[Refactor] Remove unused dead code (#40640)
Signed-off-by: yewentao256 <zhyanwentao@126.com>
This commit is contained in:
@@ -3362,7 +3362,6 @@ class GPUModelRunner(
|
||||
logits: torch.Tensor | None,
|
||||
hidden_states: torch.Tensor,
|
||||
num_scheduled_tokens: int,
|
||||
spec_decode_metadata: SpecDecodeMetadata | None,
|
||||
) -> tuple[
|
||||
dict[str, int],
|
||||
LogprobsLists | None,
|
||||
@@ -3630,7 +3629,6 @@ class GPUModelRunner(
|
||||
allow_microbatching=allow_microbatching,
|
||||
num_tokens_padded=num_tokens_padded,
|
||||
uniform_decode=uniform_decode,
|
||||
num_scheduled_tokens_per_request=num_scheduled_tokens_np,
|
||||
cudagraph_mode=cudagraph_mode.value,
|
||||
)
|
||||
)
|
||||
@@ -4308,7 +4306,6 @@ class GPUModelRunner(
|
||||
logits,
|
||||
hidden_states,
|
||||
scheduler_output.total_num_scheduled_tokens,
|
||||
spec_decode_metadata,
|
||||
)
|
||||
|
||||
if propose_drafts_after_bookkeeping:
|
||||
@@ -6540,7 +6537,6 @@ class GPUModelRunner(
|
||||
|
||||
def _reshape_kv_cache_tensors(
|
||||
self,
|
||||
kv_cache_config: KVCacheConfig,
|
||||
kv_cache_raw_tensors: dict[str, torch.Tensor],
|
||||
kernel_block_sizes: list[int],
|
||||
) -> dict[str, torch.Tensor]:
|
||||
@@ -6548,7 +6544,6 @@ class GPUModelRunner(
|
||||
Reshape the KV cache tensors to the desired shape and dtype.
|
||||
|
||||
Args:
|
||||
kv_cache_config: The KV cache config
|
||||
kv_cache_raw_tensors: The KV cache buffer of each layer, with
|
||||
correct size but uninitialized shape.
|
||||
kernel_block_sizes: The kernel block sizes for each KV cache group.
|
||||
@@ -6712,7 +6707,7 @@ class GPUModelRunner(
|
||||
|
||||
# Change the memory buffer to the desired shape
|
||||
kv_caches = self._reshape_kv_cache_tensors(
|
||||
kv_cache_config, kv_cache_raw_tensors, kernel_block_sizes
|
||||
kv_cache_raw_tensors, kernel_block_sizes
|
||||
)
|
||||
|
||||
# Set up cross-layer KV cache sharing
|
||||
|
||||
Reference in New Issue
Block a user