Evaluation Results

Benchmark	Score
GSM8K (test)	2.00%
MMLU-Pro (test)	4.00%

Results obtained via local evaluation. Given the model size (0.2B parameters), low benchmark scores are expected.

Model Usage

python

import torch
from transformers import AutoTokenizer, AutoModelForCausalLM

model_path = "FlameF0X/Qwen2-0.2B-it"

tokenizer = AutoTokenizer.from_pretrained(model_path, trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained(
    model_path,
    torch_dtype="auto",
    device_map="auto",
    trust_remote_code=True
)

messages = [
    {"role": "system", "content": "You are a helpful assistant."},
    {"role": "user", "content": "Explain how a transformer model works in one sentence."}
]

text = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True
)

model_inputs = tokenizer([text], return_tensors="pt").to(model.device)

generated_ids = model.generate(
    **model_inputs,
    max_new_tokens=128,
    do_sample=True,
    temperature=0.7
)

generated_ids = [
    output_ids[len(input_ids):]
    for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)
]

response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
print(f"--- Assistant Response ---\n{response}")

Training Data

This model was instruction-tuned on a mixture of:

Salesforce/wikitext — General text
roneneldan/TinyStories — Short story generation
FlameF0X/arXiv-AI-ML — AI/ML research papers
Skylion007/openwebtext — Web text
flytech/python-codes-25k — Python code
bookcorpus/bookcorpus — Books
HuggingFaceH4/ultrachat_200k — Instruction following
openai/gsm8k — Math reasoning
microsoft/orca-math-word-problems-200k — Math word problems
laion/OIG — Open instruction generalist
microsoft/wiki_qa — Question answering