-
Notifications
You must be signed in to change notification settings - Fork 2.9k
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.
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
[LLM] Add deepseekv2 #9061
base: develop
Are you sure you want to change the base?
[LLM] Add deepseekv2 #9061
Conversation
Thanks for your contribution! |
Codecov ReportAttention: Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## develop #9061 +/- ##
===========================================
- Coverage 53.07% 52.73% -0.35%
===========================================
Files 656 659 +3
Lines 106095 107051 +956
===========================================
+ Hits 56309 56449 +140
- Misses 49786 50602 +816 ☔ View full report in Codecov by Sentry. |
if version != "0.0.0" and version <= "2.5.2": | ||
attn_output, attn_weights = flash_attention( | ||
query_states, | ||
key_states, | ||
value_states, | ||
causal=True, | ||
return_softmax=output_attentions, | ||
) | ||
attn_output *= (head_dim ** (0.5)) * softmax_scale | ||
attn_weights *= (head_dim ** (0.5)) * softmax_scale | ||
else: |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
这个分支干掉吧,现在应该不需要了。可以统一清理
if config.sequence_parallel and use_sequence_parallel: | ||
mark_as_sequence_parallel_parameter(self.weight) | ||
|
||
def forward(self, hidden_states): |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
fuse_rms_norm 可以考虑支持
|
…PaddleNLP into dev_20240819_add_deepseekv2
PR types
New features
PR changes
Models
Description
Add DeepSeekV2.