Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

update for recent transformers updates #636

Merged
merged 3 commits into from
Sep 27, 2023
Merged

update for recent transformers updates #636

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
9dbdcec
update for recent transformers updates
winglian Sep 26, 2023
bb87204
fix checkpoint forward kwargs
winglian Sep 26, 2023
8076bba
just pass args into torch checkpoint
winglian Sep 26, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
18 changes: 16 additions & 2 deletions src/axolotl/monkeypatch/llama_attn_hijack_flash.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -99,6 +99,7 @@ def flashattn_forward(
past_key_value: Optional[Tuple[torch.Tensor]] = None,
output_attentions: bool = False,
use_cache: bool = False,
padding_mask: Optional[torch.LongTensor] = None, # pylint: disable=unused-argument
cu_seqlens: Optional[torch.Tensor] = None,
max_seqlen: Optional[torch.Tensor] = None,
) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
Expand Down Expand Up @@ -476,6 +477,13 @@ def llama_model_forward(
dtype=torch.bool,
device=inputs_embeds.device,
)
padding_mask = None
else:
if 0 in attention_mask:
padding_mask = attention_mask
else:
padding_mask = None

attention_mask = (
self._prepare_decoder_attention_mask( # pylint: disable=protected-access
attention_mask,
Expand Down Expand Up @@ -510,7 +518,9 @@ def llama_model_forward(
def create_custom_forward(module):
def custom_forward(*inputs):
# None for past_key_value
return module(*inputs)
return module(
*inputs,
)

return custom_forward

Expand All @@ -519,9 +529,10 @@ def custom_forward(*inputs):
hidden_states,
attention_mask,
position_ids,
None,
past_key_value,
output_attentions,
None,
padding_mask,
cu_seqlens,
max_seqlen,
)
Expand All @@ -533,6 +544,7 @@ def custom_forward(*inputs):
past_key_value=past_key_value,
output_attentions=output_attentions,
use_cache=use_cache,
padding_mask=padding_mask,
cu_seqlens=cu_seqlens,
max_seqlen=max_seqlen,
)
Expand Down Expand Up @@ -579,6 +591,7 @@ def forward(
past_key_value: Optional[Tuple[torch.Tensor]] = None,
output_attentions: Optional[bool] = False,
use_cache: Optional[bool] = False,
padding_mask: Optional[torch.LongTensor] = None,
cu_seqlens: Optional[torch.Tensor] = None,
max_seqlen: Optional[torch.Tensor] = None,
) -> Tuple[
Expand Down Expand Up @@ -611,6 +624,7 @@ def forward(
past_key_value=past_key_value,
output_attentions=output_attentions,
use_cache=use_cache,
padding_mask=padding_mask,
cu_seqlens=cu_seqlens,
max_seqlen=max_seqlen,
)
Expand Down