Update flax requirement from <0.6,>=0.5 to >=0.5,<0.7
Created by: dependabot[bot]
Updates the requirements on flax to permit the latest version.
Release notes
Sourced from flax's releases.
Version 0.6.0
What's Changed
- Add
on_commit_callback
to put the responsibility of renaming the directories on the users of the serialization library. This will also fix the GCS atomic rename issue where the users can write a success file when the commit is successful and check the existence of that file before deserialization. by@copybara-service
in google/flax#2328- RDT Redesign by
@cgarciae
in google/flax#2177- Fix stale URLs to read the docs site. by
@levskaya
in google/flax#2338- Replace all jax.tree_* calls with jax.tree_util.tree_* by
@levskaya
in google/flax#2325- Further fix the singular-leaf checkpointing, and add tests. by
@copybara-service
in google/flax#2336- Forward unroll argument in scan_with_axes by
@sanchit-gandhi
in google/flax#2339- Document repo analytics by
@cgarciae
in google/flax#2317- Improve flax basics by
@cgarciae
in google/flax#2291- Split build into multiple jobs by
@cgarciae
in google/flax#2277- Fix type annotation for
step
intraining.checkpoints
by@Chuxiaof
in google/flax#2343- Add test for writing and restoring empty checkpoints. by
@copybara-service
in google/flax#2345- Allow all processes to checkpoint when not using GDA by
@copybara-service
in google/flax#2350- Make importing tensorstore optional and move related type hints to comments. by
@copybara-service
in google/flax#2348- Use jax.named_scope for name stack rather than named_call. by
@copybara-service
in google/flax#2349- Internal change by
@copybara-service
in google/flax#2356- Fix sphinx CI errors by
@cgarciae
in google/flax#2361- Forward path to rewound Scope by
@jheek
in google/flax#2360- Make link a link on the getting started by
@Davidnet
in google/flax#2340- Fix colab & github links by
@cgarciae
in google/flax#2363- Fix ConvTranspose with circular padding by
@cgarciae
in google/flax#2364- Add some docstrings to the old flax.training.common_utils module. by
@levskaya
in google/flax#2373- Correct state variable name by
@Jeevesh8
in google/flax#2369- Allow linen's Conv layer to operate on arbitrary-rank inputs. by
@copybara-service
in google/flax#2308- Copies
dynamic_scale.py
fromoptim/
totraining/
. by@copybara-service
in google/flax#2375- Add option
auto_flush
toflax.metrics.tensorboard.SummaryWriter
by@copybara-service
in google/flax#2376- updated supported transforms in lifting docs. by
@levskaya
in google/flax#2374- Test docstrings with autodoc on CI by
@cgarciae
in google/flax#2372- skip remat test that fails with autodiff by
@mattjj
in google/flax#2389- Removes
flax.optim.dynamic_scale
. by@copybara-service
in google/flax#2314- Plumb spmd_axis_name from vmap_with_axes through to JAX vmap by
@copybara-service
in google/flax#2390- fixed math expressions by
@banda-larga
in google/flax#2392New Contributors
@sanchit-gandhi
made their first contribution in google/flax#2339@Chuxiaof
made their first contribution in google/flax#2343@Davidnet
made their first contribution in google/flax#2340@Jeevesh8
made their first contribution in google/flax#2369@banda-larga
made their first contribution in google/flax#2392Full Changelog: https://github.com/google/flax/compare/v0.5.3...v0.6.0
Changelog
Sourced from flax's changelog.
0.6.0
- Removed deprecated optimizers in
flax.optim
package.- Moved
flax.optim.dynamic_scale
toflax.training.dynamic_scale
.- Switched to using
jax.named_scope
for all profile naming, cut some pointless stack traces out.0.5.3
New features:
- Added
nn.switch
as a lifted version ofjax.lax.switch
.- Added a method for detecting the use of "init" functions.
- Added checkpointing support for
jax.experimental.GlobalDeviceArray
, a useful array type for multiprocess/multihost computing.- Added async option to
save_checkpoints()
on single-process scenario.- Improved documentation pages.
Bug fixes:
- Fixed variable aliasing in put_variable
- Fixed missing passthrough of nn.scan unroll arg
- Fixed the MNIST example
0.5.2
- Fixes missing PyYAML dependecy.
0.5.1
New features:
- Added
nn.tabulate
andModule.tabulate
to generate rich representations of the network structure.0.5.0
- Added
flax.jax_utils.ad_shard_unpad()
by@lucasb-eyer
- Implemented default dtype FLIP. This means the default dtype is now inferred from inputs and params rather than being hard-coded to float32. This is especially useful for dealing with complex numbers because the standard Modules will no longer truncate complex numbers to their real component by default. Instead the complex dtype is preserved by default.
Bug fixes:
- Fix support for JAX's experimental_name_stack.
Breaking changes:
- In rare cases the dtype of a layer can change due to default dtype FLIP. See the "Backward compatibility" section of the proposal for more information.
0.4.2
New features:
... (truncated)
Commits
-
cda7a4c
Removesflax.optim
& bumps version to 0.6.0 -
47418b3
Merge pull request #2392 from banda-larga:patch-1 -
99294fb
Fix minor typos in struct.py docstring. -
43c3c29
fixed jax_for_the_impatient line -
8fe8f5e
Plumb spmd_axis_name from vmap_with_axes through to JAX vmap -
d9897c5
Update minimal JAX version to latest (0.3.16). -
3d77db1
single commit -
517b763
Removesflax.optim.dynamic_scale
. -
28291a8
Update minimal JAX version -
890a391
Merge pull request #2389 from mattjj:skip-buggy-test - Additional commits viewable in compare view
Dependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting @dependabot rebase
.
Dependabot commands and options
You can trigger Dependabot actions by commenting on this PR:
-
@dependabot rebase
will rebase this PR -
@dependabot recreate
will recreate this PR, overwriting any edits that have been made to it -
@dependabot merge
will merge this PR after your CI passes on it -
@dependabot squash and merge
will squash and merge this PR after your CI passes on it -
@dependabot cancel merge
will cancel a previously requested merge and block automerging -
@dependabot reopen
will reopen this PR if it is closed -
@dependabot close
will close this PR and stop Dependabot recreating it. You can achieve the same result by closing it manually -
@dependabot ignore this major version
will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself) -
@dependabot ignore this minor version
will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself) -
@dependabot ignore this dependency
will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)