data2json’s gists

data2json / _myatari.py

Created September 23, 2025 04:51 — forked from qpwo/_myatari.py

atari realtime rl runner

	#!/usr/bin/env python3
	import torch, gymnasium as gym, numpy as np, time, sys, threading, os, random
	import torch.multiprocessing as mp
	from torch import Tensor

	from bg_record import log_step, bind_logger, log_close

	# torch.set_num_threads(1)

	NUM_PROCS = 16

data2json / SWP-SFT-PEFT-LORA-Qwen3-4B.py

Created May 27, 2025 22:44

	#!/usr/bin/env python3
	"""
	Complete training pipeline for Qwen3-4B with Super Weight Preservation
	Supports HuggingFace datasets with automatic chat template conversion
	"""

	import torch
	import torch.nn as nn
	from torch.utils.data import DataLoader, Dataset
	import json

data2json / delta.py

Created May 2, 2025 16:50

Parameter Deltas enables zero-cost transfer of post-training capabilities to new base models

	import torch
	from transformers import AutoModelForCausalLM

	# Load models
	llama3_base = AutoModelForCausalLM.from_pretrained("meta-llama/Meta-Llama-3-8B")
	llama3_inst = AutoModelForCausalLM.from_pretrained("meta-llama/Meta-Llama-3-8B-Instruct")
	llama31_base = AutoModelForCausalLM.from_pretrained("meta-llama/Llama-3.1-8B")

	# Calculate Δθ = θpost - θbase
	delta_params = {}

data2json / GSM-OK.py

Created February 12, 2025 04:25

VLLM GSM-8K Test Dataset Evaluation

	import argparse
	from typing import List, Dict, Any, Optional
	from vllm import EngineArgs, LLMEngine, RequestOutput, SamplingParams
	from datasets import load_dataset
	from tqdm import tqdm
	import json
	import re
	from dataclasses import dataclass
	from vllm.utils import FlexibleArgumentParser

data2json / GRPO-Function_Calling_Qwen2.5-0.5B-Instruct.py

Created February 1, 2025 22:04

GRPO-Function_Calling_Qwen2.5-0.5B-Instruct

	import re
	import torch
	import numpy as np
	from datasets import Dataset
	from transformers import AutoTokenizer, AutoModelForCausalLM
	from peft import LoraConfig
	from trl import GRPOConfig, GRPOTrainer

	class SimpleCalculator:
	"""Simple calculator tool implementation"""

data2json / grpo_demo.py

Created February 1, 2025 18:19 — forked from willccbb/grpo_demo.py

GRPO Llama-1B

	# train_grpo.py
	import re
	import torch
	from datasets import load_dataset, Dataset
	from transformers import AutoTokenizer, AutoModelForCausalLM
	from peft import LoraConfig
	from trl import GRPOConfig, GRPOTrainer

	# Load and prep dataset

data2json / OpenAI_Asyncio.py

Last active August 27, 2024 04:24

	import asyncio
	import os
	import json
	from openai import AsyncOpenAI
	from datetime import datetime
	import random

	# Set up the client with custom base URL
	client = AsyncOpenAI(base_url="http://0.0.0.0:8000/v1/")

data2json / LLAMA-3.1-8B-Transformer-Tool-calling.py

Created August 27, 2024 04:02

LLAMA-3.1-8B Template Bug

	essobi@thegreenbox:~/LLaMA-Factory$ more OKG.py
	import torch
	from transformers import AutoModelForCausalLM, AutoTokenizer

	# Load model and tokenizer
	checkpoint = "meta-llama/Meta-Llama-3.1-8B-Instruct"
	tokenizer = AutoTokenizer.from_pretrained(checkpoint)
	model = AutoModelForCausalLM.from_pretrained(
	checkpoint, torch_dtype=torch.bfloat16, device_map="auto"
	)

data2json / not_quite.sh

Created July 27, 2024 02:58


	curl -X POST http://192.168.50.146:8000/v1/chat/completions \
	-H 'accept: application/json' \
	-H 'Content-Type: application/json' \
	-d '{
	"messages": [
	{
	"content": "You are a helpful assistant that can check the weather. Use the get_weather tool when asked about weather conditions. If you choose to call a function ONLY reply in the following format: <{sta
	rt_tag}={function_name}>{parameters}{end_tag} where start_tag => <function parameters => a JSON dict with the function argument name as key and function argument value as value. end_tag => </function>",
	"role": "system"

data2json / l3.sh

Created July 27, 2024 02:22

	curl -s -X POST 'http://0.0.0.0:8000/v1/chat/completions' -H "Content-Type: application/json" -d '{
	"model": "gpt-3.5-turbo",
	"messages": [
	{
	"role": "system",
	"content": "Environment: ipython\nTools: brave_search, wolfram_alpha\n\nCutting Knowledge Date: December 2023\nToday Date: 23 Jul 2024\n\nYou are a helpful Assistant."
	},
	{
	"role": "user",
	"content": "Can you help me solve this equation: x^3 - 4x^2 + 6x - 24 = 0"