[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
@@ -319,7 +319,7 @@ class MambaMixer2(CustomOp):
n_groups == 1, # if there was only one group
)
intermediate_settings = (intermediate_size, 0, False)
head_setings = (self.num_heads, 0, False)
head_settings = (self.num_heads, 0, False)
# - the weight already has a "weight_loader" attribute
# which set_weight_attrs will raise if we do not
@@ -372,7 +372,7 @@ class MambaMixer2(CustomOp):
intermediate_settings,
group_shard_settings,
group_shard_settings,
head_setings, # for dt
head_settings, # for dt
],
self.tp_size,
tp_rank,
@@ -516,7 +516,7 @@ def _chunk_state_varlen_kernel(
offs_n[None, :] * stride_chunk_states_dstate)
else:
# - this seems repetitve, buts its to help the compiler
# - this seems repetitive, buts its to help the compiler
if start_idx < pid_c * chunk_size:
past_states_ptrs = chunk_states_ptr + (
offs_m[:, None] * stride_chunk_states_hdim +
@@ -219,7 +219,7 @@ def per_token_group_quant_int8(
quantized tensor along with the scaling factor used for quantization.
Args:
x: The input tenosr with ndim >= 2.
x: The input tensor with ndim >= 2.
group_size: The group size used for quantization.
eps: The minimum to avoid dividing zero.
dtype: The dype of output tensor. Note that only `torch.int8`