carmocca
Follow
Pinned
3,103 contributions in the last year
Activity overview
Contributed to
Lightning-AI/lightning,
Lightning-AI/utilities,
Lightning-Universe/Finetune-miniLM
and 32 other
repositories
Contribution activity
February 2023
Created 39 commits in 1 repository
Created a pull request in Lightning-AI/lightning that received 6 comments
Fix TPU tests
What does this PR do? #16571 was merged with failing TPU tests cc @Borda @carmocca @justusschock @awaelchli
+35
−148
•
6
comments
Opened 25 other pull requests in 2 repositories
Lightning-AI/lightning
5
open
19
merged
-
Refactor
CombinedLoaderusing pytrees -
Move
Trainer._log_hyperparamsto an utility -
Group
torch.compileutilities together - Remove custom log events
- Always run standalone tests
-
Move
inference_modelogic to the loops -
Remove the
trainer.data_parallelproperty - Group trainer call methods as functions
- Poptorch 3.1.0 supports dictionary inputs
-
Remove the unused
utilities.finite_checks -
Remove the unused
utilities.metrics.metrics_to_scalars - Move logger utilities with the loggers
- Move fetchers classes with the loops
- Flatten fetching abstract interface
- Refactor supporters
- Remove metric reset after checkpoint load
-
Remove outputs from
on_predict_epoch_end - Refactor deterministic and benchmark logic
- Clarify default epoch and step values in ModelCheckpoint
- [WIP] Remove duplicate import checks
- Fix TPU CI
- Fix broken links after reverse mirror changes
- Install colossalai==0.1.12 in CI
- Only run TPU standalone tests
Lightning-AI/utilities
1
open
Reviewed 48 pull requests in 2 repositories
Lightning-AI/lightning
25 pull requests
- ci: freeze pypi releasing
- ci/hotfix: pypi releasing
-
Refactor
CombinedLoaderusing pytrees - preparing 1.9.2dev
- Group trainer call methods as functions
-
move
_HPU_AVAILABLE - Add .git-blame-ignore-revs
-
Group
torch.compileutilities together - Update examples with multiple optimizers
-
Move
inference_modelogic to the loops - Always run standalone tests
-
Remove the
trainer.data_parallelproperty - Split train- and val progress into separate bars
- ci: parameterize GPU testing
- Update pytorch_lightning imports in examples
- Fix strategy type validation in connectors
- tests: drop slow flag
- ci: merge pytest with slow
-
Add
Strategy.on_exception - rename docs/source-app & adjust docs links for lightning
- Update Fabric introduction
- Refactor supporters
- Fix import from torch.distributed when distributed not available
- Fix a typo in transfer_batch_to_device function's comment
- Use base version when comparing torch versions
- Some pull request reviews not shown.
Lightning-AI/utilities
1 pull request
Created an issue in Lightning-AI/lightning that received 5 comments
Drop the numpy requirement
Outline & Motivation
We barely use numpy. We should replace its uses with torch.
Reduce the package footprint
Pitch
Remove lightning/requirement…
5
comments
2
contributions
in private repositories
Feb 1