Refactor execution device & cpu offload #4114
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| return torch.device(module._hf_hook.execution_device) | ||
| return self.device | ||
|
|
||
| def enable_sequential_cpu_offload(self, gpu_id: int = 0, device: Union[torch.device, str] = "cuda"): |
Contributor
Author
There was a problem hiding this comment.
Is general enough to work with all models. Also applies suggestion from #3990
|
The documentation is not available anymore as the PR was closed or merged. |
williamberman
approved these changes
Jul 17, 2023
Contributor
williamberman
left a comment
williamberman
left a comment
There was a problem hiding this comment.
Nice, this makes a ton of sense! We should be sure to monitor the integration tests
sayakpaul
reviewed
Jul 18, 2023
| def _execution_device(self): | ||
| r""" | ||
| Returns the device on which the pipeline's models will be executed. After calling | ||
| `pipeline.enable_sequential_cpu_offload()` the execution device can only be inferred from Accelerate's module |
Member
There was a problem hiding this comment.
Can we format pipeline.enable_sequential_cpu_offload() such that the hyperlink generates properly?
sayakpaul
reviewed
Jul 18, 2023
sayakpaul
reviewed
Jul 18, 2023
|
|
||
| # set these parameters to False in the child class if the pipeline does not support the corresponding functionality | ||
| test_attention_slicing = True | ||
| test_cpu_offload = True |
Contributor
Author
There was a problem hiding this comment.
We always test it
Contributor
Author
There was a problem hiding this comment.
We completely remove the test_cpu_offload flag because every pipeline is now tested
sayakpaul
approved these changes
Jul 18, 2023
6 tasks
orpatashnik
pushed a commit
to orpatashnik/diffusers
that referenced
this pull request
Aug 1, 2023
* create general cpu offload & execution device * Remove boiler plate * finish * kp * Correct offload more pipelines * up * Update src/diffusers/pipelines/pipeline_utils.py * make style * up
orpatashnik
pushed a commit
to orpatashnik/diffusers
that referenced
this pull request
Aug 1, 2023
* create general cpu offload & execution device * Remove boiler plate * finish * kp * Correct offload more pipelines * up * Update src/diffusers/pipelines/pipeline_utils.py * make style * up
orpatashnik
pushed a commit
to orpatashnik/diffusers
that referenced
this pull request
Aug 1, 2023
* create general cpu offload & execution device * Remove boiler plate * finish * kp * Correct offload more pipelines * up * Update src/diffusers/pipelines/pipeline_utils.py * make style * up
Merged
yoonseokjin
pushed a commit
to yoonseokjin/diffusers
that referenced
this pull request
Dec 25, 2023
* create general cpu offload & execution device * Remove boiler plate * finish * kp * Correct offload more pipelines * up * Update src/diffusers/pipelines/pipeline_utils.py * make style * up
AmericanPresidentJimmyCarter
pushed a commit
to AmericanPresidentJimmyCarter/diffusers
that referenced
this pull request
Apr 26, 2024
* create general cpu offload & execution device * Remove boiler plate * finish * kp * Correct offload more pipelines * up * Update src/diffusers/pipelines/pipeline_utils.py * make style * up
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Refactor
enable_sequential_cpu_offloadand_execution_deviceEvery pipeline should be able to use
cpu_offloadin a very non-customized way as essentially we just need to wrap alltorch.nn.Modulecomponets into thecpu_offloadfunction. Similarly can we determine the_execution_deviceby looping over all torch.nn.Module components.Therefore this PR:
enable_sequantial_cpu_offloadand_execution_devicetoDiffusionPipeline_exclude_from_cpu_offloadclass list to classes where certain specific components cannot be offloaded because of weird parameter structureI tested this PR on GPU by running: