CQ3-Qwen3-8B-K8-CodeQuant / codequant_baseline.json
sy128's picture
CodeQuant baseline qwen3_8b K8, single-GPU build
2675974 verified
Raw History Blame Contribute Delete
346 Bytes
{
"baseline": "codequant",
"model": "Qwen/Qwen3-8B",
"bits": 3,
"cluster_num": 8,
"calibration_dataset": "mixed",
"max_samples": 512,
"rotation_max_samples": 512,
"sequence_length": 512,
"seconds": 8988.209409672767,
"rotation_seconds": 1182.2467327425256,
"cluster_seconds": 7805.962676930241,
"folded_projections": 180
}