experiments/STT-tensorflow
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity
nightly
STT-tensorflow/tensorflow/compiler/xla/pjrt
History
Skye Wanderman-Milne 67ff36f392 [PJRT] Expose TPU runtime version via new PjRtClient::platform_version string. 
On Cloud TPU VMs, the returned platform version will look like:

libtpu version 0.0.1
Built on Mar 4 2021 15:25:57 (1614900357) cl/360760169

On all other platform, the platform version will be "<unknown>".

[XLA:Python] Add xla_client.Client.platform_version

PiperOrigin-RevId: 361215574
Change-Id: Ie65ce509caa550277972751cfb58027ad58c1d23
2021-03-05 13:45:22 -08:00
..
distributed Replace deprecated thread annotations macros. 2021-01-25 06:29:34 -08:00
BUILD [XLA:Python] Expose buffer addresses in Python. 2021-03-02 16:57:49 -08:00
cpu_device.cc [PJRT] Rename PjRtDevice::host_id and PjRtClient::host_id to ...::task_id instead. 2021-01-26 17:32:20 -08:00
cpu_device.h Split StreamExecutor impl of pjrt_client into its own files. 2020-12-10 10:07:54 -08:00
event_pool.cc [PJRT] Move host literal cache out of PjRtBuffer::ToLiteral() and into the XLA Python bindings. Change ToLiteral() to have an asynchronous API that writes its output into a caller-provided buffer. Delete CopyToHostAsync() because it now serves no purpose. 2021-01-20 13:17:12 -08:00
event_pool.h [PJRT] Move host literal cache out of PjRtBuffer::ToLiteral() and into the XLA Python bindings. Change ToLiteral() to have an asynchronous API that writes its output into a caller-provided buffer. Delete CopyToHostAsync() because it now serves no purpose. 2021-01-20 13:17:12 -08:00
gpu_device.cc - Fix PjRt GPU Client to allow num_partitions > 1 2021-02-01 12:33:28 -08:00
gpu_device.h Split StreamExecutor impl of pjrt_client into its own files. 2020-12-10 10:07:54 -08:00
gpu_multistream_test.cc [PJRT] Change PjRtClient::BufferFromHostBuffer to take an on_done_with_host_buffer callback rather than a std::shared_ptr<void> that must be kept alive. 2021-01-27 07:49:33 -08:00
interpreter_device.cc [PJRT] Rename PjRtDevice::host_id and PjRtClient::host_id to ...::task_id instead. 2021-01-26 17:32:20 -08:00
interpreter_device.h Split StreamExecutor impl of pjrt_client into its own files. 2020-12-10 10:07:54 -08:00
local_device_state.cc Avoid crashing in LocalDeviceState::ReturnStreamToPool and BorrowStreamFromPool when stream fails with "ABORTED: Bad connection". 2021-02-02 22:11:16 -08:00
local_device_state.h Use perfect forwarding for ThenRelease. 2021-02-25 12:16:41 -08:00
pjrt_client.cc [XLA:Python] Expose buffer addresses in Python. 2021-03-02 16:57:49 -08:00
pjrt_client.h [PJRT] Expose TPU runtime version via new PjRtClient::platform_version string. 2021-03-05 13:45:22 -08:00
pjrt_stream_executor_client.cc [PJRT] Extract utils for getting MaxParallelism level. 2021-02-28 12:42:03 -08:00
pjrt_stream_executor_client.h [PJRT] Expose TPU runtime version via new PjRtClient::platform_version string. 2021-03-05 13:45:22 -08:00
semaphore_test.cc
semaphore.cc
semaphore.h
tpu_client.cc [PJRT] Expose TPU runtime version via new PjRtClient::platform_version string. 2021-03-05 13:45:22 -08:00
tpu_client.h [PJRT] Rename PjRtDevice::host_id and PjRtClient::host_id to ...::task_id instead. 2021-01-26 17:32:20 -08:00
tracked_device_buffer_test.cc [PJRT] Remove on_host_shape from PjRtBuffer API. 2021-01-28 15:12:53 -08:00
tracked_device_buffer.cc [PJRT] Remove on_host_shape from PjRtBuffer API. 2021-01-28 15:12:53 -08:00
tracked_device_buffer.h [PJRT] Remove on_host_shape from PjRtBuffer API. 2021-01-28 15:12:53 -08:00
utils.cc [PJRT] Extract utils for getting MaxParallelism level. 2021-02-28 12:42:03 -08:00
utils.h [PJRT] Extract utils for getting MaxParallelism level. 2021-02-28 12:42:03 -08:00
worker_thread.cc
worker_thread.h