spencerkittleson

Local Qwen on RTX 3090 Ti (24GB) via llama.cpp

Running uncensored / reasoning Qwen3.6 models on a single RTX 3090 Ti (24GB VRAM) via llama.cpp's OpenAI-compatible server, wired into OpenCode.

Current primary: Qwopus3.6-27B (CoT fine-tune, working MTP) at 128K context — more accurate on code, ~68-108 tok/s. Fast fallback: Qwen3.6-35B-A3B Wasserstein (uncensored MoE) at 224K — ~174 tok/s.

Note: Qwopus is the one Qwen3.6 GGUF where MTP speculative decoding works in

spencerkittleson

from machine import Pin,I2C,SPI,PWM,ADC
import framebuf
import time
import math

DC = 8
CS = 9
SCK = 10
MOSI = 11

spencerkittleson

import requests
import json
import jwt
from dotenv import dotenv_values


def get_json(endpoint: str):
    response = requests.get(endpoint)
    response.raise_for_status()
    return json.loads(response.content)

spencerkittleson

public async Task<DataResponse> GetAsync(string uri, List<KeyValuePair<string, string>> formValues) {
            try {
                var auth = new AuthenticationHeaderValue("Basic",
                    Convert.ToBase64String(Encoding.UTF8.GetBytes(
                        $"{HttpUtility.UrlEncode(_clientId)}:{HttpUtility.UrlEncode(_clientSecret)}")));
                var request = new HttpRequestMessage(HttpMethod.Post, uri) {
                    Headers = { Authorization = auth },
                    Content = new FormUrlEncodedContent(formValues)
                };
                using var response = await HttpClient.SendAsync(request);

spencerkittleson

import asyncio
import aiohttp
import time


async def get(url, session):
    try:
        async with session.get(url=url) as response:
            resp = await response.read()
            print("Successfully got url {} with resp of length {}.".format(url, len(resp)))

spencerkittleson

# https://python.langchain.com/docs/modules/model_io/models/llms/integrations/llamacpp
import os

from langchain import PromptTemplate, LLMChain
from langchain.llms import LlamaCpp
from langchain import PromptTemplate, LLMChain
from langchain.callbacks.manager import CallbackManager
from langchain.callbacks.streaming_stdout import StreamingStdOutCallbackHandler

template = """Question: {question}

spencerkittleson

from langchain.llms import GPT4All
from langchain.vectorstores import Chroma
from langchain.indexes import VectorstoreIndexCreator
from langchain.indexes.vectorstore import VectorStoreIndexWrapper
from langchain.document_loaders import DirectoryLoader, TextLoader, UnstructuredMarkdownLoader
from langchain.chains import RetrievalQA
import sys
import os
from langchain.callbacks.streaming_stdout import StreamingStdOutCallbackHandler
from langchain.embeddings import HuggingFaceEmbeddings, SentenceTransformerEmbeddings

spencerkittleson

import requests
import xml.etree.ElementTree as ET

def get_current_weather(uri: str, client = requests) -> str:
    result = client.get(uri)
    weather = ET.ElementTree(ET.fromstring(result.text)).getroot().findall("./weather")[0].text
    return weather

spencerkittleson

from simple_http_server import route, server
from threading import Thread
import spotipy
from spotipy.oauth2 import SpotifyClientCredentials
import spotipy
from spotipy.oauth2 import SpotifyOAuth
import os
from simple_http_server import route, server

@route("/")

spencerkittleson

using System;
using System.IO;
using System.Security.Cryptography;
using System.Net.Http;

// Server API Token key and secret are available from API token management dashboard when Server API Token is generated
var key = "da28b6ec3ea350db524d80099f8c9f40f2fab2f4caf91a8d49d4cb4d659a9785";
var secret = "60a63916a77ef70c0ecf42b134f44fbb9732b395ae33dc6d5165aad2b5668bb4";

// include username prefixed with '='