Revert "[Bugfix] Limit profiling run sequence length by max_model_len (#14785) (#14892)

Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
This commit is contained in:
Cyrus Leung
2025-03-16 09:13:46 -07:00
committed by GitHub
parent e53b1350f2
commit f6137adbcb
5 changed files with 0 additions and 9 deletions
-1
View File
@@ -281,7 +281,6 @@ class EncoderDecoderModelRunner(GPUModelRunnerBase[EncoderDecoderModelInput]):
for group_id in range(max_num_seqs):
seq_len = (max_num_batched_tokens // max_num_seqs +
(group_id < max_num_batched_tokens % max_num_seqs))
seq_len = min(seq_len, self.model_config.max_model_len)
batch_size += seq_len
decoder_dummy_data = self.input_registry \
-1
View File
@@ -1302,7 +1302,6 @@ class GPUModelRunnerBase(ModelRunnerBase[TModelInputForGPU]):
for group_id in range(max_num_seqs):
seq_len = (max_num_batched_tokens // max_num_seqs +
(group_id < max_num_batched_tokens % max_num_seqs))
seq_len = min(seq_len, self.model_config.max_model_len)
batch_size += seq_len
dummy_data = self.input_registry \
-1
View File
@@ -148,7 +148,6 @@ class OpenVINOModelRunner(ModelRunnerBase):
seq_len = min(
seq_data.get_len(),
computed_len + seq_group_metadata.token_chunk_size,
self.model_config.max_model_len,
)
if is_prompt:
tokens = seq_data.get_token_ids()[computed_len:seq_len]
-1
View File
@@ -466,7 +466,6 @@ class XPUModelRunner(ModelRunnerBase[ModelInputForXPUWithSamplingMetadata]):
for group_id in range(max_num_seqs):
seq_len = (max_num_batched_tokens // max_num_seqs +
(group_id < max_num_batched_tokens % max_num_seqs))
seq_len = min(seq_len, self.model_config.max_model_len)
batch_size += seq_len
dummy_data = self.input_registry \