EngineX-Cambricon/enginex-mlu370-vllm
10
0
Fork 1
Code Issues Pull Requests Actions Projects Releases Wiki Activity

add deepseekv3 and llama4

Browse Source
This commit is contained in:
Chranos
2026-02-11 15:27:19 +08:00
parent 9b05d7285e
commit f6d6f69abc
2 changed files with 9 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
vllm-v0.6.2/vllm/spec_decode/mlu_spec_decode_worker.py
View File

@@ -159,9 +159,11 @@ class MLUSpecDecodeWorker(LoraNotSupportedWorkerBase):
draft_worker_kwargs[
"model_runner_cls"] = MLUTP1DraftModelRunner
else:
if draft_model_config.hf_config.model_type == "eagle":
if draft_model_config.hf_config.model_type in (
"eagle", "deepseek_mtp"):
raise NotImplementedError(
"EAGLE does not support TP > 1 yet")
f"{draft_model_config.hf_config.model_type} "
"does not support TP > 1 yet")
allow_zero_draft_token_step = False
proposer_worker = MLUMultiStepWorker(**draft_worker_kwargs)