Commit Graph
33 Commits
Author SHA1 Message Date
JieXin Liang 69af3ec35f [doc] add note for get_num_kv_splits in triton_backend (#6444) 2025-05-19 21:40:21 -07:00
JieXin LiangandStefan He 1f30c05d4a [fix] fix fa3 forward_decode with spec_decode (#6395)
Co-authored-by: Stefan He <hebiaobuaa@gmail.com>
2025-05-18 12:50:15 -07:00
JieXin Liang 9a405274e2 [misc] remove redundant platform codes (#6298) 2025-05-15 00:51:30 -07:00
JieXin Liang 17299f088a [misc] deep_gemm fallback to NVRTC when NVCC not found (#6252) 2025-05-13 01:41:35 -07:00
JieXin Liang e9bebafb19 [fix] remove mixtral from is_fa3_default_architecture (#6191) 2025-05-11 00:15:54 -07:00
JieXin Liang c178abdabc [fix] fix determine_n_share_experts_fusion (#6118) 2025-05-10 01:19:09 -07:00
JieXin Liang 5e02330137 [perf] dsv3 bmm fallback to bf16 (#5662) 2025-05-08 11:43:39 -07:00
JieXin Liang f1ff736d68 [fix] fix pyproject.toml dependencies (#6119) 2025-05-08 02:14:36 -07:00
JieXin Liang b70957fcf8 [refactor] slightly tidy fp8 module (#5993) 2025-05-07 17:28:24 -07:00
JieXin Liang 8ebde73f7d [perf] H100 DeepSeek-V3 fused moe tuned config (#5998) 2025-05-03 14:02:26 -07:00
JieXin Liang 3cff963335 [fix] kimi-vl test in test_vision_openai_server.py (#5910) 2025-04-29 23:59:10 -07:00
JieXin Liangandalcanerian e4b6133b78 [fix] relax mem_fraction_static for h200 (#5893)
Co-authored-by: alcanerian <alcanerian@gmail.com>
2025-04-29 17:01:12 -07:00
JieXin Liang f55933e1cc [misc] more decode step log for batch_one_batch (#5565) 2025-04-26 19:50:28 -07:00
JieXin Liang 3c4dc38a9a [fix] fix bench_one_batch_server (#5607) 2025-04-26 18:49:45 -07:00
JieXin Liang eebfdb9459 [fix] fix potential bumpy throughtput with deepgemm (#5722) 2025-04-26 18:12:48 -07:00
JieXin Liangandalcanerian 917324862e [fix] reduce dp capture bs (#5634)
Co-authored-by: alcanerian <alcanerian@gmail.com>
2025-04-22 11:08:45 -07:00
JieXin Liangandispobock 4418f599a5 Fix FA3 DeepSeek prefill performance regression (#5624)
Co-authored-by: ispobock <ispobaoke@gmail.com>
2025-04-22 01:41:41 -07:00
JieXin Liang 506be6b892 [fix] fix compile_deep_gemm missing kv_b_proj (#5620) 2025-04-22 00:06:36 -07:00
JieXin Liang 2343d8df7d [fix] force use deepgemm in compile_deep_gemm (#5618) 2025-04-21 21:36:02 -07:00
JieXin Liang c2942907d5 [feature] enable pre compile jit deep_gemm (#5580) 2025-04-21 16:52:53 -07:00
JieXin Liang 97cb762bb6 [misc] remove is_cuda_available (#5319) 2025-04-20 18:16:51 -07:00
JieXin Liang bca832c7c6 [Fix] fix outlines and xgrammar (#4947) 2025-04-20 13:31:25 -07:00
JieXin Liang 99456bcacb [perf] introduce deep gemm group_gemm_masked as bmm (#5432) 2025-04-20 00:38:27 -07:00
JieXin Liang f88f7e1943 [misc] fix ci flaky case (#5352) 2025-04-15 01:37:16 -07:00
JieXin Liang bdde237562 [perf] experimental enhance fp8 per-tensor quant (#5370) 2025-04-14 12:35:43 -07:00
JieXin Liang a995a773a0 [fix] remove cuda_device_count_stateless (#5060) 2025-04-04 00:18:26 -07:00
JieXin Liang 51ac297ace [feat] interface for platforms abstraction (#4928) 2025-03-31 00:04:21 -07:00
JieXin Liang 9e93ef3f8e [fix] fix illegal mem access and clean up triton attention backend (#4571) 2025-03-20 02:01:52 -07:00
JieXin Liang 4942074174 [fix] fix initialization of _ENABLE_TORCH_INFERENCE_MODE (#4549) 2025-03-19 09:57:59 -07:00
JieXin Liang c0e9a36c5f Optimize Triton decoding kernel for dynamic workload (#4553) 2025-03-18 21:25:38 -07:00
JieXin Liang 0212d2e288 [Fix] use torch.inference_mode() instead of torch.no_grad() (#4372) 2025-03-16 22:54:16 -07:00
JieXin Liang 1a3fa75f2f [Fix] use torch.cat instead of torch.concat to prevent entering the Autograd backends. (#4466) 2025-03-16 00:02:47 -07:00
JieXin Liang 0540fef7a1 [Fix] fix _yarn_linear_ramp_mask with device parameter (#4337) 2025-03-12 02:28:19 -07:00