Skip to content

[benchmark] add option to enable CompiledAutograd #1536

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

Closed
wants to merge 2 commits into from
Closed

[benchmark] add option to enable CompiledAutograd #1536

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
b772a75
add option to enable CompiledAutograd
crcrpar Dec 10, 2024
ed9c92e
Merge branch 'main' into crpa/litgptbench_compiledautograd
crcrpar Feb 6, 2025
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
20 changes: 12 additions & 8 deletions thunder/benchmarks/benchmark_litgpt.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -251,6 +251,7 @@ def __init__(
use_torchao_fp8_allgather: bool = False,
use_torchao_fp8_precompute_scale_for_fsdp: bool = False,
fp8_shard_intermediate_activation: bool = False,
enable_compiled_autograd: bool = False,
use_sdpa: bool = False,
):
seed = 1337
Expand Down Expand Up @@ -286,6 +287,7 @@ def __init__(
self.dump_thunder_traces = dump_thunder_traces
self.dump_memory_snapshot = dump_memory_snapshot
self.fp8_shard_intermediate_activation = fp8_shard_intermediate_activation
self.enable_compiled_autograd = enable_compiled_autograd

self.use_sdpa = use_sdpa

Expand Down Expand Up @@ -685,11 +687,12 @@ def train(self):
input_ids, targets = next(self.train_data_iter)
input_ids = input_ids.to(self.device)
targets = targets.to(self.device)
if self.use_te_fp8_autocast:
with te.fp8_autocast():
with torch._dynamo.utils.maybe_enable_compiled_autograd(self.enable_compiled_autograd):
if self.use_te_fp8_autocast:
with te.fp8_autocast():
logits = self.model(input_ids)
else:
logits = self.model(input_ids)
else:
logits = self.model(input_ids)
logits = logits.reshape(-1, logits.size(-1))
targets = targets.reshape(-1)
loss = (
Expand All @@ -701,11 +704,12 @@ def train(self):
input_ids, targets = next(self.train_data_iter)
input_ids = input_ids.to(self.device)
targets = targets.to(self.device)
if self.use_te_fp8_autocast:
with te.fp8_autocast():
with torch._dynamo.utils.maybe_enable_compiled_autograd(self.enable_compiled_autograd):
if self.use_te_fp8_autocast:
with te.fp8_autocast():
logits = self.model(input_ids)
else:
logits = self.model(input_ids)
else:
logits = self.model(input_ids)
# This information is accurate only in the case when torch.compile
# uses a single graph for the entire forward pass In the case of
# torch.compile using multiple graphs, the saved_tensors will be
Expand Down
Loading