Skip to content
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

handle rando device for exported model in model builder #759

Merged
merged 2 commits into from
May 13, 2024
Merged

handle rando device for exported model in model builder #759

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
ea71e2b
handle rando device for exported model in model builder
May 12, 2024
9c51e12
typo
May 12, 2024
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: 13 additions & 7 deletions build/builder.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -19,7 +19,7 @@
from quantize import quantize_model

from build.model import Transformer
from build.utils import device_sync, name_to_dtype
from build.utils import device_sync, is_cpu_device, is_cuda_or_cpu_device, name_to_dtype


@dataclass
Expand Down Expand Up @@ -371,6 +371,12 @@ def _initialize_model(
_set_gguf_kwargs(builder_args, is_et=is_pte, context="generate")

if builder_args.dso_path:
if not is_cuda_or_cpu_device(builder_args.device):
print(
f"Cannot load specified DSO to {builder_args.device}. Attempting to load model to CPU instead"
)
builder_args.device = "cpu"

# assert (
# quantize is None or quantize == "{ }"
# ), "quantize not valid for exported DSO model. Specify quantization during export."
Expand All @@ -381,12 +387,6 @@ def _initialize_model(
print(f"Time to load model: {time.time() - t0:.02f} seconds")

try:
if "mps" in builder_args.device:
print(
"Cannot load specified DSO to MPS. Attempting to load model to CPU instead"
)
builder_args.device = "cpu"

# Replace model forward with the AOT-compiled forward
# This is a hacky way to quickly demo AOTI's capability.
# model is still a Python object, and any mutation to its
Expand All @@ -399,6 +399,12 @@ def _initialize_model(
except:
raise RuntimeError(f"Failed to load AOTI compiled {builder_args.dso_path}")
elif builder_args.pte_path:
if not is_cpu_device(builder_args.device):
print(
f"Cannot load specified PTE to {builder_args.device}. Attempting to load model to CPU instead"
)
builder_args.device = "cpu"

# assert (
# quantize is None or quantize == "{ }"
# ), "quantize not valid for exported PTE model. Specify quantization during export."
Expand Down
8 changes: 8 additions & 0 deletions build/utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -255,3 +255,11 @@ def get_device(device) -> str:
else "mps" if is_mps_available() else "cpu"
)
return torch.device(device)


def is_cuda_or_cpu_device(device) -> bool:
return device == "" or str(device) == "cpu" or ("cuda" in str(device))


def is_cpu_device(device) -> bool:
return device == "" or str(device) == "cpu"
Loading