AMCT ConvTranspose2dQAT API

ConvTranspose2dQAT

【免费下载链接】amctAMCT是CANN提供的昇腾AI处理器亲和的模型压缩工具仓。项目地址: https://gitcode.com/cann/amct

产品支持情况

功能说明

构造ConvTranspose2d的QAT算子。

函数原型

• 直接构造接口：

  qat = amct_pytorch.nn.module.quantization.conv_transpose_2d.ConvTranspose2dQAT(in_channels, out_channels, kernel_size, stride, padding, dilation, groups, bias, padding_mode, device, dtype, config)

• 基于原生算子构造接口：

  qat = amct_pytorch.nn.module.quantization.conv_transpose_2d.ConvTranspose2dQAT.from_float(mod, config)

参数说明

表 1直接构造接口参数说明

config = { "retrain_enable":true, "retrain_data_config": { "dst_type": "INT8", "batch_num": 10, "fixed_min": False, "clip_min": -1.0, "clip_max": 1.0 }, "retrain_weight_config": { "dst_type": "INT8", "weights_retrain_algo": "arq_retrain", "channel_wise": False } }

表 2基于原生算子构造接口

config = { "retrain_enable":true, "retrain_data_config": { "dst_type": "INT8", "batch_num": 10, "fixed_min": False, "clip_min": -1.0, "clip_max": 1.0 }, "retrain_weight_config": { "dst_type": "INT8", "weights_retrain_algo": "arq_retrain", "channel_wise": False } }

返回值说明

• 直接构造：返回构造的QAT单算子实例。
• 基于原生算子构造：torch.nn.Module转化后的QAT单算子。

调用示例

• 直接构造：

  from amct_pytorch.nn.module.quantization.conv_transpose_2d import ConvTranspose2dQAT ConvTranspose2dQAT(in_channels=1, out_channels=1, kernel_size=1, stride=1, padding=0, dilation=1, groups=1, bias=True, padding_mode='zeros', device=None, dtype=None, config=None)

• 基于原生算子构造：

  import torch from amct_pytorch.nn.module.quantization.conv_transpose_2d import ConvTranspose2dQAT conv_transpose2d_op = torch.nn.ConvTranspose2d(in_channels=1, out_channels=1, kernel_size=1, stride=1, padding=0, dilation=1, groups=1, bias=True, padding_mode='zeros', device=None, dtype=None) ConvTranspose2dQAT.from_float(mod=conv_transpose2d_op, config=None)

【免费下载链接】amctAMCT是CANN提供的昇腾AI处理器亲和的模型压缩工具仓。项目地址: https://gitcode.com/cann/amct