Fix torch.compile
with fullgraph=True
when attention_mask
input is used
#36885
Job | Run time |
---|---|
18m 18s | |
18m 18s |
torch.compile
with fullgraph=True
when attention_mask
input is used
#36885
Job | Run time |
---|---|
18m 18s | |
18m 18s |