Jack cao g/xrt cherrypick 0721 #5334
JackCaoG
merged
20 commits into
xrt
from
JackCaoG/xrt_cherrypick_0721
Update inline style code to multiline (#5291)
db7c973d
Fix typo in _test.yml (#5172)
e8e66d40
[SPMD][Virtual Device]All tensors should be in SPMD:0 C++ device (#5284)
848c00d7
Revert pr https://github.com/pytorch/xla/pull/2682 (#5215)
07d6f7f5
Make README more actionable (#5262)
9a0e24de
[SPMD] Use xs.Mesh in test_2d_tensor_3d_mesh (#5295)
cc4f3043
[SPMD] Add FSDP sharding for test_train_spmd_linear_model.py (#5299)
af0e0c39
[SPMD] Avoid recompilations in xs.mark_sharding() (#5300)
a27382a3
[SPMD] Support mark_sharding on IRs (#5301)
ff974276
[SPMD] Allow dumping post optimizations hlo (#5302)
21784cef
Add `_sharded_cpu_state_dict` for distributed checkpointing (#5288)
67ab9750
Supoort unordered sharding spec correctly (#5305)
080fdcf6
Support unordered sharding spec for partial replication (#5316)
aac03da0
Fix mismatched GPU docker image in the doc. (#5319)
def08b43
quick refactor on _get_group_assignment (#5318)
37b85187
Add tf independent serialization (#5308)
d9c92bd3
Disable coverage for now (#5321)
ebb51208
Enable Some input output aliasing under SPMD (#5320)
9a995403
Use `_sharded_cpu_state_dict` functionality to Write Items for SPMD S…
8c20cbd6
handle single tensor for method send_to_device_single (#5317)
d507067a
will-cromar
approved these changes
on 2023-07-24
Login to write a write a comment.
Login via GitHub