[Feature] reset_parameters for multiagent nets
#1970
Merged
Conversation
This file contains 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
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/rl/1970
Note: Links to docs will display an error until the docs builds have been completed. ❌ 6 New Failures, 21 Unrelated FailuresAs of commit 48757d0 with merge base 3d65083 ( NEW FAILURES - The following jobs have failed:
FLAKY - The following jobs failed but were likely due to flakiness present on trunk:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
facebook-github-bot
added
the
CLA Signed
matteobettini
commented
Feb 27, 2024
| return torch.vmap(reset_module, *args, **kwargs) | ||
|
|
||
| if not self.share_params: | ||
| vmap_reset_module(self._empty_net, randomness="different")(self.params) |
There was a problem hiding this comment.
I would like to discuss the vmap randomness of this class
@property
def _vmap_randomness(self):
if self.initialized:
return "error"
return "same"Why would this be a class property and why are we having those values?
For me this should be
@property
def _vmap_randomness(self):
return "different"There was a problem hiding this comment.
In this case it needs to be different to have a different reset values for each agent.
But also in the forward pass I feel like it should be "different"
There was a problem hiding this comment.
IMO it should be
@property
def _vmap_randomness(self):
if self.initialized:
return self.vmap_randomness
return "different"and users are in charge of telling the module what randomness they want.
There was a problem hiding this comment.
Could you explain this a bit? Why do we have a switch on initialization?
There was a problem hiding this comment.
For init you need "different" because you must have different weights for each net.
But in other settings you can't tell, and the best is to let the user choose.
They may as well want the same random number for each element of the batch
There was a problem hiding this comment.
but what I do not understand is why before we had
@property
def _vmap_randomness(self):
if self.initialized:
return "error"
return "same"There was a problem hiding this comment.
if I try to change this to
@property
def _vmap_randomness(self):
if self.initialized:
return self.vmap_randomness
return "different"the lazy layers will crash
There was a problem hiding this comment.
the lazy layers will crash
this is a statement that is hard to reproduce, can you share more?
For instance this works fine on my end:
from torchrl.modules import MLP
from tensordict import TensorDict
import torch
from functorch import dim
d0 = dim.dims(1)
modules = [torch.nn.Linear(2, 3) for _ in range(3)]
td = TensorDict.from_modules(*modules, as_module=True)
def reset(td):
with td.to_module(modules[0]):
modules[0].reset_parameters()
return td
td = torch.vmap(reset, randomness="same")(td)
print(td["weight"])
td = torch.vmap(reset, randomness="different")(td)
print(td["weight"])the first produces a stack of identical tensors, the second different
matteobettini
changed the title
reset_params for multiagent netsreset_parameters for multiagent nets
Co-authored-by: Vincent Moens <vincentmoens@gmail.com>
vmoens
reviewed
Feb 27, 2024
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.
Fixes #1967
Todo: