[CI] change spell checker from codespell to typos (#18711)

Signed-off-by: Andy Xie <andy.xning@gmail.com>
This commit is contained in:
Ning Xie
2025-06-11 19:57:10 -07:00
committed by GitHub
parent 42f52cc95b
commit 2f1c19b245
57 changed files with 335 additions and 163 deletions
+1 -1
View File
@@ -886,7 +886,7 @@ class HPUModelRunnerBase(ModelRunnerBase[TModelInputForHPU]):
num_decode_tokens=0,
slot_mapping=slot_mapping,
multi_modal_placeholder_index_maps=
None, # FIXME(kzawora): mutli-modality will not work here
None, # FIXME(kzawora): multi-modality will not work here
enable_kv_scales_calculation=False,
)
multi_modal_kwargs = MultiModalKwargs.batch(multi_modal_kwargs_list)
+1 -1
View File
@@ -277,7 +277,7 @@ class StatefulModelInput(BroadcastableModelInput):
assert fmi.input_tokens.shape[0] >= self.num_seqs
fmi_new_input_tokens: torch.Tensor = fmi.input_tokens[:self.num_seqs]
# Update frozen_model_input::input_positons.
# Update frozen_model_input::input_positions.
assert fmi.input_positions is not None
assert fmi.input_positions.shape[0] >= self.num_seqs
fmi_new_input_positions: torch.Tensor = fmi.input_positions[:self.
+8 -8
View File
@@ -798,9 +798,9 @@ class ModelWrapper(nn.Module):
"""
batch_size, seq_len = token_ids.shape
# Calculate the positions to sample from.
start_indicies = torch.arange(
start_indices = torch.arange(
batch_size, dtype=torch.int32, device=input_lens.device) * seq_len
logits_indices = start_indicies + input_lens - 1
logits_indices = start_indices + input_lens - 1
attn_metadata = get_forward_context().attn_metadata
# FIXME(woosuk): This is a temporary hack to avoid using the existing
@@ -822,14 +822,14 @@ class ModelWrapper(nn.Module):
num_kv_heads, num_blocks, block_size, _ = kv_caches[0][0].shape
slot_mapping = attn_metadata.slot_mapping
slot_mapping = slot_mapping.flatten()
head_indicies = torch.arange(0,
num_kv_heads,
device=slot_mapping.device,
dtype=slot_mapping.dtype)
head_indicies *= block_size * num_blocks
head_indices = torch.arange(0,
num_kv_heads,
device=slot_mapping.device,
dtype=slot_mapping.dtype)
head_indices *= block_size * num_blocks
slot_mapping = slot_mapping.repeat_interleave(num_kv_heads).view(
-1, num_kv_heads)
slot_mapping = slot_mapping + head_indicies.view(1, -1)
slot_mapping = slot_mapping + head_indices.view(1, -1)
slot_mapping = slot_mapping.flatten()
attn_metadata.slot_mapping = slot_mapping