neuro-ml/thunder issues and pull requests

#89 - Dev

Pull Request - State: open - Opened by arseniybelkov 10 months ago

#89 - Dev

Pull Request - State: open - Opened by arseniybelkov 10 months ago

#88 - Support for pydantic<3.0

Pull Request - State: open - Opened by arseniybelkov 10 months ago

#88 - Support for pydantic<3.0

Pull Request - State: open - Opened by arseniybelkov 10 months ago

#87 - `thunder backend list` raises exception when executed with no added backends.

Issue - State: open - Opened by arseniybelkov 10 months ago
Labels: bug, cli

#87 - `thunder backend list` raises exception when executed with no added backends.

Issue - State: open - Opened by arseniybelkov 10 months ago
Labels: bug, cli

#86 - Checkpoint loading outside of experiment is inconvinent now. We need `load_checkpoint` to fetch all the dependencies from the config.

Issue - State: open - Opened by arseniybelkov 10 months ago - 1 comment
Labels: feature, good first issue, invalid, configs

#86 - Checkpoint loading outside of experiment is inconvinent now. We need `load_checkpoint` to fetch all the dependencies from the config.

Issue - State: open - Opened by arseniybelkov 10 months ago - 1 comment
Labels: feature, good first issue, invalid, configs

#85 - Callbacks for configs

Pull Request - State: closed - Opened by arseniybelkov 11 months ago - 1 comment

#85 - Callbacks for configs

Pull Request - State: closed - Opened by arseniybelkov 11 months ago - 1 comment

#84 - Fix `last_checkpoint`

Pull Request - State: closed - Opened by kirpoly 12 months ago - 2 comments

#84 - Fix `last_checkpoint`

Pull Request - State: closed - Opened by kirpoly 12 months ago - 2 comments

#83 - `group_metrics` consume too much memory

Issue - State: closed - Opened by arseniybelkov 12 months ago
Labels: invalid, callback

#83 - `group_metrics` consume too much memory

Issue - State: closed - Opened by arseniybelkov 12 months ago
Labels: invalid, callback

#82 - Dev

Pull Request - State: closed - Opened by arseniybelkov 12 months ago - 1 comment

#82 - Dev

Pull Request - State: closed - Opened by arseniybelkov 12 months ago - 1 comment

#81 - fixed last checkpoint loading

Pull Request - State: closed - Opened by shevtsovalexey about 1 year ago

#81 - fixed last checkpoint loading

Pull Request - State: closed - Opened by shevtsovalexey about 1 year ago

#80 - `last_checkpoint` doesn't work with symlinks

Issue - State: closed - Opened by arseniybelkov about 1 year ago
Labels: bug

#80 - `last_checkpoint` doesn't work with symlinks

Issue - State: closed - Opened by arseniybelkov about 1 year ago
Labels: bug

#79 - accumalation of gradient (at least with 4, 8) leads to Nones in time stamps in TimeProfiler

Issue - State: open - Opened by arseniybelkov about 1 year ago
Labels: bug, callback

#79 - accumalation of gradient (at least with 4, 8) leads to Nones in time stamps in TimeProfiler

Issue - State: open - Opened by arseniybelkov about 1 year ago
Labels: bug, callback

#78 - AMP during inference

Issue - State: open - Opened by arseniybelkov about 1 year ago
Labels: invalid

#78 - AMP during inference

Issue - State: open - Opened by arseniybelkov about 1 year ago
Labels: invalid

#77 - Customizable logging in `MetricMonitor`

Issue - State: open - Opened by arseniybelkov about 1 year ago
Labels: good first issue, invalid, callback

#77 - Customizable logging in `MetricMonitor`

Issue - State: open - Opened by arseniybelkov about 1 year ago
Labels: good first issue, invalid, callback

#76 - `ThunderModule` does not convert targets to the same type as outputs of the model.

Issue - State: open - Opened by arseniybelkov about 1 year ago
Labels: invalid, questionable

#76 - `ThunderModule` does not convert targets to the same type as outputs of the model.

Issue - State: open - Opened by arseniybelkov about 1 year ago
Labels: invalid, questionable

#75 - Make convenient `fix_seed` function for experiments.

Issue - State: closed - Opened by arseniybelkov about 1 year ago - 1 comment
Labels: invalid, configs

#75 - Make convenient `fix_seed` function for experiments.

Issue - State: closed - Opened by arseniybelkov about 1 year ago - 1 comment
Labels: invalid, configs

#74 - breaking changes incoming

Pull Request - State: closed - Opened by arseniybelkov about 1 year ago - 1 comment

#74 - breaking changes incoming

Pull Request - State: closed - Opened by arseniybelkov about 1 year ago - 1 comment

#73 - `MetricMonitor` api changes and fixes

Pull Request - State: closed - Opened by arseniybelkov about 1 year ago - 1 comment

#73 - `MetricMonitor` api changes and fixes

Pull Request - State: closed - Opened by arseniybelkov about 1 year ago - 1 comment

#72 - Learning Rate schedulers are not restored from checkpoint

Issue - State: closed - Opened by arseniybelkov about 1 year ago
Labels: bug

#72 - Learning Rate schedulers are not restored from checkpoint

Issue - State: closed - Opened by arseniybelkov about 1 year ago
Labels: bug

#71 - Default backends `cli` and `slurm` are not recognized by `thunder backend set`

Issue - State: open - Opened by arseniybelkov about 1 year ago
Labels: bug, good first issue, cli

#71 - Default backends `cli` and `slurm` are not recognized by `thunder backend set`

Issue - State: open - Opened by arseniybelkov about 1 year ago
Labels: bug, good first issue, cli

#70 - Named dataloaders for MetricLogger

Issue - State: open - Opened by arseniybelkov about 1 year ago
Labels: feature, invalid, callback

#70 - Named dataloaders for MetricLogger

Issue - State: open - Opened by arseniybelkov about 1 year ago
Labels: feature, invalid, callback

#69 - CLI speedup

Issue - State: open - Opened by arseniybelkov about 1 year ago
Labels: invalid, cli

#69 - CLI speedup

Issue - State: open - Opened by arseniybelkov about 1 year ago
Labels: invalid, cli

#68 - preprocessing for group metrics works only on `all_predictions` which is inconvenient

Issue - State: closed - Opened by arseniybelkov about 1 year ago
Labels: callback

#68 - preprocessing for group metrics works only on `all_predictions` which is inconvenient

Issue - State: closed - Opened by arseniybelkov about 1 year ago
Labels: callback

#67 - add prediction as part of `thunder run`

Issue - State: closed - Opened by arseniybelkov over 1 year ago
Labels: feature, cli

#67 - add prediction as part of `thunder run`

Issue - State: closed - Opened by arseniybelkov over 1 year ago
Labels: feature, cli

#66 - `thunder run NAME` does not fail when NAME does not exist

Issue - State: closed - Opened by arseniybelkov over 1 year ago
Labels: bug, cli

#66 - `thunder run NAME` does not fail when NAME does not exist

Issue - State: closed - Opened by arseniybelkov over 1 year ago
Labels: bug, cli

#65 - `TimeProfiler` and `MetricLogger` don't save and load state after failed training.

Issue - State: closed - Opened by arseniybelkov over 1 year ago
Labels: bug, callback

#65 - `TimeProfiler` and `MetricLogger` don't save and load state after failed training.

Issue - State: closed - Opened by arseniybelkov over 1 year ago
Labels: bug, callback

#64 - Dev

Pull Request - State: closed - Opened by arseniybelkov over 1 year ago - 1 comment

#64 - Dev

Pull Request - State: closed - Opened by arseniybelkov over 1 year ago - 1 comment

#63 - add `FixedLayout` layout

Issue - State: closed - Opened by arseniybelkov over 1 year ago
Labels: feature, layout

#63 - add `FixedLayout` layout

Issue - State: closed - Opened by arseniybelkov over 1 year ago
Labels: feature, layout

#62 - SingleSplit splits 255 cases with in relation 0.7 and 0.3 into 254 total cases.

Issue - State: closed - Opened by arseniybelkov over 1 year ago
Labels: bug, layout

#62 - SingleSplit splits 255 cases with in relation 0.7 and 0.3 into 254 total cases.

Issue - State: closed - Opened by arseniybelkov over 1 year ago
Labels: bug, layout

#61 - WandbLogger can't find project while deleting duplicate runs.

Issue - State: open - Opened by arseniybelkov over 1 year ago - 1 comment
Labels: bug, logger, distributed training

#61 - WandbLogger can't find project while deleting duplicate runs.

Issue - State: open - Opened by arseniybelkov over 1 year ago - 1 comment
Labels: bug, logger, distributed training

#60 - test js

Pull Request - State: closed - Opened by arseniybelkov over 1 year ago - 1 comment

#60 - test js

Pull Request - State: closed - Opened by arseniybelkov over 1 year ago - 1 comment

#59 - WandbLogger creates new run after experiment was stopped.

Issue - State: closed - Opened by arseniybelkov over 1 year ago - 1 comment
Labels: invalid, logger

#59 - WandbLogger creates new run after experiment was stopped.

Issue - State: closed - Opened by arseniybelkov over 1 year ago - 1 comment
Labels: invalid, logger

#58 - add support for other configs except lazycon

Issue - State: open - Opened by arseniybelkov over 1 year ago
Labels: feature, cli, configs

#58 - add support for other configs except lazycon

Issue - State: open - Opened by arseniybelkov over 1 year ago
Labels: feature, cli, configs

#57 - add tests for wandb

Issue - State: open - Opened by arseniybelkov over 1 year ago
Labels: good first issue, tests

#57 - add tests for wandb

Issue - State: open - Opened by arseniybelkov over 1 year ago
Labels: good first issue, tests

#56 - dev

Pull Request - State: closed - Opened by arseniybelkov over 1 year ago - 1 comment

#55 - Dev

Pull Request - State: closed - Opened by arseniybelkov over 1 year ago - 1 comment

#55 - Dev

Pull Request - State: closed - Opened by arseniybelkov over 1 year ago - 1 comment

#54 - fix (#53)

Pull Request - State: closed - Opened by arseniybelkov over 1 year ago - 1 comment

#54 - fix (#53)

Pull Request - State: closed - Opened by arseniybelkov over 1 year ago - 1 comment

#53 - `WandbLogger` spams experiments

Issue - State: closed - Opened by arseniybelkov over 1 year ago
Labels: feature, logger

#53 - `WandbLogger` spams experiments

Issue - State: closed - Opened by arseniybelkov over 1 year ago
Labels: feature, logger

#52 - Asynchronous validation

Issue - State: open - Opened by arseniybelkov over 1 year ago
Labels: feature, questionable

#52 - Asynchronous validation

Issue - State: open - Opened by arseniybelkov over 1 year ago
Labels: feature, questionable

#51 - Add framework overview to docs.

Issue - State: open - Opened by arseniybelkov over 1 year ago
Labels: documentation

#51 - Add framework overview to docs.

Issue - State: open - Opened by arseniybelkov over 1 year ago
Labels: documentation

#50 - cli-features

Pull Request - State: closed - Opened by arseniybelkov over 1 year ago - 1 comment

#50 - cli-features

Pull Request - State: closed - Opened by arseniybelkov over 1 year ago - 1 comment

#49 - `MetricLogger` can't cast tensors to other devices.

Issue - State: closed - Opened by arseniybelkov over 1 year ago - 1 comment
Labels: bug, wontfix, callback

#49 - `MetricLogger` can't cast tensors to other devices.

Issue - State: closed - Opened by arseniybelkov over 1 year ago - 1 comment
Labels: bug, wontfix, callback

#48 - `MetricLogger` fails if outputs are None.

Issue - State: open - Opened by arseniybelkov over 1 year ago
Labels: bug, callback, questionable

#48 - `MetricLogger` fails if outputs are None.

Issue - State: open - Opened by arseniybelkov over 1 year ago
Labels: bug, callback, questionable

#47 - Add flag for overwriting experiment directory for `thunder build`.

Issue - State: closed - Opened by arseniybelkov over 1 year ago
Labels: feature, cli

#47 - Add flag for overwriting experiment directory for `thunder build`.

Issue - State: closed - Opened by arseniybelkov over 1 year ago
Labels: feature, cli

#46 - TimeProfiler fixes

Pull Request - State: closed - Opened by arseniybelkov over 1 year ago - 1 comment

#46 - TimeProfiler fixes

Pull Request - State: closed - Opened by arseniybelkov over 1 year ago - 1 comment

#45 - `avg val downtime` can sometimes be negative during first epoch.

Issue - State: open - Opened by arseniybelkov over 1 year ago
Labels: bug, callback

#45 - `avg val downtime` can sometimes be negative during first epoch.

Issue - State: open - Opened by arseniybelkov over 1 year ago
Labels: bug, callback

#44 - Remove batch_size dependency from `TimeProfiler`.

Issue - State: closed - Opened by arseniybelkov over 1 year ago
Labels: invalid, callback

#44 - Remove batch_size dependency from `TimeProfiler`.

Issue - State: closed - Opened by arseniybelkov over 1 year ago
Labels: invalid, callback

#43 - Subtract `optimizer step` and `backward` from `avg train downtime` in TimeProfiler.

Issue - State: closed - Opened by arseniybelkov over 1 year ago - 1 comment
Labels: invalid, callback

#43 - Subtract `optimizer step` and `backward` from `avg train downtime` in TimeProfiler.

Issue - State: closed - Opened by arseniybelkov over 1 year ago - 1 comment
Labels: invalid, callback

#42 - Make preprocessing outputs shareable.

Issue - State: open - Opened by arseniybelkov over 1 year ago
Labels: feature, callback

#42 - Make preprocessing outputs shareable.

Issue - State: open - Opened by arseniybelkov over 1 year ago
Labels: feature, callback

#41 - pydantic>2.0 causes `<Result AttributeError("'FieldInfo' object has no attribute 'outer_type_'")>.exit_code`

Issue - State: closed - Opened by arseniybelkov over 1 year ago - 1 comment
Labels: bug, invalid, cli

#41 - pydantic>2.0 causes `<Result AttributeError("'FieldInfo' object has no attribute 'outer_type_'")>.exit_code`

Issue - State: closed - Opened by arseniybelkov over 1 year ago - 1 comment
Labels: bug, invalid, cli

#40 - Add flag for setting default backend

Issue - State: closed - Opened by arseniybelkov over 1 year ago
Labels: feature, cli

#40 - Add flag for setting default backend

Issue - State: closed - Opened by arseniybelkov over 1 year ago
Labels: feature, cli

#39 - SLURM flag `-r` receives memory in MBs (GBs are preferred)

Issue - State: closed - Opened by arseniybelkov over 1 year ago - 1 comment
Labels: feature, cli, questionable

GitHub / neuro-ml/thunder issues and pull requests