fzyzcjy 8b8f2e7463 Support new DeepGEMM input format in silu_and_mul_masked_post_quant_fwd (#7153)

2025-06-13 20:40:24 -07:00

..

configs

Support VILA models (#6106)

2025-06-11 11:47:25 -07:00

connector

feat(remote_model): support variable remote backend for model loader (#3964)

2025-03-14 00:40:44 -07:00

constrained

Correctly abort the failed grammar requests & Improve the handling of abort (#6803)

2025-06-01 19:00:07 -07:00

disaggregation

[Fix] Reduce busy polling when scheduler is idle (#6026)

2025-06-12 14:58:22 -07:00

distributed

support 1 shot allreduce in 1-node and 2-node using mscclpp (#6277)

2025-06-04 22:11:24 -07:00

entrypoints

Support new DeepGEMM format in per token group quant (part 2: srt) (#7155)

2025-06-13 14:25:40 -07:00

eplb_simulator

Minor add utility to read expert distribution recorder output (#7134)

2025-06-12 16:50:05 -07:00

function_call

feat(tool call): Enhance Llama32Detector for improved JSON parsing in non-stream (#6784)

2025-05-30 17:05:17 -07:00

layers

Support new DeepGEMM input format in silu_and_mul_masked_post_quant_fwd (#7153)

2025-06-13 20:40:24 -07:00

lora

[Perf] Refactor LoRAManager to eliminate stream syncs and redundant computations (#6994)

2025-06-11 16:18:57 -07:00

managers

[Fix] Reduce busy polling when scheduler is idle (#6026)

2025-06-12 14:58:22 -07:00

mem_cache

[feat]: Emit fixed-size KV blocks events (#6824)

2025-06-11 13:07:58 -07:00

metrics

Correctly abort the failed grammar requests & Improve the handling of abort (#6803)

2025-06-01 19:00:07 -07:00

model_executor

Simplify the heuristics for setting --mem-fraction-static (#7054)

2025-06-10 19:01:39 -07:00

model_loader

Fix GGuf and add back test_gguf.py (#7067)

2025-06-10 21:07:20 -07:00

models

Re-quantize DeepSeek model weights to support DeepGEMM new input format (#7156)

2025-06-13 15:57:45 -07:00

openai_api

[Feature] Add Logit Bias (#6579)

2025-06-10 15:39:25 -07:00

sampling

[Feature] Add Logit Bias (#6579)

2025-06-10 15:39:25 -07:00

speculative

Open AI API hidden states (#6716)

2025-06-10 14:37:29 -07:00

_custom_ops.py

support 1 shot allreduce in 1-node and 2-node using mscclpp (#6277)

2025-06-04 22:11:24 -07:00

aio_rwlock.py

A better aio rwlock that guarantees the order (#2547)

2024-12-22 15:44:32 -08:00

code_completion_parser.py

Add typo checker in pre-commit (#6179)

2025-05-11 12:55:00 +08:00

conversation.py

Support VILA models (#6106)

2025-06-11 11:47:25 -07:00

custom_op.py

[fix] recover auto-dispatch for rmsnorm and rope (#6745)

2025-06-03 21:44:20 -07:00

debug_utils.py

Add simple utility to dump tensors for debugging (#6815)

2025-06-02 11:15:31 -07:00

hf_transformers_utils.py

[New Model] Devstral support (#6547)

2025-05-26 19:27:48 -07:00

mm_utils.py

refactor: minor refactors regarding multimodal processing (#6187)

2025-05-17 22:53:20 -07:00

model_parallel.py

Make torch TP composable with torchao (#2436)

2024-12-11 04:21:42 -08:00

operations_strategy.py

Overlap two kernels in DeepSeek with communication (#6711)

2025-05-28 15:53:51 -07:00

operations.py

Support overlapping two batches (#4068)

2025-05-24 17:39:07 -07:00

patch_torch.py

Fix torch.compile cacheing (#5259)

2025-04-10 18:08:45 -07:00

reasoning_parser.py

Add typo checker in pre-commit (#6179)

2025-05-11 12:55:00 +08:00

server_args.py

[Fix] Reduce busy polling when scheduler is idle (#6026)

2025-06-12 14:58:22 -07:00

torch_memory_saver_adapter.py

[CI] Fix ci tests (#5769)

2025-04-27 07:18:10 -07:00

two_batch_overlap.py

Remove unnecessary kernels of num_token_non_padded (#6965)

2025-06-08 05:09:17 -07:00

utils.py

vlm: adapt internvl to VisionAttention (#6870)

2025-06-11 01:16:04 -07:00

warmup.py

Support penalty in overlap mode; return logprob with chunked prefill; improve benchmark scripts (#3988)

2025-03-03 00:12:04 -08:00