deepseek-ai/Janus
1
0
Fork 0
mirror of https://github.com/deepseek-ai/Janus.git synced 2026-11-05 00:57:18 +00:00
Code Issues 183 Projects Packages Wiki Activity

[GH-ISSUE #89] Error in pip install flash-attn in colab #69

New issue
Open
opened 2026-10-26 21:11:28 +00:00 by bsakaguchi · 5 comments
bsakaguchi commented 2026-10-26 21:11:28 +00:00
Owner
Copy link

Originally created by @strikertushar19 on GitHub (Jan 29, 2025).
Original GitHub issue: https://github.com/deepseek-ai/Janus/issues/89

After I ran !pip install flash-attn in google collab in environment Tpu V28 I got error

Image

Originally created by @strikertushar19 on GitHub (Jan 29, 2025). Original GitHub issue: https://github.com/deepseek-ai/Janus/issues/89 After I ran !pip install flash-attn in google collab in environment Tpu V28 I got error ![Image](https://github.com/user-attachments/assets/fee1b14a-b44b-4846-9db8-ea951d4c912b)
bsakaguchi commented 2026-10-26 21:11:30 +00:00
Author
Owner
Copy link

@Surya443 commented on GitHub (Jan 29, 2025):

It seems that FlashAttention(>2.0.0) is not fully compatible with TPU v2, or there may be missing or misconfigured dependencies related to the TPU architecture. The same installation process works successfully on a T4 GPU or higher GPU models. This issue may be specific to TPU v2. Worked for me when I ran the installation on T4 runtime.

<!-- gh-comment-id:2621737402 --> @Surya443 commented on GitHub (Jan 29, 2025): It seems that FlashAttention(>2.0.0) is not fully compatible with TPU v2, or there may be missing or misconfigured dependencies related to the TPU architecture. The same installation process works successfully on a T4 GPU or higher GPU models. This issue may be specific to TPU v2. Worked for me when I ran the installation on T4 runtime.
bsakaguchi commented 2026-10-26 21:11:30 +00:00
Author
Owner
Copy link

@ukumar0 commented on GitHub (Jan 29, 2025):

flash-attn is optimized for GPUs with CUDA and may not be compatible with TPU. If you are using a TPU runtime in Colab, try switching to a GPU runtime

<!-- gh-comment-id:2621768947 --> @ukumar0 commented on GitHub (Jan 29, 2025): `flash-attn` is optimized for GPUs with CUDA and may not be compatible with TPU. If you are using a TPU runtime in Colab, try switching to a GPU runtime
bsakaguchi commented 2026-10-26 21:11:31 +00:00
Author
Owner
Copy link

@strikertushar19 commented on GitHub (Jan 29, 2025):

It seems that FlashAttention(>2.0.0) is not fully compatible with TPU v2, or there may be missing or misconfigured dependencies related to the TPU architecture. The same installation process works successfully on a T4 GPU or higher GPU models. This issue may be specific to TPU v2. Worked for me when I ran the installation on T4 runtime.

Yeah i figured out that it will not work with tpu there are cpu specific not gpu

<!-- gh-comment-id:2622010113 --> @strikertushar19 commented on GitHub (Jan 29, 2025): > It seems that FlashAttention(>2.0.0) is not fully compatible with TPU v2, or there may be missing or misconfigured dependencies related to the TPU architecture. The same installation process works successfully on a T4 GPU or higher GPU models. This issue may be specific to TPU v2. Worked for me when I ran the installation on T4 runtime. Yeah i figured out that it will not work with tpu there are cpu specific not gpu
bsakaguchi commented 2026-10-26 21:11:31 +00:00
Author
Owner
Copy link

@strikertushar19 commented on GitHub (Jan 29, 2025):

flash-attn is optimized for GPUs with CUDA and may not be compatible with TPU. If you are using a TPU runtime in Colab, try switching to a GPU runtime
I did but unfortunately I do not have premium version of colab , system is crashing for lower version Gpu's .

<!-- gh-comment-id:2622013860 --> @strikertushar19 commented on GitHub (Jan 29, 2025): > `flash-attn` is optimized for GPUs with CUDA and may not be compatible with TPU. If you are using a TPU runtime in Colab, try switching to a GPU runtime I did but unfortunately I do not have premium version of colab , system is crashing for lower version Gpu's .
bsakaguchi commented 2026-10-26 21:11:31 +00:00
Author
Owner
Copy link

@strikertushar19 commented on GitHub (Jan 29, 2025):

Flashattn used is flash-attn2 so i figured out that minimum gpu in order to run the 1.3B param model of janus i required
Compute : Ampere, Ada, or Hopper GPUs (e.g., A100, RTX 3090, RTX 4090, H100).

<!-- gh-comment-id:2622021740 --> @strikertushar19 commented on GitHub (Jan 29, 2025): Flashattn used is flash-attn2 so i figured out that minimum gpu in order to run the 1.3B param model of janus i required Compute : Ampere, Ada, or Hopper GPUs (e.g., A100, RTX 3090, RTX 4090, H100).
Sign in to join this conversation.
No Branch/Tag specified
Branches Tags
main
No results found.
Labels
Clear labels   
pull-request
Mirrored from GitHub Pull Request

No labels
pull-request
Milestone
Clear milestone
No items
No milestone
Projects
Clear projects
No items
No project
Assignees
Clear assignees
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
deepseek-ai/Janus#69
No description provided.