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.

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

fix falcon-40b accuracy issue #39

Merged
merged 1 commit into from
Jan 4, 2024

Conversation

Yejing-Lai
Copy link

"FalconDecoderLayer" module will choose the "glmtype" fused_qkv_type. But the Falcon model should use the "bloomtype".

@delock
Copy link
Owner

delock commented Jan 3, 2024

@Yejing-Lai can you add text output before/after your fix for documentation purpose? Thanks!

@Yejing-Lai
Copy link
Author

before fix output: ('DeepSpeed is a machine learning framework for deep neural networks (DNNs) and deep reinforcement learning (DRL). It is written in C++ and is e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e', 32)

after fix output:('DeepSpeed is a machine learning framework for deep neural networks (DNNs) and deep reinforcement learning (DRL). It is written in C++ and is based on PyTorch.\nDeepSpeed is designed to scale DNNs and DRL to large numbers of GPUs and TPUs.\nDeepSpeed is', 32)

@delock
Copy link
Owner

delock commented Jan 4, 2024

before fix output: ('DeepSpeed is a machine learning framework for deep neural networks (DNNs) and deep reinforcement learning (DRL). It is written in C++ and is e e e e e e e e e e e e e e e e e e e e e e e e e e e e e e', 32)

after fix output:('DeepSpeed is a machine learning framework for deep neural networks (DNNs) and deep reinforcement learning (DRL). It is written in C++ and is based on PyTorch.\nDeepSpeed is designed to scale DNNs and DRL to large numbers of GPUs and TPUs.\nDeepSpeed is', 32)

Thanks!

@delock delock merged commit 092b0f2 into delock:gma/run-opt-branch Jan 4, 2024
7 of 16 checks passed
@delock
Copy link
Owner

delock commented Jan 4, 2024

@Yejing-Lai is this PR already submitted to upstream?

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants