Skip to content

support sharding in fp16 on xpu, #48897

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.

Sign up for GitHub

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

Merged
merged 3 commits into from
Dec 12, 2022
Merged

support sharding in fp16 on xpu, #48897

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
16b2257
support sharding in fp16 on xpu, change reduce_max to reduce_sum for …
sljlp Dec 8, 2022
9e6ef4b
Merge remote-tracking branch 'origin/develop' into sharding_xpu_fp16
sljlp Dec 9, 2022
0737c55
update
sljlp Dec 9, 2022
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
6 changes: 4 additions & 2 deletions python/paddle/distributed/fleet/meta_parallel/sharding/group_sharded_utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -220,7 +220,8 @@ def unscale_method(self, optimizer):
temp_found_inf_fp16 = to_variable(np.array([0]).astype(np.bool_))
temp_found_inf_fp32 = to_variable(np.array([0]).astype(np.bool_))

device = "cpu" if optimizer.offload else "gpu"
device = paddle.get_device().split(":")[0]
device = "cpu" if optimizer.offload else device
dev_id = (
0 if device == "cpu" else int(paddle.get_device().split(":")[1])
)
Expand All @@ -245,8 +246,9 @@ def unscale_method(self, optimizer):
is_found_inf = paddle.to_tensor([self._found_inf], dtype="int32")

paddle.distributed.all_reduce(
is_found_inf, op=paddle.distributed.ReduceOp.MAX, group=None
is_found_inf, op=paddle.distributed.ReduceOp.SUM, group=None
)

self._found_inf = is_found_inf.numpy()[0]

scaler._unscale = MethodType(unscale_method, scaler)
Expand Down
2 changes: 1 addition & 1 deletion python/paddle/fluid/tests/unittests/collective/fleet/test_imperative_auto_mixed_precision.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -344,7 +344,7 @@ def nan_inf(self):
scaled_loss = scaler.scale(loss)
scaled_loss.backward()
optimize_ops, params_grads = scaler.minimize(optimizer, scaled_loss)
self.assertEqual(scaler._found_inf.numpy() == 1, True)
self.assertEqual(scaler._found_inf.numpy() >= 1, True)

for param in model.parameters():
# param not update when tensor contains nan or inf
Expand Down
2 changes: 1 addition & 1 deletion .../fluid/tests/unittests/collective/fleet/test_imperative_auto_mixed_precision_for_eager.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -343,7 +343,7 @@ def nan_inf(self):
scaled_loss = scaler.scale(loss)
scaled_loss.backward()
optimize_ops, params_grads = scaler.minimize(optimizer, scaled_loss)
self.assertEqual(scaler._found_inf.numpy() == 1, True)
self.assertEqual(scaler._found_inf.numpy() >= 1, True)

for param in model.parameters():
# param not update when tensor contains nan or inf
Expand Down