Add typo checker in pre-commit (#6179)

Co-authored-by: Brayden Zhong <b8zhong@uwaterloo.ca>
This commit is contained in:
applesaucethebun
2025-05-11 12:55:00 +08:00
committed by GitHub
co-authored by Brayden Zhong
parent de167cf5fa
commit 2ce8793519
99 changed files with 154 additions and 144 deletions
@@ -116,7 +116,7 @@ def deepep_run_moe_deep_preprocess(topk_ids: torch.Tensor, num_experts: int):
seg_indptr = torch.empty(num_experts + 1, device=topk_ids.device, dtype=torch.int64)
src2dst = torch.empty(topk_ids.numel(), device=topk_ids.device, dtype=torch.int64)
# Find offet
# Find offset
expert_ids = torch.arange(
num_experts + 1, device=topk_ids.device, dtype=reorder_topk_ids.dtype
)
+1 -1
View File
@@ -611,7 +611,7 @@ class Fp8EPMoEMethod(Fp8MoEMethod):
self.quant_config.weight_block_size[1],
)
# NOTE(HandH1998): To ensure proper alignment of the block-wise quantization scales, the output_size of the weights for both the gate and up layers must be divisible by block_n.
# Required by collum parallel or enabling merged weights
# Required by column parallel or enabling merged weights
if intermediate_size % block_n != 0:
raise ValueError(
f"The output_size of gate's and up's weight = "
@@ -994,7 +994,7 @@ def get_default_config(
"num_stages": 2 if _is_hip else 4,
}
else:
# Block-wise quant: BLOCK_SIZE_K must be divisable by block_shape[1]
# Block-wise quant: BLOCK_SIZE_K must be divisible by block_shape[1]
config = {
"BLOCK_SIZE_M": 64,
"BLOCK_SIZE_N": block_shape[0],
+1 -1
View File
@@ -270,7 +270,7 @@ def select_experts(
routed_scaling_factor: Optional[float] = None,
):
n_share_experts_fusion = global_server_args_dict["n_share_experts_fusion"]
# DeekSeek V2/V3/R1 serices models uses grouped_top_k
# DeepSeek V2/V3/R1 series models use grouped_top_k
if use_grouped_topk:
assert topk_group is not None
assert num_expert_group is not None