Skip to content

[shardformer]fix, test gpt2 for AMP+TP #4403

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
ver217 merged 2 commits into from
Aug 11, 2023
Merged

[shardformer]fix, test gpt2 for AMP+TP #4403

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
b2766ca
[shardformer] gpt2 tests fix
flybird11111 Aug 10, 2023
00a3634
[shardformer] gpt2 tests fix
flybird11111 Aug 10, 2023
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
8 changes: 3 additions & 5 deletions tests/test_shardformer/test_model/_utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -206,7 +206,7 @@ def check_weight(org_model: Module,

if is_distributed_tensor(sharded_weight) or is_customized_distributed_tensor(sharded_weight):
sharded_weight_list = [
torch.zeros([*sharded_weight.shape]).to('cuda') for _ in range(dist.get_world_size(tp_group))
torch.zeros_like(sharded_weight).to('cuda') for _ in range(dist.get_world_size(tp_group))
]
dist.all_gather(sharded_weight_list, sharded_weight, tp_group)
sharded_weight = torch.cat(sharded_weight_list, dim=dim)
Expand All @@ -215,7 +215,7 @@ def check_weight(org_model: Module,
print(f"'{suffix}' weight: {org_weight}, {sharded_weight}")

assert torch.allclose(org_weight.float(), sharded_weight.float(), atol=atol, rtol=rtol), \
f"shard model weight is not equal to origin model weight\n{org_weight}\n{sharded_weight}"
f"shard model weight {suffix} is not equal to origin model weight\n{org_weight}\n{sharded_weight}"


def check_grad(org_model: Module,
Expand All @@ -233,9 +233,7 @@ def check_grad(org_model: Module,
shard_weight = getattr_(sharded_model, suffix).weight

if is_distributed_tensor(shard_weight) or is_customized_distributed_tensor(shard_weight):
shard_grad_list = [
torch.zeros([*shard_grad.shape]).to('cuda') for _ in range(dist.get_world_size(tp_group))
]
shard_grad_list = [torch.zeros_like(shard_grad).to('cuda') for _ in range(dist.get_world_size(tp_group))]
dist.all_gather(shard_grad_list, shard_grad, tp_group)
shard_grad = torch.cat(shard_grad_list, dim=dim)

Expand Down
8 changes: 3 additions & 5 deletions tests/test_shardformer/test_model/test_shard_gpt2.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -23,7 +23,6 @@ def check_forward_backward(model_fn, data_gen_fn, output_transform_fn, loss_fn,
org_model, org_optimizer, sharded_model, sharded_optimizer, criterion, booster = \
build_model_from_hybrid_plugin(model_fn, loss_fn, test_config)


org_loss, org_output, sharded_loss, sharded_output = \
run_forward_backward_with_hybrid_plugin(
org_model,
Expand All @@ -47,7 +46,6 @@ def check_forward_backward(model_fn, data_gen_fn, output_transform_fn, loss_fn,
if org_model.__class__.__name__ == 'GPT2Model':
check_output_hidden_state(org_output, sharded_output, stage_manager, atol=atol, rtol=rtol)

# check loss
check_loss(org_loss, sharded_loss, atol=atol, rtol=rtol)

def unwrap(module):
Expand Down Expand Up @@ -92,13 +90,14 @@ def unwrap(module):
'num_microbatches': 4,
'enable_all_optimization': True,
'use_lazy_init': True,
'precision': 'fp32',
'precision': 'fp16',
'initial_scale': 1,
}, {
'tp_size': 1,
'pp_size': 2,
'num_microbatches': 4,
'enable_all_optimization': True,
'use_lazy_init': False,
'use_lazy_init': True,
'precision': 'fp16',
'initial_scale': 1,
}, {
Expand All @@ -112,7 +111,6 @@ def unwrap(module):
def run_gpt2_test(test_config):

# TODO: add test_config for TP+DP after supporting & debugging it
# TODO: check and debug TP+AMP

sub_model_zoo = model_zoo.get_sub_registry('transformers_gpt')

Expand Down