EngineX-Hygon/sglang
5
0
Fork 0
Code Issues Pull Requests Actions 7 Projects Releases Wiki Activity

Deprecate enable-flashinfer-mla and enable-flashmla (#5480)

Browse Source
This commit is contained in:
Baizhou Zhang
2025-04-17 01:43:33 -07:00
committed by GitHub
parent 4fb05583ef
commit 6fb29ffd9e
6 changed files with 18 additions and 31 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
python/sglang/srt/managers/schedule_batch.py
View File

@@ -76,7 +76,6 @@ global_server_args_dict = {
"device": ServerArgs.device,
"speculative_accept_threshold_single": ServerArgs.speculative_accept_threshold_single,
"speculative_accept_threshold_acc": ServerArgs.speculative_accept_threshold_acc,
"enable_flashmla": ServerArgs.enable_flashmla,
"disable_radix_cache": ServerArgs.disable_radix_cache,
"flashinfer_mla_disable_ragged": ServerArgs.flashinfer_mla_disable_ragged,
"chunked_prefill_size": ServerArgs.chunked_prefill_size,
@@ -1480,7 +1479,7 @@ class ScheduleBatch(ScheduleBatchDisaggregationDecodeMixin):
global_server_args_dict["use_mla_backend"]
and global_server_args_dict["attention_backend"] == "flashinfer"
)
or global_server_args_dict["enable_flashmla"]
or global_server_args_dict["attention_backend"] == "flashmla"
or global_server_args_dict["attention_backend"] == "fa3"
):
seq_lens_cpu = self.seq_lens.cpu()