Skip to content

add extra_options use_channel_wised_quantization to builder.py #1362

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

Open
wants to merge 2 commits into
base: main
Choose a base branch
from
Open

add extra_options use_channel_wised_quantization to builder.py #1362

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
531a712
add extra_options use_channel_wised_quantization to builder.py, quant…
bopeng1234 Mar 31, 2025
260e02f
add int_ prefix to use_channel_wised_quantization and fix typo
bopeng1234 Jul 29, 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
6 changes: 5 additions & 1 deletion src/python/py/models/builder.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -300,6 +300,7 @@ def __init__(
"op_types_to_quantize": extra_options.get("int4_op_types_to_quantize", ("MatMul", )),
"nodes_to_exclude": extra_options.get("int4_nodes_to_exclude", []),
"algo_config": int4_algo_config,
"use_channel_wised_quantization": extra_options.get("int4_use_channel_wised_quantization", False),
},
"use_qdq": extra_options.get("use_qdq", False),
}
Expand Down Expand Up @@ -482,6 +483,7 @@ def to_int4(self) -> ir.Model:
quant_format=QuantFormat.QDQ if self.quant_attrs["use_qdq"] else QuantFormat.QOperator,
op_types_to_quantize=self.quant_attrs["int4"]["op_types_to_quantize"],
algo_config=self.quant_attrs["int4"]["algo_config"],
channel_wised_quantize = self.quant_attrs["int4"]["use_channel_wised_quantization"],
)
quant.process()
return ir.from_proto(quant.model.model)
Expand Down Expand Up @@ -3637,7 +3639,7 @@ def check_extra_options(kv_pairs):
"""
Check key-value pairs and set values correctly
"""
bools = ["int4_is_symmetric", "exclude_embeds", "exclude_lm_head", "include_hidden_states", "enable_cuda_graph", "use_8bits_moe", "use_qdq", "use_webgpu_fp32"]
bools = ["int4_is_symmetric", "int4_use_channel_wised_quantization", "exclude_embeds", "exclude_lm_head", "include_hidden_states", "enable_cuda_graph", "use_8bits_moe", "use_qdq", "use_webgpu_fp32"]
for key in bools:
if key in kv_pairs:
if kv_pairs[key] in {"false", "False", "0"}:
Expand Down Expand Up @@ -3916,6 +3918,8 @@ def get_args():
Currently supported options are: 'default', 'rtn', 'k_quant_mixed', 'k_quant_last'.
k_quant_mixed = k_quant algorithm with mixed precision (int4 + int8).
k_quant_last = k_quant algorithm where only the last MatMul (/lm_head/MatMul) is quantized as int8. Other MatMuls are quantized as int4.
int4_use_channel_wised_quantization = Use channel wised quantization, in which block size = rows (K)
Use this option when you want use K as block size. Default is False
num_hidden_layers = Manually specify the number of layers in your ONNX model (for unit testing purposes).
filename = Filename for ONNX model (default is 'model.onnx').
For models with multiple components, each component is exported to its own ONNX model.
Expand Down
Loading