Sophiex/dev ssl/main#2511
Open
sophie-xhonneux wants to merge 11 commits into
Open
Conversation
Fix was the dataset length computation in the workers
shmh40
reviewed
Jun 16, 2026
| } | ||
|
|
||
| if channels is not None: | ||
| # Respect the order given in the stream config so the channel layout is identical |
Contributor
There was a problem hiding this comment.
Do you think we need this? Are we unsure at the moment whether this is necessary?
Contributor
Author
There was a problem hiding this comment.
We are unsure if it is necessary
shmh40
reviewed
Jun 16, 2026
| output = self.predict_decoders(model_params, step, tokens, batch, output) | ||
| # latent predictions (raw and with SSL heads) | ||
| output = self.predict_latent(model_params, step, tokens, batch, output, intermediates) | ||
| if "masking" in self.cf.training_config.training_mode: |
shmh40
reviewed
Jun 16, 2026
|
|
||
| for module in model.encoder.ae_local_engine.ae_local_blocks.modules(): | ||
| if isinstance(module, modules_to_shard): | ||
| if isinstance(module, modules_to_shard) and _has_trainable_params(module): |
Contributor
There was a problem hiding this comment.
This is the fsdp fix from this morning?
shmh40
reviewed
Jun 16, 2026
| continue # set disabled, e.g. by a train_continue override | ||
| stage_label = f"val_{name}" | ||
| extra_cfg = get_active_stage_config(self.validation_cfg, overrides, cfg_keys_to_filter) | ||
| # extra sets never write sample output files (would collide with primary val output) |
Contributor
There was a problem hiding this comment.
So we only look at these in plot_train? And then if we run inference (using test that I suppose inherits from validation) we only ever write out one of the validation periods? Not important, just want to check the mechanics here.
Contributor
Author
There was a problem hiding this comment.
not sure, didn't check to be honest
shmh40
reviewed
Jun 16, 2026
| self.validate(0, self.test_cfg, self.batch_size_test_per_gpu) | ||
| logger.info(f"Finished inference run with id: {cf.general.run_id}") | ||
|
|
||
| def _check_channel_order_consistency( |
Contributor
There was a problem hiding this comment.
Can we keep this guard and remove the reordering you did above if we don't think we need it maybe?
shmh40
reviewed
Jun 16, 2026
| self.dataset = MultiStreamDataSampler(cf, self.training_cfg, stage=TRAIN) | ||
| self.dataset_val = MultiStreamDataSampler(cf, self.validation_cfg, stage=VAL) | ||
|
|
||
| if run_id_contd is not None: |
shmh40
approved these changes
Jun 16, 2026
github-actions
Bot
added
infra
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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.
Description
This has several small fixes, including:
./scripts/actions.sh lint./scripts/actions.sh unit-test./scripts/actions.sh integration-testlaunch-slurm.py --time 60