yangjianfengo1 93fcf7e4ec 【New Feature】W4afp8 supports per group quantization (#4272) ...

* w4afp8 支持per group

* code style

* 精度完成

* revert append attn utils

* ffn1 动态量化

* ffn2 支持动态量化

* code style

* code style

* 修改单测

* 修改单测

* fix bug

* Implement conditional parameter creation for layers

Add parameter creation for up_gate_proj_in_scale when ep_size > 1.

* code style

* fix conflict

* code style

* code style

* 修复w4aint8 精度

* fix ci

---------

Co-authored-by: yuanxiaolan <yuanxiaolan01@baidu.com>

2025-11-05 21:00:23 +08:00

..

benchmarks

[CLI]Update parameters in bench latecy cli tool and fix collect-env cli tool (#4558)

2025-10-24 16:46:45 +08:00

cache_manager

[bugfix] kill cache_transfer_manager process (#4401)

2025-10-16 20:45:24 +08:00

ce

[CI] fix some ci yaml (#4747)

2025-11-02 21:28:04 +08:00

ci_use

support ep+tp at op layer (#4688)

2025-11-05 11:15:57 +08:00

distributed

[BugFix]Dev fix custom ar unstable result (#4437)

2025-10-17 11:47:16 +08:00

e2e

Update Unit Test for PaddleOCR-VL (#4802)

2025-11-04 22:40:15 +08:00

engine

fix total_block_num init error in worker_process (#4687)

2025-10-30 19:53:09 +08:00

entrypoints

fix parser register name (#4795)

2025-11-05 11:27:30 +08:00

graph_optimization

remove input_ids from ForwardMeta (#4793)

2025-11-05 11:55:51 +08:00

input

Update Unit Test for PaddleOCR-VL (#4802)

2025-11-04 22:40:15 +08:00

inter_communicator

perf: Optimize task queue communication from engine to worker (#4531)

2025-10-25 22:45:38 +08:00

layers

remove input_ids from ForwardMeta (#4793)

2025-11-05 11:55:51 +08:00

logger

[BugFix] PaddleOCR-VL fix FD_DEBUG type and support v1 loader (#4605)

2025-10-28 09:47:47 +08:00

metrics

[fix] Fixed the issue of excessive/redundant spans being returned for streaming requests. (#4375)

2025-10-15 11:47:47 +08:00

model_executor

remove input_ids from ForwardMeta (#4793)

2025-11-05 11:55:51 +08:00

model_loader

[CI]delete test_common_model (#4794)

2025-11-04 13:57:55 +08:00

multimodal

[Feature] mm support prefix cache (#4134)

2025-10-27 17:39:51 +08:00

operators

【New Feature】W4afp8 supports per group quantization (#4272)

2025-11-05 21:00:23 +08:00

output

[Feature] Add timestamp for profiler (#4726)

2025-11-05 12:04:59 +08:00

platforms

【Hackathon 9th No.70】supplementary unit test for CPUPlatform and CUDAPlatform (#3580)

2025-08-29 10:34:05 +08:00

plugins

[Feature] support pool (#3827)

2025-09-22 14:09:09 +08:00

pooling

[Feature] support pooling model runner (#4590)

2025-10-31 22:32:05 +08:00

quantization

[BugFix]Fix load kv cache quant scale (#4077)

2025-09-12 17:44:03 +08:00

reasoning

[Feature] Unify the registration name recognition for tool_parser and reasoning_parser to “-” (#4668)

2025-10-31 10:45:27 +08:00

utils

【feature】support n parameter (#4273)

2025-10-17 20:51:59 +08:00

v1

[Feature] mm support prefix cache (#4134)

2025-10-27 17:39:51 +08:00

conftest.py

[Executor] Default use CUDAGraph (#3594)

2025-10-21 14:25:45 +08:00

cov_pytest.ini

[CI] Handle unit test issues (#4483)

2025-10-20 10:13:21 +08:00

utils.py

[FDConfig]Remove max_model_len in FDConfig (#4350)

2025-10-11 14:04:17 +08:00