[Bugfix][Quantization] Fix FP8 + EP (#13784)
Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com>
This commit is contained in:
@@ -174,7 +174,7 @@ class QuarkW8A8Fp8MoEMethod(QuarkMoEMethod):
|
||||
assert layer.w13_weight_scale is not None
|
||||
shard_size = layer.intermediate_size_per_partition
|
||||
max_w13_scales = layer.w13_weight_scale.max(dim=1).values
|
||||
for expert_id in range(layer.num_experts):
|
||||
for expert_id in range(layer.local_num_experts):
|
||||
start = 0
|
||||
for shard_id in range(2):
|
||||
dq_weight = per_tensor_dequantize(
|
||||
|
||||
Reference in New Issue
Block a user