Trevor Morris 58c468f404 Fix FP4 MoE accuracy from missing routed_scaling_factor (#8333)

2025-07-25 16:40:23 -07:00

..

configs

fix bug when eos_ids==0 (#8315)

2025-07-23 23:18:47 -07:00

connector

…

constrained

…

disaggregation

[P/D] Support ipv6 in P/D scenario (#7858)

2025-07-25 08:53:30 -07:00

distributed

[P/D] Support ipv6 in P/D scenario (#7858)

2025-07-25 08:53:30 -07:00

entrypoints

Clean up server_args, triton cache manager (#8332)

2025-07-25 14:14:51 -07:00

eplb

fix moe gate dtype, fix tbo, fix fake dispatch (#7825)

2025-07-19 22:13:46 -07:00

function_call

[Bugfix][Feat] Add XML-ish grammar in EBNFComposer and fix misc bugs in Qwen3 detector (#8357)

2025-07-25 12:03:16 -07:00

layers

Fix FP4 MoE accuracy from missing routed_scaling_factor (#8333)

2025-07-25 16:40:23 -07:00

lora

Introduce Stable LoRA ID System for Overlapped Updates and Prefix Caching (#8261)

2025-07-23 00:32:16 -07:00

managers

fix: kimi k2 xgrammar crash (#8367)

2025-07-25 15:44:01 -07:00

mem_cache

HiCache Storage TP Refinement (#8307)

2025-07-25 08:31:47 +08:00

metrics

feat: add production metric for retracted requests due to insufficient kvcache (#7030)

2025-07-17 11:59:05 -07:00

model_executor

Clean up server_args, triton cache manager (#8332)

2025-07-25 14:14:51 -07:00

model_loader

Clean up server args (#8161)

2025-07-19 11:32:52 -07:00

models

DP Enhancement (#8280)

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

multimodal

fix: retrieve mm token by modality, raise error if none (#8221)

2025-07-22 08:06:35 +08:00

sampling

fix: prevent crashes due to logit bias dimension mismatch (#7685)

2025-07-23 15:30:55 -07:00

speculative

Clean up server_args, triton cache manager (#8332)

2025-07-25 14:14:51 -07:00

_custom_ops.py

[Feature] Integrate quick allreduce and select the best allreduce implementation (#6619)

2025-07-24 20:48:42 -07:00

aio_rwlock.py

…

code_completion_parser.py

feat(oai refactor): Replace openai_api with entrypoints/openai (#7351)

2025-06-21 13:21:06 -07:00

constants.py

Multi-Stage Awake: Support Resume and Pause KV Cache and Weights separately (#7099)

2025-06-19 00:56:37 -07:00

conversation.py

fix: match chat-template for internvl3 (#8262)

2025-07-23 13:29:03 -07:00

custom_op.py

[1/N] MoE Refactor: refactor select_experts (#7966)

2025-07-19 00:51:15 -07:00

debug_utils.py

…

hf_transformers_utils.py

[Minor] Remove redundant print (#8005)

2025-07-14 10:55:13 -07:00

jinja_template_utils.py

vlm: support video as an input modality (#5888)

2025-07-09 23:48:35 -07:00

model_parallel.py

…

operations_strategy.py

feat: support compatibility between MTP and two-batch-overlap (#7225)

2025-06-27 01:10:27 -07:00

operations.py

…

patch_torch.py

…

reasoning_parser.py

[Feat] Add reasoning parser for Qwen/Qwen3-235B-A22B-Thinking-2507 (#8363)

2025-07-25 14:59:42 -07:00

server_args.py

Fix FP4 MoE accuracy from missing routed_scaling_factor (#8333)

2025-07-25 16:40:23 -07:00

torch_memory_saver_adapter.py

Multi-Stage Awake: Support Resume and Pause KV Cache and Weights separately (#7099)

2025-06-19 00:56:37 -07:00

two_batch_overlap.py

DP Enhancement (#8280)

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

utils.py

Clean up server_args, triton cache manager (#8332)

2025-07-25 14:14:51 -07:00

warmup.py

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

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