[GCU] Update post_process (#3012)

This commit is contained in:
EnflameGCU
2025-07-25 11:03:03 +08:00
committed by GitHub
parent 7634ffb709
commit 8c167e130c

View File

@@ -772,6 +772,8 @@ class GCUModelRunner(ModelRunnerBase):
post_process( post_process(
sampler_output=sampler_output, sampler_output=sampler_output,
model_output=model_output_data, model_output=model_output_data,
share_inputs=self.share_inputs,
block_size=self.parallel_config.block_size,
speculative_decoding=self.speculative_decoding, speculative_decoding=self.speculative_decoding,
skip_save_output=True, skip_save_output=True,
) )
@@ -989,6 +991,8 @@ class GCUModelRunner(ModelRunnerBase):
post_process( post_process(
sampler_output=sampler_output, sampler_output=sampler_output,
model_output=model_output_data, model_output=model_output_data,
share_inputs=self.share_inputs,
block_size=self.parallel_config.block_size,
save_each_rank=self.parallel_config.use_ep, save_each_rank=self.parallel_config.use_ep,
speculative_decoding=self.speculative_decoding, speculative_decoding=self.speculative_decoding,
skip_save_output=skip_save_output, skip_save_output=skip_save_output,