apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2025-10-05 16:48:03 +08:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

[CP]Glm45 air 2.2 (#4073)
Some checks failed
CE Compile Job / ce_job_pre_check (push) Has been cancelled
Details
CE Compile Job / print_ce_job_pre_check_outputs (push) Has been cancelled
Details
CE Compile Job / FD-Clone-Linux (push) Has been cancelled
Details
CE Compile Job / Show Code Archive Output (push) Has been cancelled
Details
CE Compile Job / BUILD_SM8090 (push) Has been cancelled
Details
CE Compile Job / BUILD_SM8689 (push) Has been cancelled
Details
CE Compile Job / CE_UPLOAD (push) Has been cancelled
Details

Browse Source 
* [Feature] Support zai-org/GLM-4.5-Air BF16 model (#3928)

* support glm45_air

* [Feature] GLM-45-AIR Support Mix Quantization(Dense wfp8afp8 and wint8 triton_moe_backend) (#4051)

* check

* fix v1 load for mix and wint8

* check --quantizations 'None'

* check

* support RL rollout

* check v1 loader

* check glm rollout_model, change wfp8afp8 per_token_cast_to_fp8 to native impl

* check rollout moe gate begin layer_id

* check rollout e_score_correction_bias

* delete infer_to_train_mapping={}

* code check
This commit is contained in:
chen
2025-09-15 18:52:58 +08:00
committed by GitHub
parent 4e8ba62241
commit fbb4e0f8d1
25 changed files with 1505 additions and 170 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
fastdeploy/config.py
View File

@@ -132,6 +132,7 @@ class ModelConfig:
self.eos_tokens_lens: int = 2
self.lm_head_fp32: bool = False
self.model_format = "auto"
self.partial_rotary_factor: float = 1.0
for key, value in args.items():
if hasattr(self, key):
setattr(self, key, value)
@@ -396,7 +397,7 @@ class SpeculativeConfig:
# model for mtp/eagle/draft_model
self.model: Optional[str] = None
# quantization of model
self.quantization: Optional[str] = None
self.quantization: Optional[Dict[str, Any]] = None
# allocate more blocks to prevent mtp from finishing the block earlier than the main model
# Fixed now
self.num_gpu_block_expand_ratio: Optional[float] = 1