extend operation support for conv and gemm
add more tests in validation
This commit is contained in:
NiccoloN
@@ -0,0 +1,47 @@
|
||||
# Validation Operations
|
||||
|
||||
ONNX test models used by `validate.py` to verify the Raptor compiler + PIM simulator pipeline.
|
||||
|
||||
Generated tests can be regenerated with:
|
||||
```
|
||||
python3 validation/operations/gen_tests.py
|
||||
```
|
||||
|
||||
## Conv
|
||||
|
||||
| Test | Directory | Input | Output | Kernel | Stride | Padding | Bias | Notes |
|
||||
|------|-----------|-------|--------|--------|--------|---------|------|-------|
|
||||
| Simple | `conv/simple` | [1,3,3,3] | [1,1,2,2] | 2x2 | 1 | none | no | Basic conv, hand-crafted |
|
||||
| With constant | `conv/with_constant` | [1,3,3,3] | [1,1,3,3] | 2x2 | 1 | SAME_UPPER | yes | Hand-crafted, constant weight+bias |
|
||||
| Batch 2 | `conv/batch_2` | [2,3,3,3] | [2,1,3,3] | 2x2 | 1 | SAME_UPPER | yes | Batched input |
|
||||
| Kernel 3x3 | `conv/kernel_3x3` | [1,1,5,5] | [1,1,3,3] | 3x3 | 1 | none | no | Larger kernel |
|
||||
| Stride 2 | `conv/stride_2` | [1,1,6,6] | [1,1,2,2] | 3x3 | 2 | none | no | Strided convolution |
|
||||
| Multi channel | `conv/multi_channel` | [1,3,5,5] | [1,4,3,3] | 3x3 | 1 | none | no | 3 in channels, 4 out channels |
|
||||
| Pointwise 1x1 | `conv/pointwise_1x1` | [1,8,4,4] | [1,4,4,4] | 1x1 | 1 | none | no | Channel mixing |
|
||||
| SAME padding 3x3 | `conv/same_padding_3x3` | [1,1,5,5] | [1,1,5,5] | 3x3 | 1 | SAME_UPPER | no | Spatial dims preserved |
|
||||
| Explicit padding | `conv/explicit_padding` | [1,1,4,4] | [1,1,4,4] | 3x3 | 1 | [1,1,1,1] | no | Symmetric explicit pads |
|
||||
| With bias 3x3 | `conv/with_bias_3x3` | [1,3,5,5] | [1,2,3,3] | 3x3 | 1 | none | yes | Multi-channel with bias |
|
||||
| Large spatial | `conv/large_spatial` | [1,1,8,8] | [1,1,6,6] | 3x3 | 1 | none | no | Larger spatial input |
|
||||
|
||||
## Gemm
|
||||
|
||||
| Test | Directory | A (input) | W (weight) | Output | transB | alpha | beta | Bias | Notes |
|
||||
|------|-----------|-----------|------------|--------|--------|-------|------|------|-------|
|
||||
| Default | `gemm/` | [10,132] | [132,132] | [10,132] | no | 1 | 1 | no | Hand-crafted, square weights |
|
||||
| Non-square | `gemm/non_square` | [4,128] | [128,64] | [4,64] | no | 1 | 1 | no | K != N |
|
||||
| With bias | `gemm/with_bias` | [4,128] | [128,128] | [4,128] | no | 1 | 1 | [128] | Bias vector |
|
||||
| transB | `gemm/transB` | [4,128] | [64,128] | [4,64] | yes | 1 | 1 | no | Transposed weight |
|
||||
| Alpha/beta | `gemm/alpha_beta` | [4,64] | [64,64] | [4,64] | no | 0.5 | 0.25 | [64] | Scaled matmul + bias |
|
||||
| Small | `gemm/small` | [2,8] | [8,4] | [2,4] | no | 1 | 1 | no | Tiny matrices |
|
||||
| Large | `gemm/large` | [8,256] | [256,128] | [8,128] | no | 1 | 1 | no | Larger matrices |
|
||||
| transB + bias | `gemm/transB_with_bias` | [4,128] | [64,128] | [4,64] | yes | 1 | 1 | [64] | Combined |
|
||||
|
||||
## Gemv
|
||||
|
||||
| Test | Directory | Input | W (weight) | Output | Bias | Notes |
|
||||
|------|-----------|-------|------------|--------|------|-------|
|
||||
| Simple | `gemv/simple` | [1,132] | [132,132] | [1,132] | no | Single-sample matmul |
|
||||
| Constant | `gemv/constant` | _(none)_ | [132,132] | [1,132] | no | All inputs constant |
|
||||
| Homogeneous const | `gemv/with_homogeneous_constant` | [1,132] | [132,132] | [1,132] | [1,132] | Bias matches output shape |
|
||||
| Heterogeneous const | `gemv/with_heterogeneous_constant` | [1,132] | [132,132] | [1,132] | [1,132] | Different constant pattern |
|
||||
| Scalar const | `gemv/with_scalar_constant` | [1,132] | [132,132] | [1,132] | [1,1] | Scalar bias, broadcast |
|
||||
Binary file not shown.
Binary file not shown.
Binary file not shown.
Binary file not shown.
Binary file not shown.
Binary file not shown.
Binary file not shown.
Binary file not shown.
Binary file not shown.
Binary file not shown.
Binary file not shown.
Binary file not shown.
Binary file not shown.
Binary file not shown.
Binary file not shown.
Binary file not shown.
Binary file not shown.
@@ -0,0 +1,276 @@
|
||||
#!/usr/bin/env python3
|
||||
"""Generate ONNX test models for validating GEMM and Conv implementations."""
|
||||
|
||||
import numpy as np
|
||||
import onnx
|
||||
from onnx import helper, TensorProto, numpy_helper
|
||||
from pathlib import Path
|
||||
|
||||
OPERATIONS_DIR = Path(__file__).parent
|
||||
|
||||
|
||||
def save_model(model, directory, filename):
|
||||
"""Save an ONNX model, creating the directory if needed."""
|
||||
d = OPERATIONS_DIR / directory
|
||||
d.mkdir(parents=True, exist_ok=True)
|
||||
path = d / filename
|
||||
onnx.checker.check_model(model)
|
||||
onnx.save(model, str(path))
|
||||
print(f" {path.relative_to(OPERATIONS_DIR)}")
|
||||
|
||||
|
||||
# ---------------------------------------------------------------------------
|
||||
# GEMM tests
|
||||
# ---------------------------------------------------------------------------
|
||||
|
||||
def gemm_non_square():
|
||||
"""GEMM with non-square weight matrix: [B, K] @ [K, N], K != N."""
|
||||
B, K, N = 4, 128, 64
|
||||
W = numpy_helper.from_array(np.random.default_rng(42).uniform(-1, 1, (K, N)).astype(np.float32), name="W")
|
||||
A = helper.make_tensor_value_info("A", TensorProto.FLOAT, [B, K])
|
||||
Y = helper.make_tensor_value_info("Y", TensorProto.FLOAT, [B, N])
|
||||
node = helper.make_node("Gemm", ["A", "W"], ["Y"])
|
||||
graph = helper.make_graph([node], "gemm_non_square", [A], [Y], initializer=[W])
|
||||
model = helper.make_model(graph, opset_imports=[helper.make_opsetid("", 13)])
|
||||
save_model(model, "gemm/non_square", "gemm_non_square.onnx")
|
||||
|
||||
|
||||
def gemm_with_bias():
|
||||
"""GEMM with bias: Y = A @ W + C."""
|
||||
B, K, N = 4, 128, 128
|
||||
rng = np.random.default_rng(43)
|
||||
W = numpy_helper.from_array(rng.uniform(-1, 1, (K, N)).astype(np.float32), name="W")
|
||||
C = numpy_helper.from_array(rng.uniform(-1, 1, (N,)).astype(np.float32), name="C")
|
||||
A = helper.make_tensor_value_info("A", TensorProto.FLOAT, [B, K])
|
||||
Y = helper.make_tensor_value_info("Y", TensorProto.FLOAT, [B, N])
|
||||
node = helper.make_node("Gemm", ["A", "W", "C"], ["Y"])
|
||||
graph = helper.make_graph([node], "gemm_with_bias", [A], [Y], initializer=[W, C])
|
||||
model = helper.make_model(graph, opset_imports=[helper.make_opsetid("", 13)])
|
||||
save_model(model, "gemm/with_bias", "gemm_with_bias.onnx")
|
||||
|
||||
|
||||
def gemm_transB():
|
||||
"""GEMM with transB=1: Y = A @ W^T."""
|
||||
B, K, N = 4, 128, 64
|
||||
rng = np.random.default_rng(44)
|
||||
# W stored as [N, K], transposed during computation
|
||||
W = numpy_helper.from_array(rng.uniform(-1, 1, (N, K)).astype(np.float32), name="W")
|
||||
A = helper.make_tensor_value_info("A", TensorProto.FLOAT, [B, K])
|
||||
Y = helper.make_tensor_value_info("Y", TensorProto.FLOAT, [B, N])
|
||||
node = helper.make_node("Gemm", ["A", "W"], ["Y"], transB=1)
|
||||
graph = helper.make_graph([node], "gemm_transB", [A], [Y], initializer=[W])
|
||||
model = helper.make_model(graph, opset_imports=[helper.make_opsetid("", 13)])
|
||||
save_model(model, "gemm/transB", "gemm_transB.onnx")
|
||||
|
||||
|
||||
def gemm_alpha_beta():
|
||||
"""GEMM with alpha and beta: Y = 0.5 * A @ W + 0.25 * C."""
|
||||
B, K, N = 4, 64, 64
|
||||
rng = np.random.default_rng(45)
|
||||
W = numpy_helper.from_array(rng.uniform(-1, 1, (K, N)).astype(np.float32), name="W")
|
||||
C = numpy_helper.from_array(rng.uniform(-1, 1, (N,)).astype(np.float32), name="C")
|
||||
A = helper.make_tensor_value_info("A", TensorProto.FLOAT, [B, K])
|
||||
Y = helper.make_tensor_value_info("Y", TensorProto.FLOAT, [B, N])
|
||||
node = helper.make_node("Gemm", ["A", "W", "C"], ["Y"], alpha=0.5, beta=0.25)
|
||||
graph = helper.make_graph([node], "gemm_alpha_beta", [A], [Y], initializer=[W, C])
|
||||
model = helper.make_model(graph, opset_imports=[helper.make_opsetid("", 13)])
|
||||
save_model(model, "gemm/alpha_beta", "gemm_alpha_beta.onnx")
|
||||
|
||||
|
||||
def gemm_small():
|
||||
"""Small GEMM: [2, 8] @ [8, 4]."""
|
||||
B, K, N = 2, 8, 4
|
||||
rng = np.random.default_rng(46)
|
||||
W = numpy_helper.from_array(rng.uniform(-1, 1, (K, N)).astype(np.float32), name="W")
|
||||
A = helper.make_tensor_value_info("A", TensorProto.FLOAT, [B, K])
|
||||
Y = helper.make_tensor_value_info("Y", TensorProto.FLOAT, [B, N])
|
||||
node = helper.make_node("Gemm", ["A", "W"], ["Y"])
|
||||
graph = helper.make_graph([node], "gemm_small", [A], [Y], initializer=[W])
|
||||
model = helper.make_model(graph, opset_imports=[helper.make_opsetid("", 13)])
|
||||
save_model(model, "gemm/small", "gemm_small.onnx")
|
||||
|
||||
|
||||
def gemm_large():
|
||||
"""Larger GEMM: [8, 256] @ [256, 128]."""
|
||||
B, K, N = 8, 256, 128
|
||||
rng = np.random.default_rng(47)
|
||||
W = numpy_helper.from_array(rng.uniform(-1, 1, (K, N)).astype(np.float32), name="W")
|
||||
A = helper.make_tensor_value_info("A", TensorProto.FLOAT, [B, K])
|
||||
Y = helper.make_tensor_value_info("Y", TensorProto.FLOAT, [B, N])
|
||||
node = helper.make_node("Gemm", ["A", "W"], ["Y"])
|
||||
graph = helper.make_graph([node], "gemm_large", [A], [Y], initializer=[W])
|
||||
model = helper.make_model(graph, opset_imports=[helper.make_opsetid("", 13)])
|
||||
save_model(model, "gemm/large", "gemm_large.onnx")
|
||||
|
||||
|
||||
def gemm_transB_with_bias():
|
||||
"""GEMM with transB and bias: Y = A @ W^T + C."""
|
||||
B, K, N = 4, 128, 64
|
||||
rng = np.random.default_rng(48)
|
||||
W = numpy_helper.from_array(rng.uniform(-1, 1, (N, K)).astype(np.float32), name="W")
|
||||
C = numpy_helper.from_array(rng.uniform(-1, 1, (N,)).astype(np.float32), name="C")
|
||||
A = helper.make_tensor_value_info("A", TensorProto.FLOAT, [B, K])
|
||||
Y = helper.make_tensor_value_info("Y", TensorProto.FLOAT, [B, N])
|
||||
node = helper.make_node("Gemm", ["A", "W", "C"], ["Y"], transB=1)
|
||||
graph = helper.make_graph([node], "gemm_transB_with_bias", [A], [Y], initializer=[W, C])
|
||||
model = helper.make_model(graph, opset_imports=[helper.make_opsetid("", 13)])
|
||||
save_model(model, "gemm/transB_with_bias", "gemm_transB_with_bias.onnx")
|
||||
|
||||
|
||||
# ---------------------------------------------------------------------------
|
||||
# Conv tests
|
||||
# ---------------------------------------------------------------------------
|
||||
|
||||
def conv_3x3_kernel():
|
||||
"""Conv with 3x3 kernel, no padding."""
|
||||
# Input: [1, 1, 5, 5], Kernel: [1, 1, 3, 3] -> Output: [1, 1, 3, 3]
|
||||
X = helper.make_tensor_value_info("X", TensorProto.FLOAT, [1, 1, 5, 5])
|
||||
Y = helper.make_tensor_value_info("Y", TensorProto.FLOAT, [1, 1, 3, 3])
|
||||
W = numpy_helper.from_array(
|
||||
np.random.default_rng(50).uniform(-1, 1, (1, 1, 3, 3)).astype(np.float32), name="W")
|
||||
node = helper.make_node("Conv", ["X", "W"], ["Y"],
|
||||
kernel_shape=[3, 3], strides=[1, 1], pads=[0, 0, 0, 0])
|
||||
graph = helper.make_graph([node], "conv_3x3", [X], [Y], initializer=[W])
|
||||
model = helper.make_model(graph, opset_imports=[helper.make_opsetid("", 13)])
|
||||
save_model(model, "conv/kernel_3x3", "conv_kernel_3x3.onnx")
|
||||
|
||||
|
||||
def conv_stride2():
|
||||
"""Conv with 3x3 kernel and stride 2."""
|
||||
# Input: [1, 1, 6, 6], Kernel: [1, 1, 3, 3], stride 2 -> Output: [1, 1, 2, 2]
|
||||
X = helper.make_tensor_value_info("X", TensorProto.FLOAT, [1, 1, 6, 6])
|
||||
Y = helper.make_tensor_value_info("Y", TensorProto.FLOAT, [1, 1, 2, 2])
|
||||
W = numpy_helper.from_array(
|
||||
np.random.default_rng(51).uniform(-1, 1, (1, 1, 3, 3)).astype(np.float32), name="W")
|
||||
node = helper.make_node("Conv", ["X", "W"], ["Y"],
|
||||
kernel_shape=[3, 3], strides=[2, 2], pads=[0, 0, 0, 0])
|
||||
graph = helper.make_graph([node], "conv_stride2", [X], [Y], initializer=[W])
|
||||
model = helper.make_model(graph, opset_imports=[helper.make_opsetid("", 13)])
|
||||
save_model(model, "conv/stride_2", "conv_stride_2.onnx")
|
||||
|
||||
|
||||
def conv_multi_channel():
|
||||
"""Conv with multiple input and output channels."""
|
||||
# Input: [1, 3, 5, 5], Kernel: [4, 3, 3, 3] -> Output: [1, 4, 3, 3]
|
||||
X = helper.make_tensor_value_info("X", TensorProto.FLOAT, [1, 3, 5, 5])
|
||||
Y = helper.make_tensor_value_info("Y", TensorProto.FLOAT, [1, 4, 3, 3])
|
||||
W = numpy_helper.from_array(
|
||||
np.random.default_rng(52).uniform(-1, 1, (4, 3, 3, 3)).astype(np.float32), name="W")
|
||||
node = helper.make_node("Conv", ["X", "W"], ["Y"],
|
||||
kernel_shape=[3, 3], strides=[1, 1], pads=[0, 0, 0, 0])
|
||||
graph = helper.make_graph([node], "conv_multi_channel", [X], [Y], initializer=[W])
|
||||
model = helper.make_model(graph, opset_imports=[helper.make_opsetid("", 13)])
|
||||
save_model(model, "conv/multi_channel", "conv_multi_channel.onnx")
|
||||
|
||||
|
||||
def conv_1x1():
|
||||
"""1x1 pointwise convolution (channel mixing)."""
|
||||
# Input: [1, 8, 4, 4], Kernel: [4, 8, 1, 1] -> Output: [1, 4, 4, 4]
|
||||
X = helper.make_tensor_value_info("X", TensorProto.FLOAT, [1, 8, 4, 4])
|
||||
Y = helper.make_tensor_value_info("Y", TensorProto.FLOAT, [1, 4, 4, 4])
|
||||
W = numpy_helper.from_array(
|
||||
np.random.default_rng(53).uniform(-1, 1, (4, 8, 1, 1)).astype(np.float32), name="W")
|
||||
node = helper.make_node("Conv", ["X", "W"], ["Y"],
|
||||
kernel_shape=[1, 1], strides=[1, 1], pads=[0, 0, 0, 0])
|
||||
graph = helper.make_graph([node], "conv_1x1", [X], [Y], initializer=[W])
|
||||
model = helper.make_model(graph, opset_imports=[helper.make_opsetid("", 13)])
|
||||
save_model(model, "conv/pointwise_1x1", "conv_1x1.onnx")
|
||||
|
||||
|
||||
def conv_same_padding_3x3():
|
||||
"""Conv 3x3 with SAME_UPPER padding, preserving spatial dimensions."""
|
||||
# Input: [1, 1, 5, 5], Kernel: [1, 1, 3, 3], SAME_UPPER -> Output: [1, 1, 5, 5]
|
||||
X = helper.make_tensor_value_info("X", TensorProto.FLOAT, [1, 1, 5, 5])
|
||||
Y = helper.make_tensor_value_info("Y", TensorProto.FLOAT, [1, 1, 5, 5])
|
||||
W = numpy_helper.from_array(
|
||||
np.random.default_rng(54).uniform(-1, 1, (1, 1, 3, 3)).astype(np.float32), name="W")
|
||||
node = helper.make_node("Conv", ["X", "W"], ["Y"],
|
||||
kernel_shape=[3, 3], strides=[1, 1], auto_pad="SAME_UPPER")
|
||||
graph = helper.make_graph([node], "conv_same_3x3", [X], [Y], initializer=[W])
|
||||
model = helper.make_model(graph, opset_imports=[helper.make_opsetid("", 13)])
|
||||
save_model(model, "conv/same_padding_3x3", "conv_same_padding_3x3.onnx")
|
||||
|
||||
|
||||
def conv_explicit_padding():
|
||||
"""Conv 3x3 with explicit asymmetric padding."""
|
||||
# Input: [1, 1, 4, 4], Kernel: [1, 1, 3, 3], pads=[1,1,1,1] -> Output: [1, 1, 4, 4]
|
||||
X = helper.make_tensor_value_info("X", TensorProto.FLOAT, [1, 1, 4, 4])
|
||||
Y = helper.make_tensor_value_info("Y", TensorProto.FLOAT, [1, 1, 4, 4])
|
||||
W = numpy_helper.from_array(
|
||||
np.random.default_rng(55).uniform(-1, 1, (1, 1, 3, 3)).astype(np.float32), name="W")
|
||||
node = helper.make_node("Conv", ["X", "W"], ["Y"],
|
||||
kernel_shape=[3, 3], strides=[1, 1], pads=[1, 1, 1, 1])
|
||||
graph = helper.make_graph([node], "conv_explicit_pad", [X], [Y], initializer=[W])
|
||||
model = helper.make_model(graph, opset_imports=[helper.make_opsetid("", 13)])
|
||||
save_model(model, "conv/explicit_padding", "conv_explicit_padding.onnx")
|
||||
|
||||
|
||||
def conv_with_bias_3x3():
|
||||
"""Conv 3x3 with bias."""
|
||||
# Input: [1, 3, 5, 5], Kernel: [2, 3, 3, 3], Bias: [2] -> Output: [1, 2, 3, 3]
|
||||
rng = np.random.default_rng(56)
|
||||
X = helper.make_tensor_value_info("X", TensorProto.FLOAT, [1, 3, 5, 5])
|
||||
Y = helper.make_tensor_value_info("Y", TensorProto.FLOAT, [1, 2, 3, 3])
|
||||
W = numpy_helper.from_array(rng.uniform(-1, 1, (2, 3, 3, 3)).astype(np.float32), name="W")
|
||||
B = numpy_helper.from_array(rng.uniform(-1, 1, (2,)).astype(np.float32), name="B")
|
||||
node = helper.make_node("Conv", ["X", "W", "B"], ["Y"],
|
||||
kernel_shape=[3, 3], strides=[1, 1], pads=[0, 0, 0, 0])
|
||||
graph = helper.make_graph([node], "conv_with_bias_3x3", [X], [Y], initializer=[W, B])
|
||||
model = helper.make_model(graph, opset_imports=[helper.make_opsetid("", 13)])
|
||||
save_model(model, "conv/with_bias_3x3", "conv_with_bias_3x3.onnx")
|
||||
|
||||
|
||||
def conv_batch_2():
|
||||
"""Batched conv (batch=2) with SAME_UPPER padding and bias."""
|
||||
# Input: [2, 3, 3, 3], Kernel: [1, 3, 2, 2], Bias: [1] -> Output: [2, 1, 3, 3]
|
||||
rng = np.random.default_rng(57)
|
||||
X = helper.make_tensor_value_info("X", TensorProto.FLOAT, [2, 3, 3, 3])
|
||||
Y = helper.make_tensor_value_info("Y", TensorProto.FLOAT, [2, 1, 3, 3])
|
||||
W = numpy_helper.from_array(rng.uniform(-1, 1, (1, 3, 2, 2)).astype(np.float32), name="W")
|
||||
B = numpy_helper.from_array(rng.uniform(-1, 1, (1,)).astype(np.float32), name="B")
|
||||
node = helper.make_node("Conv", ["X", "W", "B"], ["Y"],
|
||||
kernel_shape=[2, 2], strides=[1, 1], auto_pad="SAME_UPPER")
|
||||
graph = helper.make_graph([node], "conv_batch_2", [X], [Y], initializer=[W, B])
|
||||
model = helper.make_model(graph, opset_imports=[helper.make_opsetid("", 13)])
|
||||
save_model(model, "conv/batch_2", "conv_batch_2.onnx")
|
||||
|
||||
|
||||
def conv_large_spatial():
|
||||
"""Conv on larger spatial input: [1, 1, 8, 8] with 3x3 kernel."""
|
||||
X = helper.make_tensor_value_info("X", TensorProto.FLOAT, [1, 1, 8, 8])
|
||||
Y = helper.make_tensor_value_info("Y", TensorProto.FLOAT, [1, 1, 6, 6])
|
||||
W = numpy_helper.from_array(
|
||||
np.random.default_rng(58).uniform(-1, 1, (1, 1, 3, 3)).astype(np.float32), name="W")
|
||||
node = helper.make_node("Conv", ["X", "W"], ["Y"],
|
||||
kernel_shape=[3, 3], strides=[1, 1], pads=[0, 0, 0, 0])
|
||||
graph = helper.make_graph([node], "conv_large_spatial", [X], [Y], initializer=[W])
|
||||
model = helper.make_model(graph, opset_imports=[helper.make_opsetid("", 13)])
|
||||
save_model(model, "conv/large_spatial", "conv_large_spatial.onnx")
|
||||
|
||||
|
||||
# ---------------------------------------------------------------------------
|
||||
# Main
|
||||
# ---------------------------------------------------------------------------
|
||||
|
||||
if __name__ == "__main__":
|
||||
print("Generating GEMM tests:")
|
||||
gemm_non_square()
|
||||
gemm_with_bias()
|
||||
gemm_transB()
|
||||
gemm_alpha_beta()
|
||||
gemm_small()
|
||||
gemm_large()
|
||||
gemm_transB_with_bias()
|
||||
|
||||
print("\nGenerating Conv tests:")
|
||||
conv_3x3_kernel()
|
||||
conv_stride2()
|
||||
conv_multi_channel()
|
||||
conv_1x1()
|
||||
conv_same_padding_3x3()
|
||||
conv_explicit_padding()
|
||||
conv_with_bias_3x3()
|
||||
conv_batch_2()
|
||||
conv_large_spatial()
|
||||
|
||||
print("\nDone.")
|
||||
Reference in New Issue
Block a user