Skip to content

GitLab

Commit cba9a28c authored by Casper Hansen's avatar Casper Hansen
Browse files

Clean code

parent 707315a6
Hide whitespace changes
Inline Side-by-side
Showing with 2 additions and 9 deletions
awq/modules/fused/attn.py
View file @ cba9a28c
......@@ -190,15 +190,8 @@ class QuantAttentionFused(nn.Module):
.contiguous()
)
try:
self.cache_v[:bsz, :, self.start_pos : self.start_pos + seqlen, :] = values_store
self.cache_k[:bsz, :, :, self.start_pos : self.start_pos + seqlen, :] = keys_store
except Exception as ex:
print(seqlen, self.max_seq_len)
print(self.cache_v.shape, self.cache_v[:bsz, :, self.start_pos : self.start_pos + seqlen, :].shape, values_store.shape)
print(self.cache_k.shape, self.cache_k[:bsz, :, :, self.start_pos : self.start_pos + seqlen, :].shape, keys_store.shape)
print(ex)
exit(0)
self.cache_v[:bsz, :, self.start_pos : self.start_pos + seqlen, :] = values_store
self.cache_k[:bsz, :, :, self.start_pos : self.start_pos + seqlen, :] = keys_store
if seqlen == 1:
xv = self.cache_v[:bsz, :, : self.start_pos + seqlen, :].transpose(1, 2).contiguous()
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment