{ "created_at": "2026-08-11T04:18:07.460954+00:00", "source_model": "/inspire/qb-ilm/project/video-understanding/public/train/moss_vl_streaming/8B/final_release/MOSS-VL-0708-Instruct-sglang", "output_dir": "/inspire/qb-ilm/project/video-understanding/public/train/moss_vl_streaming/8B/final_release/quant/MOSS-VL-0708-Instruct-SGLang-NF4-KV8", "runtime": "native SGLang MOSS-VL", "weight_format": "bitsandbytes NF4 W4A16 keep-first-last-4", "quantized_linear_count": 240, "kv_cache": "native SGLang FP8 E4M3 via runtime argument", "tp_support": "prequantized bitsandbytes is TP1 only in current SGLang", "peak_gpu_memory_gib_during_conversion": 11.038, "versions": { "torch": "2.8.0+cu128", "transformers": "4.57.1", "bitsandbytes": "0.49.2" } }