| 6d95b6c | jax authors | 14 October 2020, 06:25:32 UTC | Merge pull request #4573 from google:update-pypi PiperOrigin-RevId: 337028688 | 14 October 2020, 06:25:32 UTC |
| 503bc36 | Matthew Johnson | 14 October 2020, 06:02:49 UTC | fix typo in comment | 14 October 2020, 06:02:49 UTC |
| 4128240 | Matthew Johnson | 14 October 2020, 06:01:30 UTC | update version and changelog for pypi | 14 October 2020, 06:01:30 UTC |
| cb4a014 | jax authors | 14 October 2020, 00:41:21 UTC | Merge pull request #4567 from skye:cuda111 PiperOrigin-RevId: 336992058 | 14 October 2020, 00:41:21 UTC |
| 86e113c | jax authors | 13 October 2020, 23:34:58 UTC | Merge pull request #4563 from jakevdp:pmap-error PiperOrigin-RevId: 336980117 | 13 October 2020, 23:34:58 UTC |
| dfe0526 | jax authors | 13 October 2020, 23:31:41 UTC | Merge pull request #4570 from google:jaxpr-util-test PiperOrigin-RevId: 336979966 | 13 October 2020, 23:31:41 UTC |
| 0daf4c0 | Roy Frostig | 13 October 2020, 22:48:04 UTC | assume less about source locations in jaxpr_util_test | 13 October 2020, 22:48:04 UTC |
| 484ec3e | Jake VanderPlas | 13 October 2020, 21:36:16 UTC | Internal change PiperOrigin-RevId: 336957861 | 13 October 2020, 21:36:45 UTC |
| 01e113d | Skye Wanderman-Milne | 13 October 2020, 21:31:15 UTC | Update jaxlib build scripts to build CUDA 11.1 wheels. | 13 October 2020, 21:31:43 UTC |
| 4624603 | Jake VanderPlas | 13 October 2020, 18:22:55 UTC | Improve pmap axis error in the presence of pytrees | 13 October 2020, 18:22:59 UTC |
| 83d0115 | jax authors | 13 October 2020, 17:49:04 UTC | Merge pull request #4557 from jakevdp:fix-tests PiperOrigin-RevId: 336909706 | 13 October 2020, 17:49:04 UTC |
| 76fe180 | jax authors | 13 October 2020, 17:09:06 UTC | Merge pull request #4556 from jakevdp:index-fix PiperOrigin-RevId: 336900642 | 13 October 2020, 17:09:06 UTC |
| db7f1f4 | Jake VanderPlas | 13 October 2020, 17:08:20 UTC | Increase test coverage for indexing ops | 13 October 2020, 17:08:20 UTC |
| 9c48502 | jax authors | 13 October 2020, 16:48:32 UTC | Merge pull request #4562 from bchetioui:expand_dot_general PiperOrigin-RevId: 336895192 | 13 October 2020, 16:48:32 UTC |
| 5981fc3 | jax authors | 13 October 2020, 16:45:13 UTC | Merge pull request #4530 from jakevdp:cleanup PiperOrigin-RevId: 336895154 | 13 October 2020, 16:45:13 UTC |
| be149f4 | Benjamin Chetioui | 09 October 2020, 15:59:04 UTC | [jax2tf] Added conversion paths without einsum for dot_general. | 13 October 2020, 16:13:34 UTC |
| f077139 | jax authors | 13 October 2020, 16:08:28 UTC | Merge pull request #4451 from alexminnaar:master PiperOrigin-RevId: 336888044 | 13 October 2020, 16:08:28 UTC |
| 4f7aec8 | jax authors | 13 October 2020, 14:14:12 UTC | Merge pull request #4561 from hawkinsp:truncnorm PiperOrigin-RevId: 336870048 | 13 October 2020, 14:14:12 UTC |
| 0b8eb92 | Peter Hawkins | 13 October 2020, 13:16:29 UTC | Add stop_gradients around lax.nextafter to fix TFP gradient errors for jax.random.truncated_normal. | 13 October 2020, 13:16:29 UTC |
| 8b44131 | jax authors | 13 October 2020, 09:00:09 UTC | Merge pull request #4522 from bchetioui:expand_conv_conversion PiperOrigin-RevId: 336833753 | 13 October 2020, 09:00:09 UTC |
| c674c19 | Benjamin Chetioui | 08 October 2020, 15:46:37 UTC | [jax2tf] Add paths that do not use XLA in conv_general_dilated. This adds some amount of support for people who want to run convolutions without having XLA linked in. These paths can seemingly be converted for TFJS as well. Due to a so far unknown bug in some of the conversions, the paths are disabled by default and the "ENABLE_TF_CONVOLUTION" global variable in jax2tf.py must be explictly toggled to use them. See the comment associated with ENABLE_TF_CONVOLUTION for context. | 13 October 2020, 08:00:34 UTC |
| e064451 | jax authors | 12 October 2020, 23:40:55 UTC | Merge pull request #4543 from google:remove-double-dtype-warning PiperOrigin-RevId: 336771011 | 12 October 2020, 23:40:55 UTC |
| dbebc9a | Matthew Johnson | 12 October 2020, 21:30:32 UTC | remove double warning with asarray dtype='int' | 12 October 2020, 22:00:35 UTC |
| 3ca9bba | Jake VanderPlas | 12 October 2020, 20:58:49 UTC | BUG: fix indexing error | 12 October 2020, 20:58:49 UTC |
| d1ca3b3 | jax authors | 12 October 2020, 20:51:21 UTC | Merge pull request #4548 from hawkinsp:truncnorm PiperOrigin-RevId: 336735846 | 12 October 2020, 20:51:21 UTC |
| 080007a | Peter Hawkins | 12 October 2020, 16:41:58 UTC | Ensure values returned by jax.random.truncated_normal() are in range. A user observed -inf values being returned by truncated_normal(), which occur if the uniform random value passed to erfinv() is out of range, e.g., due to rounding. Do more of the computation using jax.random.uniform(), which promises correct behavior in the face of rounding. As an added security measure, also clamp the outputs of the function to the open interval. | 12 October 2020, 20:27:54 UTC |
| 5a3929e | jax authors | 12 October 2020, 20:19:33 UTC | Merge pull request #4547 from jblespiau:changelist/336647290 PiperOrigin-RevId: 336728597 | 12 October 2020, 20:19:33 UTC |
| 0660939 | Jean-Baptiste Lespiau | 12 October 2020, 19:17:16 UTC | Enable the C++ jax.jit fast code-path by default. Here is the list of things I think we should do/revisit at some point in the future: 1. Most importantly, think about Lazy expressions. I think they are powerful, and they can likely enable some optimization impossible otherwise (e.g. stack a ShardedDeviceArray and split it). We likely want to add support for them in the C++ jax.jit. 2. Remove trivial computation support (currently fallback to Python), at least in the omnistaging code-path, as it will no longer necessary with omnistaging, and that the complexity do not justify the feature. 3. Not sure what to do about jax.jit(pmap). It currently fallback to Python when the Executable has more than one device. I understand it executes this on all cores (as a pmap), but will return the first result (from the first core). I am tempted to think this is a non-feature, that no one is looking for and that it can be achieved by doing pmap(f)(x)[0] instead. Having a single simple way of doing one thing is usually beneficial. 4. Not sure what to do about `DeviceConstant` support (currently fallback to Python). I do not understand yet why they exist and problem they solve. 5. Revisit the stikiness for jax.jit(f, device=device)(x). I think it should fail if x is stiky to another device, to let the user aware of the copies. My humble opinion is that, with an efficient device_put, we should write jitted_f(device_put(x, device)). Improving performance is key for JAX users, and the way JAX let you control what lives where is wonderful. Helping the user be aware of their copies is helping in this direction/Doing copies for the users hurts performance. (The support for `jax_debug_nans` could be improved.) With the C++ jit: └──╼ benchy //third_party/py/jax/benchmarks:api_benchmark name time/op jit_trivial_dispatch 35.2µs ± 2% jit_trivial 36.8µs ± 2% jit_simple_dispatch 16.1µs ± 5% jit_simple 29.3µs ± 8% jit_simple_many_args_dispatch 149µs ± 8% jit_simple_many_args 153µs ± 9% jit_dispatch_without_transfer 144µs ± 7% jit_dispatch_with_transfer 150µs ± 5% Without: └──╼ benchy //third_party/py/jax/benchmarks:api_benchmark name time/op jit_trivial_dispatch 29.9µs ± 2% jit_trivial 31.8µs ± 4% jit_simple_dispatch 59.7µs ± 1% jit_simple 66.9µs ± 5% jit_simple_many_args_dispatch 368µs ± 3% jit_simple_many_args 367µs ± 3% jit_dispatch_without_transfer 346µs ± 9% jit_dispatch_with_transfer 392µs ± 9% See also https://github.com/google/jax/pull/4169 for context. PiperOrigin-RevId: 336716707 | 12 October 2020, 19:18:04 UTC |
| 639cda2 | jax authors | 12 October 2020, 19:05:27 UTC | Merge pull request #4554 from skye:workspace PiperOrigin-RevId: 336713864 | 12 October 2020, 19:05:27 UTC |
| b8cac03 | Skye Wanderman-Milne | 12 October 2020, 18:57:59 UTC | Update XLA in WORKSPACE | 12 October 2020, 18:57:59 UTC |
| b13775f | Jean-Baptiste Lespiau | 12 October 2020, 12:52:47 UTC | Enrich the error messages with the bound names that are available. The user often do not know whether it's not the correct name, or whether it was not defined, etc. It's easier to get this information when debugging. | 12 October 2020, 18:33:27 UTC |
| 7c4b935 | jax authors | 12 October 2020, 18:26:34 UTC | Merge pull request #4550 from jblespiau:changelist/336694640 PiperOrigin-RevId: 336705548 | 12 October 2020, 18:26:34 UTC |
| ee9ca56 | Jean-Baptiste Lespiau | 12 October 2020, 18:04:24 UTC | Copybara import of the project: -- d70782e0a1c6568eab6ef1573fa277850cc0da97 by Jean-Baptiste Lespiau <jblespiau@google.com>: Gate some jax_jit test with a version check. COPYBARA_INTEGRATE_REVIEW=https://github.com/google/jax/pull/4550 from jblespiau:changelist/336694640 d70782e0a1c6568eab6ef1573fa277850cc0da97 PiperOrigin-RevId: 336700683 | 12 October 2020, 18:04:58 UTC |
| 5a097f5 | Jean-Baptiste Lespiau | 12 October 2020, 17:39:03 UTC | Gate some jax_jit test with a version check. | 12 October 2020, 18:04:19 UTC |
| c1e2595 | Jean-Baptiste Lespiau | 12 October 2020, 15:49:14 UTC | Add support for jax_debug_nans and fix the last few glitches with the C++ jax.jit. - Sorting the keyword arguments must be done on the string, because we go through the Python path which uses flatten() which sort them by string. - Some error with obj == obj which is the same as obj.is(obj) and not obj.equal(obj). - Moves all the Python tests to the C++ tests (which also run on the _python_jit). PiperOrigin-RevId: 336671123 | 12 October 2020, 15:50:13 UTC |
| 8f64715 | jax authors | 12 October 2020, 13:48:35 UTC | Merge pull request #4546 from hawkinsp:flakes PiperOrigin-RevId: 336653760 | 12 October 2020, 13:48:35 UTC |
| e1adbcd | Peter Hawkins | 12 October 2020, 13:35:39 UTC | Fix batching_test flakiness on GPU. | 12 October 2020, 13:35:39 UTC |
| 9d8139e | jax authors | 12 October 2020, 11:49:55 UTC | Merge pull request #4440 from google:long-line PiperOrigin-RevId: 336638751 | 12 October 2020, 11:49:55 UTC |
| 7320fc7 | jax authors | 12 October 2020, 11:46:26 UTC | Merge pull request #4542 from google:fix-jax2tf-typo PiperOrigin-RevId: 336638720 | 12 October 2020, 11:46:26 UTC |
| bcf3209 | Matthew Johnson | 11 October 2020, 17:17:46 UTC | remove trailing whitespace in jax2tf readme lines | 11 October 2020, 17:17:46 UTC |
| e1708f5 | Matthew Johnson | 11 October 2020, 17:15:48 UTC | fix typo "differentiaion" | 11 October 2020, 17:15:48 UTC |
| e6ab45e | jax authors | 11 October 2020, 05:14:45 UTC | Merge pull request #4538 from google:fix-mypy PiperOrigin-RevId: 336507262 | 11 October 2020, 05:14:45 UTC |
| 2839b2b | Matthew Johnson | 11 October 2020, 05:01:43 UTC | ignore mypy error in jax2tf | 11 October 2020, 05:01:53 UTC |
| 20e11a0 | jax authors | 11 October 2020, 04:34:44 UTC | Merge pull request #4509 from malmaud:pmin_multihost PiperOrigin-RevId: 336504945 | 11 October 2020, 04:34:44 UTC |
| 0485fd8 | Alex Minnaar | 04 October 2020, 21:46:13 UTC | adding histogram2d implementation | 10 October 2020, 16:03:46 UTC |
| cff6c0b | Jake VanderPlas | 02 October 2020, 22:08:21 UTC | Improve errors for failed compilations w/ core.concrete_or_error | 10 October 2020, 16:03:46 UTC |
| a97cb87 | jax authors | 10 October 2020, 15:00:02 UTC | Merge pull request #4532 from gnecula:tf_omnistaging PiperOrigin-RevId: 336456477 | 10 October 2020, 15:00:02 UTC |
| 9d9762b | George Necula | 10 October 2020, 07:30:23 UTC | [jax2tf] Ensure jax2tf still works without omnistaging This refines changes from PR 4470 to ensure that jax2tf works even without omnistaging. Some tests would fail though, e.g., converting a function with no arguments | 10 October 2020, 07:43:17 UTC |
| 0d69a84 | jax authors | 09 October 2020, 21:39:49 UTC | Merge pull request #4065 from skye:wheels PiperOrigin-RevId: 336368043 | 09 October 2020, 21:39:49 UTC |
| cacb017 | Skye Wanderman-Milne | 14 August 2020, 00:30:08 UTC | Use local version identifiers to distribute cuda jaxlib wheels. This change: * Updates our jaxlib build scripts to add `+cudaXXX` to the wheel version, where XXX is the CUDA version number (e.g. `110`). nocuda builds remain unchanged and do not have this extra identifier. * Adds `generate_release_index.py`, which writes an html page that pip can use to find the cuda wheels. (I based this format off of wheel PyTorch's index). * Updates the README to use the new local version identifier + wheel index. The end result is that the command to install cuda wheels is now much simpler. I manually made copies of the latest jaxlib 0.1.55 wheels that have the local version identifiers, so the new installation commands already work (as well as the old ones, until the next jaxlib release using the new tooling). Fow now, I put the html index to the GCP bucket with the wheels. We can move it to a prettier URL if/when we have one. | 09 October 2020, 20:47:54 UTC |
| ae3d310 | jax authors | 09 October 2020, 20:05:37 UTC | Merge pull request #4524 from jakevdp:fix-mean PiperOrigin-RevId: 336350279 | 09 October 2020, 20:05:37 UTC |
| a57243c | Jake VanderPlas | 09 October 2020, 19:57:17 UTC | Cleanup: remove extraneous device_put | 09 October 2020, 19:57:17 UTC |
| cd03bea | jax authors | 09 October 2020, 18:31:47 UTC | Merge pull request #4526 from google:4510-2 PiperOrigin-RevId: 336332245 | 09 October 2020, 18:31:47 UTC |
| 78fe038 | Matthew Johnson | 09 October 2020, 18:01:58 UTC | fix x64 issue | 09 October 2020, 18:01:58 UTC |
| 0aeeb63 | Matthew Johnson | 09 October 2020, 17:35:47 UTC | add test for scan input forwarding | 09 October 2020, 17:35:47 UTC |
| 5355776 | Matthew Johnson | 09 October 2020, 17:26:28 UTC | make scan fwd raw extensive inputs as DeviceArray follow-up on #4517 | 09 October 2020, 17:26:28 UTC |
| fc69d45 | jax authors | 09 October 2020, 17:06:22 UTC | Merge pull request #4470 from gnecula:tf_tracers PiperOrigin-RevId: 336314262 | 09 October 2020, 17:06:22 UTC |
| 67a77c7 | Jake VanderPlas | 09 October 2020, 16:13:21 UTC | Fix extraneous dtype warning in jnp.mean | 09 October 2020, 16:13:21 UTC |
| 0213efd | George Necula | 07 October 2020, 06:18:15 UTC | [jax2tf] Port jax2tf to use omnistaging The main change is that we use `core.new_base_main` to use an omnistaging-based tracer. This has the benefit that we can convert to TF even functions with no arguments (previously they would be constant-folded by JAX prior to the conversion). We also add an explicit error if the jax2tf.convert transformation is nested under other JAX transformations. | 09 October 2020, 15:42:28 UTC |
| e194dff | jax authors | 09 October 2020, 15:37:52 UTC | Merge pull request #4518 from gnecula:tf_debug PiperOrigin-RevId: 336298684 | 09 October 2020, 15:37:52 UTC |
| a1e67be | George Necula | 09 October 2020, 09:39:21 UTC | [jax2tf] Fix the behavior for shift when the shift amount is negative or large. It turns out that TF shift behavior is "implementation defined" if the shift amount is negative or larger or equal to the bitsize of the operand. This is different than XLA. We add conditionals to check and fix the corner cases, we expand the tests. A better solution would be to use an XlaShift operation. This also revealed some bugs with shift_right on TPU (for JAX). The jax2tf behavior is what we believe is the correct behavior, which means that it differs from JAX results on TPU. | 09 October 2020, 15:09:08 UTC |
| e6399cd | jax authors | 09 October 2020, 05:24:59 UTC | Merge pull request #4517 from google:issue4510 PiperOrigin-RevId: 336230439 | 09 October 2020, 05:24:59 UTC |
| 7691be9 | Matthew Johnson | 09 October 2020, 04:04:44 UTC | make pytype happy | 09 October 2020, 04:04:44 UTC |
| 52fe026 | Matthew Johnson | 09 October 2020, 03:31:39 UTC | optimize scan partial_eval to fix #4510 fixes #4510 | 09 October 2020, 03:34:34 UTC |
| d4da9cc | jax authors | 08 October 2020, 23:49:22 UTC | Merge pull request #4514 from google:remove-unneeded-code PiperOrigin-RevId: 336191840 | 08 October 2020, 23:49:22 UTC |
| ab462ba | jax authors | 08 October 2020, 23:17:38 UTC | Merge pull request #4515 from google:enable-conv-test PiperOrigin-RevId: 336186047 | 08 October 2020, 23:17:38 UTC |
| 4994b82 | jax authors | 08 October 2020, 22:46:50 UTC | Merge pull request #4512 from jakevdp:raise-to-shaped PiperOrigin-RevId: 336180373 | 08 October 2020, 22:46:50 UTC |
| 1614e28 | Roy Frostig | 08 October 2020, 22:43:15 UTC | re-enable lax autodiff test xla:cpu bug was due to a change in llvm, now reverted | 08 October 2020, 22:43:15 UTC |
| 55be9dd | Matthew Johnson | 08 October 2020, 22:36:42 UTC | remove mysterious code that is no longer needed | 08 October 2020, 22:36:42 UTC |
| 91492ca | jax authors | 08 October 2020, 22:33:02 UTC | Merge pull request #4511 from google:debug-nans-test-move PiperOrigin-RevId: 336178046 | 08 October 2020, 22:33:02 UTC |
| ef25da8 | Jake VanderPlas | 08 October 2020, 22:21:57 UTC | lax_control_flow: retry via function rather than via loop | 08 October 2020, 22:22:16 UTC |
| 008eae6 | Jean-Baptiste Lespiau | 08 October 2020, 21:39:06 UTC | Fallback to Python for float0 in the C++ jax.jit. PiperOrigin-RevId: 336167762 | 08 October 2020, 21:39:35 UTC |
| 24de811 | Matthew Johnson | 08 October 2020, 20:34:56 UTC | move a debug_nans test into debug_nans test file | 08 October 2020, 20:34:56 UTC |
| ddcfd45 | jax authors | 08 October 2020, 20:15:48 UTC | Merge pull request #4161 from jakevdp:raise-to-shaped PiperOrigin-RevId: 336150089 | 08 October 2020, 20:15:48 UTC |
| 3be8188 | jax authors | 08 October 2020, 20:12:35 UTC | Merge pull request #4483 from google:debug-nans-no-store-error PiperOrigin-RevId: 336149914 | 08 October 2020, 20:12:35 UTC |
| 57ec9dd | jax authors | 08 October 2020, 20:01:26 UTC | Merge pull request #4483 from google:debug-nans-no-store-error PiperOrigin-RevId: 336147286 | 08 October 2020, 20:01:26 UTC |
| 09f2be1 | Matthew Johnson | 08 October 2020, 20:00:32 UTC | wait for result in debug_nans_test | 08 October 2020, 20:00:32 UTC |
| 2b76bcc | jax authors | 08 October 2020, 19:53:33 UTC | Merge pull request #4508 from zhangqiaorjc:expm_test_slow PiperOrigin-RevId: 336146064 | 08 October 2020, 19:53:33 UTC |
| 4dd802c | Matthew Johnson | 08 October 2020, 19:28:11 UTC | Merge branch 'master' into debug-nans-no-store-error | 08 October 2020, 19:28:11 UTC |
| a2a9409 | Jonathan Malmaud | 08 October 2020, 19:12:15 UTC | Mark pmin and pmax as multi-host supported. | 08 October 2020, 19:12:15 UTC |
| afef644 | Qiao Zhang | 08 October 2020, 19:09:21 UTC | Skip 2nd order grad in expm tests. | 08 October 2020, 19:09:21 UTC |
| a2a6dca | jax authors | 08 October 2020, 19:06:15 UTC | Merge pull request #4507 from hawkinsp:xla PiperOrigin-RevId: 336135890 | 08 October 2020, 19:06:15 UTC |
| a20b00a | Jake VanderPlas | 08 October 2020, 18:59:56 UTC | fix weak_type issues in while_loop | 08 October 2020, 18:59:56 UTC |
| 72be97c | Peter Hawkins | 08 October 2020, 18:59:06 UTC | Update XLA. | 08 October 2020, 18:59:06 UTC |
| 0f0fa53 | Jake VanderPlas | 08 October 2020, 18:52:07 UTC | fix weak_type issues in while_loop | 08 October 2020, 18:53:52 UTC |
| 3ca6e5a | Jake VanderPlas | 07 October 2020, 22:25:01 UTC | fix weak type issues in scan | 08 October 2020, 18:53:52 UTC |
| 6393349 | Jake VanderPlas | 07 October 2020, 18:41:22 UTC | raise_to_shaped: preserve weak_type by default | 08 October 2020, 18:53:52 UTC |
| ab48a9a | jax authors | 08 October 2020, 18:49:06 UTC | Merge pull request #4498 from jblespiau:changelist/333713171 PiperOrigin-RevId: 336132107 | 08 October 2020, 18:49:06 UTC |
| 145ac40 | jax authors | 08 October 2020, 17:45:01 UTC | Merge pull request #4502 from hawkinsp:linalg PiperOrigin-RevId: 336117857 | 08 October 2020, 17:45:01 UTC |
| 684a584 | jax authors | 08 October 2020, 17:20:34 UTC | Merge pull request #4500 from LenaMartens:changelist/336082045 PiperOrigin-RevId: 336112168 | 08 October 2020, 17:20:34 UTC |
| e3d622c | Lena Martens | 08 October 2020, 14:36:05 UTC | Recast int/bool tangents to float0 in custom_jvp/vjps (also in the initial_style path). | 08 October 2020, 16:37:35 UTC |
| 95c0fea | jax authors | 08 October 2020, 16:06:43 UTC | Merge pull request #4501 from gnecula:tf_tests_cf1 PiperOrigin-RevId: 336096721 | 08 October 2020, 16:06:43 UTC |
| 2fca917 | Peter Hawkins | 08 October 2020, 14:59:52 UTC | Enable TPU linalg tests that now pass. | 08 October 2020, 14:59:52 UTC |
| 6fec497 | Jean-Baptiste Lespiau | 25 September 2020, 12:14:41 UTC | Add a flag to control the C++ jax.jit behavior. | 08 October 2020, 14:51:07 UTC |
| dc9168b | George Necula | 08 October 2020, 09:28:42 UTC | [jax2tf] Ensure that in tests TF does not constant-fold in eager mode before compiling | 08 October 2020, 14:44:00 UTC |
| 8df116b | jax authors | 08 October 2020, 14:27:06 UTC | Merge pull request #4475 from bchetioui:fix_static_tests PiperOrigin-RevId: 336080718 | 08 October 2020, 14:27:06 UTC |
| 8299598 | Benjamin Chetioui | 08 October 2020, 13:37:20 UTC | Fix failing tests on TPU by avoiding using scalars | 08 October 2020, 13:37:20 UTC |
| 83060cc | jax authors | 08 October 2020, 13:06:56 UTC | Merge pull request #4491 from google:remove-test PiperOrigin-RevId: 336069850 | 08 October 2020, 13:06:56 UTC |
| e65115b | jax authors | 08 October 2020, 02:55:59 UTC | Merge pull request #4489 from google:test-fixes PiperOrigin-RevId: 336003413 | 08 October 2020, 02:55:59 UTC |
| 30b8cc5 | Roy Frostig | 08 October 2020, 02:49:06 UTC | remove an always-skipped test, redundant by now | 08 October 2020, 02:49:06 UTC |
