Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Develop upstream sync 230717 #2160

Merged
merged 400 commits into from
Jul 19, 2023
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
400 commits
Select commit Hold shift + click to select a range
ad14d68
Internal Code Change
rickeylev Jul 11, 2023
138866e
Ensure scatter_dims_to_operand_dims is not an out of range index.
BlaziusMaximus Jul 11, 2023
2b1cf8a
[xla:gpu] Add an option to instantiate all CUDA graphs before running…
ezhulenev Jul 11, 2023
9af07b5
Move compiler imports from python/__init__.py to python/modules_with_…
fionalang Jul 11, 2023
bfe68b5
Integrate LLVM at llvm/llvm-project@be29fe2f987b
tensorflower-gardener Jul 11, 2023
89cae4a
[NFC] Change uses of get_compatible_with_cloud to get_compatible_with…
ddunl Jul 11, 2023
4ca4a8d
This is an initial commit to introduce a MHLO custom op that leverage…
tensorflower-gardener Jul 11, 2023
57dd47d
Update TFRT dependency to use revision
tensorflower-gardener Jul 11, 2023
e449b10
[TF:PJRT] Returns an error if the compilation result is TensorList.
Jul 11, 2023
f687e26
Enable to set fdo_profile through XLA python client.
wang12tao Jul 11, 2023
d1e3c9c
Fix typo in comment
tensorflower-gardener Jul 11, 2023
abae3ee
Disable test that breaks tensorflow.gpu.pascal
w-xinyi Jul 11, 2023
930a384
Update rules_python version to 0.23.1
tensorflower-gardener Jul 11, 2023
6229f74
Update TFRT dependency to use revision
tensorflower-gardener Jul 11, 2023
eff4a80
[XLA:GPU] Rollback cl/547196631.
Jul 11, 2023
6df4c01
Update ops.Tensor references to //third_party/tensorflow/python/frame…
fionalang Jul 11, 2023
dac9af0
[TF:PJRT] Use ShapeUtil::Compatible when checking the compatibility b…
Jul 11, 2023
e713567
[XLA] Extended associative reordering to work with arbitrary contract…
tensorflower-gardener Jul 11, 2023
e87d1c9
Internal Code Change
rickeylev Jul 11, 2023
41f4706
Rollback 8b116e21d125efe69764316a86af4f109bb3d5b6.
reedwm Jul 12, 2023
1295151
[XLA] Fix `IsPerIdOffset` to check IsEffectiveScalar
jurahul Jul 12, 2023
6c7ebc9
Benchmarks from MLIR for tfl tensorlists
LukeBoyer Jul 12, 2023
ce78c0d
[xla:runtime] Use volatile store to encode args/rets
ezhulenev Jul 12, 2023
221af6a
[PJRT C API] Add host_layout to ToHostBufferArg.
Jul 12, 2023
ee4da7b
Add an e2e test for a SAX model with streaming using TFRT
cky9301 Jul 12, 2023
2821ab8
[PJRT C API] Support passing allow_devices as an option in PJRT GPU p…
Jul 12, 2023
e07f254
Internal Code Change
tensorflower-gardener Jul 12, 2023
61df417
Internal Code Change
tensorflower-gardener Jul 12, 2023
bc3e83d
Remove unused include (NFC)
akuegel Jul 12, 2023
21eb514
Refactor xla_call_module_loader to share refine_polymorphic_shapes
gnecula Jul 12, 2023
941c1ee
Internal Code Change
tensorflower-gardener Jul 12, 2023
94eeda9
Integrate LLVM at llvm/llvm-project@5671f023042b
tensorflower-gardener Jul 12, 2023
08ecf8a
Update TFRT dependency to use revision
tensorflower-gardener Jul 12, 2023
aeb69db
Add gpu_asm_compiler and gpu_asm_opts_util deps behind a guard.
akuegel Jul 12, 2023
2eec7c5
Also simplify Bitcast(Broadcast) -> Broadcast if possible.
akuegel Jul 12, 2023
0ee90b9
compat: Update forward compatibility horizon to 2023-07-12
tensorflower-gardener Jul 12, 2023
f5ca7ef
Update GraphDef version to 1555.
tensorflower-gardener Jul 12, 2023
121ae36
[DelegatePerformance] Move the MiniBenchmark C APIs to shims.
shiqing117 Jul 12, 2023
1e8ed6d
Replace quantfork.stats to quantfork.q/dq instead of stablehlo.q/dq
doyeonkim0 Jul 12, 2023
43e7f26
[DelegatePerformance] Updated the accuracy benchmark flow.
shiqing117 Jul 12, 2023
22abfba
[XLA:GPU] [NFC] Generalize autotuner_compile_util to support multiple…
cheshire Jul 12, 2023
bb0f461
WHILE: Set subgraphs_prepared to false during lazy initialization
alankelly Jul 12, 2023
a2c7600
[XLA:GPU] Enable --xla_gpu_enable_experimental_block_size by default.
Jul 12, 2023
7ef08c3
Don't depend on @llvm-project//mlir:ConversionPasses, instead depend …
hawkinsp Jul 12, 2023
4b81db9
Add support for shape assertions to RefinePolymorphicShapes
gnecula Jul 12, 2023
ffe0b91
Import openai/triton from GitHub.
chsigg Jul 12, 2023
bca2121
Make upload condition more specific
MichaelHudgins Jul 12, 2023
395d942
[XLA:GPU] Minor refactoring of autotuning utils
cheshire Jul 12, 2023
3a5c137
Do not run `concat_ops_test` in asan config due to timeouts.
tensorflower-gardener Jul 12, 2023
ff8abca
[XLA:GPU] [NFC] Consolidate the logic to create RedzoneAllocator to a…
cheshire Jul 12, 2023
b853df2
[XLA StreamExecutor TPU] Unit tests for c_api_conversions.h
tensorflower-gardener Jul 12, 2023
dbec0c0
[Linaro:ARM_CI] Add broken test to skip list
elfringham Jul 12, 2023
38a9f7d
Correct layout order on top-level multi-device call op.
jszaday Jul 12, 2023
58e8887
Mirror conditional structure between send and recv lowering paths.
jszaday Jul 12, 2023
3a8a467
Fix AVX512 builds involving XLA CPU conv2d.
cantonios Jul 12, 2023
ad5b2ea
[XLA/GPU] Change PGLE to support either text or binary proto files
jurahul Jul 12, 2023
3efa230
[KernelGen] JIT-compile most the MLIR-generated GPU kernels
rainwoodman Jul 12, 2023
a4d063f
Fix a copy-and-paste error
junwhanahn Jul 12, 2023
2e86aa7
Add CompatibilityStatusToString and StringToCompatibilityStatus util …
shiqing117 Jul 12, 2023
24f3230
[XLA:GPU] Support Conv-Bias-Relu6/LeakyRelu fusion in XLA using cuDNN…
Jul 12, 2023
fcb3f4f
Merge pull request #60723 from renato-arantes:remedial_perf
tensorflower-gardener Jul 12, 2023
05f947c
[xla:gpu] Remove async passes from default Gpu pipeline
ezhulenev Jul 12, 2023
729bf53
#tf-data Set up `file_locality` experiment.
mpcallanan Jul 12, 2023
bd2188b
[XLA] Add missing argument `shard_count` to the table for AllGather.
ghpvnist Jul 12, 2023
6e95771
Merge pull request #60528 from AsakusaRinne:add_c_apis
tensorflower-gardener Jul 12, 2023
9d0bf90
Extract HloRematerialization options out into a new struct called Opt…
SandSnip3r Jul 12, 2023
30813d0
Integrate StableHLO at openxla/stablehlo@4add5f0
ghpvnist Jul 12, 2023
d4b760a
Add input type inference logic and WeakTensor input and output suppor…
tensorflower-gardener Jul 12, 2023
63bfa44
Integrate LLVM at llvm/llvm-project@b10899d86995
tensorflower-gardener Jul 12, 2023
0108f44
Delete get_compatible_with_cloud. Update all users to get_compatible_…
ddunl Jul 12, 2023
c35b2d5
Merge pull request #61206 from angerson:master
tensorflower-gardener Jul 12, 2023
b9e1aff
Update TFRT dependency to use revision
tensorflower-gardener Jul 12, 2023
3ce365a
Add experimental APIs to signature runner to set custom allocation fo…
sirakiin Jul 12, 2023
59db071
[XLA:GPU] Enable relu6 fusion on Turing.
Jul 12, 2023
45a552b
[XLA:CPU] Shard runtime matmul kernels.
hawkinsp Jul 12, 2023
30da06d
Handle multiple sets of inferred resource indices and layouts.
jszaday Jul 12, 2023
def8728
Remove autoclustering in TFRT.
tensorflower-gardener Jul 12, 2023
ab731c0
Add error logging for internal error statistics. Logging invocation…
tensorflower-gardener Jul 12, 2023
74a6afa
Increase type support for DoInplace and DoCopy to uint8, int8 and uint64
rainwoodman Jul 12, 2023
8ed1178
Integrate StableHLO at openxla/stablehlo@41bad51
ghpvnist Jul 12, 2023
dae72e3
Code reformatting for readibility.
rainwoodman Jul 12, 2023
268090e
Move tf debug imports from python/__init__.py to python/modules_with_…
fionalang Jul 12, 2023
cb73612
[XLA:GPU] Disable leaky-relu fusion.
Jul 12, 2023
c4d31b0
Adds check for Optional Tensors before increasing the reference count…
tensorflower-gardener Jul 12, 2023
bd41e62
[xla:gpu] Support exporting dataflow graph in DOT format for debugging
anlunx Jul 12, 2023
1ef2540
Add more context to the incorrect tensor num error
faizan-m Jul 12, 2023
d6aba8d
Allow RegisterClientFactory() and GetClient() in xla::ifrt::test_util…
tensorflower-gardener Jul 12, 2023
27a9389
[PJRT C API] Add a C API to query plugin attributes.
Jul 12, 2023
c0b6aa8
Update Tensor dunder methods and add numpy methods to WeakTensor.
tensorflower-gardener Jul 12, 2023
1459cc7
[XLA] Make sure that we reset the main loop of whle_loop_all_reduce_m…
Jul 12, 2023
f819dbd
Merge pull request #57143 from luliyucoordinate:fix_tensor
tensorflower-gardener Jul 12, 2023
928ae03
Update build rule for WeightWatcher
terryheo Jul 13, 2023
21829ff
[IFRT] Add serialization/deserialization for shardings
hyeontaek Jul 13, 2023
726d6cf
Add TfLiteInterpreterOptionsSetOpResolverExternalWithFallback.
fergushenderson Jul 13, 2023
2d938b3
Addressed review comments
mahmoud-abuzaina Jul 13, 2023
177aa64
Delete `tsl_gpu_cc_test` and replace all users with `xla_cc_test`.
ddunl Jul 13, 2023
a310a86
Remove use of jitrt, as it is no longer used.
tensorflower-gardener Jul 13, 2023
1c0b86b
Fix segfault in `XlaCallModule` shape inference
junwhanahn Jul 13, 2023
7b2adc1
Merge pull request #61262 from elfringham:broken_fusion
tensorflower-gardener Jul 13, 2023
16d2b01
Internal Code Change
tensorflower-gardener Jul 13, 2023
22f177b
Internal Code Change
tensorflower-gardener Jul 13, 2023
edeacae
Merge pull request #60393 from VijayR19:my-feature-branch
tensorflower-gardener Jul 13, 2023
7af32ab
Revert: Attempt to be less restrictive in FusionCanShareBufferHint().
akuegel Jul 13, 2023
1268853
Internal Code Change
tensorflower-gardener Jul 13, 2023
70c60ac
Merge pull request #60498 from Intel-tensorflow:yuqingding/fix_null_bug
tensorflower-gardener Jul 13, 2023
b9df58c
Internal Code Change
tensorflower-gardener Jul 13, 2023
8181a7f
Integrate LLVM at llvm/llvm-project@a69b2e3d1c1a
tensorflower-gardener Jul 13, 2023
a09d9ae
Fixed string comparison.
roserg Jul 13, 2023
e07594e
[XLA:GPU] Roll-forward cl/543680393: Fuse more inputs into Triton GEMMs.
Jul 13, 2023
dfd5e6e
Update GraphDef version to 1556.
tensorflower-gardener Jul 13, 2023
0afb542
compat: Update forward compatibility horizon to 2023-07-13
tensorflower-gardener Jul 13, 2023
0db2637
Clear the patch file, it is not needed anymore.
akuegel Jul 13, 2023
d87b3cb
Add HloVerifier to run_hlo_module.
akuegel Jul 13, 2023
08a1c52
Remove unused autograph converter code in `tensorflow/python/autograp…
ashishenoyp Jul 13, 2023
f31e1b5
[XLA:GPU] Roll forward cl/543697810: Fuse outputs into Triton GEMMs.
Jul 13, 2023
4e5c09c
Avoid using invalidated iterator after erase.
akuegel Jul 13, 2023
83bbee7
[XLA:GPU] Disable cudnn runtime fusion.
Jul 13, 2023
25c87db
Fix typo: `s/the the/the /`.
tensorflower-gardener Jul 13, 2023
cf4fb63
Remove unused autograph converter in `tensorflow/python/autograph/con…
ashishenoyp Jul 13, 2023
6635250
Fix typo: `s/the the/the /`.
tensorflower-gardener Jul 13, 2023
651876f
Remove a potential duplicate of `RETURN_IF_ERROR`.
jblespiau Jul 13, 2023
ade1ff0
[XLA:GPU] Support specific kind of output transposes in Triton GEMM.
Jul 13, 2023
9fa5e77
[XLA:GPU] [NFC] In Triton Gemm autotuning, do not copy the returned b…
cheshire Jul 13, 2023
aa7fbf1
Merge pull request #61123 from davsva01:acl2305_acl_reorder_upstream
tensorflower-gardener Jul 13, 2023
3e0503c
Delete DeviceCompiler when a new PjRtClient is created for DEVICE_GPU.
swachhandl Jul 13, 2023
fff678b
[IFRT] Rollback for serialization/deserialization for shardings
hyeontaek Jul 13, 2023
b57c866
Merge pull request #61261 from MichaelHudgins:arm-64-release-branch-b…
tensorflower-gardener Jul 13, 2023
e72477c
Collect target processing time in ClientHeartbeat
armandouv Jul 13, 2023
8d4e219
Catch another type of error that indicates coordination service is do…
jamesmullenbach Jul 13, 2023
59e4c69
Add logging information to Ph1 call sites
tensorflower-gardener Jul 13, 2023
0ed8e90
Delete unused tensor_priority_test.py.
fionalang Jul 13, 2023
f861397
Fix the propagation of the errors through TF_ASSIGN_OR_RETURN.
jblespiau Jul 13, 2023
24c79b6
Log ParameterServerStrategy variable placements at a higher verbosity…
jamesmullenbach Jul 13, 2023
f49705d
Fix typo that causes CPU kernel to be registered twice.
tensorflower-gardener Jul 13, 2023
253dab8
Build changes for AArch64.
tensorflower-gardener Jul 13, 2023
182bec8
Pattern to fuse/fold the reshape ops around TFL_BatchMatMulOp
tensorflower-gardener Jul 13, 2023
8f37b93
#tf-data-service Enable bufferedio for loading tf.data snapshots.
yangustc07 Jul 13, 2023
2c4a76e
Adds check for Optional Tensors before increasing the reference count…
tensorflower-gardener Jul 13, 2023
7d2345d
Correct shader generation for different cases of the MUL operation.
tensorflower-gardener Jul 13, 2023
8f04131
Internal change only.
BlaziusMaximus Jul 13, 2023
26d212e
Delete DeviceCompiler when a new PjRtClient is created for DEVICE_GPU.
swachhandl Jul 13, 2023
29b9430
Support custom dataclasses in TensorFlowTestCase.evaluate.
tensorflower-gardener Jul 13, 2023
06dc7c3
[xla:gpu] NFC: Add graph exec id to cuda graphs logging
ezhulenev Jul 13, 2023
6f7c908
Explicitly disable use of tfrt for failing test.
sizhit2 Jul 13, 2023
15be3fb
[xla:gpu] Improve stream assignment for concurrent regions
anlunx Jul 13, 2023
b94792a
Replaces absl::string_view to llvm::StringRef to make it compatible…
chhe7 Jul 13, 2023
9335670
Do not run pjrt_c_api_gpu_test in debug mode due to failure.
tensorflower-gardener Jul 13, 2023
58d49ef
Integrate LLVM at llvm/llvm-project@1936bb81aafd
tensorflower-gardener Jul 13, 2023
f71f31c
Make FunctionType.flat_inputs and .flat_captures thread safe.
jamesmullenbach Jul 13, 2023
a2ab482
Change IFRT's kPred DType to have 1-byte width instead of 1-bit.
tensorflower-gardener Jul 13, 2023
3a8b4ff
PR #3980: [python:xla_extension] Handle unbounded recursion in cyclic…
tttc3 Jul 13, 2023
8708a23
Add dataclass support in tf.data data processing
tensorflower-gardener Jul 13, 2023
a03af28
[XLA:GPU] Implement the dimension analysis of broadcasts in Triton GE…
Jul 13, 2023
b8fecf0
[NFC] Unify the 2 variant of MatchReduceScatter by using default values.
jurahul Jul 13, 2023
fc6e00d
[xla:gpu] Pass valid pointer for CUDA graph instantiation
ezhulenev Jul 13, 2023
306fb28
Update TFRT dependency to use revision
tensorflower-gardener Jul 13, 2023
4e7a7c4
Pattern to fuse/fold TFL_TransposeOp into TFL_BatchMatMulOp
tensorflower-gardener Jul 13, 2023
65d75a7
Internal change only.
tkoeppe Jul 13, 2023
c6dcf8e
Update ops.Tensor references to //third_party/tensorflow/python/frame…
fionalang Jul 13, 2023
125e130
Update ops.Tensor references to //third_party/tensorflow/python/frame…
fionalang Jul 13, 2023
52b3b68
Fix heap use after free for local variable c_options.
Jul 14, 2023
b27da3b
Remove GPU specific logic in tfrt_graph_execution_state.
changhuilin Jul 14, 2023
e468d20
[xla:gpu] Instantiate all cuda graphs ahead of time
ezhulenev Jul 14, 2023
be1c8f6
[NFC] Cleanup unused headers in HLO Rematerialization
jurahul Jul 14, 2023
9102418
Pass criticality and queue option for low priority to support priorit…
tensorflower-gardener Jul 14, 2023
4499c96
[XLA] Fix masking for pad uneven sharding.
Jul 14, 2023
cd35adf
[IFRT] Roll forward with fix: Add serialization/deserialization for s…
hyeontaek Jul 14, 2023
4d911a7
Merge `GetWindowedOutputSizeVerboseV2` with `GetWindowedOutputSizeVer…
ashishenoyp Jul 14, 2023
a76f255
Fix typo
yimeisun123 Jul 14, 2023
c1998c8
Merge pull request #60410 from Intel-tensorflow:tenglu/fix_plugin_info
tensorflower-gardener Jul 14, 2023
f63aefe
[XLA:GPU] Do not enable debug info manager for recursive compilation …
akuegel Jul 14, 2023
85844e7
Update auto assignment list
tensorflower-gardener Jul 14, 2023
d046765
Merge pull request #60376 from Intel-tensorflow:mabuzain/onednn-3.1-I…
tensorflower-gardener Jul 14, 2023
ceb2910
Update GraphDef version to 1557.
tensorflower-gardener Jul 14, 2023
562db36
compat: Update forward compatibility horizon to 2023-07-14
tensorflower-gardener Jul 14, 2023
4c6b4dc
[XLA:GPU] Add row length filter for matching normalization diamond in…
bchetioui Jul 14, 2023
d02daa3
Flex: Cache ValidateOutputTensorShapeConsistency
alankelly Jul 14, 2023
cdeabc5
[XLA:GPU] Minor fix: do not trigger Triton GEMM on narrowing but unsu…
Jul 14, 2023
70e59f8
Check whether there is a module before calling RegisterModule()
akuegel Jul 14, 2023
458c94a
HloFusionAnalysis: Don't return statuses where it's not necessary.
jreiffers Jul 14, 2023
6e15332
[XLA:GPU] Handle edge case in Triton Softmax rewriter where bitcast p…
bchetioui Jul 14, 2023
391b868
Forward declare AsyncBundleTypeStorage for mhlo/IR/hlo_ops_typedefs.h…
Jul 14, 2023
1b780c5
Some minor cleanup
AspirinSJL Jul 14, 2023
78f425b
Separates ValidateStaticShapes from RefineDynamicShapes.
gnecula Jul 14, 2023
9743226
[XLA:GPU] Unify two fusion traversals in the Triton GEMM rewriter.
Jul 14, 2023
257a5fc
[PJRT C API] Move PlatformName test to the test factory (pjrt_c_api_t…
tensorflower-gardener Jul 14, 2023
a003528
Make all Python targets under tensorflow/compiler/mlir/tfrt/jit/pytho…
jam14j Jul 14, 2023
d254eab
[XLA:GPU][NFC] Remove unused includes
tyb0807 Jul 14, 2023
aca9fe6
Fix some tests that were broken with XLA:GPU.
reedwm Jul 14, 2023
ea054a4
[XLA] Remove To/FromAbslStatus functions from XLA.
Jul 14, 2023
9b31c22
[KernelGen] JIT-compile most the MLIR-generated GPU kernels
frgossen Jul 14, 2023
6596126
Update cpuinfo dependency
Maratyszcza Jul 14, 2023
446d38d
Explicitly include stdlib.h in lite/util.h
tensorflower-gardener Jul 14, 2023
873e387
Explicitly disable use of tfrt for failing test.
sizhit2 Jul 14, 2023
02f2474
Explicitly disable use of tfrt for failing test.
sizhit2 Jul 14, 2023
9647183
Add logging information to Ph1 call sites
tensorflower-gardener Jul 14, 2023
2a69423
Fix multi-GPU FP8 crash.
reedwm Jul 14, 2023
4bc6cde
[xla:gpu] Compute the transitive reduction of the dependency graph
anlunx Jul 14, 2023
335bc7e
[XLA:GPU][NFC] Use more concise TF_CHECK_OK macro
tyb0807 Jul 14, 2023
28372ef
Check `indices` are in range in `tf.TensorArray.gather` method
tensorflower-gardener Jul 14, 2023
3803e23
Remove the distinction between rematerialization which happens before…
SandSnip3r Jul 14, 2023
d9de2bc
Ensure there is a Model instance in standalone::Iterator's context
armandouv Jul 14, 2023
24ddf0a
Disable cluster_coordinator_test with XLA.
reedwm Jul 14, 2023
c509b9a
Return Floating-Point Tensor for Dot-Like Hybrid Ops In ConvertMHLOQu…
tensorflower-gardener Jul 14, 2023
387f5f4
Add fingerprints generated by different compilation modes to chunked …
BlaziusMaximus Jul 14, 2023
40f0167
Deprecate `experimental_from_jax` in favor of `jax2tf` + `from_saved_…
arfaian Jul 14, 2023
5006a30
Improve ragged_cross_op input ragged splits check and fix flaky ragge…
tensorflower-gardener Jul 14, 2023
4af196b
Add metrics in LegalizeTfTypesPass to help debugging
tensorflower-gardener Jul 14, 2023
c060992
Added attributes within hlo and xla data proto to keep track of desir…
tensorflower-gardener Jul 14, 2023
e9dc450
Merge branch 'master' into agramesh/refactor-threadpool
agramesh1 Jul 15, 2023
282f113
legalize mhlo.dynamic_broadcast_in_dim to tf.broadcast_to and tf.expa…
yishuangP Jul 15, 2023
bf19194
Expose ShapeVerifier::CheckParameterCount() as a protected static fun…
tensorflower-gardener Jul 15, 2023
46b9830
Remove the flag use_bridge_for_gpu.
changhuilin Jul 15, 2023
afb11fa
Moved code to get threadpool to a function and fixed a few typos
agramesh1 Jul 15, 2023
b2b41f4
Internal Code Change
tensorflower-gardener Jul 15, 2023
b259da3
[IFRT] Clean up sharding tests to use a mock client
hyeontaek Jul 15, 2023
46896ea
[IFRT] Use `= True` instead of `= 1` in BUILD
hyeontaek Jul 15, 2023
dddee9e
[xla:gpu] NFC: Prepare graph instances cache for adding eviction policy
ezhulenev Jul 15, 2023
00fa3ec
compat: Update forward compatibility horizon to 2023-07-15
tensorflower-gardener Jul 15, 2023
69533b1
Update GraphDef version to 1558.
tensorflower-gardener Jul 15, 2023
03a0795
Rollback of PR #61123
penpornk Jul 15, 2023
3627fb0
[XLA] Bump up the number of inline tiles
majnemer Jul 16, 2023
45829c3
[IFRT] Make Sharding deserialization to use a function to lookup devices
hyeontaek Jul 16, 2023
d554f77
compat: Update forward compatibility horizon to 2023-07-16
tensorflower-gardener Jul 16, 2023
61302d8
Update GraphDef version to 1559.
tensorflower-gardener Jul 16, 2023
af20981
Add runtime config when creating KernelFallbackCompatRequest
deqiangc Jul 16, 2023
db4c8f4
Merge pull request #60611 from Intel-tensorflow:yimei/onednn3x_instno…
tensorflower-gardener Jul 16, 2023
3e69e98
[XLA:GPU] Re-enable fusion of broadcasts of scalar constants in Trito…
Jul 16, 2023
0dc2402
[XLA:GPU] Implement dimension analysis of output fusions in Triton GE…
Jul 16, 2023
52dc796
[XLA] Add option to pattern match true only scalars in offset computa…
jurahul Jul 17, 2023
95d7020
[XLA:GPU][NFC] Add FindHeroReduction util function
tyb0807 Jul 17, 2023
9f66a76
Merge pull request #60189 from linux-on-ibm-z:enable_secure_grpc_on_s…
tensorflower-gardener Jul 17, 2023
ccc45ff
Modify the signature of `QuantizeModel` to accept the buffer directly…
dansuh17 Jul 17, 2023
c35ac7f
[XLA] Fix a bug in required assignment matching that caused buffers i…
berkinilbeyi Jul 17, 2023
bdfbde0
compat: Update forward compatibility horizon to 2023-07-17
tensorflower-gardener Jul 17, 2023
5848201
Update GraphDef version to 1560.
tensorflower-gardener Jul 17, 2023
396f8ab
Support all fusion kinds except Triton in GetLaunchDimensions.
jreiffers Jul 17, 2023
1a66ca3
Extend c_api_opaque to support reading and writing strings values fro…
Jul 17, 2023
fa791fc
[XLA:GPU] Switch Triton GEMM to block pointers.
Jul 17, 2023
76d0af9
[XLA:GPU] Prevent matching converts from/to bf16 in Triton Softmax re…
bchetioui Jul 17, 2023
e1ad3b7
[XLA:GPU] Handle edge case in Triton Softmax rewriter where bitcast i…
bchetioui Jul 17, 2023
65b5bc5
Merge pull request #61236 from Intel-tensorflow:agramesh/refactor-thr…
tensorflower-gardener Jul 17, 2023
8c1827c
[XLA:GPU] Flip default for --xla_gpu_enable_triton_softmax_fusion flag.
bchetioui Jul 17, 2023
90513cc
[XLA:GPU] Only use nvlink for linking if it is at least as new as the…
hawkinsp Jul 17, 2023
26ad97d
Compute MakeEmbeddedComputationsList() iteratively.
akuegel Jul 17, 2023
e790a98
weekly sync 230717 before solving conflicts
weihanmines Jul 17, 2023
d803377
weekly sync 230717 after solving conflicts
weihanmines Jul 17, 2023
44ec8a3
fixed API changes in a few places
weihanmines Jul 17, 2023
4d7ba9d
attemp to fix xla_sharding_serdes_test failure
weihanmines Jul 18, 2023
36719ed
disable hlo-llvm ir tests and unaray op gpu test
weihanmines Jul 18, 2023
File filter

Filter by extension

Filter by extension


Conversations
Failed to load comments.
Loading
Jump to
The table of contents is too big for display.
Diff view
Diff view
  •  
  •  
  •  
1 change: 1 addition & 0 deletions .github/bot_config.yml
Original file line number Diff line number Diff line change
Expand Up @@ -18,6 +18,7 @@ assignees:
- sushreebarsa
- SuryanarayanaY
- tilakrayal
- Varsha-anjanappa
# A list of assignees for compiler folder
compiler_assignees:
- joker-eph
Expand Down
5 changes: 4 additions & 1 deletion .github/workflows/arm-cd.yml
Original file line number Diff line number Diff line change
Expand Up @@ -19,6 +19,8 @@ on:
push:
tags:
- v2.**
branches:
- r2.**
schedule:
- cron: '0 8 * * *'

Expand All @@ -30,7 +32,7 @@ jobs:
strategy:
fail-fast: false
matrix:
pyver: ['3.8', '3.9', '3.10']
pyver: ['3.9', '3.10']
experimental: [false]
include:
- pyver: '3.11'
Expand Down Expand Up @@ -66,5 +68,6 @@ jobs:
CI_DOCKER_BUILD_EXTRA_PARAMS="--build-arg py_major_minor_version=${{ matrix.pyver }} --build-arg is_nightly=${is_nightly} --build-arg tf_project_name=${tf_project_name}" \
./tensorflow/tools/ci_build/ci_build.sh cpu.arm64 bash tensorflow/tools/ci_build/rel/ubuntu/cpu_arm64_pip.sh
- name: Upload pip wheel to PyPI
if: github.event_name == 'schedule' || (github.event_name == 'push' && startsWith(github.ref, 'refs/tags/v2')) # only if it is a scheduled nightly or tagged
shell: bash
run: python3 -m twine upload --verbose /home/ubuntu/actions-runner/_work/tensorflow/tensorflow/whl/* -u "__token__" -p ${{ secrets.AWS_PYPI_ACCOUNT_TOKEN }}
2 changes: 1 addition & 1 deletion .github/workflows/arm-ci-extended.yml
Original file line number Diff line number Diff line change
Expand Up @@ -29,7 +29,7 @@ jobs:
strategy:
fail-fast: false
matrix:
pyver: ['3.8', '3.9', '3.10', '3.11']
pyver: ['3.9', '3.10', '3.11']
steps:
- name: Stop old running containers (if any)
shell: bash
Expand Down
6 changes: 3 additions & 3 deletions WORKSPACE
Original file line number Diff line number Diff line change
Expand Up @@ -14,9 +14,9 @@ http_archive(

http_archive(
name = "rules_python",
sha256 = "29a801171f7ca190c543406f9894abf2d483c206e14d6acbd695623662320097",
strip_prefix = "rules_python-0.18.1",
url = "https://github.com/bazelbuild/rules_python/releases/download/0.18.1/rules_python-0.18.1.tar.gz",
sha256 = "84aec9e21cc56fbc7f1335035a71c850d1b9b5cc6ff497306f84cced9a769841",
strip_prefix = "rules_python-0.23.1",
url = "https://github.com/bazelbuild/rules_python/releases/download/0.23.1/rules_python-0.23.1.tar.gz",
)

load("@rules_python//python:repositories.bzl", "python_register_toolchains")
Expand Down
51 changes: 51 additions & 0 deletions ci/official/any.sh
Original file line number Diff line number Diff line change
@@ -0,0 +1,51 @@
#!/bin/bash
# Copyright 2023 The TensorFlow Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# ==============================================================================
source "${BASH_SOURCE%/*}/utilities/setup.sh"

# Parse options and build targets into arrays, so that shelllint doesn't yell
# about readability. We can't pipe into 'read -ra' to create an array because
# piped commands run in subshells, which can't store variables outside of the
# subshell environment.
# Ignore grep failures since we're using it for basic filtering
set +e
filtered_build_targets=( $(echo "$BUILD_TARGETS" | tr ' ' '\n' | grep .) )
nonpip_targets=( $(echo "$TEST_TARGETS" | tr ' ' '\n' | grep -E "^//tensorflow/" ) )
config=( $(echo "$CONFIG_OPTIONS" ) )
test_flags=( $(echo "$TEST_FLAGS" ) )
set -e

if [[ "$TFCI_NVIDIA_SMI_ENABLE" == 1 ]]; then
tfrun nvidia-smi
fi

if [[ "${#filtered_build_targets[@]}" -ne 0 ]]; then
tfrun bazel "${TFCI_BAZEL_BAZELRC_ARGS[@]}" "${config[@]}" "${filtered_build_targets[@]}"
fi

if [[ "${PIP_WHEEL}" -eq "1" ]]; then
# Update the version numbers to build a "nightly" package
if [[ "$TFCI_NIGHTLY_UPDATE_VERSION_ENABLE" == 1 ]]; then
tfrun python3 tensorflow/tools/ci_build/update_version.py --nightly
fi

tfrun bazel "${TFCI_BAZEL_BAZELRC_ARGS[@]}" build "${TFCI_BAZEL_COMMON_ARGS[@]}" tensorflow/tools/pip_package:build_pip_package
tfrun ./bazel-bin/tensorflow/tools/pip_package/build_pip_package build "${TFCI_BUILD_PIP_PACKAGE_ARGS[@]}"
tfrun ./ci/official/utilities/rename_and_verify_wheels.sh
fi

if [[ "${#nonpip_targets[@]}" -ne 0 ]]; then
tfrun bazel "${TFCI_BAZEL_BAZELRC_ARGS[@]}" test "${config[@]}" "${test_flags[@]}" "${nonpip_targets[@]}"
fi
110 changes: 110 additions & 0 deletions ci/official/bazelrcs/cpu.bazelrc
Original file line number Diff line number Diff line change
@@ -0,0 +1,110 @@
# Copyright 2023 The TensorFlow Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# ==============================================================================
# This bazelrc can build a CPU-supporting TF package.

# Convenient cache configurations
# Use a cache directory mounted to /tf/cache. Very useful!
build:sigbuild_local_cache --disk_cache=/tf/cache
# Use the public-access TF DevInfra cache (read only)
build:sigbuild_remote_cache --remote_cache="https://storage.googleapis.com/tensorflow-devinfra-bazel-cache/september2022" --remote_upload_local_results=false
# Write to the TF DevInfra cache (only works for internal TF CI)
build:sigbuild_remote_cache_push --remote_cache="https://storage.googleapis.com/tensorflow-devinfra-bazel-cache/september2022" --google_default_credentials
# Change the value of CACHEBUSTER when upgrading the toolchain, or when testing
# different compilation methods. E.g. for a PR to test a new CUDA version, set
# the CACHEBUSTER to the PR number.
build --action_env=CACHEBUSTER=501872366

# Use Python 3.X as installed in container image
build --action_env PYTHON_BIN_PATH="/usr/bin/python3"
build --action_env PYTHON_LIB_PATH="/usr/lib/tf_python"
build --python_path="/usr/bin/python3"

# Build TensorFlow v2
build --define=tf_api_version=2 --action_env=TF2_BEHAVIOR=1

# Target the AVX instruction set
build --copt=-mavx --host_copt=-mavx

# Use lld as the linker
build --linkopt="-fuse-ld=lld"
build --linkopt="-lm"

# Disable clang extention that rejects type definitions within offsetof.
# This was added in clang-16 by https://reviews.llvm.org/D133574.
# Can be removed once upb is updated, since a type definition is used within
# offset of in the current version of ubp.
# See https://github.com/protocolbuffers/upb/blob/9effcbcb27f0a665f9f345030188c0b291e32482/upb/upb.c#L183.
build --copt=-Wno-gnu-offsetof-extensions

# Store performance profiling log in the mounted artifact directory.
# The profile can be viewed by visiting chrome://tracing in a Chrome browser.
# See https://docs.bazel.build/versions/main/skylark/performance.html#performance-profiling
build --profile=build/profile.json.gz

# Use the NVCC toolchain to compile for manylinux2014
build --crosstool_top="@sigbuild-r2.14-clang_config_cuda//crosstool:toolchain"

# Test-related settings below this point.
test --build_tests_only --keep_going --test_output=errors --verbose_failures=true
test --local_test_jobs=HOST_CPUS
test --test_env=LD_LIBRARY_PATH
# Give only the list of failed tests at the end of the log
test --test_summary=short

# "nonpip" tests are regular py_test tests.
# Pass --config=nonpip to run the same suite of tests. If you want to run just
# one test for investigation, you don't need --config=nonpip; just run the
# bazel test invocation as normal.
test:nonpip_filters --test_tag_filters=-no_oss,-oss_excluded,-oss_serial,-gpu,-tpu,-benchmark-test,-v1only,-no_oss_py38,-no_oss_py39,-no_oss_py310
test:nonpip_filters --build_tag_filters=-no_oss,-oss_excluded,-oss_serial,-gpu,-tpu,-benchmark-test,-v1only,-no_oss_py38,-no_oss_py39,-no_oss_py310
test:nonpip_filters --test_lang_filters=py --test_size_filters=small,medium
test:nonpip --config=nonpip_filters -- //tensorflow/... -//tensorflow/python/integration_testing/... -//tensorflow/compiler/tf2tensorrt/... -//tensorflow/compiler/xrt/... -//tensorflow/core/tpu/... -//tensorflow/lite/... -//tensorflow/tools/toolchains/...

# For building libtensorflow archives
test:libtensorflow_test -- //tensorflow/tools/lib_package:libtensorflow_test //tensorflow/tools/lib_package:libtensorflow_java_test
build:libtensorflow_build -- //tensorflow/tools/lib_package:libtensorflow.tar.gz //tensorflow/tools/lib_package:libtensorflow_jni.tar.gz //tensorflow/java:libtensorflow.jar //tensorflow/java:libtensorflow-src.jar //tensorflow/tools/lib_package:libtensorflow_proto.zip

# For outputting Build Event Protocol files
build:build_event_export --build_event_json_file=build/bep.json

# For Remote Build Execution.
build:rbe --google_default_credentials
build:rbe --bes_backend=buildeventservice.googleapis.com
build:rbe --bes_results_url="https://source.cloud.google.com/results/invocations"
build:rbe --bes_timeout=600s
build:rbe --define=EXECUTOR=remote
build:rbe --jobs=800
build:rbe --remote_executor=grpcs://remotebuildexecution.googleapis.com
build:rbe --remote_timeout=3600
build:rbe --spawn_strategy=remote,worker,standalone,local
build:rbe --remote_download_toplevel
build:rbe --action_env=PATH="/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/local/go/bin"
build:rbe --linkopt=-lrt --host_linkopt=-lrt --linkopt=-lm --host_linkopt=-lm # Unclear why this is here
build:rbe --host_crosstool_top="@sigbuild-r2.14-clang_config_cuda//crosstool:toolchain"
build:rbe --crosstool_top="@sigbuild-r2.14-clang_config_cuda//crosstool:toolchain"
build:rbe --extra_toolchains="@sigbuild-r2.14-clang_config_cuda//crosstool:toolchain-linux-x86_64"
build:rbe --extra_execution_platforms="@sigbuild-r2.14-clang_config_platform//:platform"
build:rbe --host_platform="@sigbuild-r2.14-clang_config_platform//:platform"
build:rbe --platforms="@sigbuild-r2.14-clang_config_platform//:platform"
# Python config is the same across all containers because the binary is the same
build:rbe --repo_env=TF_PYTHON_CONFIG_REPO="@sigbuild-r2.14-clang_config_python"
build:rbe --remote_instance_name=projects/tensorflow-testing/instances/default_instance
build:rbe --project_id="tensorflow-testing"

# For continuous builds
test:pycpp_filters --test_tag_filters=-no_oss,-oss_excluded,-oss_serial,-gpu,-tpu,-benchmark-test,-v1only
test:pycpp_filters --build_tag_filters=-no_oss,-oss_excluded,-oss_serial,-gpu,-tpu,-benchmark-test,-v1only
test:pycpp_filters --test_lang_filters=cc,py --test_size_filters=small,medium
test:pycpp --config=pycpp_filters -- //tensorflow/... -//tensorflow/python/integration_testing/... -//tensorflow/compiler/tf2tensorrt/... -//tensorflow/compiler/xrt/... -//tensorflow/core/tpu/... -//tensorflow/lite/... -//tensorflow/tools/toolchains/...
99 changes: 99 additions & 0 deletions ci/official/bazelrcs/cpu_gcc.bazelrc
Original file line number Diff line number Diff line change
@@ -0,0 +1,99 @@
# Copyright 2023 The TensorFlow Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# ==============================================================================
# This bazelrc can build a CPU-supporting TF package.

# Convenient cache configurations
# Use a cache directory mounted to /tf/cache. Very useful!
build:sigbuild_local_cache --disk_cache=/tf/cache
# Use the public-access TF DevInfra cache (read only)
build:sigbuild_remote_cache --remote_cache="https://storage.googleapis.com/tensorflow-devinfra-bazel-cache/september2022" --remote_upload_local_results=false
# Write to the TF DevInfra cache (only works for internal TF CI)
build:sigbuild_remote_cache_push --remote_cache="https://storage.googleapis.com/tensorflow-devinfra-bazel-cache/september2022" --google_default_credentials
# Change the value of CACHEBUSTER when upgrading the toolchain, or when testing
# different compilation methods. E.g. for a PR to test a new CUDA version, set
# the CACHEBUSTER to the PR number.
build --action_env=CACHEBUSTER=501872366

# Use Python 3.X as installed in container image
build --action_env PYTHON_BIN_PATH="/usr/bin/python3"
build --action_env PYTHON_LIB_PATH="/usr/lib/tf_python"
build --python_path="/usr/bin/python3"

# Build TensorFlow v2
build --define=tf_api_version=2 --action_env=TF2_BEHAVIOR=1

# Target the AVX instruction set
build --copt=-mavx --host_copt=-mavx

# Store performance profiling log in the mounted artifact directory.
# The profile can be viewed by visiting chrome://tracing in a Chrome browser.
# See https://docs.bazel.build/versions/main/skylark/performance.html#performance-profiling
build --profile=build/profile.json.gz

# Use the NVCC toolchain to compile for manylinux2014
build --crosstool_top="@sigbuild-r2.14_config_cuda//crosstool:toolchain"

# Test-related settings below this point.
test --build_tests_only --keep_going --test_output=errors --verbose_failures=true
test --local_test_jobs=HOST_CPUS
test --test_env=LD_LIBRARY_PATH
# Give only the list of failed tests at the end of the log
test --test_summary=short

# "nonpip" tests are regular py_test tests.
# Pass --config=nonpip to run the same suite of tests. If you want to run just
# one test for investigation, you don't need --config=nonpip; just run the
# bazel test invocation as normal.
test:nonpip_filters --test_tag_filters=-no_oss,-oss_excluded,-oss_serial,-gpu,-tpu,-benchmark-test,-v1only,-no_oss_py38,-no_oss_py39,-no_oss_py310
test:nonpip_filters --build_tag_filters=-no_oss,-oss_excluded,-oss_serial,-gpu,-tpu,-benchmark-test,-v1only,-no_oss_py38,-no_oss_py39,-no_oss_py310
test:nonpip_filters --test_lang_filters=py --test_size_filters=small,medium
test:nonpip --config=nonpip_filters -- //tensorflow/... -//tensorflow/python/integration_testing/... -//tensorflow/compiler/tf2tensorrt/... -//tensorflow/compiler/xrt/... -//tensorflow/core/tpu/... -//tensorflow/lite/... -//tensorflow/tools/toolchains/...

# For building libtensorflow archives
test:libtensorflow_test -- //tensorflow/tools/lib_package:libtensorflow_test //tensorflow/tools/lib_package:libtensorflow_java_test
build:libtensorflow_build -- //tensorflow/tools/lib_package:libtensorflow.tar.gz //tensorflow/tools/lib_package:libtensorflow_jni.tar.gz //tensorflow/java:libtensorflow.jar //tensorflow/java:libtensorflow-src.jar //tensorflow/tools/lib_package:libtensorflow_proto.zip

# For outputting Build Event Protocol files
build:build_event_export --build_event_json_file=build/bep.json

# For Remote Build Execution.
build:rbe --google_default_credentials
build:rbe --bes_backend=buildeventservice.googleapis.com
build:rbe --bes_results_url="https://source.cloud.google.com/results/invocations"
build:rbe --bes_timeout=600s
build:rbe --define=EXECUTOR=remote
build:rbe --jobs=800
build:rbe --remote_executor=grpcs://remotebuildexecution.googleapis.com
build:rbe --remote_timeout=3600
build:rbe --spawn_strategy=remote,worker,standalone,local
build:rbe --remote_download_toplevel
build:rbe --action_env=PATH="/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/local/go/bin"
build:rbe --linkopt=-lrt --host_linkopt=-lrt --linkopt=-lm --host_linkopt=-lm # Unclear why this is here
build:rbe --host_crosstool_top="@sigbuild-r2.14_config_cuda//crosstool:toolchain"
build:rbe --crosstool_top="@sigbuild-r2.14_config_cuda//crosstool:toolchain"
build:rbe --extra_toolchains="@sigbuild-r2.14_config_cuda//crosstool:toolchain-linux-x86_64"
build:rbe --extra_execution_platforms="@sigbuild-r2.14_config_platform//:platform"
build:rbe --host_platform="@sigbuild-r2.14_config_platform//:platform"
build:rbe --platforms="@sigbuild-r2.14_config_platform//:platform"
# Python config is the same across all containers because the binary is the same
build:rbe --repo_env=TF_PYTHON_CONFIG_REPO="@sigbuild-r2.14_config_python"
build:rbe --remote_instance_name=projects/tensorflow-testing/instances/default_instance
build:rbe --project_id="tensorflow-testing"

# For continuous builds
test:pycpp_filters --test_tag_filters=-no_oss,-oss_excluded,-oss_serial,-gpu,-tpu,-benchmark-test,-v1only
test:pycpp_filters --build_tag_filters=-no_oss,-oss_excluded,-oss_serial,-gpu,-tpu,-benchmark-test,-v1only
test:pycpp_filters --test_lang_filters=cc,py --test_size_filters=small,medium
test:pycpp --config=pycpp_filters -- //tensorflow/... -//tensorflow/python/integration_testing/... -//tensorflow/compiler/tf2tensorrt/... -//tensorflow/compiler/xrt/... -//tensorflow/core/tpu/... -//tensorflow/lite/... -//tensorflow/tools/toolchains/...
Loading