YouJiacheng

Credit: How to write ML papers by Neel Nanda

You are chatbot that gives constructive analysis of the following work. Specifically, you care about the following criteria:

## Core Narrative Quality
- **Clear Claims**: Contains 1-3 specific, concrete claims that fit within a cohesive theme
- **Strong Motivation**: Clearly explains why readers should care ("so what?")

YouJiacheng

def cdf(t: Tensor, thresholds: list[float]):
    level = torch.bucketize(t, t.new_tensor(thresholds, dtype=torch.float32), out_int32=True) # sum(x > v for v in thresholds)
    count = t.new_zeros(len(thresholds) + 1, dtype=torch.int32)
    count.index_put_((level,), count.new_ones(1), True)
    return count.cumsum(0) / t.numel()
  

YouJiacheng

import time
import unicodedata

import regex as re

regex_pattern = r"(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\r\n\p{L}\p{N}]?\p{L}+|\p{N}| ?[^\s\p{L}\p{N}]+[\r\n]*|\s*[\r\n]+|\s+(?!\S)|\s+"

# --- Helper Functions for Character Properties ---

YouJiacheng

with open(__file__) as f:
    code = f.read() # read the code of this file ASAP, for logging
import copy
from dataclasses import dataclass
from functools import lru_cache
import os
from pathlib import Path
import time
import uuid

YouJiacheng

[project]
name = "modded-nanogpt"
version = "0.1.0"
description = "Add your description here"
readme = "README.md"
requires-python = "==3.12.*"
dependencies = [
    "numpy>=2.1.3",
    "torch",
    "pytorch-triton",

YouJiacheng

from torch._logging._internal import trace_structured # noqa: E402
import torch._inductor.codecache # noqa: E402
import torch._inductor.graph # noqa: E402
def _patched_trace_structured(name, *args, **kwargs):
    if name == "inductor_output_code":
        match args, kwargs:
            case (metadata_fn, *_), _:
                filename = metadata_fn().get("filename", "Unknown")
            case _, {"metadata_fn": metadata_fn}:
                filename = metadata_fn().get("filename", "Unknown")

YouJiacheng

from functools import partial

import jax
import jax.numpy as jnp
import optax


def poly(x: jnp.ndarray, w: jnp.ndarray):
    assert w.shape == (3,)
    w = w.astype(jnp.float32)

YouJiacheng

import os
import sys

with open(sys.argv[0]) as f:
    code = f.read() # read the code of this file ASAP, for logging
os.environ["PYTORCH_CUDA_ALLOC_CONF"] = "expandable_segments:True"
import time
import uuid
from dataclasses import dataclass
from functools import lru_cache, partial

YouJiacheng

import os
import sys

with open(sys.argv[0]) as f:
    code = f.read() # read the code of this file ASAP, for logging
os.environ["PYTORCH_CUDA_ALLOC_CONF"] = "expandable_segments:True"
import time
import uuid
from dataclasses import dataclass
from functools import lru_cache, partial

YouJiacheng

import os
import sys
from typing import override

with open(sys.argv[0]) as f:
    code = f.read() # read the code of this file ASAP, for logging
os.environ["PYTORCH_CUDA_ALLOC_CONF"] = "expandable_segments:True"
import contextlib
import time
import uuid