Skip to content

replace FbgemmConfig with Int4WeightOnlyConfig #2727

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 1 commit into from
Closed

replace FbgemmConfig with Int4WeightOnlyConfig #2727

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
28 changes: 5 additions & 23 deletions torchao/prototype/awq/example.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -17,6 +17,7 @@
from torchao.quantization import (
quantize_,
)
from torchao.quantization.quant_api import Int4WeightOnlyConfig


# adapted from: https://github.com/mit-han-lab/llm-awq/blob/main/awq/entry.py#L255
Expand Down Expand Up @@ -223,18 +224,7 @@ def quantize_and_eval(
if quant.startswith("awq-int4wo"):
group_size = int(quant.split("-")[2])
print(f"running {quant} quantization with group size {group_size}")
# TODO: this is temporary, we'll be using Int4WeightOnlyConfig soon
from torchao.quantization import FbgemmConfig

# use_hqq = True
# base_config = Int4WeightOnlyConfig(group_size=group_size, use_hqq=use_hqq)
base_config = FbgemmConfig(
input_dtype=torch.bfloat16,
weight_dtype=torch.int4,
output_dtype=torch.bfloat16,
block_size=[1, group_size],
preshuffle=False,
)
base_config = Int4WeightOnlyConfig(group_size=group_size, use_hqq=True)
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

need to remove use_hqq and add version=2 for now, we still want to use the fbgemm kernel

base_config = Int4WeightOnlyConfig(group_size=group_size, version=2)

print(f"running {quant} prepare and calibrate")
t0 = time.time()
quant_config = AWQConfig(base_config, step="prepare")
Expand Down Expand Up @@ -267,17 +257,9 @@ def quantize_and_eval(
elif quant.startswith("int4wo"):
group_size = int(quant.split("-")[1])
print(f"running {quant} quantization with group size {group_size}")
# TODO: enable after refactor: https://github.com/pytorch/ao/pull/2474
# use_hqq = "hqq" in quant
# base_config = Int4WeightOnlyConfig(group_size=group_size, use_hqq=use_hqq)
int4_weight_only_config = FbgemmConfig(
input_dtype=torch.bfloat16,
weight_dtype=torch.int4,
output_dtype=torch.bfloat16,
block_size=[1, group_size],
preshuffle=False,
)
quantize_(model, int4_weight_only_config)
use_hqq = "hqq" in quant
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

same here

base_config = Int4WeightOnlyConfig(group_size=group_size, use_hqq=use_hqq)
quantize_(model, base_config)

if model_save_path is not None:
print(f"Saving model to {model_save_path}")
Expand Down