos))\n",
" print(\"vocab docs:\", per_corpus_docs)\n",
" with open(os.path.join(CFG.work_dir, \"vocab.json\"), \"w\", encoding=\"utf-8\") as f:\n",
" json.dump(\n",
" {\"itos\": itos, \"label2id\": LABEL2ID, \"id2label\": ID2LABEL, \"per_corpus_docs\": per_corpus_docs},\n",
" f,\n",
" ensure_ascii=False,\n",
" indent=2,\n",
" )\n",
" return stoi, itos\n",
"\n",
"stoi: Dict[str, int] = {}\n",
"itos: List[str] = []\n",
"\n",
"def char_to_id(ch: str) -> int:\n",
" return stoi.get(ch, UNK_ID)\n",
"\n",
"def stable_is_val(corpus_key: str, text: str) -> bool:\n",
" h = hashlib.md5((corpus_key + \"\\n\" + text[:512]).encode(\"utf-8\", errors=\"ignore\")).hexdigest()\n",
" x = int(h[:8], 16) / 0xFFFFFFFF\n",
" return x < CFG.val_doc_ratio\n",
"\n",
"def encode_chunk(chars: List[str], punct_labels: List[int], newline_labels: List[int]) -> Dict[str, torch.Tensor]:\n",
" n = min(len(chars), CFG.max_len)\n",
" x = np.zeros((CFG.max_len,), dtype=np.int32)\n",
" y_punct = np.full((CFG.max_len,), -100, dtype=np.int64)\n",
" y_newline = np.full((CFG.max_len,), -100, dtype=np.int64)\n",
" m = np.zeros((CFG.max_len,), dtype=np.bool_)\n",
"\n",
" x[:n] = [char_to_id(ch) for ch in chars[:n]]\n",
" y_punct[:n] = punct_labels[:n]\n",
" y_newline[:n] = newline_labels[:n]\n",
" m[:n] = True\n",
"\n",
" return {\n",
" \"input_ids\": torch.from_numpy(x),\n",
" \"punct_labels\": torch.from_numpy(y_punct),\n",
" \"newline_labels\": torch.from_numpy(y_newline),\n",
" \"attention_mask\": torch.from_numpy(m),\n",
" }"
]
},
{
"cell_type": "markdown",
"id": "223851cd",
"metadata": {
"id": "223851cd"
},
"source": [
"## 8. Streaming Dataset"
]
},
{
"cell_type": "code",
"execution_count": 17,
"id": "8f6b89cd",
"metadata": {
"executionInfo": {
"elapsed": 15,
"status": "ok",
"timestamp": 1777267940131,
"user": {
"displayName": "kz",
"userId": "04068570917018885044"
},
"user_tz": -540
},
"id": "8f6b89cd",
"lines_to_next_cell": 1
},
"outputs": [],
"source": [
"\n",
"def iter_corpus_chunks(spec: Dict, split: str, rng: random.Random, repeat: bool) -> Iterator[Dict[str, torch.Tensor]]:\n",
" paths = list(spec.get(\"local_paths\", spec[\"gitlab_paths\"]))\n",
" if not paths:\n",
" return\n",
"\n",
" while True:\n",
" rng.shuffle(paths)\n",
" yielded_this_pass = 0\n",
"\n",
" for gitlab_path in paths:\n",
" try:\n",
" text_iter = iter_jsonl_gz_texts_from_gitlab_path(gitlab_path)\n",
" for raw_text in text_iter:\n",
" text = clean_text(raw_text)\n",
" if not text:\n",
" continue\n",
"\n",
" is_val = stable_is_val(spec[\"key\"], text)\n",
" if split == \"train\" and is_val:\n",
" continue\n",
" if split == \"val\" and not is_val:\n",
" continue\n",
"\n",
" chars, punct_labels, newline_labels = text_to_pairs(text)\n",
" chunks = make_chunks(chars, punct_labels, newline_labels)\n",
" if not chunks:\n",
" continue\n",
" rng.shuffle(chunks)\n",
" for c, y_punct, y_newline in chunks:\n",
" yielded_this_pass += 1\n",
" yield encode_chunk(c, y_punct, y_newline)\n",
" except Exception as e:\n",
" print(\"SKIP file\", gitlab_path, repr(e))\n",
" continue\n",
"\n",
" if not repeat or yielded_this_pass == 0:\n",
" break\n",
"\n",
"def buffer_shuffle(iterator: Iterator[Dict[str, torch.Tensor]], buffer_size: int, rng: random.Random) -> Iterator[Dict[str, torch.Tensor]]:\n",
" if buffer_size <= 1:\n",
" yield from iterator\n",
" return\n",
"\n",
" buf: List[Dict[str, torch.Tensor]] = []\n",
" for item in iterator:\n",
" if len(buf) < buffer_size:\n",
" buf.append(item)\n",
" continue\n",
" j = rng.randrange(len(buf))\n",
" yield buf[j]\n",
" buf[j] = item\n",
"\n",
" while buf:\n",
" j = rng.randrange(len(buf))\n",
" yield buf.pop(j)\n",
"\n",
"class StreamingGzipPunctDataset(IterableDataset):\n",
" def __init__(self, corpora: List[Dict], shuffle_buffer_chunks: int):\n",
" super().__init__()\n",
" self.corpora = corpora\n",
" self.shuffle_buffer_chunks = shuffle_buffer_chunks\n",
" self.epoch = 0\n",
"\n",
" def set_epoch(self, epoch: int):\n",
" self.epoch = int(epoch)\n",
"\n",
" def __iter__(self):\n",
" worker_info = torch.utils.data.get_worker_info()\n",
" worker_id = worker_info.id if worker_info else 0\n",
" num_workers = worker_info.num_workers if worker_info else 1\n",
" rng = random.Random(SEED + 1009 * self.epoch + 17 * worker_id)\n",
"\n",
" worker_corpora = []\n",
" for spec in self.corpora:\n",
" paths = list(spec[\"gitlab_paths\"])\n",
" part = paths[worker_id::num_workers]\n",
" if not part:\n",
" part = paths\n",
" item = dict(spec)\n",
" item[\"gitlab_paths\"] = part\n",
" worker_corpora.append(item)\n",
"\n",
" keys = [c[\"key\"] for c in worker_corpora]\n",
" weights = [c[\"ratio\"] for c in worker_corpora]\n",
" generators = {\n",
" c[\"key\"]: iter_corpus_chunks(\n",
" c,\n",
" split=\"train\",\n",
" rng=random.Random(rng.randint(0, 10**9)),\n",
" repeat=True,\n",
" )\n",
" for c in worker_corpora\n",
" }\n",
"\n",
" def mixed_stream():\n",
" while True:\n",
" key = rng.choices(keys, weights=weights, k=1)[0]\n",
" try:\n",
" yield next(generators[key])\n",
" except StopIteration:\n",
" idx = keys.index(key)\n",
" keys.pop(idx)\n",
" weights.pop(idx)\n",
" generators.pop(key, None)\n",
" if not keys:\n",
" raise RuntimeError(\"All corpus streams are empty after filtering.\")\n",
"\n",
" yield from buffer_shuffle(mixed_stream(), self.shuffle_buffer_chunks, rng)\n",
"\n",
"class CachedPunctDataset(Dataset):\n",
" def __init__(self, rows: List[Dict[str, torch.Tensor]]):\n",
" self.rows = rows\n",
"\n",
" def __len__(self):\n",
" return len(self.rows)\n",
"\n",
" def __getitem__(self, idx):\n",
" return self.rows[idx]\n",
"\n",
"def compact_cache_item(item: Dict[str, torch.Tensor]) -> Dict[str, torch.Tensor]:\n",
" return {\n",
" \"input_ids\": item[\"input_ids\"].short(),\n",
" \"punct_labels\": item[\"punct_labels\"].short(),\n",
" \"newline_labels\": item[\"newline_labels\"].short(),\n",
" \"attention_mask\": item[\"attention_mask\"].bool(),\n",
" }\n",
"\n",
"def describe_chunk_rows(rows: List[Dict[str, torch.Tensor]], name: str) -> None:\n",
" punct_counts = Counter()\n",
" newline_counts = Counter()\n",
" valid_tokens = 0\n",
" for row in rows:\n",
" punct = row[\"punct_labels\"]\n",
" newline = row[\"newline_labels\"]\n",
" mask = punct != -100\n",
" if mask.any():\n",
" punct_counts.update(punct[mask].long().tolist())\n",
" valid_tokens += int(mask.sum())\n",
" newline_mask = newline != -100\n",
" if newline_mask.any():\n",
" newline_counts.update(newline[newline_mask].long().tolist())\n",
" print(\n",
" f\"{name} label stats:\",\n",
" {\n",
" \"chunks\": len(rows),\n",
" \"tokens\": valid_tokens,\n",
" \"punct\": {ID2LABEL[int(k)]: int(v) for k, v in sorted(punct_counts.items())},\n",
" \"newline\": {int(k): int(v) for k, v in sorted(newline_counts.items())},\n",
" },\n",
" )\n",
"\n",
"def build_or_load_chunk_cache(\n",
" split: str,\n",
" corpora: List[Dict],\n",
" cache_path: str,\n",
" max_chunks: Optional[int] = None,\n",
") -> List[Dict[str, torch.Tensor]]:\n",
" if os.path.exists(cache_path):\n",
" print(\"load chunk cache:\", cache_path)\n",
" try:\n",
" rows = torch.load(cache_path, map_location=\"cpu\", weights_only=False)\n",
" except TypeError:\n",
" rows = torch.load(cache_path, map_location=\"cpu\")\n",
" describe_chunk_rows(rows, f\"{split} cache\")\n",
" return rows\n",
"\n",
" os.makedirs(os.path.dirname(cache_path) or \".\", exist_ok=True)\n",
" rows: List[Dict[str, torch.Tensor]] = []\n",
" per_corpus_count = defaultdict(int)\n",
" per_corpus_target = None\n",
" if split == \"val\":\n",
" per_corpus_target = {\n",
" c[\"key\"]: max(1, int(round(CFG.val_chunks * c[\"ratio\"])))\n",
" for c in corpora\n",
" }\n",
"\n",
" for spec in corpora:\n",
" rng = random.Random(SEED + 3000 + len(spec[\"key\"]) + (0 if split == \"val\" else 100_000))\n",
" target = per_corpus_target[spec[\"key\"]] if per_corpus_target else None\n",
" print(\"build chunk cache:\", split, spec[\"key\"], \"target\", target or max_chunks or \"all\")\n",
" iterator = iter_corpus_chunks(spec, split=split, rng=rng, repeat=False)\n",
" progress_total = target if target is not None else max_chunks\n",
" for item in tqdm(iterator, total=progress_total, desc=f\"{split} {spec['key']}\"):\n",
" rows.append(compact_cache_item(item))\n",
" per_corpus_count[spec[\"key\"]] += 1\n",
" if target is not None and per_corpus_count[spec[\"key\"]] >= target:\n",
" break\n",
" if max_chunks is not None and len(rows) >= max_chunks:\n",
" break\n",
" if max_chunks is not None and len(rows) >= max_chunks:\n",
" break\n",
"\n",
" if split == \"val\" and len(rows) < 256:\n",
" raise RuntimeError(\"Validation cache is too small. Loosen filters or increase val_doc_ratio/max_files_per_corpus.\")\n",
" if not rows:\n",
" raise RuntimeError(f\"{split} chunk cache is empty. Loosen filters or check corpus paths.\")\n",
"\n",
" rng = random.Random(SEED + (4444 if split == \"val\" else 5555))\n",
" rng.shuffle(rows)\n",
" if split == \"val\":\n",
" rows = rows[: CFG.val_chunks]\n",
" elif max_chunks is not None:\n",
" rows = rows[:max_chunks]\n",
" describe_chunk_rows(rows, f\"{split} cache\")\n",
" torch.save(rows, cache_path)\n",
" print(\"saved chunk cache:\", cache_path, \"chunks\", len(rows), dict(per_corpus_count))\n",
" return rows\n",
"\n",
"def default_chunk_cache_path(split: str) -> str:\n",
" if split == \"train\":\n",
" chunk_part = \"all\" if CFG.train_cache_chunks is None else str(CFG.train_cache_chunks)\n",
" else:\n",
" chunk_part = str(CFG.val_chunks)\n",
" return os.path.join(\n",
" CFG.cache_dir,\n",
" f\"{CFG.experiment_name}_{split}_chunks_{chunk_part}_len{CFG.max_len}_stride{CFG.stride}_vocab{CFG.vocab_max_size}.pt\",\n",
" )\n",
"\n",
"def build_val_dataset(corpora: List[Dict]) -> CachedPunctDataset:\n",
" if CFG.use_val_chunk_cache:\n",
" cache_path = CFG.val_chunk_cache_path or default_chunk_cache_path(\"val\")\n",
" return CachedPunctDataset(build_or_load_chunk_cache(\"val\", corpora, cache_path, CFG.val_chunks))\n",
"\n",
" rows: List[Dict[str, torch.Tensor]] = []\n",
" per_corpus_target = {\n",
" c[\"key\"]: max(1, int(round(CFG.val_chunks * c[\"ratio\"])))\n",
" for c in corpora\n",
" }\n",
" per_corpus_count = defaultdict(int)\n",
"\n",
" for spec in corpora:\n",
" target = per_corpus_target[spec[\"key\"]]\n",
" rng = random.Random(SEED + 3000 + len(spec[\"key\"]))\n",
" print(\"build val:\", spec[\"key\"], \"target\", target)\n",
" iterator = iter_corpus_chunks(spec, split=\"val\", rng=rng, repeat=False)\n",
" for item in tqdm(iterator, total=target, desc=f\"val {spec['key']}\"):\n",
" rows.append(item)\n",
" per_corpus_count[spec[\"key\"]] += 1\n",
" if per_corpus_count[spec[\"key\"]] >= target or len(rows) >= CFG.val_chunks:\n",
" break\n",
"\n",
" if len(rows) < 256:\n",
" raise RuntimeError(\"Validation cache is too small. Loosen filters or increase val_doc_ratio/max_files_per_corpus.\")\n",
"\n",
" random.Random(SEED + 4444).shuffle(rows)\n",
" rows = rows[: CFG.val_chunks]\n",
" print(\"val chunks:\", len(rows), dict(per_corpus_count))\n",
" return CachedPunctDataset(rows)"
]
},
{
"cell_type": "markdown",
"id": "a7dc9f25",
"metadata": {
"id": "a7dc9f25"
},
"source": [
"## 9. Build DataLoaders"
]
},
{
"cell_type": "code",
"execution_count": 18,
"id": "c68df681",
"metadata": {
"executionInfo": {
"elapsed": 3,
"status": "ok",
"timestamp": 1777267940172,
"user": {
"displayName": "kz",
"userId": "04068570917018885044"
},
"user_tz": -540
},
"id": "c68df681",
"lines_to_next_cell": 1
},
"outputs": [],
"source": [
"\n",
"def build_dataloaders() -> Tuple[Dataset, Dataset, DataLoader, DataLoader]:\n",
" if CFG.use_train_chunk_cache:\n",
" train_cache_path = CFG.train_chunk_cache_path or default_chunk_cache_path(\"train\")\n",
" train_ds = CachedPunctDataset(\n",
" build_or_load_chunk_cache(\"train\", AVAILABLE_CORPORA, train_cache_path, CFG.train_cache_chunks)\n",
" )\n",
" else:\n",
" train_ds = StreamingGzipPunctDataset(AVAILABLE_CORPORA, CFG.shuffle_buffer_chunks)\n",
" val_ds = build_val_dataset(AVAILABLE_CORPORA)\n",
"\n",
" loader_kwargs = dict(\n",
" batch_size=CFG.batch_size,\n",
" num_workers=CFG.num_workers,\n",
" pin_memory=(DEVICE == \"cuda\"),\n",
" )\n",
" if CFG.num_workers > 0:\n",
" loader_kwargs.update(dict(persistent_workers=False, prefetch_factor=2))\n",
"\n",
" train_loader = DataLoader(\n",
" train_ds,\n",
" shuffle=isinstance(train_ds, CachedPunctDataset),\n",
" drop_last=True,\n",
" **loader_kwargs,\n",
" )\n",
" val_loader = DataLoader(val_ds, shuffle=False, drop_last=False, **loader_kwargs)\n",
" return train_ds, val_ds, train_loader, val_loader"
]
},
{
"cell_type": "markdown",
"id": "3d534deb",
"metadata": {
"id": "3d534deb"
},
"source": [
"## 10. Model"
]
},
{
"cell_type": "code",
"execution_count": 24,
"id": "8ceb6b96",
"metadata": {
"executionInfo": {
"elapsed": 54,
"status": "ok",
"timestamp": 1777280464378,
"user": {
"displayName": "kz",
"userId": "04068570917018885044"
},
"user_tz": -540
},
"id": "8ceb6b96"
},
"outputs": [],
"source": [
"\n",
"class SwiGLUFFN(nn.Module):\n",
" def __init__(self, hidden: int, ffn_hidden: int, dropout: float):\n",
" super().__init__()\n",
" self.w1 = nn.Linear(hidden, ffn_hidden, bias=True)\n",
" self.w2 = nn.Linear(hidden, ffn_hidden, bias=True)\n",
" self.w3 = nn.Linear(ffn_hidden, hidden, bias=True)\n",
" self.drop = nn.Dropout(dropout)\n",
"\n",
" def forward(self, x):\n",
" x = torch.nn.functional.silu(self.w1(x)) * self.w2(x)\n",
" x = self.drop(x)\n",
" return self.w3(x)\n",
"\n",
"class CustomEncoderLayer(nn.Module):\n",
" def __init__(\n",
" self,\n",
" hidden: int,\n",
" heads: int,\n",
" dropout: float,\n",
" ffn_type: str,\n",
" swiglu_hidden: int,\n",
" norm_first: bool,\n",
" ):\n",
" super().__init__()\n",
" self.norm_first = norm_first\n",
" self.self_attn = nn.MultiheadAttention(hidden, heads, dropout=dropout, batch_first=True)\n",
" self.attn_drop = nn.Dropout(dropout)\n",
" self.ffn_drop = nn.Dropout(dropout)\n",
" self.norm1 = nn.LayerNorm(hidden)\n",
" self.norm2 = nn.LayerNorm(hidden)\n",
"\n",
" if ffn_type == \"swiglu\":\n",
" self.ffn = SwiGLUFFN(hidden, swiglu_hidden, dropout)\n",
" elif ffn_type == \"gelu\":\n",
" self.ffn = nn.Sequential(\n",
" nn.Linear(hidden, hidden * 4),\n",
" nn.GELU(),\n",
" nn.Dropout(dropout),\n",
" nn.Linear(hidden * 4, hidden),\n",
" )\n",
" else:\n",
" raise ValueError(f\"Unsupported ffn_type: {ffn_type!r}\")\n",
"\n",
" def _sa_block(self, x, key_padding_mask):\n",
" x, _ = self.self_attn(\n",
" x,\n",
" x,\n",
" x,\n",
" key_padding_mask=key_padding_mask,\n",
" need_weights=False,\n",
" )\n",
" return self.attn_drop(x)\n",
"\n",
" def _ff_block(self, x):\n",
" return self.ffn_drop(self.ffn(x))\n",
"\n",
" def forward(self, x, key_padding_mask):\n",
" if self.norm_first:\n",
" x = x + self._sa_block(self.norm1(x), key_padding_mask)\n",
" x = x + self._ff_block(self.norm2(x))\n",
" else:\n",
" x = self.norm1(x + self._sa_block(x, key_padding_mask))\n",
" x = self.norm2(x + self._ff_block(x))\n",
" return x\n",
"\n",
"class CustomEncoder(nn.Module):\n",
" def __init__(\n",
" self,\n",
" hidden: int,\n",
" heads: int,\n",
" layers: int,\n",
" dropout: float,\n",
" ffn_type: str,\n",
" swiglu_hidden: int,\n",
" norm_first: bool,\n",
" ):\n",
" super().__init__()\n",
" self.layers = nn.ModuleList(\n",
" [\n",
" CustomEncoderLayer(hidden, heads, dropout, ffn_type, swiglu_hidden, norm_first)\n",
" for _ in range(layers)\n",
" ]\n",
" )\n",
" self.norm = nn.LayerNorm(hidden)\n",
"\n",
" def forward(self, x, src_key_padding_mask):\n",
" for layer in self.layers:\n",
" x = layer(x, src_key_padding_mask)\n",
" return self.norm(x)\n",
"\n",
"class CharTransformerPunct(nn.Module):\n",
" def __init__(\n",
" self,\n",
" vocab_size,\n",
" hidden=384,\n",
" heads=6,\n",
" layers=4,\n",
" max_len=192,\n",
" dropout=0.12,\n",
" num_labels=len(LABEL2ID),\n",
" ffn_type=\"gelu\",\n",
" swiglu_hidden=1024,\n",
" norm_first=True,\n",
" ):\n",
" super().__init__()\n",
" self.token_emb = nn.Embedding(vocab_size, hidden, padding_idx=PAD_ID)\n",
" self.pos_emb = nn.Embedding(max_len, hidden)\n",
" self.emb_norm = nn.LayerNorm(hidden)\n",
" self.drop = nn.Dropout(dropout)\n",
" self.ffn_type = ffn_type\n",
"\n",
" self.encoder = CustomEncoder(\n",
" hidden=hidden,\n",
" heads=heads,\n",
" layers=layers,\n",
" dropout=dropout,\n",
" ffn_type=ffn_type,\n",
" swiglu_hidden=swiglu_hidden,\n",
" norm_first=norm_first,\n",
" )\n",
"\n",
" self.punct_head = nn.Linear(hidden, num_labels)\n",
" self.newline_head = nn.Linear(hidden, 1)\n",
"\n",
" def forward(self, input_ids, attention_mask):\n",
" input_ids = input_ids.long()\n",
" bsz, seqlen = input_ids.shape\n",
" pos = torch.arange(seqlen, device=input_ids.device).unsqueeze(0).expand(bsz, seqlen)\n",
" x = self.token_emb(input_ids) + self.pos_emb(pos)\n",
" x = self.emb_norm(x)\n",
" x = self.drop(x)\n",
"\n",
" # TransformerEncoderLayer に渡す boolean mask\n",
" # (Trueの場所がパディングとして無視されます)\n",
" x = self.encoder(x, src_key_padding_mask=(~attention_mask))\n",
" return {\n",
" \"punct_logits\": self.punct_head(x),\n",
" \"newline_logits\": self.newline_head(x).squeeze(-1),\n",
" }\n",
"\n",
"def build_model():\n",
" model = CharTransformerPunct(\n",
" vocab_size=len(itos),\n",
" hidden=CFG.hidden,\n",
" heads=CFG.heads,\n",
" layers=CFG.layers,\n",
" max_len=CFG.max_len,\n",
" dropout=CFG.dropout,\n",
" num_labels=CFG.num_labels,\n",
" ffn_type=CFG.ffn_type,\n",
" swiglu_hidden=CFG.swiglu_hidden,\n",
" norm_first=CFG.norm_first,\n",
" ).to(DEVICE)\n",
" if CFG.use_torch_compile and hasattr(torch, \"compile\"):\n",
" model = torch.compile(model)\n",
" print(\"parameters:\", sum(p.numel() for p in model.parameters()))\n",
" return model\n",
"\n",
"model = None\n",
"if not CFG.auto_tune:\n",
" stoi, itos = build_vocab()\n",
" train_ds, val_ds, train_loader, val_loader = build_dataloaders()\n",
" model = build_model()"
]
},
{
"cell_type": "markdown",
"id": "8bdacfe7",
"metadata": {
"id": "8bdacfe7"
},
"source": [
"## 11. Loss / Optimizer / Scheduler"
]
},
{
"cell_type": "code",
"execution_count": 20,
"id": "9ebab425",
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"executionInfo": {
"elapsed": 222,
"status": "ok",
"timestamp": 1777267940421,
"user": {
"displayName": "kz",
"userId": "04068570917018885044"
},
"user_tz": -540
},
"id": "9ebab425",
"lines_to_next_cell": 1,
"outputId": "61b15062-8273-402c-d7ed-55700f78d32c"
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"class_weights: {'NONE': 0.3658536672592163, '、': 1.1707316637039185, '。': 1.4634146690368652}\n",
"start_epoch: 1 best_score: -1.0 history_rows: 0\n"
]
}
],
"source": [
"\n",
"weights = np.asarray(CFG.class_weights, dtype=np.float32)\n",
"weights = weights / weights.mean()\n",
"class_weights = torch.tensor(weights, dtype=torch.float32, device=DEVICE)\n",
"print(\"class_weights:\", {ID2LABEL[i]: float(class_weights[i].detach().cpu()) for i in range(CFG.num_labels)})\n",
"\n",
"punct_criterion = nn.CrossEntropyLoss(\n",
" weight=class_weights,\n",
" ignore_index=-100,\n",
" label_smoothing=CFG.label_smoothing,\n",
")\n",
"newline_criterion = nn.BCEWithLogitsLoss(\n",
" pos_weight=torch.tensor(CFG.newline_pos_weight, dtype=torch.float32, device=DEVICE),\n",
" reduction=\"none\",\n",
")\n",
"\n",
"def compute_loss(outputs, punct_labels, newline_labels):\n",
" punct_logits = outputs[\"punct_logits\"]\n",
" newline_logits = outputs[\"newline_logits\"]\n",
" punct_labels = punct_labels.long()\n",
" newline_labels = newline_labels.long()\n",
" punct_loss = punct_criterion(punct_logits.view(-1, CFG.num_labels), punct_labels.view(-1))\n",
"\n",
" newline_mask = newline_labels != -100\n",
" if newline_mask.any():\n",
" newline_loss_raw = newline_criterion(\n",
" newline_logits[newline_mask],\n",
" newline_labels[newline_mask].float(),\n",
" )\n",
" newline_loss = newline_loss_raw.mean()\n",
" else:\n",
" newline_loss = newline_logits.sum() * 0.0\n",
"\n",
" return punct_loss + CFG.newline_loss_weight * newline_loss, punct_loss, newline_loss\n",
"\n",
"def build_optimizer_scheduler_scaler(model):\n",
" try:\n",
" optimizer = torch.optim.AdamW(model.parameters(), lr=CFG.lr, weight_decay=CFG.weight_decay, fused=(DEVICE == \"cuda\"))\n",
" except TypeError:\n",
" optimizer = torch.optim.AdamW(model.parameters(), lr=CFG.lr, weight_decay=CFG.weight_decay)\n",
"\n",
" total_updates = max(1, CFG.steps_per_epoch * CFG.epochs)\n",
" warmup_updates = max(1, int(total_updates * CFG.warmup_ratio))\n",
"\n",
" def lr_lambda(step):\n",
" if step < warmup_updates:\n",
" return max(1e-8, step / warmup_updates)\n",
" progress = (step - warmup_updates) / max(1, total_updates - warmup_updates)\n",
" cosine = 0.5 * (1.0 + math.cos(math.pi * progress))\n",
" return CFG.min_lr_ratio + (1.0 - CFG.min_lr_ratio) * cosine\n",
"\n",
" scheduler = torch.optim.lr_scheduler.LambdaLR(optimizer, lr_lambda)\n",
" scaler = torch.amp.GradScaler(\"cuda\", enabled=(DEVICE == \"cuda\"))\n",
" return optimizer, scheduler, scaler\n",
"\n",
"optimizer = scheduler = scaler = None\n",
"if not CFG.auto_tune:\n",
" optimizer, scheduler, scaler = build_optimizer_scheduler_scaler(model)\n",
"\n",
"# Resume support. Old checkpoints that contain only model weights can still be used;\n",
"# newer checkpoints also restore optimizer/scheduler/scaler and continue at the next epoch.\n",
"def _strip_state_dict_prefixes(state_dict):\n",
" cleaned = {}\n",
" for key, value in state_dict.items():\n",
" for prefix in (\"module.\", \"_orig_mod.\"):\n",
" if key.startswith(prefix):\n",
" key = key[len(prefix):]\n",
" cleaned[key] = value\n",
" return cleaned\n",
"\n",
"\n",
"def find_resume_checkpoint() -> Optional[str]:\n",
" if not CFG.resume_from_checkpoint:\n",
" return None\n",
" candidates = []\n",
" if CFG.resume_ckpt_path:\n",
" candidates.append(CFG.resume_ckpt_path)\n",
" candidates.extend(CFG.resume_ckpt_candidates)\n",
" for candidate in candidates:\n",
" if candidate and os.path.exists(candidate):\n",
" return candidate\n",
" return None\n",
"\n",
"\n",
"def torch_load_checkpoint(path: str):\n",
" try:\n",
" return torch.load(path, map_location=DEVICE, weights_only=False)\n",
" except TypeError:\n",
" return torch.load(path, map_location=DEVICE)\n",
"\n",
"\n",
"def load_training_checkpoint(path: str):\n",
" print(\"resume checkpoint:\", path)\n",
" ckpt = torch_load_checkpoint(path)\n",
" state_dict = ckpt.get(\"model\", ckpt)\n",
" model_to_load = model.module if hasattr(model, \"module\") else model\n",
" missing, unexpected = model_to_load.load_state_dict(_strip_state_dict_prefixes(state_dict), strict=False)\n",
" if missing or unexpected:\n",
" print(\"checkpoint load notes:\", {\"missing\": missing, \"unexpected\": unexpected})\n",
"\n",
" if \"optimizer\" in ckpt:\n",
" try:\n",
" optimizer.load_state_dict(ckpt[\"optimizer\"])\n",
" print(\"optimizer restored\")\n",
" except ValueError as e:\n",
" print(\"optimizer state not restored; architecture changed:\", repr(e))\n",
" else:\n",
" print(\"optimizer state not found; continuing with a fresh optimizer\")\n",
"\n",
" if \"scheduler\" in ckpt:\n",
" try:\n",
" scheduler.load_state_dict(ckpt[\"scheduler\"])\n",
" print(\"scheduler restored\")\n",
" except Exception as e:\n",
" print(\"scheduler state not restored; architecture changed:\", repr(e))\n",
" else:\n",
" print(\"scheduler state not found; continuing with a fresh scheduler\")\n",
"\n",
" if \"scaler\" in ckpt and DEVICE == \"cuda\":\n",
" scaler.load_state_dict(ckpt[\"scaler\"])\n",
" print(\"AMP scaler restored\")\n",
"\n",
" resume_history = list(ckpt.get(\"history\", []))\n",
" if \"epoch\" in ckpt:\n",
" completed_epoch = int(ckpt[\"epoch\"])\n",
" elif resume_history:\n",
" completed_epoch = max(int(row.get(\"epoch\", idx + 1)) for idx, row in enumerate(resume_history))\n",
" print(\"epoch key not found; inferred completed_epoch from history:\", completed_epoch)\n",
" else:\n",
" completed_epoch = 0\n",
"\n",
" resume_epoch = completed_epoch + 1\n",
" resume_best_score = float(ckpt.get(\"best_score\", -1.0))\n",
" if resume_best_score < 0.0 and resume_history:\n",
" resume_best_score = max(float(row.get(\"score\", -1.0)) for row in resume_history)\n",
" return resume_epoch, resume_best_score, resume_history\n",
"\n",
"resume_ckpt = find_resume_checkpoint() if not CFG.auto_tune else None\n",
"if resume_ckpt:\n",
" start_epoch, best_score, history = load_training_checkpoint(resume_ckpt)\n",
"else:\n",
" start_epoch, best_score, history = 1, -1.0, []\n",
"\n",
"print(\"start_epoch:\", start_epoch, \"best_score:\", best_score, \"history_rows:\", len(history))"
]
},
{
"cell_type": "markdown",
"id": "279bed0e",
"metadata": {
"id": "279bed0e"
},
"source": [
"## 12. Evaluation"
]
},
{
"cell_type": "code",
"execution_count": 26,
"id": "c9aef193",
"metadata": {
"executionInfo": {
"elapsed": 56,
"status": "ok",
"timestamp": 1777280630563,
"user": {
"displayName": "kz",
"userId": "04068570917018885044"
},
"user_tz": -540
},
"id": "c9aef193"
},
"outputs": [],
"source": [
"\n",
"@torch.no_grad()\n",
"def evaluate(model, loader):\n",
" model.eval()\n",
" total_loss = 0.0\n",
" total_punct_loss = 0.0\n",
" total_newline_loss = 0.0\n",
" total_batches = 0\n",
" punct_ys = []\n",
" punct_preds = []\n",
" newline_ys = []\n",
" newline_preds = []\n",
"\n",
" for batch in tqdm(loader, desc=\"eval\", leave=False):\n",
" input_ids = batch[\"input_ids\"].to(DEVICE, non_blocking=True)\n",
" attention_mask = batch[\"attention_mask\"].to(DEVICE, non_blocking=True)\n",
" punct_labels = batch[\"punct_labels\"].to(DEVICE, non_blocking=True)\n",
" newline_labels = batch[\"newline_labels\"].to(DEVICE, non_blocking=True)\n",
"\n",
" with torch.amp.autocast(\"cuda\", enabled=(DEVICE == \"cuda\")):\n",
" outputs = model(input_ids, attention_mask)\n",
" loss, punct_loss, newline_loss = compute_loss(outputs, punct_labels, newline_labels)\n",
"\n",
" total_loss += loss.item()\n",
" total_punct_loss += punct_loss.item()\n",
" total_newline_loss += newline_loss.item()\n",
" total_batches += 1\n",
" punct_pred = outputs[\"punct_logits\"].argmax(dim=-1)\n",
" punct_mask = punct_labels != -100\n",
" punct_ys.append(punct_labels[punct_mask].detach().cpu().numpy())\n",
" punct_preds.append(punct_pred[punct_mask].detach().cpu().numpy())\n",
"\n",
" newline_pred = (torch.sigmoid(outputs[\"newline_logits\"]) >= 0.5).long()\n",
" newline_mask = newline_labels != -100\n",
" newline_ys.append(newline_labels[newline_mask].detach().cpu().numpy())\n",
" newline_preds.append(newline_pred[newline_mask].detach().cpu().numpy())\n",
"\n",
" y_true = np.concatenate(punct_ys)\n",
" y_pred = np.concatenate(punct_preds)\n",
" newline_true = np.concatenate(newline_ys)\n",
" newline_pred = np.concatenate(newline_preds)\n",
"\n",
" label_ids = list(range(CFG.num_labels))\n",
" target_names = [ID2LABEL[i] for i in label_ids]\n",
" report = classification_report(\n",
" y_true,\n",
" y_pred,\n",
" labels=label_ids,\n",
" target_names=target_names,\n",
" zero_division=0,\n",
" )\n",
"\n",
" punctuation_label_ids = [\n",
" LABEL2ID[\"、\"], LABEL2ID[\"。\"],\n",
" ]\n",
"\n",
" def binary_f1(true_ids, pred_ids, positive_ids):\n",
" _, _, f1, _ = precision_recall_fscore_support(\n",
" np.isin(true_ids, positive_ids).astype(np.int32),\n",
" np.isin(pred_ids, positive_ids).astype(np.int32),\n",
" labels=[1],\n",
" average=None,\n",
" zero_division=0,\n",
" )\n",
" return float(f1[0])\n",
"\n",
" punct_f1 = binary_f1(y_true, y_pred, punctuation_label_ids)\n",
" comma_f1 = binary_f1(y_true, y_pred, [LABEL2ID[\"、\"]])\n",
" period_f1 = binary_f1(y_true, y_pred, [LABEL2ID[\"。\"]])\n",
" newline_f1 = binary_f1(newline_true, newline_pred, [1])\n",
" punct_macro_f1 = float((comma_f1 + period_f1) / 2)\n",
" score = float(0.4 * comma_f1 + 0.4 * period_f1 + 0.2 * newline_f1)\n",
"\n",
" return {\n",
" \"loss\": total_loss / max(1, total_batches),\n",
" \"punct_loss\": total_punct_loss / max(1, total_batches),\n",
" \"newline_loss\": total_newline_loss / max(1, total_batches),\n",
" \"punct_binary_f1\": punct_f1,\n",
" \"comma_f1\": comma_f1,\n",
" \"period_f1\": period_f1,\n",
" \"punct_macro_f1\": punct_macro_f1,\n",
" \"newline_f1\": newline_f1,\n",
" \"newline_binary_f1\": newline_f1,\n",
" \"score\": score,\n",
" \"report\": report,\n",
" }\n",
"\n",
"@torch.no_grad()\n",
"def search_thresholds(model, loader):\n",
" model.eval()\n",
" punct_logits_all = []\n",
" punct_labels_all = []\n",
" newline_logits_all = []\n",
" newline_labels_all = []\n",
"\n",
" for batch in tqdm(loader, desc=\"threshold search\", leave=False):\n",
" input_ids = batch[\"input_ids\"].to(DEVICE, non_blocking=True)\n",
" attention_mask = batch[\"attention_mask\"].to(DEVICE, non_blocking=True)\n",
" punct_labels = batch[\"punct_labels\"].to(DEVICE, non_blocking=True)\n",
" newline_labels = batch[\"newline_labels\"].to(DEVICE, non_blocking=True)\n",
"\n",
" with torch.amp.autocast(\"cuda\", enabled=(DEVICE == \"cuda\")):\n",
" outputs = model(input_ids, attention_mask)\n",
"\n",
" punct_mask = punct_labels != -100\n",
" newline_mask = newline_labels != -100\n",
" punct_logits_all.append(outputs[\"punct_logits\"][punct_mask].float().detach().cpu())\n",
" punct_labels_all.append(punct_labels[punct_mask].long().detach().cpu())\n",
" newline_logits_all.append(outputs[\"newline_logits\"][newline_mask].float().detach().cpu())\n",
" newline_labels_all.append(newline_labels[newline_mask].long().detach().cpu())\n",
"\n",
" punct_logits = torch.cat(punct_logits_all, dim=0).numpy()\n",
" punct_true = torch.cat(punct_labels_all, dim=0).numpy()\n",
" newline_logits = torch.cat(newline_logits_all, dim=0).numpy()\n",
" newline_true = torch.cat(newline_labels_all, dim=0).numpy()\n",
"\n",
" def binary_f1(true_ids, pred_ids, positive_ids):\n",
" _, _, f1, _ = precision_recall_fscore_support(\n",
" np.isin(true_ids, positive_ids).astype(np.int32),\n",
" np.isin(pred_ids, positive_ids).astype(np.int32),\n",
" labels=[1],\n",
" average=None,\n",
" zero_division=0,\n",
" )\n",
" return float(f1[0])\n",
"\n",
" best = None\n",
" none_id = LABEL2ID[\"NONE\"]\n",
" for punct_margin in CFG.punct_margin_candidates:\n",
" raw_best = punct_logits.argmax(axis=-1)\n",
" best_logits = punct_logits[np.arange(len(raw_best)), raw_best]\n",
" none_logits = punct_logits[:, none_id]\n",
" punct_pred = np.where(\n",
" (raw_best != none_id) & ((best_logits - none_logits) >= punct_margin),\n",
" raw_best,\n",
" none_id,\n",
" )\n",
" comma_f1 = binary_f1(punct_true, punct_pred, [LABEL2ID[\"、\"]])\n",
" period_f1 = binary_f1(punct_true, punct_pred, [LABEL2ID[\"。\"]])\n",
" punct_macro_f1 = float((comma_f1 + period_f1) / 2)\n",
"\n",
" for newline_threshold in CFG.newline_threshold_candidates:\n",
" newline_pred = (1.0 / (1.0 + np.exp(-newline_logits)) >= newline_threshold).astype(np.int32)\n",
" newline_f1 = binary_f1(newline_true, newline_pred, [1])\n",
" score = float(0.4 * comma_f1 + 0.4 * period_f1 + 0.2 * newline_f1)\n",
" row = {\n",
" \"newline_threshold\": float(newline_threshold),\n",
" \"punct_margin\": float(punct_margin),\n",
" \"comma_f1\": comma_f1,\n",
" \"period_f1\": period_f1,\n",
" \"punct_macro_f1\": punct_macro_f1,\n",
" \"newline_f1\": newline_f1,\n",
" \"score\": score,\n",
" }\n",
" if best is None or row[\"score\"] > best[\"score\"]:\n",
" best = row\n",
"\n",
" return best\n",
"\n",
"def train_for_epochs(model, train_ds, train_loader, val_loader, save_best: bool):\n",
" optimizer, scheduler, scaler = build_optimizer_scheduler_scaler(model)\n",
" proc = psutil.Process(os.getpid())\n",
" local_history = []\n",
" local_best_score = -1.0\n",
" local_best_metrics = None\n",
" local_best_epoch = 0\n",
"\n",
" for epoch in range(1, CFG.epochs + 1):\n",
" if hasattr(train_ds, \"set_epoch\"):\n",
" train_ds.set_epoch(epoch)\n",
" model.train()\n",
" t0 = time.perf_counter()\n",
" running = 0.0\n",
" seen_tokens = 0\n",
"\n",
" pbar = tqdm(enumerate(train_loader, start=1), total=CFG.steps_per_epoch, desc=f\"{CFG.experiment_name} epoch {epoch}\")\n",
" for step, batch in pbar:\n",
" if step > CFG.steps_per_epoch:\n",
" break\n",
"\n",
" input_ids = batch[\"input_ids\"].to(DEVICE, non_blocking=True)\n",
" attention_mask = batch[\"attention_mask\"].to(DEVICE, non_blocking=True)\n",
" punct_labels = batch[\"punct_labels\"].to(DEVICE, non_blocking=True)\n",
" newline_labels = batch[\"newline_labels\"].to(DEVICE, non_blocking=True)\n",
"\n",
" optimizer.zero_grad(set_to_none=True)\n",
" with torch.amp.autocast(\"cuda\", enabled=(DEVICE == \"cuda\")):\n",
" outputs = model(input_ids, attention_mask)\n",
" loss, punct_loss, newline_loss = compute_loss(outputs, punct_labels, newline_labels)\n",
"\n",
" scaler.scale(loss).backward()\n",
" scaler.unscale_(optimizer)\n",
" torch.nn.utils.clip_grad_norm_(model.parameters(), CFG.grad_clip)\n",
" scaler.step(optimizer)\n",
" scaler.update()\n",
" scheduler.step()\n",
"\n",
" running += loss.item()\n",
" seen_tokens += int(attention_mask.sum().detach().cpu())\n",
"\n",
" if step % 25 == 0:\n",
" elapsed = max(1e-6, time.perf_counter() - t0)\n",
" postfix = {\n",
" \"loss\": f\"{running / 25:.4f}\",\n",
" \"p_loss\": f\"{punct_loss.item():.4f}\",\n",
" \"n_loss\": f\"{newline_loss.item():.4f}\",\n",
" \"lr\": f\"{scheduler.get_last_lr()[0]:.2e}\",\n",
" \"tok/s\": f\"{seen_tokens / elapsed:,.0f}\",\n",
" \"rssGB\": f\"{proc.memory_info().rss / 1024**3:.2f}\",\n",
" }\n",
" if DEVICE == \"cuda\":\n",
" postfix[\"vramGB\"] = f\"{torch.cuda.max_memory_allocated() / 1024**3:.2f}\"\n",
" pbar.set_postfix(postfix)\n",
" running = 0.0\n",
"\n",
" metrics = evaluate(model, val_loader)\n",
" threshold_metrics = search_thresholds(model, val_loader)\n",
" if threshold_metrics:\n",
" CFG.newline_threshold = threshold_metrics[\"newline_threshold\"]\n",
" CFG.punct_margin = threshold_metrics[\"punct_margin\"]\n",
" metrics[\"threshold_search\"] = threshold_metrics\n",
" metrics[\"score\"] = threshold_metrics[\"score\"]\n",
" epoch_time = time.perf_counter() - t0\n",
" row = {\"epoch\": epoch, \"time_sec\": epoch_time, **metrics}\n",
" local_history.append({k: v for k, v in row.items() if k != \"report\"})\n",
"\n",
" print(\n",
" f\"epoch {epoch} val_loss={metrics['loss']:.4f} \"\n",
" f\"comma_f1={metrics['comma_f1']:.4f} \"\n",
" f\"period_f1={metrics['period_f1']:.4f} \"\n",
" f\"punct_macro_f1={metrics['punct_macro_f1']:.4f} \"\n",
" f\"newline_f1={metrics['newline_f1']:.4f} \"\n",
" f\"score={metrics['score']:.4f} \"\n",
" f\"thr=({CFG.newline_threshold:.2f}, margin={CFG.punct_margin:.2f}) \"\n",
" f\"time={epoch_time:.1f}s\"\n",
" )\n",
"\n",
" if metrics[\"score\"] > local_best_score:\n",
" local_best_score = metrics[\"score\"]\n",
" local_best_metrics = metrics\n",
" local_best_epoch = epoch\n",
" if save_best:\n",
" save_obj = {\n",
" \"model\": model.module.state_dict() if hasattr(model, \"module\") else model.state_dict(),\n",
" \"cfg\": asdict(CFG),\n",
" \"label2id\": LABEL2ID,\n",
" \"id2label\": ID2LABEL,\n",
" \"heads\": {\"punct\": len(LABEL2ID), \"newline\": 1},\n",
" \"history\": local_history,\n",
" \"corpora\": [{k: v for k, v in c.items() if k != \"gitlab_paths\"} | {\"num_files\": len(c[\"gitlab_paths\"])} for c in AVAILABLE_CORPORA],\n",
" \"itos\": itos,\n",
" \"vocab_size\": len(itos),\n",
" \"inference\": {\n",
" \"mode\": CFG.inference_mode,\n",
" \"newline_threshold\": CFG.newline_threshold,\n",
" \"punct_margin\": CFG.punct_margin,\n",
" },\n",
" \"epoch\": epoch,\n",
" \"best_score\": local_best_score,\n",
" \"optimizer\": optimizer.state_dict(),\n",
" \"scheduler\": scheduler.state_dict(),\n",
" \"scaler\": scaler.state_dict(),\n",
" }\n",
" torch.save(save_obj, CFG.ckpt_path)\n",
" with open(CFG.bundle_path, \"w\", encoding=\"utf-8\") as f:\n",
" json.dump({k: v for k, v in save_obj.items() if k != \"model\"}, f, ensure_ascii=False, indent=2, default=str)\n",
" print(\"saved:\", CFG.ckpt_path, \"best_score\", local_best_score)\n",
"\n",
" return local_best_epoch, local_best_score, local_best_metrics, local_history\n",
"\n",
"def apply_auto_tune_candidate(candidate: Dict[str, object]) -> None:\n",
" CFG.vocab_max_size = int(candidate[\"vocab_max_size\"])\n",
" CFG.ffn_type = str(candidate[\"ffn_type\"])\n",
" CFG.swiglu_hidden = int(candidate.get(\"swiglu_hidden\", CFG.swiglu_hidden))\n",
" CFG.experiment_name = str(candidate[\"name\"])\n",
" CFG.max_files_per_corpus = CFG.auto_tune_max_files_per_corpus\n",
" CFG.steps_per_epoch = CFG.auto_tune_steps_per_epoch\n",
" CFG.epochs = CFG.auto_tune_epochs\n",
" CFG.val_chunks = CFG.auto_tune_val_chunks\n",
" CFG.train_cache_chunks = CFG.auto_tune_train_cache_chunks\n",
" CFG.resume_from_checkpoint = False\n",
" CFG.train_chunk_cache_path = None\n",
" CFG.val_chunk_cache_path = None\n",
" CFG.ckpt_path = os.path.join(CFG.work_dir, f\"{CFG.experiment_name}.pt\")\n",
" CFG.bundle_path = os.path.join(CFG.work_dir, f\"{CFG.experiment_name}_bundle.json\")\n",
"\n",
"def candidate_summary(candidate: Dict[str, object], best_epoch: int, best_score: float, metrics: Dict, history_rows: List[Dict]) -> Dict:\n",
" threshold = metrics.get(\"threshold_search\", {}) if metrics else {}\n",
" return {\n",
" \"name\": str(candidate[\"name\"]),\n",
" \"vocab_max_size\": int(candidate[\"vocab_max_size\"]),\n",
" \"actual_vocab_size\": len(itos),\n",
" \"ffn_type\": str(candidate[\"ffn_type\"]),\n",
" \"swiglu_hidden\": int(candidate.get(\"swiglu_hidden\", CFG.swiglu_hidden)),\n",
" \"best_epoch\": int(best_epoch),\n",
" \"best_score\": float(best_score),\n",
" \"comma_f1\": float(metrics[\"comma_f1\"]),\n",
" \"period_f1\": float(metrics[\"period_f1\"]),\n",
" \"punct_macro_f1\": float(metrics[\"punct_macro_f1\"]),\n",
" \"newline_f1\": float(metrics[\"newline_f1\"]),\n",
" \"newline_threshold\": float(threshold.get(\"newline_threshold\", CFG.newline_threshold)),\n",
" \"punct_margin\": float(threshold.get(\"punct_margin\", CFG.punct_margin)),\n",
" \"history\": history_rows,\n",
" }\n",
"\n",
"def select_auto_tune_config(results: List[Dict]) -> Dict:\n",
" by_name = {row[\"name\"]: row for row in results}\n",
" baseline = by_name.get(\"A1_vocab16000_gelu\")\n",
" if baseline is None:\n",
" baseline = max(results, key=lambda row: row[\"punct_macro_f1\"])\n",
"\n",
" punct_floor = baseline[\"punct_macro_f1\"] - 0.003\n",
" acceptable = [row for row in results if row[\"punct_macro_f1\"] >= punct_floor]\n",
" if not acceptable:\n",
" acceptable = [baseline]\n",
"\n",
" best = max(\n",
" acceptable,\n",
" key=lambda row: (\n",
" row[\"best_score\"],\n",
" -row[\"vocab_max_size\"],\n",
" 1 if row[\"ffn_type\"] == \"swiglu\" else 0,\n",
" ),\n",
" )\n",
"\n",
" return {\n",
" \"selected\": {\n",
" \"vocab_max_size\": best[\"vocab_max_size\"],\n",
" \"ffn_type\": best[\"ffn_type\"],\n",
" \"swiglu_hidden\": best[\"swiglu_hidden\"],\n",
" \"newline_threshold\": best[\"newline_threshold\"],\n",
" \"punct_margin\": best[\"punct_margin\"],\n",
" \"source_candidate\": best[\"name\"],\n",
" },\n",
" \"baseline_candidate\": baseline[\"name\"],\n",
" \"punct_macro_floor\": float(punct_floor),\n",
" \"selection_rule\": \"highest score among candidates whose punct_macro_f1 is within 0.003 of A1 baseline; ties prefer smaller vocab and SwiGLU\",\n",
" }\n",
"\n",
"def run_auto_tune() -> None:\n",
" global stoi, itos\n",
"\n",
" base_candidates = [dict(x) for x in CFG.auto_tune_candidates]\n",
" pending = list(base_candidates)\n",
" results: List[Dict] = []\n",
" ran_names = set()\n",
"\n",
" while pending:\n",
" candidate = pending.pop(0)\n",
" if candidate[\"name\"] in ran_names:\n",
" continue\n",
" ran_names.add(candidate[\"name\"])\n",
" print(\"\\n===== AUTO TUNE\", candidate[\"name\"], \"=====\")\n",
" apply_auto_tune_candidate(candidate)\n",
" stoi, itos = build_vocab()\n",
" train_ds, val_ds, train_loader, val_loader = build_dataloaders()\n",
" model = build_model()\n",
" best_epoch, best_score, best_metrics, history_rows = train_for_epochs(\n",
" model,\n",
" train_ds,\n",
" train_loader,\n",
" val_loader,\n",
" save_best=False,\n",
" )\n",
" row = candidate_summary(candidate, best_epoch, best_score, best_metrics, history_rows)\n",
" results.append(row)\n",
" print(\"AUTO TUNE RESULT:\", json.dumps({k: v for k, v in row.items() if k != \"history\"}, ensure_ascii=False, indent=2))\n",
"\n",
" baseline = next((x for x in results if x[\"name\"] == \"A1_vocab16000_gelu\"), None)\n",
" vocab8000 = next((x for x in results if x[\"name\"] == \"A3_vocab8000_gelu\"), None)\n",
" if baseline and vocab8000 and vocab8000[\"punct_macro_f1\"] >= baseline[\"punct_macro_f1\"] - 0.003:\n",
" pending.append(\n",
" {\n",
" \"name\": \"A5_vocab8000_swiglu\",\n",
" \"vocab_max_size\": 8_000,\n",
" \"ffn_type\": \"swiglu\",\n",
" \"swiglu_hidden\": 1_024,\n",
" }\n",
" )\n",
"\n",
" del model, train_loader, val_loader, train_ds, val_ds\n",
" gc.collect()\n",
" if DEVICE == \"cuda\":\n",
" torch.cuda.empty_cache()\n",
"\n",
" results_path = os.path.join(CFG.work_dir, CFG.auto_tune_results_path)\n",
" with open(results_path, \"w\", encoding=\"utf-8\") as f:\n",
" json.dump(results, f, ensure_ascii=False, indent=2, default=str)\n",
" print(\"auto tune partial results:\", results_path)\n",
"\n",
" selection = select_auto_tune_config(results)\n",
" out = {\"results\": results, **selection}\n",
" results_path = os.path.join(CFG.work_dir, CFG.auto_tune_results_path)\n",
" select_path = os.path.join(CFG.work_dir, CFG.auto_tune_select_path)\n",
" with open(results_path, \"w\", encoding=\"utf-8\") as f:\n",
" json.dump(out, f, ensure_ascii=False, indent=2, default=str)\n",
" with open(select_path, \"w\", encoding=\"utf-8\") as f:\n",
" json.dump(selection, f, ensure_ascii=False, indent=2, default=str)\n",
" print(\"auto tune results:\", results_path)\n",
" print(\"auto tune selected:\", select_path)\n",
" print(\"SELECTED CONFIG:\", json.dumps(selection[\"selected\"], ensure_ascii=False, indent=2))\n",
"\n",
"#if CFG.auto_tune:\n",
"# run_auto_tune()\n",
"# raise SystemExit(\"auto_tune completed\")"
]
},
{
"cell_type": "markdown",
"id": "0b2be684",
"metadata": {
"id": "0b2be684"
},
"source": [
"## 13. Train"
]
},
{
"cell_type": "code",
"execution_count": 35,
"id": "gwEngYSgLlir",
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/",
"height": 408
},
"executionInfo": {
"elapsed": 114562,
"status": "error",
"timestamp": 1777286360202,
"user": {
"displayName": "kz",
"userId": "04068570917018885044"
},
"user_tz": -540
},
"id": "gwEngYSgLlir",
"outputId": "ed79e63c-4249-4d29-8f2d-5c7c621902b5"
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Configuration updated to A2 selection.\n",
"vocab scan: ja_wiki\n",
"vocab scan: ja_kokkai_giji\n"
]
},
{
"ename": "KeyboardInterrupt",
"evalue": "",
"output_type": "error",
"traceback": [
"\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
"\u001b[0;31mKeyboardInterrupt\u001b[0m Traceback (most recent call last)",
"\u001b[0;32m/tmp/ipykernel_12856/1266080359.py\u001b[0m in \u001b[0;36m<cell line: 0>\u001b[0;34m()\u001b[0m\n\u001b[1;32m 10\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 11\u001b[0m \u001b[0;31m# Re-initialize vocab and data loaders based on updated CFG\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 12\u001b[0;31m \u001b[0mstoi\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mitos\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mbuild_vocab\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 13\u001b[0m \u001b[0mtrain_ds\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mval_ds\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtrain_loader\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mval_loader\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mbuild_dataloaders\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 14\u001b[0m \u001b[0mmodel\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mbuild_model\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
"\u001b[0;32m/tmp/ipykernel_12856/2069738902.py\u001b[0m in \u001b[0;36mbuild_vocab\u001b[0;34m()\u001b[0m\n\u001b[1;32m 64\u001b[0m \u001b[0;32mtry\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 65\u001b[0m \u001b[0;32mfor\u001b[0m \u001b[0mraw_text\u001b[0m \u001b[0;32min\u001b[0m \u001b[0miter_jsonl_gz_texts_from_gitlab_path\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mgitlab_path\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 66\u001b[0;31m \u001b[0mtext\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mclean_text\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mraw_text\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 67\u001b[0m \u001b[0;32mif\u001b[0m \u001b[0;32mnot\u001b[0m \u001b[0mtext\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 68\u001b[0m \u001b[0;32mcontinue\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
"\u001b[0;32m/tmp/ipykernel_12856/759019609.py\u001b[0m in \u001b[0;36mclean_text\u001b[0;34m(raw_text)\u001b[0m\n\u001b[1;32m 111\u001b[0m \u001b[0;32mcontinue\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 112\u001b[0m \u001b[0;32mfor\u001b[0m \u001b[0msegment\u001b[0m \u001b[0;32min\u001b[0m \u001b[0msplit_long_line\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mline\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 113\u001b[0;31m \u001b[0msegment\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mnormalize_line\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0msegment\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 114\u001b[0m \u001b[0;32mif\u001b[0m \u001b[0mlooks_like_noise_line\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0msegment\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 115\u001b[0m \u001b[0;32mcontinue\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
"\u001b[0;32m/tmp/ipykernel_12856/759019609.py\u001b[0m in \u001b[0;36mnormalize_line\u001b[0;34m(line)\u001b[0m\n\u001b[1;32m 46\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 47\u001b[0m \u001b[0;32mdef\u001b[0m \u001b[0mnormalize_line\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mline\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0mstr\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;34m->\u001b[0m \u001b[0mstr\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 48\u001b[0;31m \u001b[0mline\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mline\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtranslate\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mTRANSLATION\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 49\u001b[0m \u001b[0mline\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mline\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mreplace\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m\"\\u3000\"\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m\" \"\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mreplace\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m\"\\t\"\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m\" \"\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 50\u001b[0m \u001b[0mline\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mre\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msub\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34mr\"\\s+\"\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m\" \"\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mline\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
"\u001b[0;31mKeyboardInterrupt\u001b[0m: "
]
}
],
"source": [
"# Apply selected auto-tune configuration\n",
"CFG.vocab_max_size = 12000\n",
"CFG.ffn_type = \"gelu\"\n",
"CFG.swiglu_hidden = 1024\n",
"CFG.newline_threshold = 0.3\n",
"CFG.punct_margin = 0.0\n",
"CFG.experiment_name = \"A2_vocab12000_gelu\"\n",
"\n",
"print(\"Configuration updated to A2 selection.\")\n",
"\n",
"# Re-initialize vocab and data loaders based on updated CFG\n",
"stoi, itos = build_vocab()\n",
"train_ds, val_ds, train_loader, val_loader = build_dataloaders()\n",
"model = build_model()\n",
"\n",
"# Re-initialize optimizer, scheduler, and scaler\n",
"optimizer, scheduler, scaler = build_optimizer_scheduler_scaler(model)"
]
},
{
"cell_type": "code",
"execution_count": 30,
"id": "4R3YPijCW0yn",
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"executionInfo": {
"elapsed": 75,
"status": "ok",
"timestamp": 1777283594960,
"user": {
"displayName": "kz",
"userId": "04068570917018885044"
},
"user_tz": -540
},
"id": "4R3YPijCW0yn",
"outputId": "0b053c2c-31c5-42da-a9e1-b140a1e990f3"
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"parameters: 10369668\n"
]
}
],
"source": [
"# Apply selected auto-tune configuration\n",
"CFG.vocab_max_size = 12000\n",
"CFG.ffn_type = \"swiglu\"\n",
"CFG.swiglu_hidden = 1024\n",
"CFG.newline_threshold = 0.3\n",
"CFG.punct_margin = 0.0\n",
"CFG.experiment_name = \"A2_vocab12000_swiglu\"\n",
"model = build_model()\n",
"\n",
"# Re-initialize optimizer, scheduler, and scaler\n",
"optimizer, scheduler, scaler = build_optimizer_scheduler_scaler(model)"
]
},
{
"cell_type": "code",
"execution_count": 32,
"id": "PQJyNzbZXDPb",
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"executionInfo": {
"elapsed": 12,
"status": "ok",
"timestamp": 1777283664998,
"user": {
"displayName": "kz",
"userId": "04068570917018885044"
},
"user_tz": -540
},
"id": "PQJyNzbZXDPb",
"outputId": "a325af00-49ae-4ad1-facd-5a8b067b206b"
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Reverted to default: epochs=5, steps_per_epoch=2000\n"
]
}
],
"source": [
"# Revert training duration parameters to default\n",
"CFG.epochs = 5\n",
"CFG.steps_per_epoch = 2000\n",
"CFG.max_files_per_corpus = 32\n",
"CFG.train_cache_chunks = 1000000\n",
"\n",
"print(f\"Reverted to default: epochs={CFG.epochs}, steps_per_epoch={CFG.steps_per_epoch}\")"
]
},
{
"cell_type": "code",
"execution_count": 33,
"id": "2e28eb0f",
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/",
"height": 1000,
"referenced_widgets": [
"a0f5901d9d494ce4b2c3c3c1fcabd06e",
"09f8dcbd7cc544108e43fead639ce277",
"fd81a797f70a4e5a9043750586774fed",
"554a8cfc16b647a7b6e68a43bb13ef4a",
"7b95b36d83c44f5cb6d270aaab448883",
"a45644ce46e44a7abc1cd9d701ae66fc",
"b5033174abbc4eb8871ecaa764231ba7",
"aec866f26d7147c6ad2c9175a3d2f076",
"70e92fcc70e34d969ac7722f9ec0882b",
"21e39a7f2fc641f5923f096425ba5fbd",
"ef7995f9babf447f959df314b89d54a2",
"4776891f42a84f4aafbdbe3f74cd66a2",
"7b7cf83dbd214c6ca29340a376ba0188",
"2137415ed4064c1bb1c69856af0c22b7",
"041a636352994ebc870fd076fc6ccd69",
"502cb323941b4fa48995979b96e169d3",
"7d4b47b3692142279f7082bc7e3a62df",
"45586a223595426990ea412755f7a7c7",
"2e5bebd32d994437a75aab26549aaa64",
"e1cb8d90b9d841c4adb5d2696208c169",
"fdac6f566099458fa038d11e7aa44452",
"1e288dfa575c41f198ca0cb09235a721",
"180f3d66814d4888a7f1cfa5d5efd683",
"aab889a6f5114d15a1572b4a96de6434",
"08a1c054d81d4ed08147017c3026fa77",
"49b8fe32e5f84104a9fcb9e32ac1ee21",
"a42f75516bbb47ed9290bb4a3c75379c",
"39d94eae6b534d04946f534afc3250ab",
"1566aaff7a8c4e119103f8f76b4afb70",
"8170a2dba6594d7db272ad52436d5a37",
"19421bf7b12c40d6b788792e2914c853",
"fd0343bc7fbc4c9ea14fefac4f7c4d05",
"8f5d53603f784d19a8756b92a13a8329",
"cd3f6fff352a45fd9aa853e95cd2706c",
"a749690d6e5e48d4b9d3314e73830487",
"f49fc60609dc40dd8bc7f82380f70b5a",
"0b316a18319644108a1b2f70a5d7fe42",
"9f7eabb7ee5d42bd84c1c58f77c8a300",
"ea346f402476488fb4e199783a6ab1e0",
"e08bfb41a114424f94565f0807c43db4",
"8a0aca6587d34623b6ef752c28ac43bc",
"27a28b493e2c4942b094dde8151623f4",
"26dd110c6ef1424e983a65e551ce366d",
"8132e5f353f1440d808373a2dd6eba00",
"5ce1a0cd8c7841649d5a82b6712aa20e",
"3b81f0be353e44e4b01056e9822f009d",
"5c39a8db5353446f9563543e08fa0e64",
"c0b0d68c39db415abe7b6c92e250c451",
"4f0ba045438d47d1a39190b8478053ac",
"fefc2eba241e4adfa56ced16e098b6bf",
"6160e0acd5bd405ebf8954b9a4319c56",
"dd3eccb6c46348a69d10948abbf0022d",
"292ea043876d42b2873940f59432731b",
"aea1966bdb2143cca681bf847aab023e",
"8aa3ce282f9241639ce22de6b68bef3c",
"627850b0b6974b8993680eda6227f828",
"45c39f21acff4d78a956956c0984a16e",
"ec5985dd680e432f85bffd2b8e1f1c3c",
"7ab1270e0f0e4af8b1a54353a77ce20d",
"0484474d43a34dbabf45fa11b3391881",
"e7e889ec45d64224ab02f5bbe75a1e03",
"6483fb2129c74784b63fd5e955766161",
"e1d90740ddeb4d4d8b3c6a8fce56d3db",
"fbb0d85e685e45c190e7a51919172567",
"ee58e1eaab9146ce9a5612db0ed381ae",
"6d7db29783a848889ecb3bafb116af05",
"9abdd27c6115474886492649d0dc648e",
"71b3f6a3bff648138625e12531733a3c",
"ce34796b31bf4df1bddded3a1f487b9e",
"075d17ef882d4194a5ca7127676ed555",
"becccdae788f40de81cf7aaa5431beaf",
"1a499ab336df4bc8a04b70d3527a4af6",
"9d5feeb47a614109a5672841ea846978",
"787926cd507e4fdbaa0909cdef6965b1",
"167fc03b177e4db39b2f93b80a5cc4ff",
"36b27a2d6d9e429ebf0d5b8394cf48f4",
"17b23beda4c44f5b9c4c7128133caad0",
"e37050b0a75849588860e86fae4a3de5",
"7d6932638b67462a92160bdacef39fb6",
"32a30d4203e94ca7bdaa9f5cef984e09",
"69c363522b2848959affbe12a20a04cd",
"a9e0b849a2c549ae890181cd80aa6f89",
"d642239da7dc4df9b26ec5e64e914a3d",
"98850cebbc794eeb990b15db32ab4e96",
"a38dd16fdb7f4197b965f37a4f68a00b",
"d063aeb7be1149acaa1b7aa181f9bf56",
"0b4f70bbdc134e81bc171bc3b1238ea4",
"12ca3b0daaba42f9b6533359fdeefabc",
"1a81cd63f93c4c3ebd62901ab77ecdfe",
"3e00d721c53243489a2ae9c751d78177",
"aefaeeee1ba5490d921955a897f9541c",
"c212d6d4cbea43fa97694baed3630c3a",
"1d1f2a148d2d4d57b5ef27e224059361",
"024617caf06745eeb5048bc675721e40",
"5a32a56b0a7e4479bc4cc76c816e6761",
"a416a534366b4539a4ad20edaebca067",
"016fd7a3fff3471c8235b05d5e825378",
"c2b5e9ce07db4769ac0f51742d3da924",
"fa9613e7b6cb4ce686d2850d55cb2b2a",
"8cd218c09d544e88bd0791469aa6b91e",
"770463e9a2b9465c89c625cf34f28a7a",
"f15b2b9fab4c4693b57f9952ce8ec8b7",
"77ab293c0f724ccfb340237e0d4c3e2d",
"9549d84184a24d0cb6ce9867f7ddda9d",
"cdbce9ddd5034482b576c633b751356b",
"821ffebb5c6044409920e08051a58c58",
"8f42d0c475ef4c998730c54676dc7571",
"1473f532e6d94ae496cf13a61130644d",
"13cd0dfa941d4e438b70aab28e5d4b16",
"74c56cfe7f0f4392b3627cb2c6907057",
"625b8b0529d144d2bb35ac65a7943e7d",
"1410828c0f53465db662f06761c3e516",
"a1e63911f34a4eb99e8b143354c73177",
"0fa52099c0b747d391e0147c765c6039",
"69b78cd662c24edfa08f6b00be37b6a2",
"b02dac4c47d847079d038a86970f5604",
"69af44e6425a47c691d2218df10ede98",
"8afb205894df4e94b29b957d204a7aa2",
"36c9a51ddc9745029f2de35a79cff975",
"4a970046e4de4718931d2bbe1142d8cd",
"dca5f404bf0c4602917e147a0888a5f0",
"0e9bf51ddbdb413e9d36188189fdc4a0",
"74e7a58d3f1940a884654e645f31485f",
"1c690038190544bea9bdf78da69c9468",
"20e07c941a69499581b6f3be98867f59",
"330c414a44824bda8af6e728e91e1fc9",
"1dfbc9c6dbc7457caed5ddc869bf4869",
"af58010cf4b6458ca05d2bc75c921a7b",
"8180d73c32374a7586b1b20abb63184a",
"5840cd0e7a404fecbbb400e1cff8c4dd",
"8d284c20b9bc437390c4acc737829d80",
"be03d3e6a9c4413890f37cd1fca5cb8f",
"d73a3037907e4dc4ace253e3335e56ab",
"a281bb601c514bd9971d7e5c7573f534",
"2d25ee7b589a4d8bbebe8fee2eea44e6",
"d578407774bb4d94b9fefc4835a6dd23",
"e035a8c273764466a6f3b15500706a99",
"bbb49fca62f64f219f8623adcb2a703d",
"2a73f10fb6a847a1bb04916a80c8481d",
"46502c4bb3314c9fa28990e15144c2a0",
"23b4e0c81fd440948571377dea1e8a44",
"c82b97c7519e45f5b1066be8498d5e8d",
"d938189dfc30407baa674fb25cfda105",
"60c1071ceb3741a78d8f339ca0870b62",
"6da4960dfb5f45e4b7e6af2b416d4060",
"eeca4c0bcacf4b548219af0a8b9e5c2d",
"9f4938e677a04309a230f6a947f8ab6b",
"0c1c65ee7a00442e9308b466a8f4cbaf",
"1366eed9abb54d1e8344833e3279adf1",
"c88a3823c8474771bd96c2c0b28e91bb",
"507e6190c8f24801b6554b99272b6ff6",
"de023f071cb5492191a56e2c534e9218",
"3cfe7bfff845466195703171e3eeaa7a",
"7a1175be5ea947229f1b446d8bdadca0",
"30e9a35d9a2c4408997a34419521d652",
"d39f45b7d83a4f2c87b6c957990958aa",
"211b37cbf7244ec7980442fe844be24a",
"ee941a0b8d3a45d58e960c907b64d9b3",
"6c81a449918d4390943bdcb75f36c839",
"3f67b975573d4cb8b19b5eefbddb06b1",
"4cd16dfccd85470f8f743a8d932a13bf",
"51cfe5d1309e4cabaddc41d90a7ffe59",
"ac8d0e89705c48ae8d2b82fa6fb615e2",
"c5c231ac276c496891531cf58b7bb21c",
"6668ab0d821e4711bb580d5c7315ed1a"
]
},
"executionInfo": {
"elapsed": 2362634,
"status": "ok",
"timestamp": 1777286029275,
"user": {
"displayName": "kz",
"userId": "04068570917018885044"
},
"user_tz": -540
},
"id": "2e28eb0f",
"lines_to_next_cell": 1,
"outputId": "b67865d2-05c7-4119-bc32-2814cdf8f76b"
},
"outputs": [
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "a0f5901d9d494ce4b2c3c3c1fcabd06e",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"epoch 1: 0%| | 0/2000 [00:00<?, ?it/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "4776891f42a84f4aafbdbe3f74cd66a2",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"eval: 0%| | 0/16 [00:00<?, ?it/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "180f3d66814d4888a7f1cfa5d5efd683",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"threshold search: 0%| | 0/16 [00:00<?, ?it/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"epoch 1 val_loss=0.3042 comma_f1=0.5096 period_f1=0.6356 punct_macro_f1=0.5726 newline_f1=0.4579 score=0.5574 thr=(0.35, margin=0.25) time=451.1s\n",
" precision recall f1-score support\n",
"\n",
" NONE 0.98 0.96 0.97 704715\n",
" 、 0.46 0.57 0.51 23677\n",
" 。 0.56 0.73 0.64 18901\n",
"\n",
" accuracy 0.95 747293\n",
" macro avg 0.67 0.76 0.71 747293\n",
"weighted avg 0.95 0.95 0.95 747293\n",
"\n",
"saved: punct_train_llmjp_gitlab_work/A4_vocab12000_swiglu.pt best_score 0.5574186101254461\n"
]
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "cd3f6fff352a45fd9aa853e95cd2706c",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"epoch 2: 0%| | 0/2000 [00:00<?, ?it/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "5ce1a0cd8c7841649d5a82b6712aa20e",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"eval: 0%| | 0/16 [00:00<?, ?it/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "627850b0b6974b8993680eda6227f828",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"threshold search: 0%| | 0/16 [00:00<?, ?it/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"epoch 2 val_loss=0.2589 comma_f1=0.5481 period_f1=0.6923 punct_macro_f1=0.6202 newline_f1=0.5227 score=0.6094 thr=(0.45, margin=0.50) time=450.9s\n",
" precision recall f1-score support\n",
"\n",
" NONE 0.99 0.96 0.97 704715\n",
" 、 0.48 0.64 0.55 23677\n",
" 。 0.60 0.81 0.69 18901\n",
"\n",
" accuracy 0.95 747293\n",
" macro avg 0.69 0.81 0.74 747293\n",
"weighted avg 0.96 0.95 0.95 747293\n",
"\n",
"saved: punct_train_llmjp_gitlab_work/A4_vocab12000_swiglu.pt best_score 0.6094140399627694\n"
]
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "9abdd27c6115474886492649d0dc648e",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"epoch 3: 0%| | 0/2000 [00:00<?, ?it/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "e37050b0a75849588860e86fae4a3de5",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"eval: 0%| | 0/16 [00:00<?, ?it/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "1a81cd63f93c4c3ebd62901ab77ecdfe",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"threshold search: 0%| | 0/16 [00:00<?, ?it/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"epoch 3 val_loss=0.2403 comma_f1=0.5803 period_f1=0.7286 punct_macro_f1=0.6544 newline_f1=0.5294 score=0.6401 thr=(0.35, margin=0.50) time=481.1s\n",
" precision recall f1-score support\n",
"\n",
" NONE 0.99 0.97 0.98 704715\n",
" 、 0.51 0.67 0.58 23677\n",
" 。 0.65 0.83 0.73 18901\n",
"\n",
" accuracy 0.96 747293\n",
" macro avg 0.72 0.82 0.76 747293\n",
"weighted avg 0.96 0.96 0.96 747293\n",
"\n",
"saved: punct_train_llmjp_gitlab_work/A4_vocab12000_swiglu.pt best_score 0.6400807038803825\n"
]
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "8cd218c09d544e88bd0791469aa6b91e",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"epoch 4: 0%| | 0/2000 [00:00<?, ?it/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "625b8b0529d144d2bb35ac65a7943e7d",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"eval: 0%| | 0/16 [00:00<?, ?it/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "0e9bf51ddbdb413e9d36188189fdc4a0",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"threshold search: 0%| | 0/16 [00:00<?, ?it/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"epoch 4 val_loss=0.2257 comma_f1=0.5981 period_f1=0.7436 punct_macro_f1=0.6708 newline_f1=0.5588 score=0.6578 thr=(0.40, margin=0.50) time=451.0s\n",
" precision recall f1-score support\n",
"\n",
" NONE 0.99 0.97 0.98 704715\n",
" 、 0.53 0.69 0.60 23677\n",
" 。 0.66 0.85 0.74 18901\n",
"\n",
" accuracy 0.96 747293\n",
" macro avg 0.73 0.84 0.77 747293\n",
"weighted avg 0.96 0.96 0.96 747293\n",
"\n",
"saved: punct_train_llmjp_gitlab_work/A4_vocab12000_swiglu.pt best_score 0.6577902480877615\n"
]
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "d73a3037907e4dc4ace253e3335e56ab",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"epoch 5: 0%| | 0/2000 [00:00<?, ?it/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "60c1071ceb3741a78d8f339ca0870b62",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"eval: 0%| | 0/16 [00:00<?, ?it/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "30e9a35d9a2c4408997a34419521d652",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"threshold search: 0%| | 0/16 [00:00<?, ?it/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"epoch 5 val_loss=0.2181 comma_f1=0.6117 period_f1=0.7602 punct_macro_f1=0.6859 newline_f1=0.5668 score=0.6726 thr=(0.40, margin=0.75) time=500.6s\n",
" precision recall f1-score support\n",
"\n",
" NONE 0.99 0.97 0.98 704715\n",
" 、 0.54 0.70 0.61 23677\n",
" 。 0.68 0.86 0.76 18901\n",
"\n",
" accuracy 0.96 747293\n",
" macro avg 0.74 0.84 0.78 747293\n",
"weighted avg 0.97 0.96 0.96 747293\n",
"\n",
"saved: punct_train_llmjp_gitlab_work/A4_vocab12000_swiglu.pt best_score 0.6726245187304916\n",
"training history:\n",
"[\n",
" {\n",
" \"epoch\": 1,\n",
" \"time_sec\": 451.0605950160025,\n",
" \"loss\": 0.30415573716163635,\n",
" \"punct_loss\": 0.23817488178610802,\n",
" \"newline_loss\": 0.13196171168237925,\n",
" \"punct_binary_f1\": 0.6050871807592529,\n",
" \"comma_f1\": 0.509586257502493,\n",
" \"period_f1\": 0.6355838913861114,\n",
" \"punct_macro_f1\": 0.5725850744443022,\n",
" \"newline_f1\": 0.4578520809678556,\n",
" \"newline_binary_f1\": 0.4578520809678556,\n",
" \"score\": 0.5574186101254461,\n",
" \"threshold_search\": {\n",
" \"newline_threshold\": 0.35,\n",
" \"punct_margin\": 0.25,\n",
" \"comma_f1\": 0.5085482350047148,\n",
" \"period_f1\": 0.6424271495383139,\n",
" \"punct_macro_f1\": 0.5754876922715144,\n",
" \"newline_f1\": 0.4851422815411735,\n",
" \"score\": 0.5574186101254461\n",
" }\n",
" },\n",
" {\n",
" \"epoch\": 2,\n",
" \"time_sec\": 450.9349462930004,\n",
" \"loss\": 0.25894941575825214,\n",
" \"punct_loss\": 0.20197359006851912,\n",
" \"newline_loss\": 0.11395165137946606,\n",
" \"punct_binary_f1\": 0.6463850894333384,\n",
" \"comma_f1\": 0.5480802086143333,\n",
" \"period_f1\": 0.6922833220185562,\n",
" \"punct_macro_f1\": 0.6201817653164448,\n",
" \"newline_f1\": 0.5227272727272727,\n",
" \"newline_binary_f1\": 0.5227272727272727,\n",
" \"score\": 0.6094140399627694,\n",
" \"threshold_search\": {\n",
" \"newline_threshold\": 0.45,\n",
" \"punct_margin\": 0.5,\n",
" \"comma_f1\": 0.548344203127968,\n",
" \"period_f1\": 0.7115328608872925,\n",
" \"punct_macro_f1\": 0.6299385320076303,\n",
" \"newline_f1\": 0.5273160717833256,\n",
" \"score\": 0.6094140399627694\n",
" }\n",
" },\n",
" {\n",
" \"epoch\": 3,\n",
" \"time_sec\": 481.1308304359991,\n",
" \"loss\": 0.24031874351203442,\n",
" \"punct_loss\": 0.18404116947203875,\n",
" \"newline_loss\": 0.11255514714866877,\n",
" \"punct_binary_f1\": 0.6796622515876843,\n",
" \"comma_f1\": 0.5803133465114576,\n",
" \"period_f1\": 0.7285504167347606,\n",
" \"punct_macro_f1\": 0.6544318816231092,\n",
" \"newline_f1\": 0.5294227535961143,\n",
" \"newline_binary_f1\": 0.5294227535961143,\n",
" \"score\": 0.6400807038803825,\n",
" \"threshold_search\": {\n",
" \"newline_threshold\": 0.35,\n",
" \"punct_margin\": 0.5,\n",
" \"comma_f1\": 0.5830625764800202,\n",
" \"period_f1\": 0.7445135195865479,\n",
" \"punct_macro_f1\": 0.6637880480332841,\n",
" \"newline_f1\": 0.5452513272687758,\n",
" \"score\": 0.6400807038803825\n",
" }\n",
" },\n",
" {\n",
" \"epoch\": 4,\n",
" \"time_sec\": 451.0275581269998,\n",
" \"loss\": 0.22570471093058586,\n",
" \"punct_loss\": 0.1732629807665944,\n",
" \"newline_loss\": 0.10488346265628934,\n",
" \"punct_binary_f1\": 0.6947168034716394,\n",
" \"comma_f1\": 0.5980752277402293,\n",
" \"period_f1\": 0.7435850016181977,\n",
" \"punct_macro_f1\": 0.6708301146792135,\n",
" \"newline_f1\": 0.558847825306181,\n",
" \"newline_binary_f1\": 0.558847825306181,\n",
" \"score\": 0.6577902480877615,\n",
" \"threshold_search\": {\n",
" \"newline_threshold\": 0.4,\n",
" \"punct_margin\": 0.5,\n",
" \"comma_f1\": 0.6000922896216125,\n",
" \"period_f1\": 0.7612817874467147,\n",
" \"punct_macro_f1\": 0.6806870385341637,\n",
" \"newline_f1\": 0.5662030863021528,\n",
" \"score\": 0.6577902480877615\n",
" }\n",
" },\n",
" {\n",
" \"epoch\": 5,\n",
" \"time_sec\": 500.571824947001,\n",
" \"loss\": 0.21805834583938122,\n",
" \"punct_loss\": 0.16585472878068686,\n",
" \"newline_loss\": 0.1044072350487113,\n",
" \"punct_binary_f1\": 0.7075757263760476,\n",
" \"comma_f1\": 0.6117413812093843,\n",
" \"period_f1\": 0.7601519272249836,\n",
" \"punct_macro_f1\": 0.6859466542171839,\n",
" \"newline_f1\": 0.5667818810260142,\n",
" \"newline_binary_f1\": 0.5667818810260142,\n",
" \"score\": 0.6726245187304916,\n",
" \"threshold_search\": {\n",
" \"newline_threshold\": 0.4,\n",
" \"punct_margin\": 0.75,\n",
" \"comma_f1\": 0.6092300914197599,\n",
" \"period_f1\": 0.7828472716187379,\n",
" \"punct_macro_f1\": 0.696038681519249,\n",
" \"newline_f1\": 0.5789678675754625,\n",
" \"score\": 0.6726245187304916\n",
" }\n",
" }\n",
"]\n",
"best_score: 0.6726245187304916\n",
"max VRAM allocated GB: 4.01\n",
"checkpoint: punct_train_llmjp_gitlab_work/A4_vocab12000_swiglu.pt\n",
"bundle: punct_train_llmjp_gitlab_work/A4_vocab12000_swiglu_bundle.json\n"
]
}
],
"source": [
"\n",
"proc = psutil.Process(os.getpid())\n",
"if start_epoch > CFG.epochs:\n",
" print(f\"checkpoint already reached epoch {start_epoch - 1}; CFG.epochs={CFG.epochs}. Nothing to train.\")\n",
"\n",
"for epoch in range(start_epoch, CFG.epochs + 1):\n",
" if hasattr(train_ds, \"set_epoch\"):\n",
" train_ds.set_epoch(epoch)\n",
" model.train()\n",
" t0 = time.perf_counter()\n",
" running = 0.0\n",
" seen_tokens = 0\n",
"\n",
" pbar = tqdm(enumerate(train_loader, start=1), total=CFG.steps_per_epoch, desc=f\"epoch {epoch}\")\n",
" for step, batch in pbar:\n",
" if step > CFG.steps_per_epoch:\n",
" break\n",
"\n",
" input_ids = batch[\"input_ids\"].to(DEVICE, non_blocking=True)\n",
" attention_mask = batch[\"attention_mask\"].to(DEVICE, non_blocking=True)\n",
" punct_labels = batch[\"punct_labels\"].to(DEVICE, non_blocking=True)\n",
" newline_labels = batch[\"newline_labels\"].to(DEVICE, non_blocking=True)\n",
"\n",
" optimizer.zero_grad(set_to_none=True)\n",
" with torch.amp.autocast(\"cuda\", enabled=(DEVICE == \"cuda\")):\n",
" outputs = model(input_ids, attention_mask)\n",
" loss, punct_loss, newline_loss = compute_loss(outputs, punct_labels, newline_labels)\n",
"\n",
" scaler.scale(loss).backward()\n",
" scaler.unscale_(optimizer)\n",
" torch.nn.utils.clip_grad_norm_(model.parameters(), CFG.grad_clip)\n",
" scaler.step(optimizer)\n",
" scaler.update()\n",
" scheduler.step()\n",
"\n",
" running += loss.item()\n",
" seen_tokens += int(attention_mask.sum().detach().cpu())\n",
"\n",
" if step % 25 == 0:\n",
" elapsed = max(1e-6, time.perf_counter() - t0)\n",
" postfix = {\n",
" \"loss\": f\"{running / 25:.4f}\",\n",
" \"p_loss\": f\"{punct_loss.item():.4f}\",\n",
" \"n_loss\": f\"{newline_loss.item():.4f}\",\n",
" \"lr\": f\"{scheduler.get_last_lr()[0]:.2e}\",\n",
" \"tok/s\": f\"{seen_tokens / elapsed:,.0f}\",\n",
" \"rssGB\": f\"{proc.memory_info().rss / 1024**3:.2f}\",\n",
" }\n",
" if DEVICE == \"cuda\":\n",
" postfix[\"vramGB\"] = f\"{torch.cuda.max_memory_allocated() / 1024**3:.2f}\"\n",
" pbar.set_postfix(postfix)\n",
" running = 0.0\n",
"\n",
" metrics = evaluate(model, val_loader)\n",
" threshold_metrics = search_thresholds(model, val_loader)\n",
" if threshold_metrics:\n",
" CFG.newline_threshold = threshold_metrics[\"newline_threshold\"]\n",
" CFG.punct_margin = threshold_metrics[\"punct_margin\"]\n",
" metrics[\"threshold_search\"] = threshold_metrics\n",
" metrics[\"score\"] = threshold_metrics[\"score\"]\n",
" epoch_time = time.perf_counter() - t0\n",
" row = {\"epoch\": epoch, \"time_sec\": epoch_time, **metrics}\n",
" history.append({k: v for k, v in row.items() if k != \"report\"})\n",
"\n",
" print(\n",
" f\"epoch {epoch} val_loss={metrics['loss']:.4f} \"\n",
" f\"comma_f1={metrics['comma_f1']:.4f} \"\n",
" f\"period_f1={metrics['period_f1']:.4f} \"\n",
" f\"punct_macro_f1={metrics['punct_macro_f1']:.4f} \"\n",
" f\"newline_f1={metrics['newline_f1']:.4f} \"\n",
" f\"score={metrics['score']:.4f} \"\n",
" f\"thr=({CFG.newline_threshold:.2f}, margin={CFG.punct_margin:.2f}) \"\n",
" f\"time={epoch_time:.1f}s\"\n",
" )\n",
" print(metrics[\"report\"])\n",
"\n",
" if metrics[\"score\"] > best_score:\n",
" best_score = metrics[\"score\"]\n",
" save_obj = {\n",
" \"model\": model.module.state_dict() if hasattr(model, \"module\") else model.state_dict(),\n",
" \"cfg\": asdict(CFG),\n",
" \"label2id\": LABEL2ID,\n",
" \"id2label\": ID2LABEL,\n",
" \"heads\": {\"punct\": len(LABEL2ID), \"newline\": 1},\n",
" \"history\": history,\n",
" \"corpora\": [{k: v for k, v in c.items() if k != \"gitlab_paths\"} | {\"num_files\": len(c[\"gitlab_paths\"])} for c in AVAILABLE_CORPORA],\n",
" \"itos\": itos,\n",
" \"vocab_size\": len(itos),\n",
" \"inference\": {\n",
" \"mode\": CFG.inference_mode,\n",
" \"newline_threshold\": CFG.newline_threshold,\n",
" \"punct_margin\": CFG.punct_margin,\n",
" },\n",
" \"epoch\": epoch,\n",
" \"best_score\": best_score,\n",
" \"optimizer\": optimizer.state_dict(),\n",
" \"scheduler\": scheduler.state_dict(),\n",
" \"scaler\": scaler.state_dict(),\n",
" }\n",
" torch.save(save_obj, CFG.ckpt_path)\n",
" with open(CFG.bundle_path, \"w\", encoding=\"utf-8\") as f:\n",
" json.dump({k: v for k, v in save_obj.items() if k != \"model\"}, f, ensure_ascii=False, indent=2, default=str)\n",
" print(\"saved:\", CFG.ckpt_path, \"best_score\", best_score)\n",
"\n",
"print(\"training history:\")\n",
"print(json.dumps(history, ensure_ascii=False, indent=2))\n",
"print(\"best_score:\", best_score)\n",
"if DEVICE == \"cuda\":\n",
" print(\"max VRAM allocated GB:\", round(torch.cuda.max_memory_allocated() / 1024**3, 3))\n",
"print(\"checkpoint:\", CFG.ckpt_path)\n",
"print(\"bundle:\", CFG.bundle_path)"
]
},
{
"cell_type": "markdown",
"id": "9b21ba0b",
"metadata": {
"id": "9b21ba0b"
},
"source": [
"## 14. Restore helper"
]
},
{
"cell_type": "code",
"execution_count": 34,
"id": "b5531ce8",
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"executionInfo": {
"elapsed": 75,
"status": "ok",
"timestamp": 1777286050861,
"user": {
"displayName": "kz",
"userId": "04068570917018885044"
},
"user_tz": -540
},
"id": "b5531ce8",
"outputId": "91263ef4-0394-40bd-fc58-eb2bc0a257a4"
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"これはテストです。\n",
"句読点と改行を別ヘッドで復元します。\n",
"学習が進むと、自然になります。\n"
]
}
],
"source": [
"\n",
"def center_weight(i: int, n: int) -> float:\n",
" if n <= 1:\n",
" return 1.0\n",
" center = (n - 1) / 2\n",
" dist = abs(i - center) / max(1.0, center)\n",
" return 0.25 + 0.75 * (1.0 - dist)\n",
"\n",
"@torch.no_grad()\n",
"def restore_punctuation_and_newlines(\n",
" text: str,\n",
" model,\n",
" max_len: Optional[int] = None,\n",
" newline_threshold: Optional[float] = None,\n",
" punct_margin: Optional[float] = None,\n",
") -> str:\n",
" model.eval()\n",
" chars = [ch for ch in text if ch not in BOUNDARY_CHARS and ch not in REMOVE_CHARS and ch != \" \"]\n",
" out = []\n",
"\n",
" # 修正ポイント: ストライド推論 (Stride Inference) を導入し、\n",
" # チャンク間のつなぎ目で文脈が途切れて予測精度が落ちるのを防ぎます。\n",
" model_for_shape = model.module if hasattr(model, \"module\") else model\n",
" model_for_shape = model_for_shape._orig_mod if hasattr(model_for_shape, \"_orig_mod\") else model_for_shape\n",
" model_max_len = model_for_shape.pos_emb.num_embeddings\n",
" max_len = model_max_len if max_len is None else min(max_len, model_max_len)\n",
" stride = max(1, int(max_len * 0.75))\n",
" newline_threshold = CFG.newline_threshold if newline_threshold is None else newline_threshold\n",
" punct_margin = CFG.punct_margin if punct_margin is None else punct_margin\n",
"\n",
" punct_logits_sum = np.zeros((len(chars), CFG.num_labels), dtype=np.float64)\n",
" newline_logits_sum = np.zeros((len(chars),), dtype=np.float64)\n",
" weight_sum = np.zeros((len(chars),), dtype=np.float64)\n",
"\n",
" for start in range(0, len(chars), stride):\n",
" chunk = chars[start:start + max_len]\n",
" x = np.zeros((1, max_len), dtype=np.int32)\n",
" m = np.zeros((1, max_len), dtype=np.bool_)\n",
" x[0, :len(chunk)] = [char_to_id(ch) for ch in chunk]\n",
" m[0, :len(chunk)] = True\n",
"\n",
" input_ids = torch.from_numpy(x).to(DEVICE)\n",
" attention_mask = torch.from_numpy(m).to(DEVICE)\n",
" with torch.amp.autocast(\"cuda\", enabled=(DEVICE == \"cuda\")):\n",
" outputs = model(input_ids, attention_mask)\n",
"\n",
" punct_logits = outputs[\"punct_logits\"][0, :len(chunk)].float().detach().cpu().numpy()\n",
" newline_logits = outputs[\"newline_logits\"][0, :len(chunk)].float().detach().cpu().numpy()\n",
"\n",
" for i in range(len(chunk)):\n",
" index = start + i\n",
" weight = center_weight(i, len(chunk))\n",
" punct_logits_sum[index] += weight * punct_logits[i]\n",
" newline_logits_sum[index] += weight * newline_logits[i]\n",
" weight_sum[index] += weight\n",
"\n",
" none_id = LABEL2ID[\"NONE\"]\n",
" for i in range(len(chars)):\n",
" ch = chars[i]\n",
" out.append(ch)\n",
"\n",
" if weight_sum[i] <= 0:\n",
" continue\n",
"\n",
" punct_logits = punct_logits_sum[i] / weight_sum[i]\n",
" best_punct = int(np.argmax(punct_logits))\n",
" if best_punct != none_id and punct_logits[best_punct] - punct_logits[none_id] >= punct_margin:\n",
" out.append(ID2LABEL[best_punct])\n",
"\n",
" newline_logit = newline_logits_sum[i] / weight_sum[i]\n",
" newline_prob = 1.0 / (1.0 + math.exp(-float(newline_logit)))\n",
" if newline_prob >= newline_threshold:\n",
" out.append(\"\\n\")\n",
"\n",
" return \"\".join(out)\n",
"\n",
"sample = \"これはテストです句読点と改行を別ヘッドで復元します学習が進むと自然になります\"\n",
"print(restore_punctuation_and_newlines(sample, model))"
]
},
{
"cell_type": "markdown",
"id": "8bf50b27",
"metadata": {
"id": "8bf50b27"
},
"source": [
"## 15. Tuning\n",
"\n",
"Auto Stage A sweep:\n",
"```python\n",
"CFG.auto_tune = True\n",
"CFG.auto_tune_steps_per_epoch = 600\n",
"CFG.auto_tune_epochs = 2\n",
"CFG.auto_tune_val_chunks = 4096\n",
"CFG.auto_tune_train_cache_chunks = 1_000_000\n",
"```\n",
"\n",
"Outputs:\n",
"- punct_train_llmjp_gitlab_work/stage_a_auto_tune_results.json\n",
"- punct_train_llmjp_gitlab_work/stage_a_selected_config.json\n",
"\n",
"Smaller initial test:\n",
"```python\n",
"CFG.max_files_per_corpus = 32\n",
"CFG.steps_per_epoch = 600\n",
"CFG.epochs = 2\n",
"CFG.batch_size = 256\n",
"CFG.val_chunks = 4096\n",
"CFG.train_cache_chunks = 1_000_000\n",
"```\n",
"\n",
"Stage A candidates. Set these before `stoi, itos = build_vocab()` so the\n",
"vocabulary and cache names stay aligned.\n",
"```python\n",
"# A1\n",
"CFG.vocab_max_size = 16_000; CFG.ffn_type = \"gelu\"; CFG.experiment_name = \"vocab16000_gelu\"\n",
"# A2\n",
"CFG.vocab_max_size = 12_000; CFG.ffn_type = \"gelu\"; CFG.experiment_name = \"vocab12000_gelu\"\n",
"# A3\n",
"CFG.vocab_max_size = 8_000; CFG.ffn_type = \"gelu\"; CFG.experiment_name = \"vocab8000_gelu\"\n",
"# A4\n",
"CFG.vocab_max_size = 12_000; CFG.ffn_type = \"swiglu\"; CFG.swiglu_hidden = 1024; CFG.experiment_name = \"vocab12000_swiglu\"\n",
"```\n",
"\n",
"More data:\n",
"```python\n",
"CFG.steps_per_epoch = 2000\n",
"CFG.epochs = 5\n",
"CFG.val_chunks = 8192\n",
"CFG.shuffle_buffer_chunks = 4096\n",
"```\n",
"\n",
"If GitLab throttles or fails:\n",
"```python\n",
"os.environ[\"GITLAB_TOKEN\"] = \"your_token\"\n",
"```\n",
"\n",
"Redownload corrupted cache:\n",
"```python\n",
"CFG.force_redownload = True\n",
"```\n",
"\n",
"Clear cache:\n",
"```python\n",
"!rm -rf /content/llmjp_gitlab_cache\n",
"\n",
"```"
]
}
],
"metadata": {
"accelerator": "GPU",
"colab": {
"gpuType": "L4",
"machine_shape": "hm",
"provenance": []
},
"jupytext": {
"encoding": "# -*- coding: utf-8 -*-",
"formats": "ipynb,py:percent"
},
"kernelspec": {
"display_name": "Python 3",
"name": "python3"
},
"widgets": {
"application/vnd.jupyter.widget-state+json": {
"016fd7a3fff3471c8235b05d5e825378": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "ProgressStyleModel",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "ProgressStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"bar_color": null,
"description_width": ""
}
},
"024617caf06745eeb5048bc675721e40": {
"model_module": "@jupyter-widgets/base",
"model_module_version": "1.2.0",
"model_name": "LayoutModel",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"041a636352994ebc870fd076fc6ccd69": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "HTMLModel",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HTMLModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HTMLView",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_fdac6f566099458fa038d11e7aa44452",
"placeholder": "​",
"style": "IPY_MODEL_1e288dfa575c41f198ca0cb09235a721",
"value": " 15/16 [00:01&lt;00:00, 14.62it/s]"
}
},
"0484474d43a34dbabf45fa11b3391881": {
"model_module": "@jupyter-widgets/base",
"model_module_version": "1.2.0",
"model_name": "LayoutModel",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": "hidden",
"width": null
}
},
"075d17ef882d4194a5ca7127676ed555": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "HTMLModel",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HTMLModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HTMLView",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_36b27a2d6d9e429ebf0d5b8394cf48f4",
"placeholder": "​",
"style": "IPY_MODEL_17b23beda4c44f5b9c4c7128133caad0",
"value": " 2000/2000 [07:54&lt;00:00,  4.50it/s, loss=0.1299, p_loss=0.1109, n_loss=0.0372, lr=2.06e-04, tok/s=191,949, rssGB=7.82, vramGB=4.01]"
}
},
"08a1c054d81d4ed08147017c3026fa77": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "FloatProgressModel",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "FloatProgressModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "ProgressView",
"bar_style": "",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_8170a2dba6594d7db272ad52436d5a37",
"max": 16,
"min": 0,
"orientation": "horizontal",
"style": "IPY_MODEL_19421bf7b12c40d6b788792e2914c853",
"value": 16
}
},
"09f8dcbd7cc544108e43fead639ce277": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "HTMLModel",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HTMLModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HTMLView",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_a45644ce46e44a7abc1cd9d701ae66fc",
"placeholder": "​",
"style": "IPY_MODEL_b5033174abbc4eb8871ecaa764231ba7",
"value": "epoch 1: 100%"
}
},
"0b316a18319644108a1b2f70a5d7fe42": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "HTMLModel",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HTMLModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HTMLView",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_26dd110c6ef1424e983a65e551ce366d",
"placeholder": "​",
"style": "IPY_MODEL_8132e5f353f1440d808373a2dd6eba00",
"value": " 2000/2000 [07:24&lt;00:00,  4.51it/s, loss=0.1474, p_loss=0.1211, n_loss=0.0411, lr=3.69e-04, tok/s=204,962, rssGB=7.82, vramGB=4.01]"
}
},
"0b4f70bbdc134e81bc171bc3b1238ea4": {
"model_module": "@jupyter-widgets/base",
"model_module_version": "1.2.0",
"model_name": "LayoutModel",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"0c1c65ee7a00442e9308b466a8f4cbaf": {
"model_module": "@jupyter-widgets/base",
"model_module_version": "1.2.0",
"model_name": "LayoutModel",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": "hidden",
"width": null
}
},
"0e9bf51ddbdb413e9d36188189fdc4a0": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "HBoxModel",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HBoxModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HBoxView",
"box_style": "",
"children": [
"IPY_MODEL_74e7a58d3f1940a884654e645f31485f",
"IPY_MODEL_1c690038190544bea9bdf78da69c9468",
"IPY_MODEL_20e07c941a69499581b6f3be98867f59"
],
"layout": "IPY_MODEL_330c414a44824bda8af6e728e91e1fc9"
}
},
"0fa52099c0b747d391e0147c765c6039": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "HTMLModel",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HTMLModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HTMLView",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_4a970046e4de4718931d2bbe1142d8cd",
"placeholder": "​",
"style": "IPY_MODEL_dca5f404bf0c4602917e147a0888a5f0",
"value": " 15/16 [00:01&lt;00:00, 14.76it/s]"
}
},
"12ca3b0daaba42f9b6533359fdeefabc": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "DescriptionStyleModel",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "DescriptionStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"description_width": ""
}
},
"1366eed9abb54d1e8344833e3279adf1": {
"model_module": "@jupyter-widgets/base",
"model_module_version": "1.2.0",
"model_name": "LayoutModel",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"13cd0dfa941d4e438b70aab28e5d4b16": {
"model_module": "@jupyter-widgets/base",
"model_module_version": "1.2.0",
"model_name": "LayoutModel",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"1410828c0f53465db662f06761c3e516": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "HTMLModel",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HTMLModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HTMLView",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_b02dac4c47d847079d038a86970f5604",
"placeholder": "​",
"style": "IPY_MODEL_69af44e6425a47c691d2218df10ede98",
"value": "eval:  94%"
}
},
"1473f532e6d94ae496cf13a61130644d": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "ProgressStyleModel",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "ProgressStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"bar_color": null,
"description_width": ""
}
},
"1566aaff7a8c4e119103f8f76b4afb70": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "DescriptionStyleModel",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "DescriptionStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"description_width": ""
}
},
"167fc03b177e4db39b2f93b80a5cc4ff": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "ProgressStyleModel",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "ProgressStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"bar_color": null,
"description_width": ""
}
},
"17b23beda4c44f5b9c4c7128133caad0": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "DescriptionStyleModel",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "DescriptionStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"description_width": ""
}
},
"180f3d66814d4888a7f1cfa5d5efd683": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "HBoxModel",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HBoxModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HBoxView",
"box_style": "",
"children": [
"IPY_MODEL_aab889a6f5114d15a1572b4a96de6434",
"IPY_MODEL_08a1c054d81d4ed08147017c3026fa77",
"IPY_MODEL_49b8fe32e5f84104a9fcb9e32ac1ee21"
],
"layout": "IPY_MODEL_a42f75516bbb47ed9290bb4a3c75379c"
}
},
"19421bf7b12c40d6b788792e2914c853": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "ProgressStyleModel",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "ProgressStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"bar_color": null,
"description_width": ""
}
},
"1a499ab336df4bc8a04b70d3527a4af6": {
"model_module": "@jupyter-widgets/base",
"model_module_version": "1.2.0",
"model_name": "LayoutModel",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"1a81cd63f93c4c3ebd62901ab77ecdfe": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "HBoxModel",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HBoxModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HBoxView",
"box_style": "",
"children": [
"IPY_MODEL_3e00d721c53243489a2ae9c751d78177",
"IPY_MODEL_aefaeeee1ba5490d921955a897f9541c",
"IPY_MODEL_c212d6d4cbea43fa97694baed3630c3a"
],
"layout": "IPY_MODEL_1d1f2a148d2d4d57b5ef27e224059361"
}
},
"1c690038190544bea9bdf78da69c9468": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "FloatProgressModel",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "FloatProgressModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "ProgressView",
"bar_style": "",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_8180d73c32374a7586b1b20abb63184a",
"max": 16,
"min": 0,
"orientation": "horizontal",
"style": "IPY_MODEL_5840cd0e7a404fecbbb400e1cff8c4dd",
"value": 16
}
},
"1d1f2a148d2d4d57b5ef27e224059361": {
"model_module": "@jupyter-widgets/base",
"model_module_version": "1.2.0",
"model_name": "LayoutModel",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": "hidden",
"width": null
}
},
"1dfbc9c6dbc7457caed5ddc869bf4869": {
"model_module": "@jupyter-widgets/base",
"model_module_version": "1.2.0",
"model_name": "LayoutModel",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"1e288dfa575c41f198ca0cb09235a721": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "DescriptionStyleModel",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "DescriptionStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"description_width": ""
}
},
"20e07c941a69499581b6f3be98867f59": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "HTMLModel",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HTMLModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HTMLView",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_8d284c20b9bc437390c4acc737829d80",
"placeholder": "​",
"style": "IPY_MODEL_be03d3e6a9c4413890f37cd1fca5cb8f",
"value": " 15/16 [00:01&lt;00:00, 14.58it/s]"
}
},
"211b37cbf7244ec7980442fe844be24a": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "FloatProgressModel",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "FloatProgressModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "ProgressView",
"bar_style": "",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_51cfe5d1309e4cabaddc41d90a7ffe59",
"max": 16,
"min": 0,
"orientation": "horizontal",
"style": "IPY_MODEL_ac8d0e89705c48ae8d2b82fa6fb615e2",
"value": 16
}
},
"2137415ed4064c1bb1c69856af0c22b7": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "FloatProgressModel",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "FloatProgressModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "ProgressView",
"bar_style": "",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_2e5bebd32d994437a75aab26549aaa64",
"max": 16,
"min": 0,
"orientation": "horizontal",
"style": "IPY_MODEL_e1cb8d90b9d841c4adb5d2696208c169",
"value": 16
}
},
"21e39a7f2fc641f5923f096425ba5fbd": {
"model_module": "@jupyter-widgets/base",
"model_module_version": "1.2.0",
"model_name": "LayoutModel",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"23b4e0c81fd440948571377dea1e8a44": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "ProgressStyleModel",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "ProgressStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"bar_color": null,
"description_width": ""
}
},
"26dd110c6ef1424e983a65e551ce366d": {
"model_module": "@jupyter-widgets/base",
"model_module_version": "1.2.0",
"model_name": "LayoutModel",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"27a28b493e2c4942b094dde8151623f4": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "ProgressStyleModel",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "ProgressStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"bar_color": null,
"description_width": ""
}
},
"292ea043876d42b2873940f59432731b": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "ProgressStyleModel",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "ProgressStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"bar_color": null,
"description_width": ""
}
},
"2a73f10fb6a847a1bb04916a80c8481d": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "DescriptionStyleModel",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "DescriptionStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"description_width": ""
}
},
"2d25ee7b589a4d8bbebe8fee2eea44e6": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "FloatProgressModel",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "FloatProgressModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "ProgressView",
"bar_style": "danger",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_46502c4bb3314c9fa28990e15144c2a0",
"max": 2000,
"min": 0,
"orientation": "horizontal",
"style": "IPY_MODEL_23b4e0c81fd440948571377dea1e8a44",
"value": 2000
}
},
"2e5bebd32d994437a75aab26549aaa64": {
"model_module": "@jupyter-widgets/base",
"model_module_version": "1.2.0",
"model_name": "LayoutModel",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"30e9a35d9a2c4408997a34419521d652": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "HBoxModel",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HBoxModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HBoxView",
"box_style": "",
"children": [
"IPY_MODEL_d39f45b7d83a4f2c87b6c957990958aa",
"IPY_MODEL_211b37cbf7244ec7980442fe844be24a",
"IPY_MODEL_ee941a0b8d3a45d58e960c907b64d9b3"
],
"layout": "IPY_MODEL_6c81a449918d4390943bdcb75f36c839"
}
},
"32a30d4203e94ca7bdaa9f5cef984e09": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "FloatProgressModel",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "FloatProgressModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "ProgressView",
"bar_style": "",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_a38dd16fdb7f4197b965f37a4f68a00b",
"max": 16,
"min": 0,
"orientation": "horizontal",
"style": "IPY_MODEL_d063aeb7be1149acaa1b7aa181f9bf56",
"value": 16
}
},
"330c414a44824bda8af6e728e91e1fc9": {
"model_module": "@jupyter-widgets/base",
"model_module_version": "1.2.0",
"model_name": "LayoutModel",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": "hidden",
"width": null
}
},
"36b27a2d6d9e429ebf0d5b8394cf48f4": {
"model_module": "@jupyter-widgets/base",
"model_module_version": "1.2.0",
"model_name": "LayoutModel",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"36c9a51ddc9745029f2de35a79cff975": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "ProgressStyleModel",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "ProgressStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"bar_color": null,
"description_width": ""
}
},
"39d94eae6b534d04946f534afc3250ab": {
"model_module": "@jupyter-widgets/base",
"model_module_version": "1.2.0",
"model_name": "LayoutModel",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"3b81f0be353e44e4b01056e9822f009d": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "HTMLModel",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HTMLModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HTMLView",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_fefc2eba241e4adfa56ced16e098b6bf",
"placeholder": "​",
"style": "IPY_MODEL_6160e0acd5bd405ebf8954b9a4319c56",
"value": "eval:  94%"
}
},
"3cfe7bfff845466195703171e3eeaa7a": {
"model_module": "@jupyter-widgets/base",
"model_module_version": "1.2.0",
"model_name": "LayoutModel",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"3e00d721c53243489a2ae9c751d78177": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "HTMLModel",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HTMLModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HTMLView",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_024617caf06745eeb5048bc675721e40",
"placeholder": "​",
"style": "IPY_MODEL_5a32a56b0a7e4479bc4cc76c816e6761",
"value": "threshold search:  94%"
}
},
"3f67b975573d4cb8b19b5eefbddb06b1": {
"model_module": "@jupyter-widgets/base",
"model_module_version": "1.2.0",
"model_name": "LayoutModel",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"45586a223595426990ea412755f7a7c7": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "DescriptionStyleModel",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "DescriptionStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"description_width": ""
}
},
"45c39f21acff4d78a956956c0984a16e": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "HTMLModel",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HTMLModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HTMLView",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_e7e889ec45d64224ab02f5bbe75a1e03",
"placeholder": "​",
"style": "IPY_MODEL_6483fb2129c74784b63fd5e955766161",
"value": "threshold search:  94%"
}
},
"46502c4bb3314c9fa28990e15144c2a0": {
"model_module": "@jupyter-widgets/base",
"model_module_version": "1.2.0",
"model_name": "LayoutModel",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"4776891f42a84f4aafbdbe3f74cd66a2": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "HBoxModel",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HBoxModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HBoxView",
"box_style": "",
"children": [
"IPY_MODEL_7b7cf83dbd214c6ca29340a376ba0188",
"IPY_MODEL_2137415ed4064c1bb1c69856af0c22b7",
"IPY_MODEL_041a636352994ebc870fd076fc6ccd69"
],
"layout": "IPY_MODEL_502cb323941b4fa48995979b96e169d3"
}
},
"49b8fe32e5f84104a9fcb9e32ac1ee21": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "HTMLModel",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HTMLModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HTMLView",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_fd0343bc7fbc4c9ea14fefac4f7c4d05",
"placeholder": "​",
"style": "IPY_MODEL_8f5d53603f784d19a8756b92a13a8329",
"value": " 15/16 [00:01&lt;00:00, 14.57it/s]"
}
},
"4a970046e4de4718931d2bbe1142d8cd": {
"model_module": "@jupyter-widgets/base",
"model_module_version": "1.2.0",
"model_name": "LayoutModel",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"4cd16dfccd85470f8f743a8d932a13bf": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "DescriptionStyleModel",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "DescriptionStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"description_width": ""
}
},
"4f0ba045438d47d1a39190b8478053ac": {
"model_module": "@jupyter-widgets/base",
"model_module_version": "1.2.0",
"model_name": "LayoutModel",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": "hidden",
"width": null
}
},
"502cb323941b4fa48995979b96e169d3": {
"model_module": "@jupyter-widgets/base",
"model_module_version": "1.2.0",
"model_name": "LayoutModel",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": "hidden",
"width": null
}
},
"507e6190c8f24801b6554b99272b6ff6": {
"model_module": "@jupyter-widgets/base",
"model_module_version": "1.2.0",
"model_name": "LayoutModel",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"51cfe5d1309e4cabaddc41d90a7ffe59": {
"model_module": "@jupyter-widgets/base",
"model_module_version": "1.2.0",
"model_name": "LayoutModel",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"554a8cfc16b647a7b6e68a43bb13ef4a": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "HTMLModel",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HTMLModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HTMLView",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_21e39a7f2fc641f5923f096425ba5fbd",
"placeholder": "​",
"style": "IPY_MODEL_ef7995f9babf447f959df314b89d54a2",
"value": " 2000/2000 [07:24&lt;00:00,  4.51it/s, loss=0.1870, p_loss=0.1531, n_loss=0.0479, lr=4.82e-04, tok/s=204,823, rssGB=7.82, vramGB=4.01]"
}
},
"5840cd0e7a404fecbbb400e1cff8c4dd": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "ProgressStyleModel",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "ProgressStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"bar_color": null,
"description_width": ""
}
},
"5a32a56b0a7e4479bc4cc76c816e6761": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "DescriptionStyleModel",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "DescriptionStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"description_width": ""
}
},
"5c39a8db5353446f9563543e08fa0e64": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "FloatProgressModel",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "FloatProgressModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "ProgressView",
"bar_style": "",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_dd3eccb6c46348a69d10948abbf0022d",
"max": 16,
"min": 0,
"orientation": "horizontal",
"style": "IPY_MODEL_292ea043876d42b2873940f59432731b",
"value": 16
}
},
"5ce1a0cd8c7841649d5a82b6712aa20e": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "HBoxModel",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HBoxModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HBoxView",
"box_style": "",
"children": [
"IPY_MODEL_3b81f0be353e44e4b01056e9822f009d",
"IPY_MODEL_5c39a8db5353446f9563543e08fa0e64",
"IPY_MODEL_c0b0d68c39db415abe7b6c92e250c451"
],
"layout": "IPY_MODEL_4f0ba045438d47d1a39190b8478053ac"
}
},
"60c1071ceb3741a78d8f339ca0870b62": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "HBoxModel",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HBoxModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HBoxView",
"box_style": "",
"children": [
"IPY_MODEL_6da4960dfb5f45e4b7e6af2b416d4060",
"IPY_MODEL_eeca4c0bcacf4b548219af0a8b9e5c2d",
"IPY_MODEL_9f4938e677a04309a230f6a947f8ab6b"
],
"layout": "IPY_MODEL_0c1c65ee7a00442e9308b466a8f4cbaf"
}
},
"6160e0acd5bd405ebf8954b9a4319c56": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "DescriptionStyleModel",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "DescriptionStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"description_width": ""
}
},
"625b8b0529d144d2bb35ac65a7943e7d": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "HBoxModel",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HBoxModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HBoxView",
"box_style": "",
"children": [
"IPY_MODEL_1410828c0f53465db662f06761c3e516",
"IPY_MODEL_a1e63911f34a4eb99e8b143354c73177",
"IPY_MODEL_0fa52099c0b747d391e0147c765c6039"
],
"layout": "IPY_MODEL_69b78cd662c24edfa08f6b00be37b6a2"
}
},
"627850b0b6974b8993680eda6227f828": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "HBoxModel",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HBoxModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HBoxView",
"box_style": "",
"children": [
"IPY_MODEL_45c39f21acff4d78a956956c0984a16e",
"IPY_MODEL_ec5985dd680e432f85bffd2b8e1f1c3c",
"IPY_MODEL_7ab1270e0f0e4af8b1a54353a77ce20d"
],
"layout": "IPY_MODEL_0484474d43a34dbabf45fa11b3391881"
}
},
"6483fb2129c74784b63fd5e955766161": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "DescriptionStyleModel",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "DescriptionStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"description_width": ""
}
},
"6668ab0d821e4711bb580d5c7315ed1a": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "DescriptionStyleModel",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "DescriptionStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"description_width": ""
}
},
"69af44e6425a47c691d2218df10ede98": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "DescriptionStyleModel",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "DescriptionStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"description_width": ""
}
},
"69b78cd662c24edfa08f6b00be37b6a2": {
"model_module": "@jupyter-widgets/base",
"model_module_version": "1.2.0",
"model_name": "LayoutModel",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": "hidden",
"width": null
}
},
"69c363522b2848959affbe12a20a04cd": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "HTMLModel",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HTMLModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HTMLView",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_0b4f70bbdc134e81bc171bc3b1238ea4",
"placeholder": "​",
"style": "IPY_MODEL_12ca3b0daaba42f9b6533359fdeefabc",
"value": " 15/16 [00:01&lt;00:00, 14.58it/s]"
}
},
"6c81a449918d4390943bdcb75f36c839": {
"model_module": "@jupyter-widgets/base",
"model_module_version": "1.2.0",
"model_name": "LayoutModel",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": "hidden",
"width": null
}
},
"6d7db29783a848889ecb3bafb116af05": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "DescriptionStyleModel",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "DescriptionStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"description_width": ""
}
},
"6da4960dfb5f45e4b7e6af2b416d4060": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "HTMLModel",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HTMLModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HTMLView",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_1366eed9abb54d1e8344833e3279adf1",
"placeholder": "​",
"style": "IPY_MODEL_c88a3823c8474771bd96c2c0b28e91bb",
"value": "eval:  94%"
}
},
"70e92fcc70e34d969ac7722f9ec0882b": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "ProgressStyleModel",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "ProgressStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"bar_color": null,
"description_width": ""
}
},
"71b3f6a3bff648138625e12531733a3c": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "HTMLModel",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HTMLModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HTMLView",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_1a499ab336df4bc8a04b70d3527a4af6",
"placeholder": "​",
"style": "IPY_MODEL_9d5feeb47a614109a5672841ea846978",
"value": "epoch 3: 100%"
}
},
"74c56cfe7f0f4392b3627cb2c6907057": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "DescriptionStyleModel",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "DescriptionStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"description_width": ""
}
},
"74e7a58d3f1940a884654e645f31485f": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "HTMLModel",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HTMLModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HTMLView",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_1dfbc9c6dbc7457caed5ddc869bf4869",
"placeholder": "​",
"style": "IPY_MODEL_af58010cf4b6458ca05d2bc75c921a7b",
"value": "threshold search:  94%"
}
},
"770463e9a2b9465c89c625cf34f28a7a": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "HTMLModel",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HTMLModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HTMLView",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_cdbce9ddd5034482b576c633b751356b",
"placeholder": "​",
"style": "IPY_MODEL_821ffebb5c6044409920e08051a58c58",
"value": "epoch 4: 100%"
}
},
"77ab293c0f724ccfb340237e0d4c3e2d": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "HTMLModel",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HTMLModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HTMLView",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_13cd0dfa941d4e438b70aab28e5d4b16",
"placeholder": "​",
"style": "IPY_MODEL_74c56cfe7f0f4392b3627cb2c6907057",
"value": " 2000/2000 [07:24&lt;00:00,  4.50it/s, loss=0.1187, p_loss=0.1044, n_loss=0.0350, lr=7.16e-05, tok/s=204,819, rssGB=7.82, vramGB=4.01]"
}
},
"787926cd507e4fdbaa0909cdef6965b1": {
"model_module": "@jupyter-widgets/base",
"model_module_version": "1.2.0",
"model_name": "LayoutModel",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"7a1175be5ea947229f1b446d8bdadca0": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "DescriptionStyleModel",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "DescriptionStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"description_width": ""
}
},
"7ab1270e0f0e4af8b1a54353a77ce20d": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "HTMLModel",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HTMLModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HTMLView",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_ee58e1eaab9146ce9a5612db0ed381ae",
"placeholder": "​",
"style": "IPY_MODEL_6d7db29783a848889ecb3bafb116af05",
"value": " 15/16 [00:01&lt;00:00, 14.64it/s]"
}
},
"7b7cf83dbd214c6ca29340a376ba0188": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "HTMLModel",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HTMLModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HTMLView",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_7d4b47b3692142279f7082bc7e3a62df",
"placeholder": "​",
"style": "IPY_MODEL_45586a223595426990ea412755f7a7c7",
"value": "eval:  94%"
}
},
"7b95b36d83c44f5cb6d270aaab448883": {
"model_module": "@jupyter-widgets/base",
"model_module_version": "1.2.0",
"model_name": "LayoutModel",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"7d4b47b3692142279f7082bc7e3a62df": {
"model_module": "@jupyter-widgets/base",
"model_module_version": "1.2.0",
"model_name": "LayoutModel",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"7d6932638b67462a92160bdacef39fb6": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "HTMLModel",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HTMLModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HTMLView",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_d642239da7dc4df9b26ec5e64e914a3d",
"placeholder": "​",
"style": "IPY_MODEL_98850cebbc794eeb990b15db32ab4e96",
"value": "eval:  94%"
}
},
"8132e5f353f1440d808373a2dd6eba00": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "DescriptionStyleModel",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "DescriptionStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"description_width": ""
}
},
"8170a2dba6594d7db272ad52436d5a37": {
"model_module": "@jupyter-widgets/base",
"model_module_version": "1.2.0",
"model_name": "LayoutModel",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"8180d73c32374a7586b1b20abb63184a": {
"model_module": "@jupyter-widgets/base",
"model_module_version": "1.2.0",
"model_name": "LayoutModel",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"821ffebb5c6044409920e08051a58c58": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "DescriptionStyleModel",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "DescriptionStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"description_width": ""
}
},
"8a0aca6587d34623b6ef752c28ac43bc": {
"model_module": "@jupyter-widgets/base",
"model_module_version": "1.2.0",
"model_name": "LayoutModel",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"8aa3ce282f9241639ce22de6b68bef3c": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "DescriptionStyleModel",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "DescriptionStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"description_width": ""
}
},
"8afb205894df4e94b29b957d204a7aa2": {
"model_module": "@jupyter-widgets/base",
"model_module_version": "1.2.0",
"model_name": "LayoutModel",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"8cd218c09d544e88bd0791469aa6b91e": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "HBoxModel",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HBoxModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HBoxView",
"box_style": "",
"children": [
"IPY_MODEL_770463e9a2b9465c89c625cf34f28a7a",
"IPY_MODEL_f15b2b9fab4c4693b57f9952ce8ec8b7",
"IPY_MODEL_77ab293c0f724ccfb340237e0d4c3e2d"
],
"layout": "IPY_MODEL_9549d84184a24d0cb6ce9867f7ddda9d"
}
},
"8d284c20b9bc437390c4acc737829d80": {
"model_module": "@jupyter-widgets/base",
"model_module_version": "1.2.0",
"model_name": "LayoutModel",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"8f42d0c475ef4c998730c54676dc7571": {
"model_module": "@jupyter-widgets/base",
"model_module_version": "1.2.0",
"model_name": "LayoutModel",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"8f5d53603f784d19a8756b92a13a8329": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "DescriptionStyleModel",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "DescriptionStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"description_width": ""
}
},
"9549d84184a24d0cb6ce9867f7ddda9d": {
"model_module": "@jupyter-widgets/base",
"model_module_version": "1.2.0",
"model_name": "LayoutModel",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"98850cebbc794eeb990b15db32ab4e96": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "DescriptionStyleModel",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "DescriptionStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"description_width": ""
}
},
"9abdd27c6115474886492649d0dc648e": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "HBoxModel",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HBoxModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HBoxView",
"box_style": "",
"children": [
"IPY_MODEL_71b3f6a3bff648138625e12531733a3c",
"IPY_MODEL_ce34796b31bf4df1bddded3a1f487b9e",
"IPY_MODEL_075d17ef882d4194a5ca7127676ed555"
],
"layout": "IPY_MODEL_becccdae788f40de81cf7aaa5431beaf"
}
},
"9d5feeb47a614109a5672841ea846978": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "DescriptionStyleModel",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "DescriptionStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"description_width": ""
}
},
"9f4938e677a04309a230f6a947f8ab6b": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "HTMLModel",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HTMLModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HTMLView",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_3cfe7bfff845466195703171e3eeaa7a",
"placeholder": "​",
"style": "IPY_MODEL_7a1175be5ea947229f1b446d8bdadca0",
"value": " 15/16 [00:01&lt;00:00, 14.73it/s]"
}
},
"9f7eabb7ee5d42bd84c1c58f77c8a300": {
"model_module": "@jupyter-widgets/base",
"model_module_version": "1.2.0",
"model_name": "LayoutModel",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"a0f5901d9d494ce4b2c3c3c1fcabd06e": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "HBoxModel",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HBoxModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HBoxView",
"box_style": "",
"children": [
"IPY_MODEL_09f8dcbd7cc544108e43fead639ce277",
"IPY_MODEL_fd81a797f70a4e5a9043750586774fed",
"IPY_MODEL_554a8cfc16b647a7b6e68a43bb13ef4a"
],
"layout": "IPY_MODEL_7b95b36d83c44f5cb6d270aaab448883"
}
},
"a1e63911f34a4eb99e8b143354c73177": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "FloatProgressModel",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "FloatProgressModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "ProgressView",
"bar_style": "",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_8afb205894df4e94b29b957d204a7aa2",
"max": 16,
"min": 0,
"orientation": "horizontal",
"style": "IPY_MODEL_36c9a51ddc9745029f2de35a79cff975",
"value": 16
}
},
"a281bb601c514bd9971d7e5c7573f534": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "HTMLModel",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HTMLModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HTMLView",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_bbb49fca62f64f219f8623adcb2a703d",
"placeholder": "​",
"style": "IPY_MODEL_2a73f10fb6a847a1bb04916a80c8481d",
"value": "epoch 5: 100%"
}
},
"a38dd16fdb7f4197b965f37a4f68a00b": {
"model_module": "@jupyter-widgets/base",
"model_module_version": "1.2.0",
"model_name": "LayoutModel",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"a416a534366b4539a4ad20edaebca067": {
"model_module": "@jupyter-widgets/base",
"model_module_version": "1.2.0",
"model_name": "LayoutModel",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"a42f75516bbb47ed9290bb4a3c75379c": {
"model_module": "@jupyter-widgets/base",
"model_module_version": "1.2.0",
"model_name": "LayoutModel",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": "hidden",
"width": null
}
},
"a45644ce46e44a7abc1cd9d701ae66fc": {
"model_module": "@jupyter-widgets/base",
"model_module_version": "1.2.0",
"model_name": "LayoutModel",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"a749690d6e5e48d4b9d3314e73830487": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "HTMLModel",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HTMLModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HTMLView",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_ea346f402476488fb4e199783a6ab1e0",
"placeholder": "​",
"style": "IPY_MODEL_e08bfb41a114424f94565f0807c43db4",
"value": "epoch 2: 100%"
}
},
"a9e0b849a2c549ae890181cd80aa6f89": {
"model_module": "@jupyter-widgets/base",
"model_module_version": "1.2.0",
"model_name": "LayoutModel",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": "hidden",
"width": null
}
},
"aab889a6f5114d15a1572b4a96de6434": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "HTMLModel",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HTMLModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HTMLView",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_39d94eae6b534d04946f534afc3250ab",
"placeholder": "​",
"style": "IPY_MODEL_1566aaff7a8c4e119103f8f76b4afb70",
"value": "threshold search:  94%"
}
},
"ac8d0e89705c48ae8d2b82fa6fb615e2": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "ProgressStyleModel",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "ProgressStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"bar_color": null,
"description_width": ""
}
},
"aea1966bdb2143cca681bf847aab023e": {
"model_module": "@jupyter-widgets/base",
"model_module_version": "1.2.0",
"model_name": "LayoutModel",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"aec866f26d7147c6ad2c9175a3d2f076": {
"model_module": "@jupyter-widgets/base",
"model_module_version": "1.2.0",
"model_name": "LayoutModel",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"aefaeeee1ba5490d921955a897f9541c": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "FloatProgressModel",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "FloatProgressModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "ProgressView",
"bar_style": "",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_a416a534366b4539a4ad20edaebca067",
"max": 16,
"min": 0,
"orientation": "horizontal",
"style": "IPY_MODEL_016fd7a3fff3471c8235b05d5e825378",
"value": 16
}
},
"af58010cf4b6458ca05d2bc75c921a7b": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "DescriptionStyleModel",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "DescriptionStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"description_width": ""
}
},
"b02dac4c47d847079d038a86970f5604": {
"model_module": "@jupyter-widgets/base",
"model_module_version": "1.2.0",
"model_name": "LayoutModel",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"b5033174abbc4eb8871ecaa764231ba7": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "DescriptionStyleModel",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "DescriptionStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"description_width": ""
}
},
"bbb49fca62f64f219f8623adcb2a703d": {
"model_module": "@jupyter-widgets/base",
"model_module_version": "1.2.0",
"model_name": "LayoutModel",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"be03d3e6a9c4413890f37cd1fca5cb8f": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "DescriptionStyleModel",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "DescriptionStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"description_width": ""
}
},
"becccdae788f40de81cf7aaa5431beaf": {
"model_module": "@jupyter-widgets/base",
"model_module_version": "1.2.0",
"model_name": "LayoutModel",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"c0b0d68c39db415abe7b6c92e250c451": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "HTMLModel",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HTMLModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HTMLView",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_aea1966bdb2143cca681bf847aab023e",
"placeholder": "​",
"style": "IPY_MODEL_8aa3ce282f9241639ce22de6b68bef3c",
"value": " 15/16 [00:01&lt;00:00, 14.57it/s]"
}
},
"c212d6d4cbea43fa97694baed3630c3a": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "HTMLModel",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HTMLModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HTMLView",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_c2b5e9ce07db4769ac0f51742d3da924",
"placeholder": "​",
"style": "IPY_MODEL_fa9613e7b6cb4ce686d2850d55cb2b2a",
"value": " 15/16 [00:01&lt;00:00, 14.88it/s]"
}
},
"c2b5e9ce07db4769ac0f51742d3da924": {
"model_module": "@jupyter-widgets/base",
"model_module_version": "1.2.0",
"model_name": "LayoutModel",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"c5c231ac276c496891531cf58b7bb21c": {
"model_module": "@jupyter-widgets/base",
"model_module_version": "1.2.0",
"model_name": "LayoutModel",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"c82b97c7519e45f5b1066be8498d5e8d": {
"model_module": "@jupyter-widgets/base",
"model_module_version": "1.2.0",
"model_name": "LayoutModel",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"c88a3823c8474771bd96c2c0b28e91bb": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "DescriptionStyleModel",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "DescriptionStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"description_width": ""
}
},
"cd3f6fff352a45fd9aa853e95cd2706c": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "HBoxModel",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HBoxModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HBoxView",
"box_style": "",
"children": [
"IPY_MODEL_a749690d6e5e48d4b9d3314e73830487",
"IPY_MODEL_f49fc60609dc40dd8bc7f82380f70b5a",
"IPY_MODEL_0b316a18319644108a1b2f70a5d7fe42"
],
"layout": "IPY_MODEL_9f7eabb7ee5d42bd84c1c58f77c8a300"
}
},
"cdbce9ddd5034482b576c633b751356b": {
"model_module": "@jupyter-widgets/base",
"model_module_version": "1.2.0",
"model_name": "LayoutModel",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"ce34796b31bf4df1bddded3a1f487b9e": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "FloatProgressModel",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "FloatProgressModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "ProgressView",
"bar_style": "danger",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_787926cd507e4fdbaa0909cdef6965b1",
"max": 2000,
"min": 0,
"orientation": "horizontal",
"style": "IPY_MODEL_167fc03b177e4db39b2f93b80a5cc4ff",
"value": 2000
}
},
"d063aeb7be1149acaa1b7aa181f9bf56": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "ProgressStyleModel",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "ProgressStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"bar_color": null,
"description_width": ""
}
},
"d39f45b7d83a4f2c87b6c957990958aa": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "HTMLModel",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HTMLModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HTMLView",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_3f67b975573d4cb8b19b5eefbddb06b1",
"placeholder": "​",
"style": "IPY_MODEL_4cd16dfccd85470f8f743a8d932a13bf",
"value": "threshold search:  94%"
}
},
"d578407774bb4d94b9fefc4835a6dd23": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "HTMLModel",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HTMLModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HTMLView",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_c82b97c7519e45f5b1066be8498d5e8d",
"placeholder": "​",
"style": "IPY_MODEL_d938189dfc30407baa674fb25cfda105",
"value": " 2000/2000 [08:14&lt;00:00,  4.50it/s, loss=0.1128, p_loss=0.0976, n_loss=0.0344, lr=3.14e-05, tok/s=184,309, rssGB=7.84, vramGB=4.01]"
}
},
"d642239da7dc4df9b26ec5e64e914a3d": {
"model_module": "@jupyter-widgets/base",
"model_module_version": "1.2.0",
"model_name": "LayoutModel",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"d73a3037907e4dc4ace253e3335e56ab": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "HBoxModel",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HBoxModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HBoxView",
"box_style": "",
"children": [
"IPY_MODEL_a281bb601c514bd9971d7e5c7573f534",
"IPY_MODEL_2d25ee7b589a4d8bbebe8fee2eea44e6",
"IPY_MODEL_d578407774bb4d94b9fefc4835a6dd23"
],
"layout": "IPY_MODEL_e035a8c273764466a6f3b15500706a99"
}
},
"d938189dfc30407baa674fb25cfda105": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "DescriptionStyleModel",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "DescriptionStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"description_width": ""
}
},
"dca5f404bf0c4602917e147a0888a5f0": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "DescriptionStyleModel",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "DescriptionStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"description_width": ""
}
},
"dd3eccb6c46348a69d10948abbf0022d": {
"model_module": "@jupyter-widgets/base",
"model_module_version": "1.2.0",
"model_name": "LayoutModel",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"de023f071cb5492191a56e2c534e9218": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "ProgressStyleModel",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "ProgressStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"bar_color": null,
"description_width": ""
}
},
"e035a8c273764466a6f3b15500706a99": {
"model_module": "@jupyter-widgets/base",
"model_module_version": "1.2.0",
"model_name": "LayoutModel",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"e08bfb41a114424f94565f0807c43db4": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "DescriptionStyleModel",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "DescriptionStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"description_width": ""
}
},
"e1cb8d90b9d841c4adb5d2696208c169": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "ProgressStyleModel",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "ProgressStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"bar_color": null,
"description_width": ""
}
},
"e1d90740ddeb4d4d8b3c6a8fce56d3db": {
"model_module": "@jupyter-widgets/base",
"model_module_version": "1.2.0",
"model_name": "LayoutModel",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"e37050b0a75849588860e86fae4a3de5": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "HBoxModel",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HBoxModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HBoxView",
"box_style": "",
"children": [
"IPY_MODEL_7d6932638b67462a92160bdacef39fb6",
"IPY_MODEL_32a30d4203e94ca7bdaa9f5cef984e09",
"IPY_MODEL_69c363522b2848959affbe12a20a04cd"
],
"layout": "IPY_MODEL_a9e0b849a2c549ae890181cd80aa6f89"
}
},
"e7e889ec45d64224ab02f5bbe75a1e03": {
"model_module": "@jupyter-widgets/base",
"model_module_version": "1.2.0",
"model_name": "LayoutModel",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"ea346f402476488fb4e199783a6ab1e0": {
"model_module": "@jupyter-widgets/base",
"model_module_version": "1.2.0",
"model_name": "LayoutModel",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"ec5985dd680e432f85bffd2b8e1f1c3c": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "FloatProgressModel",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "FloatProgressModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "ProgressView",
"bar_style": "",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_e1d90740ddeb4d4d8b3c6a8fce56d3db",
"max": 16,
"min": 0,
"orientation": "horizontal",
"style": "IPY_MODEL_fbb0d85e685e45c190e7a51919172567",
"value": 16
}
},
"ee58e1eaab9146ce9a5612db0ed381ae": {
"model_module": "@jupyter-widgets/base",
"model_module_version": "1.2.0",
"model_name": "LayoutModel",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"ee941a0b8d3a45d58e960c907b64d9b3": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "HTMLModel",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "HTMLModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "HTMLView",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_c5c231ac276c496891531cf58b7bb21c",
"placeholder": "​",
"style": "IPY_MODEL_6668ab0d821e4711bb580d5c7315ed1a",
"value": " 15/16 [00:01&lt;00:00, 14.68it/s]"
}
},
"eeca4c0bcacf4b548219af0a8b9e5c2d": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "FloatProgressModel",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "FloatProgressModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "ProgressView",
"bar_style": "",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_507e6190c8f24801b6554b99272b6ff6",
"max": 16,
"min": 0,
"orientation": "horizontal",
"style": "IPY_MODEL_de023f071cb5492191a56e2c534e9218",
"value": 16
}
},
"ef7995f9babf447f959df314b89d54a2": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "DescriptionStyleModel",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "DescriptionStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"description_width": ""
}
},
"f15b2b9fab4c4693b57f9952ce8ec8b7": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "FloatProgressModel",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "FloatProgressModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "ProgressView",
"bar_style": "danger",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_8f42d0c475ef4c998730c54676dc7571",
"max": 2000,
"min": 0,
"orientation": "horizontal",
"style": "IPY_MODEL_1473f532e6d94ae496cf13a61130644d",
"value": 2000
}
},
"f49fc60609dc40dd8bc7f82380f70b5a": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "FloatProgressModel",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "FloatProgressModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "ProgressView",
"bar_style": "danger",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_8a0aca6587d34623b6ef752c28ac43bc",
"max": 2000,
"min": 0,
"orientation": "horizontal",
"style": "IPY_MODEL_27a28b493e2c4942b094dde8151623f4",
"value": 2000
}
},
"fa9613e7b6cb4ce686d2850d55cb2b2a": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "DescriptionStyleModel",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "DescriptionStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"description_width": ""
}
},
"fbb0d85e685e45c190e7a51919172567": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "ProgressStyleModel",
"state": {
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "ProgressStyleModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "StyleView",
"bar_color": null,
"description_width": ""
}
},
"fd0343bc7fbc4c9ea14fefac4f7c4d05": {
"model_module": "@jupyter-widgets/base",
"model_module_version": "1.2.0",
"model_name": "LayoutModel",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"fd81a797f70a4e5a9043750586774fed": {
"model_module": "@jupyter-widgets/controls",
"model_module_version": "1.5.0",
"model_name": "FloatProgressModel",
"state": {
"_dom_classes": [],
"_model_module": "@jupyter-widgets/controls",
"_model_module_version": "1.5.0",
"_model_name": "FloatProgressModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/controls",
"_view_module_version": "1.5.0",
"_view_name": "ProgressView",
"bar_style": "danger",
"description": "",
"description_tooltip": null,
"layout": "IPY_MODEL_aec866f26d7147c6ad2c9175a3d2f076",
"max": 2000,
"min": 0,
"orientation": "horizontal",
"style": "IPY_MODEL_70e92fcc70e34d969ac7722f9ec0882b",
"value": 2000
}
},
"fdac6f566099458fa038d11e7aa44452": {
"model_module": "@jupyter-widgets/base",
"model_module_version": "1.2.0",
"model_name": "LayoutModel",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
},
"fefc2eba241e4adfa56ced16e098b6bf": {
"model_module": "@jupyter-widgets/base",
"model_module_version": "1.2.0",
"model_name": "LayoutModel",
"state": {
"_model_module": "@jupyter-widgets/base",
"_model_module_version": "1.2.0",
"_model_name": "LayoutModel",
"_view_count": null,
"_view_module": "@jupyter-widgets/base",
"_view_module_version": "1.2.0",
"_view_name": "LayoutView",
"align_content": null,
"align_items": null,
"align_self": null,
"border": null,
"bottom": null,
"display": null,
"flex": null,
"flex_flow": null,
"grid_area": null,
"grid_auto_columns": null,
"grid_auto_flow": null,
"grid_auto_rows": null,
"grid_column": null,
"grid_gap": null,
"grid_row": null,
"grid_template_areas": null,
"grid_template_columns": null,
"grid_template_rows": null,
"height": null,
"justify_content": null,
"justify_items": null,
"left": null,
"margin": null,
"max_height": null,
"max_width": null,
"min_height": null,
"min_width": null,
"object_fit": null,
"object_position": null,
"order": null,
"overflow": null,
"overflow_x": null,
"overflow_y": null,
"padding": null,
"right": null,
"top": null,
"visibility": null,
"width": null
}
}
}
}
},
"nbformat": 4,
"nbformat_minor": 5
}
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment