Fix num_token_padding support for static per-tensor scaled_fp8_quant (#20188)
Signed-off-by: mgoin <mgoin64@gmail.com>
This commit is contained in:
+1
-2
@@ -1274,8 +1274,7 @@ def scaled_fp8_quant(
|
|||||||
scale = torch.zeros(1, device=input.device, dtype=torch.float32)
|
scale = torch.zeros(1, device=input.device, dtype=torch.float32)
|
||||||
torch.ops._C.dynamic_scaled_fp8_quant(output, input, scale)
|
torch.ops._C.dynamic_scaled_fp8_quant(output, input, scale)
|
||||||
else:
|
else:
|
||||||
# num_token_padding not implemented for this case
|
assert scale.numel() == 1
|
||||||
assert (scale.numel() == 1 and num_token_padding is None)
|
|
||||||
torch.ops._C.static_scaled_fp8_quant(output, input, scale)
|
torch.ops._C.static_scaled_fp8_quant(output, input, scale)
|
||||||
|
|
||||||
return output, scale
|
return output, scale
|
||||||
|
|||||||
Reference in New Issue
Block a user