Add jax.default_device context manager
#9118
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
|
Sorry this is so huge; I can break up into smaller changes if that'd be helpful. I'll also create a CL if you prefer to review that. EDIT: once tensorflow/tensorflow#53656 goes in, which is a prerequisite. I also need to build a new jaxlib with that before this can go in. |
mattjj
approved these changes
Jan 7, 2022
google-ml-butler
bot
added
kokoro:force-run
pull ready
skye
force-pushed
the
device_context_manager
branch
2 times, most recently
from
35c1952
to
1c890d9
Compare
skye
force-pushed
the
device_context_manager
branch
2 times, most recently
from
d9a322e
to
819824c
Compare
|
👀 |
skye
force-pushed
the
device_context_manager
branch
from
819824c
to
32da2a3
Compare
This currently only supports setting a specific Device object, not a platform like "cpu". That should be added in the future. Bumps the minimum jaxlib version in order to include tensorflow/tensorflow#53656
skye
force-pushed
the
device_context_manager
branch
from
32da2a3
to
f26b866
Compare
This was referenced Sep 12, 2022
ayaka14732
added a commit
to ayaka14732/bart-base-jax
that referenced
this pull request
Oct 10, 2022
Fixed in google/jax#9118
ayaka14732
added a commit
to ayaka14732/bart-base-jax
that referenced
this pull request
Oct 13, 2022
* Update names and README * Add masking algorithm Finally solved after 6 months * Update mask algorithm * Update mask algorithm * Update mask.py * Update mask.py * Fix masking * Update random wrapper Fixed in google/jax#9118 * Add tests for random.wrapper * Update dependences * Masking using the wakong algorithm * Adapt to TPU Pods (#5) * Adapt to multihost environment * Fix multihost * Fix bug * Fix TPU initialisation Fix bug introduced by google/jax#12642, because the script is supposed to automatically fallback to CPU in subprocesses * Fix multihost environment Avoid all global variables * Fix dataloader * Fix bug * Fix bug * Use jax-smi * Scheduler * Use batch_size_per_device * Scale loss * Fix bug: no device_split * Chore * Undo scheduler and loss scaling * Tune lr * Re-apply loss scaling * Re-apply scheduler, use AdamW * Chore * Increase batch_size_per_device to 64 * Try out lamb optimiser * Increase batch size and learning rate * Decrease batch size * Formatting * Format code
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 currently only supports setting a specific Device object, not a
platform like "cpu". That should be added in the future.
Bumps the minimum jaxlib version in order to include
tensorflow/tensorflow#53656