pytorch/xla

Pull Requests Commits

ci: Migrate off of legacy GPU executor (#3278)

seemethere committed 3 years ago

893aefb1

Increase syncfree optimizer test numerical tolerance (#3326)

ymwangg committed 3 years ago

Verified 769ce2cb

Output the same dtype for aten::div.Scalar as input if input is float (#3320)

byronyi committed 3 years ago

Verified a41a04bd

Migrate to upstream tensor_util classes and functions (#3317)

wonjoo-wj committed 3 years ago

Verified f53d7431

disable test (#3319)

JackCaoG committed 3 years ago

Verified 343b475f

Add syncfree Adam and AdamW optimizers (#3294)

ymwangg committed 3 years ago

Verified 138a70f4

Basic lazy tensor notebooks for concepts (#3308)

ultrons committed 3 years ago

Verified e6439f40

Revert "Pull latest changes for TensorFlow (#3298)" (#3311)

JackCaoG committed 3 years ago

Verified ff079bb4

Pull latest changes for TensorFlow (#3298)

wonjoo-wj committed 3 years ago

Verified caaa6d8a

Circleci test using new base image (#3282)

yeounoh committed 3 years ago

Verified 91826f6a

Implement all_gather with native XLA primitive. (#3275)

hjm-aws committed 3 years ago

Verified 3413bd18

Install numpy before running test (#3300)

JackCaoG committed 3 years ago

Verified 9bac3e87

Revert "Update numpy version (#3299)" (#3301)

JackCaoG committed 3 years ago

Verified 803edefa

Update numpy version (#3299)

JackCaoG committed 3 years ago

Verified 47dc3689

Add kd_div_backward lowering back (#3261)

JackCaoG committed 3 years ago

Verified cbd77147

Update disabled test name to test_CTCLoss_no_batch_dim_xla (#3288)

wonjoo-wj committed 3 years ago

Verified 7ff6a2c7

Disable test_CTCLoss_no_batch_dim_reduction_mean_use_module_form tests (#3287)

wonjoo-wj committed 3 years ago

Verified 4aad66b2

Disable test_CTCLoss_no_batch_dim_reduction_sum_use_module_form tests (#3285)

wonjoo-wj committed 3 years ago

Verified e01996e6

Disable XLA tensors do not have storage (#3281)

JackCaoG committed 3 years ago

Verified ddbc3330

Downgrade to cuda10 and update softplus_backward op parameter (#3280)

wonjoo-wj committed 3 years ago

Verified 1f9e8d7c

.circleci: Only unlink if file exists (#3269)

seemethere committed 3 years ago

Verified 480acab3

Enabling alpha scaling for index_add op lowering (#3227)

yeounoh committed 4 years ago

Verified b5a5b39a

Modify built-in var computation to split mean and squared mean computations (#3262)

yeounoh committed 4 years ago

Verified 4188b2dc

Update CircleCI to use cuda11 (#3255)

wonjoo-wj committed 4 years ago

Verified 91b5b555

Lower sort.stable (#3246)

yeounoh committed 4 years ago

Verified 09fc189d

Make circieCI stdout non nonblocking (#3260)

JackCaoG committed 4 years ago

Verified 220e879c

Pass exponent as constant instead of device_data (#3251)

JackCaoG committed 4 years ago

Verified ca49187e

Move softplus out of tensor_ops and make it an regular op (#3250)

wonjoo-wj committed 4 years ago

Verified 4b433df1

Index add new (#3209)

JackCaoG committed 4 years ago

Verified 230f9049

Update isnan tensor method to cast to xla::PrimitiveType::PRED (#3242)

wonjoo-wj committed 4 years ago

Verified c3421393

Older