apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2025-10-04 00:06:38 +08:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

[BugFix] num_seqs (#3291)

Browse Source 
* fix num_seqs

* merge develop
This commit is contained in:
lizexu123
2025-08-11 13:38:55 +08:00
committed by GitHub
parent b88537a456
commit e0aeac58e1
3 changed files with 6 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
fastdeploy/worker/gpu_model_runner.py
View File

@@ -795,7 +795,7 @@ class GPUModelRunner(ModelRunnerBase):
output_padding_offset,
) = pre_process(
self.share_inputs["input_ids"],
self.share_inputs["seq_lens_this_time"],
getattr(self.share_inputs, "seq_lens_this_time", self.seq_lens_this_time_buffer),
self.speculative_decoding,
(self.share_inputs["draft_tokens"] if self.speculative_decoding else None),
self.share_inputs["seq_lens_encoder"],
@@ -880,7 +880,7 @@ class GPUModelRunner(ModelRunnerBase):
max_len_tensor_cpu=self.share_inputs["max_len_tensor_cpu"],
seq_lens_encoder=self.share_inputs["seq_lens_encoder"],
seq_lens_decoder=self.share_inputs["seq_lens_decoder"],
seq_lens_this_time=self.share_inputs["seq_lens_this_time"],
seq_lens_this_time=getattr(self.share_inputs, "seq_lens_this_time", self.seq_lens_this_time_buffer),
batch_id_per_token=self.share_inputs["batch_id_per_token"],
cu_seqlens_q=self.share_inputs["cu_seqlens_q"],
cu_seqlens_k=self.share_inputs["cu_seqlens_k"],