Skip to content

[BiT] Small patch fix#20657

Merged
younesbelkada merged 2 commits intohuggingface:mainfrom
younesbelkada:fix-bit-fp16
Dec 8, 2022
Merged

[BiT] Small patch fix#20657
younesbelkada merged 2 commits intohuggingface:mainfrom
younesbelkada:fix-bit-fp16

Conversation

@younesbelkada
Copy link
Contributor

@younesbelkada younesbelkada commented Dec 7, 2022 •
edited
Loading

What does this PR do?

This PR fixes a tiny issue that you can encounter if you load BiT in fp16.
diffusers uses this model under the hood for Depth Estimation inpainting and users get this error:

    593 
    594         layer_dropouts = [
--> 595             x.tolist() for x in torch.linspace(0, config.drop_path_rate, sum(config.depths), dtype=torch.float32).split(config.depths)
    596         ]
    597 

RuntimeError: "linspace_cpu" not implemented for 'Half'

However on diffusers side this can be fixed by installing accelerate and load the pipeline with low_cpu_mem_usage=True. But better to fix it to avoid any misleading issue

cc @sgugger @patil-suraj

Otherwise to reproduce:

import torch
from transformers import BitModel

model = BitModel.from_pretrained("google/bit-50", torch_dtype=torch.float16)

@younesbelkada younesbelkada requested a review from sgugger December 7, 2022 16:16
@HuggingFaceDocBuilderDev
Copy link

HuggingFaceDocBuilderDev commented Dec 7, 2022 •
edited
Loading

The documentation is not available anymore as the PR was closed or merged.

sgugger
sgugger approved these changes Dec 7, 2022
View reviewed changes
Copy link
Collaborator

@sgugger sgugger left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks for fixing! Nit: do we really need to use torch for this?

@younesbelkada younesbelkada merged commit 93b5436 into huggingface:main Dec 8, 2022
mpierrau pushed a commit to mpierrau/transformers that referenced this pull request Dec 15, 2022
@younesbelkada
[BiT] Small patch fix (huggingface#20657)
67e1953 
* patch fix for `fp16`

* use `np` instead
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

1 more reviewer

@sgugger sgugger sgugger approved these changes

Reviewers whose approvals may not affect merge requirements

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

3 participants

@younesbelkada @HuggingFaceDocBuilderDev @sgugger