Skip to content

[quant][pt2e] Relax constraints on dtype and qscheme to allow for customizations #116287

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

[quant][pt2e] Relax constraints on dtype and qscheme to allow for customizations #116287

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
f6e99d3
[quant][pt2e] Relax constraints on dtype and qscheme to allow for cus…
jerryzh168 Dec 21, 2023
6141132
Update on "[quant][pt2e] Relax constraints on dtype and qscheme to al…
jerryzh168 Dec 21, 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
25 changes: 0 additions & 25 deletions torch/ao/quantization/quantizer/quantizer.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -19,23 +19,6 @@
"QuantizationAnnotation",
]

# TODO: maybe remove torch.float32
SUPPORTED_DTYPES = [
torch.uint8,
torch.int8,
torch.int16,
torch.int32,
torch.float16,
torch.float32,
]
SUPPORTED_QSCHEMES = [
torch.per_tensor_affine,
torch.per_tensor_symmetric,
torch.per_channel_affine,
torch.per_channel_symmetric,
torch.per_channel_affine_float_qparams,
]


class QuantizationSpecBase(ABC): # noqa: B024
"""Base class for different types of quantization specs that allows users to
Expand Down Expand Up @@ -64,10 +47,6 @@ class QuantizationSpec(QuantizationSpecBase):
is_dynamic: bool = False

def __post_init__(self):
# check dtype is one of the supported types
if self.dtype not in SUPPORTED_DTYPES:
raise TypeError(f"Unsupported dtype {self.dtype}.")

# quant_min must be less than quant_max
if (
self.quant_min is not None
Expand All @@ -78,10 +57,6 @@ def __post_init__(self):
f"quant_min {self.quant_min} must be <= quant_max {self.quant_max}."
)

# check qscheme is on of the supported ones
if self.qscheme is not None and self.qscheme not in SUPPORTED_QSCHEMES:
raise ValueError(f"Unsupported qscheme {self.qscheme}.")

# ch_axis must be less than the number of channels
# but no way to check here. Just check that it is not < 0.
if self.ch_axis is not None and self.ch_axis < 0:
Expand Down