ref: 8af5c6b4a13cb66e0f3dcd465c246d2d2e4128c7
parent: b6095cf22d501cb1950685e46b334b0a2ca7e78b
author: Jan Buethe <jbuethe@amazon.de>
date: Tue Nov 7 06:54:22 EST 2023
added transposed 1d convolutions to wexchange
--- a/dnn/torch/weight-exchange/wexchange/c_export/__init__.py
+++ b/dnn/torch/weight-exchange/wexchange/c_export/__init__.py
@@ -28,4 +28,4 @@
*/
"""
-from .common import print_gru_layer, print_dense_layer, print_conv1d_layer, print_conv2d_layer, print_vector
\ No newline at end of file
+from .common import print_gru_layer, print_dense_layer, print_conv1d_layer, print_tconv1d_layer, print_conv2d_layer, print_vector
\ No newline at end of file
--- a/dnn/torch/weight-exchange/wexchange/c_export/common.py
+++ b/dnn/torch/weight-exchange/wexchange/c_export/common.py
@@ -361,3 +361,25 @@
writer.header.write(f"\n#define {name.upper()}_STATE_SIZE {N}\n")return N
+
+
+def print_tconv1d_layer(writer : CWriter,
+ name : str,
+ weight : np.ndarray,
+ bias : np.ndarray,
+ stride: int,
+ scale=1/128,
+ quantize=False):
+
+ in_channels, out_channels, kernel_size = weight.shape
+
+
+ linear_weight = weight.transpose(2, 1, 0).reshape(kernel_size * out_channels, in_channels).transpose(1, 0)
+ linear_bias = np.repeat(bias[np.newaxis, :], kernel_size, 0).flatten()
+
+ print_linear_layer(writer, name, linear_weight, linear_bias, scale=scale, quantize=quantize)
+
+ writer.header.write(f"\n#define {name.upper()}_KERNEL_SIZE {kernel_size}\n")+ writer.header.write(f"\n#define {name.upper()}_STRIDE {stride}\n")+ writer.header.write(f"\n#define {name.upper()}_IN_CHANNELS {in_channels}\n")+ writer.header.write(f"\n#define {name.upper()}_OUT_CHANNELS {out_channels}\n")\ No newline at end of file
--- a/dnn/torch/weight-exchange/wexchange/torch/torch.py
+++ b/dnn/torch/weight-exchange/wexchange/torch/torch.py
@@ -32,7 +32,7 @@
import torch
import numpy as np
-from wexchange.c_export import CWriter, print_gru_layer, print_dense_layer, print_conv1d_layer, print_conv2d_layer
+from wexchange.c_export import CWriter, print_gru_layer, print_dense_layer, print_conv1d_layer, print_tconv1d_layer, print_conv2d_layer
def dump_torch_gru_weights(where, gru, name='gru', input_sparse=False, recurrent_sparse=False, quantize=False, scale=1/128, recurrent_scale=1/128):
@@ -162,6 +162,36 @@
conv.bias.set_(torch.from_numpy(b))
+def dump_torch_tconv1d_weights(where, conv, name='conv', scale=1/128, quantize=False):
+
+ w = conv.weight.detach().cpu().numpy().copy()
+ if conv.bias is None:
+ b = np.zeros(conv.out_channels, dtype=w.dtype)
+ else:
+ b = conv.bias.detach().cpu().numpy().copy()
+
+ if isinstance(where, CWriter):
+
+ return print_tconv1d_layer(where, name, w, b, conv.stride[0], scale=scale, quantize=quantize)
+ else:
+ os.makedirs(where, exist_ok=True)
+
+ np.save(os.path.join(where, 'weight_oik.npy'), w)
+
+ np.save(os.path.join(where, 'bias.npy'), b)
+
+
+def load_torch_tconv1d_weights(where, conv):
+
+ with torch.no_grad():
+ w = np.load(os.path.join(where, 'weight_oik.npy'))
+ conv.weight.set_(torch.from_numpy(w))
+ if type(conv.bias) != type(None):
+ b = np.load(os.path.join(where, 'bias.npy'))
+ if conv.bias is not None:
+ conv.bias.set_(torch.from_numpy(b))
+
+
def dump_torch_conv2d_weights(where, conv, name='conv', scale=1/128, quantize=False):
w = conv.weight.detach().cpu().permute(0, 1, 3, 2).numpy().copy()
if conv.bias is None:
@@ -228,6 +258,8 @@
return dump_torch_conv2d_weights(where, module, name, **kwargs)
elif isinstance(module, torch.nn.Embedding):
return dump_torch_embedding_weights(where, module)
+ elif isinstance(module, torch.nn.ConvTranspose1d):
+ return dump_torch_tconv1d_weights(where, module, name, **kwargs)
else:
raise ValueError(f'dump_torch_weights: layer of type {type(module)} not supported')@@ -243,5 +275,7 @@
load_torch_conv2d_weights(where, module)
elif isinstance(module, torch.nn.Embedding):
load_torch_embedding_weights(where, module)
+ elif isinstance(module, torch.nn.ConvTranspose1d):
+ return load_torch_tconv1d_weights(where, module)
else:
- raise ValueError(f'dump_torch_weights: layer of type {type(module)} not supported')+ raise ValueError(f'load_torch_weights: layer of type {type(module)} not supported')--
⑨