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

Add mark step and inplace residual add in llama model code to reduce memory consumption #65

Merged
7 commits merged into from
Feb 29, 2024
Merged

Add mark step and inplace residual add in llama model code to reduce memory consumption #65

Changes from 1 commit
Commits
Show all changes
7 commits
Select commit Hold shift + click to select a range
389938a
Add mark step and inplace add.
Feb 23, 2024
69b5b32
Merge branch 'HabanaAI:habana-main' into llama_prefill_memoryfixes
Feb 26, 2024
094f0b7
Add lazy mode parameter
Feb 27, 2024
586e8ce
Move mark step within the loop
Feb 28, 2024
f34fae3
Move mark step before the loop
Feb 28, 2024
6c6e5da
Fix indentation
Feb 29, 2024
8eab266
update in place add only for inference
Feb 29, 2024
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
Prev Previous commit
update in place add only for inference
  • Loading branch information
Puneesh Khanna authored Feb 29, 2024
commit 8eab266b339b013aea80750b5436896952606894
15 changes: 11 additions & 4 deletions optimum/habana/transformers/models/llama/modeling_llama.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -551,8 +551,12 @@ def pre_attn(
def post_attn_pre_mlp(self, hidden_states, residual):
hidden_states = self.self_attn.post_attn_forward(hidden_states)

residual.add_(hidden_states)
hidden_states = residual
if self.training:
hidden_states = hidden_states + residual
residual = hidden_states
else:
residual.add_(hidden_states)
hidden_states = residual

hidden_states = self.post_attention_layernorm(hidden_states)

Expand All @@ -562,8 +566,11 @@ def post_attn_pre_mlp(self, hidden_states, residual):
def post_mlp(self, hidden_states, residual):
hidden_states = self.mlp.post_mlp_forward(hidden_states)

residual.add_(hidden_states)
hidden_states = residual
if self.training:
hidden_states = hidden_states + residual
else:
residual.add_(hidden_states)
hidden_states = residual

return hidden_states

Expand Down