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

qwen2_moe support w multipack #1455

Merged
merged 1 commit into from
Mar 29, 2024
Merged

qwen2_moe support w multipack #1455

merged 1 commit into from
Mar 29, 2024

Conversation

winglian
Copy link
Collaborator

@winglian winglian commented Mar 29, 2024

Description

resolves #1453

✅ multipack
✅ qwen2_moe 4-bit QLoRA
✅ qwen2_moe 16-bit LoRA
❓ qwen2_moe 8-bit LoRA

@winglian
Copy link
Collaborator Author

8-bit LoRA still doesn't work atm

  File "/root/miniconda3/envs/py3.10/lib/python3.10/site-packages/bitsandbytes/autograd/_functions.py", line 487, in backward
    grad_A = torch.matmul(grad_output, CB).view(ctx.grad_shape).to(ctx.dtype_A)
RuntimeError: mat1 and mat2 shapes cannot be multiplied (2048x1 and 32x2048)

@winglian winglian merged commit 6086be8 into main Mar 29, 2024
7 checks passed
@winglian winglian deleted the qwen2-moe branch March 29, 2024 15:04
djsaunde pushed a commit that referenced this pull request Dec 17, 2024
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.

support fine tuning of Qwen/Qwen1.5-MoE-A2.7B-Chat
1 participant