POC for mixed prec optim frontend #146640
Draft
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
[ghstack-poisoned]
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/146640
Note: Links to docs will display an error until the docs builds have been completed. ❌ 10 New FailuresAs of commit 59010b7 with merge base 83bb921 ( NEW FAILURES - The following jobs have failed:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
This PR is a prototype for what a frontend for asking for mixed precision can look like torch.optim through set_dtype_policy in optimizer.py.
This is not meant to be landable but to start some discussions on what people want/would like to see and to ask if there are things I haven't considered yet.
This currently only works with Adam(W)!
A toy script for how to use:
```
import torch
model = torch.nn.Sequential(
torch.nn.Linear(2, 3),
torch.nn.Sigmoid(),
torch.nn.Linear(3, 1),
torch.nn.Sigmoid(),
)
model.to("cuda")
optim = torch.optim.AdamW(model.named_parameters(), foreach=False)
mp_policy = {
"exp_avg": lambda _: torch.bfloat16,
"exp_avg_sq": lambda _: torch.bfloat16,
"max_exp_avg_sq": lambda _: torch.bfloat16,
}
optim.set_dtype_policy(mp_policy)
i = torch.tensor([0.1, 0.2, 0.3, 0.4, 0.5, 0.6], device="cuda").reshape(3, 2)
l = model(i).sum()
l.backward()
optim.step()
```
[ghstack-poisoned]
janeyx99
commented
Feb 8, 2025
torch/_meta_registrations.py
Outdated
| start.dtype == end.dtype, | ||
| lambda: f"expected dtype {start.dtype} for `end`, but got dtype {end.dtype}", | ||
| ) | ||
| # torch._check( |
There was a problem hiding this comment.
We should uncomment this once #146749 is fixed
This PR is a prototype for what a frontend for asking for mixed precision can look like torch.optim through set_dtype_policy in optimizer.py.
This is not meant to be landable but to start some discussions on what people want/would like to see and to ask if there are things I haven't considered yet.
This currently only works with Adam(W)!
A toy script for how to use:
```
import torch
model = torch.nn.Sequential(
torch.nn.Linear(2, 3),
torch.nn.Sigmoid(),
torch.nn.Linear(3, 1),
torch.nn.Sigmoid(),
)
model.to("cuda")
optim = torch.optim.AdamW(model.named_parameters(), foreach=False)
mp_policy = {
"exp_avg": lambda _: torch.bfloat16,
"exp_avg_sq": lambda _: torch.bfloat16,
"max_exp_avg_sq": lambda _: torch.bfloat16,
}
optim.set_dtype_policy(mp_policy)
i = torch.tensor([0.1, 0.2, 0.3, 0.4, 0.5, 0.6], device="cuda").reshape(3, 2)
l = model(i).sum()
l.backward()
optim.step()
```
[ghstack-poisoned]
This PR is a prototype for what a frontend for asking for mixed precision can look like torch.optim through set_dtype_policy in optimizer.py.
This is not meant to be landable but to start some discussions on what people want/would like to see and to ask if there are things I haven't considered yet.
This currently only works with Adam(W)!
A toy script for how to use:
```
import torch
model = torch.nn.Sequential(
torch.nn.Linear(2, 3),
torch.nn.Sigmoid(),
torch.nn.Linear(3, 1),
torch.nn.Sigmoid(),
)
model.to("cuda")
optim = torch.optim.AdamW(model.named_parameters(), foreach=False)
mp_policy = {
"exp_avg": lambda _: torch.bfloat16,
"exp_avg_sq": lambda _: torch.bfloat16,
"max_exp_avg_sq": lambda _: torch.bfloat16,
}
optim.set_dtype_policy(mp_policy)
i = torch.tensor([0.1, 0.2, 0.3, 0.4, 0.5, 0.6], device="cuda").reshape(3, 2)
l = model(i).sum()
l.backward()
optim.step()
```
[ghstack-poisoned]
desai0007
pushed a commit
to desai0007/test-repo-pytorch
that referenced
this pull request
Feb 26, 2025
ghstack-source-id: 85f4266 Pull Request resolved: pytorch/pytorch#146640
|
Looks like this PR hasn't been updated in a while so we're going to go ahead and mark this as |
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.
This PR is a prototype for what a frontend for asking for mixed precision can look like torch.optim through set_dtype_policy in optimizer.py.
This is not meant to be landable but to start some discussions on what people want/would like to see and to ask if there are things I haven't considered yet.
This currently only works with Adam(W)!
A toy script for how to use:
Stack from ghstack (oldest at bottom):