-dtype_to_name(dtype_mapping, dtype)
|
+
simple_progress_bar(total, i)
|
+Progress bar for cases where tqdm can't be used. |
+
+dtype_to_name(dtype_mapping, dtype)
|
Map data type and its string representation. |
-make_quant_node(name, inputs, outputs)
|
+
make_quant_node(name, inputs, outputs)
|
Make a QuantizeLinear node. |
-make_dquant_node(name, inputs, outputs[, axis])
|
+
make_dquant_node(name, inputs, outputs[, axis])
|
Make a DequantizeLinear node. |
-is_B_transposed(node)
|
+
is_B_transposed(node)
|
Whether inuput B is transposed. |
-split_shared_bias(model)
|
+
split_shared_bias(model)
|
Split shared tensor. |
-float_to_float16(tensor)
|
+
float_to_float16(tensor)
|
Convert float to float16. |
-float_to_bfloat16(tensor)
|
+
float_to_bfloat16(tensor)
|
Convert float to bfloat16. |
-cast_tensor(tensor, dtype[, is_large_model])
|
+
cast_tensor(tensor, dtype[, is_large_model])
|
Convert tensor float to target dtype. |
-remove_init_from_model_input(model)
|
+
remove_init_from_model_input(model)
|
Remove initializer from model input. |
-collate_preds(results)
|
+
collate_preds(results)
|
Collect model outputs. |
-quantize_data_with_scale_zero(data, qType, scheme, ...)
|
+
quantize_data_with_scale_zero(data, qType, scheme, ...)
|
Quantize data with scale and zero point. |
-calculate_scale_zp(rmin, rmax, quantize_range, qType, ...)
|
+
calculate_scale_zp(rmin, rmax, quantize_range, qType, ...)
|
Calculate scale and zero point. |
-quantize_data(data, quantize_range, qType, scheme)
|
+
quantize_data(data, quantize_range, qType, scheme)
|
Quantize data. |
-quantize_data_per_channel(data, axis, quantize_range, ...)
|
+
quantize_data_per_channel(data, axis, quantize_range, ...)
|
Quantize tensor per-channel. |
-dequantize_data_with_scale_zero(tensor_value, ...)
|
+
dequantize_data_with_scale_zero(tensor_value, ...)
|
Dequantize tensor with scale and zero point. |
-dequantize_data(tensor_value, scale_value, zo_value[, ...])
|
+
dequantize_data(tensor_value, scale_value, zo_value[, ...])
|
Dequantize tensor. |
-quantize_nparray(qtype, arr, scale, zero_point[, low, ...])
|
+
quantize_nparray(qtype, arr, scale, zero_point[, low, ...])
|
Quantize numpy array. |
-attribute_to_kwarg(attribute)
|
+
attribute_to_kwarg(attribute)
|
Convert attribute to kwarg format for use with onnx.helper.make_node. |
-find_by_name(name, item_list)
|
+
find_by_name(name, item_list)
|
Helper function to find item by name in a list. |
-trt_env_setup(model)
|
+
trt_env_setup(model)
|
Set environment variable for Tensorrt Execution Provider. |
-to_numpy(data)
|
+
to_numpy(data)
|
Convert to numpy ndarrays. |
-infer_shapes(in_mp[, int_max, auto_merge, ...])
|
+
infer_shapes(in_mp[, int_max, auto_merge, ...])
|
Symbolic shape inference. |