Qwen3-4B-Thinking-2507-llmc-awq-calib-codeforces-cots-solutions-train-w4a16-g128-n256-s1024

• method: awq
• weight format: W4A16
• group size: 128
• calibration dataset: open-r1/codeforces-cots
• calibration dataset config: solutions
• calibration split: train
• calibration samples: 256
• max sequence length: 1024