[Bugfix] Size and iterate w13 by shard count for non-gated MoE (#51125)

Signed-off-by: aoshen02 <aoshen@inferact.ai>
Co-authored-by: Matej Sirovatka <matej.sirovatka@gmail.com>
Co-authored-by: Rishi Puri <riship@nvidia.com>
Co-authored-by: Hsiao-Yuan Chen <littlecircle0730@users.noreply.github.com>
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
This commit is contained in:
aoshen02
2026-08-05 17:24:47 -07:00
committed by GitHub
co-authored by Matej Sirovatka Rishi Puri Hsiao-Yuan Chen Claude Opus 5
parent 811622c410
commit 7c77868cdf
16 changed files with 104 additions and 80 deletions
@@ -215,7 +215,7 @@ class GptOssMxfp4MoEMethod(FusedMoEMethodBase):
w13_weight = torch.nn.Parameter(
torch.zeros(
num_experts,
2 * intermediate_size_per_partition,
self.moe.w13_num_shards * intermediate_size_per_partition,
hidden_size // 2,
dtype=weight_dtype,
),
@@ -227,7 +227,7 @@ class GptOssMxfp4MoEMethod(FusedMoEMethodBase):
w13_weight_scale = torch.nn.Parameter(
torch.zeros(
num_experts,
2 * intermediate_size_per_partition,
self.moe.w13_num_shards * intermediate_size_per_partition,
hidden_size // mxfp4_block,
dtype=scale_dtype,
),
@@ -267,7 +267,7 @@ class GptOssMxfp4MoEMethod(FusedMoEMethodBase):
w13_bias = torch.nn.Parameter(
torch.zeros(
num_experts,
2 * intermediate_size_per_partition,
self.moe.w13_num_shards * intermediate_size_per_partition,
dtype=torch.bfloat16,
),
requires_grad=False,
@@ -305,13 +305,13 @@ class GptOssMxfp4MoEMethod(FusedMoEMethodBase):
assert (
w13.dim() == 3
and w13.shape[0] == num_experts
and w13.shape[1] == intermediate_size * 2
and w13.shape[1] == intermediate_size * self.moe.w13_num_shards
and w13.shape[2] == hidden_size // 2
)
assert (
w13_scale.dim() == 3
and w13_scale.shape[0] == num_experts
and w13_scale.shape[1] == intermediate_size * 2
and w13_scale.shape[1] == intermediate_size * self.moe.w13_num_shards
and w13_scale.shape[2] == hidden_size // sf_block_size
)
assert (
@@ -329,7 +329,7 @@ class GptOssMxfp4MoEMethod(FusedMoEMethodBase):
assert (
w13_bias.dim() == 2
and w13_bias.shape[0] == num_experts
and w13_bias.shape[1] == intermediate_size * 2
and w13_bias.shape[1] == intermediate_size * self.moe.w13_num_shards
)
if w2_bias is not None:
assert (
@@ -599,7 +599,7 @@ class Mxfp4MoEMethod(FusedMoEMethodBase):
w13_weight = torch.nn.Parameter(
torch.zeros(
num_experts,
2 * intermediate_size_per_partition,
self.moe.w13_num_shards * intermediate_size_per_partition,
hidden_size // 2,
dtype=weight_dtype,
),
@@ -611,7 +611,7 @@ class Mxfp4MoEMethod(FusedMoEMethodBase):
w13_weight_scale = torch.nn.Parameter(
torch.zeros(
num_experts,
2 * intermediate_size_per_partition,
self.moe.w13_num_shards * intermediate_size_per_partition,
hidden_size // mxfp4_block,
dtype=scale_dtype,
),
@@ -651,7 +651,7 @@ class Mxfp4MoEMethod(FusedMoEMethodBase):
w13_bias = torch.nn.Parameter(
torch.zeros(
num_experts,
2 * intermediate_size_per_partition,
self.moe.w13_num_shards * intermediate_size_per_partition,
dtype=torch.bfloat16,
),
requires_grad=False,
@@ -689,13 +689,13 @@ class Mxfp4MoEMethod(FusedMoEMethodBase):
assert (
w13.dim() == 3
and w13.shape[0] == num_experts
and w13.shape[1] == intermediate_size * 2
and w13.shape[1] == intermediate_size * self.moe.w13_num_shards
and w13.shape[2] == hidden_size // 2
)
assert (
w13_scale.dim() == 3
and w13_scale.shape[0] == num_experts
and w13_scale.shape[1] == intermediate_size * 2
and w13_scale.shape[1] == intermediate_size * self.moe.w13_num_shards
and w13_scale.shape[2] == hidden_size // sf_block_size
)
assert (
@@ -713,7 +713,7 @@ class Mxfp4MoEMethod(FusedMoEMethodBase):
assert (
w13_bias.dim() == 2
and w13_bias.shape[0] == num_experts
and w13_bias.shape[1] == intermediate_size * 2
and w13_bias.shape[1] == intermediate_size * self.moe.w13_num_shards
)
if w2_bias is not None:
assert (