qwen3-0.6b-codeforces-cots-test-gguf

This is a GGUF conversion of eigenben/qwen3-0.6b-codeforces-cots-test, which is a LoRA fine-tuned version of Qwen/Qwen3-0.6B.

The model was fine-tuned on the open-r1/codeforces-cots dataset for instruction following on competitive programming problems.

Model Details

• Base Model: Qwen/Qwen3-0.6B
• Fine-tuned Model: eigenben/qwen3-0.6b-codeforces-cots-test
• Training: Supervised Fine-Tuning (SFT) with TRL on 100 examples
• Dataset: open-r1/codeforces-cots (competitive programming problems)
• Format: GGUF (for llama.cpp, Ollama, LM Studio, etc.)

Available Quantization

File	Quant	Size	Description
qwen3-0.6b-codeforces-cots-test-q4_k_m.gguf	Q4_K_M	~300MB	4-bit medium - recommended balance of size and quality

Usage

With llama.cpp

# Download model
huggingface-cli download eigenben/qwen3-0.6b-codeforces-cots-test-gguf qwen3-0.6b-codeforces-cots-test-q4_k_m.gguf

# Run with llama.cpp
./llama-cli -m qwen3-0.6b-codeforces-cots-test-q4_k_m.gguf -p "Your prompt here"

With Ollama

1. Download the GGUF file:

huggingface-cli download eigenben/qwen3-0.6b-codeforces-cots-test-gguf qwen3-0.6b-codeforces-cots-test-q4_k_m.gguf

2. Create a Modelfile:

FROM ./qwen3-0.6b-codeforces-cots-test-q4_k_m.gguf

3. Create and run the model:

ollama create qwen3-codeforces -f Modelfile
ollama run qwen3-codeforces

With LM Studio

1. Download the .gguf file
2. Import into LM Studio
3. Start chatting!

Training Details

This is a test run on 100 examples from the codeforces-cots dataset:

• Training steps: 25
• Training loss: 2.68
• Mean token accuracy: 49.6%

For production use, consider training on the full dataset (35,718 examples).

License

Inherits the license from the base model: Qwen/Qwen3-0.6B

---

Converted to GGUF format using llama.cpp Fine-tuned using TRL on Hugging Face Jobs