Xiaoyu Zhang 44e86480e8 fuse allreduce and residual_rmsnorm (#8731)

2025-08-11 13:50:53 -07:00

..

attention

[Fix] Fix flashinfer cpu <-> gpu synchronization (#8340)

2025-08-10 03:11:40 +00:00

moe

fuse allreduce and residual_rmsnorm (#8731)

2025-08-11 13:50:53 -07:00

quantization

fix: w4afp8 accuracy problem and rebase (#8752)

2025-08-11 13:41:19 -07:00

activation.py

[AMD] Add silu_and_mul, gelu_and_mul, gelu_tanh_and_mul, and gelu_quick kernels for AMD GPUs (#7135)

2025-07-24 23:44:28 -07:00

amx_utils.py

[CPU] refine CPU integration code (#7647)

2025-07-03 09:51:09 -07:00

communicator.py

fuse allreduce and residual_rmsnorm (#8731)

2025-08-11 13:50:53 -07:00

dp_attention.py

Revert "[bug fix] Ensure local token and global token buffers are pointing to different storage " (#8993)

2025-08-08 21:34:17 -07:00

elementwise.py

Improve streaming, log_level, memory report, weight loading, and benchmark script (#7632)

2025-06-29 23:16:19 -07:00

flashinfer_comm_fusion.py

fuse allreduce and residual_rmsnorm (#8731)

2025-08-11 13:50:53 -07:00

layernorm.py

model: support intern-s1 (#8350)

2025-07-26 13:48:51 -07:00

linear.py

fuse allreduce and residual_rmsnorm (#8731)

2025-08-11 13:50:53 -07:00

logits_processor.py

[CUDA Graph] save cuda graph memory by using next_token_logits_buffer (#8579)

2025-08-03 03:06:47 -07:00

multimodal.py

Support precomputed_embeddings for Llama 4 (#8156)

2025-07-27 01:14:49 -07:00

parameter.py

[CPU][Qwen3 MoE] Enable fused_topk CPU fusion and enhance FP8 TP padding (#7838)

2025-07-09 02:04:21 -07:00

pooler.py

Feat/support rerank (#6058)

2025-06-16 10:50:01 -07:00

radix_attention.py

DP Enhancement (#8280)

2025-07-24 21:36:21 -07:00

rotary_embedding.py

[Feature] Optimize DeepSeek's DeepEP on Ascend NPU (#8355)

2025-08-09 01:35:00 -07:00

sampler.py

Reintroduce tiny fix sampler error when prob is not contiguous (#7354)

2025-06-19 09:39:34 -07:00

torchao_utils.py

Support loading of larger models with on-the-fly quantization (#3061)

2025-01-22 21:33:17 -08:00

utils.py

feat: support cutlass_moe_fp8 kernel for fusedmoe in sm90 (#8678)

2025-08-03 10:47:15 -07:00

vocab_parallel_embedding.py

Replace torch.jit.script with torch.compile in get_masked_input_and_mask to fix benchmark underreporting (#8733)

2025-08-03 19:02:51 -07:00