leavelet/cutlass
1
0
Fork 0
Code Issues Pull Requests Actions 25 Packages Projects Releases Wiki Activity

Make Python interface work for non-SM80 targets (#726)

Browse Source 
* Make Python interface work for non-SM80 targets

* Remove line in README
This commit is contained in:
Jack Kosaian
2022-12-07 21:53:33 -05:00
committed by GitHub
parent d6117ca362
commit df81d847d7
33 changed files with 149 additions and 20 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tools/library/scripts/pycutlass/test/conv/conv2d_dgrad_implicit_gemm_f16nhwc_f16nhwc_f16nhwc_tensor_op_f16_sm80.py
+2
View File
@@ -2,9 +2,11 @@
from pycutlass.conv2d_operation import *
from pycutlass import *
from pycutlass.test import *
from pycutlass.utils.device import device_cc
import unittest
@unittest.skipIf(device_cc() < 80, "Device compute capability is insufficient for SM80 tests.")
class Conv2dDgradImplicitGemmF16nhwcF16nhwcF16nhwcTensorOpF16SM80(unittest.TestCase):
def test_SM80_Device_Conv2d_Dgrad_Analytic_ImplicitGemm_f16nhwc_f16nhwc_f16nhwc_tensor_op_f16(self):
math_inst = MathInstruction(
tools/library/scripts/pycutlass/test/conv/conv2d_dgrad_implicit_gemm_f16nhwc_f16nhwc_f32nhwc_tensor_op_f32_sm80.py
+3
View File
@@ -2,8 +2,11 @@
import pycutlass
from pycutlass import *
from pycutlass.test import *
from pycutlass.utils.device import device_cc
import unittest
@unittest.skipIf(device_cc() < 80, "Device compute capability is insufficient for SM80 tests.")
class Conv2dDgradImplicitGemmF16nhwcF16nhwcF32nhwcTensorOpF32SM80(unittest.TestCase):
def test_SM80_Device_Conv2d_Dgrad_Optimized_ImplicitGemm_f16nhwc_f16nhwc_f32nhwc_tensor_op_f32_unity_stride_stage3(self):
math_inst = MathInstruction(
tools/library/scripts/pycutlass/test/conv/conv2d_dgrad_implicit_gemm_f32nhwc_f32nhwc_f32nhwc_simt_f32_sm80.py
+3
View File
@@ -3,8 +3,11 @@ import pycutlass
from pycutlass.conv2d_operation import *
from pycutlass import *
from pycutlass.test import *
from pycutlass.utils.device import device_cc
import unittest
@unittest.skipIf(device_cc() < 80, "Device compute capability is insufficient for SM80 tests.")
class Conv2dDgradImplicitGemmF32nhwcF32nhwcF32nhwcSimtF32SM80(unittest.TestCase):
def test_SM80_Device_Conv2d_Fprop_Analytic_ImplicitGemm_f32nhwc_f32nhwc_f32nhwc_simt_f32(self):
math_inst = MathInstruction(
tools/library/scripts/pycutlass/test/conv/conv2d_dgrad_implicit_gemm_tf32nhwc_tf32nhwc_f32nhwc_tensor_op_f32_sm80.py
+3
View File
@@ -2,8 +2,11 @@
import pycutlass
from pycutlass import *
from pycutlass.test import *
from pycutlass.utils.device import device_cc
import unittest
@unittest.skipIf(device_cc() < 80, "Device compute capability is insufficient for SM80 tests.")
class Conv2dDgradImplicitGemmTF32nhwcTF32nhwcTF32nhwcTensorOpF32SM80(unittest.TestCase):
def test_SM80_Device_Conv2d_Dgrad_Analytic_ImplicitGemm_tf32nhwc_tf32nhwc_f32nhwc_tensor_op_f32(self):
math_inst = MathInstruction(
tools/library/scripts/pycutlass/test/conv/conv2d_fprop_few_channels_f16nhwc_f16nhwc_f16nhwc_tensor_op_f32_sm80.py
+3
View File
@@ -1,8 +1,11 @@
# test/unit/conv/device/conv2d_fprop_few_channels_f16nhwc_f16nhwc_f16nhwc_tensor_op_f32_sm80.cu
import pycutlass
from pycutlass.test import *
from pycutlass.utils.device import device_cc
import unittest
@unittest.skipIf(device_cc() < 80, "Device compute capability is insufficient for SM80 tests.")
def conv2d_few_channel_problemsizes(channels):
problem_sizes = [
cutlass.conv.Conv2dProblemSize(
tools/library/scripts/pycutlass/test/conv/conv2d_fprop_fixed_channels_f16nhwc_f16nhwc_f16nhwc_tensor_op_f32_sm80.py
+3
View File
@@ -1,8 +1,11 @@
# test/unit/conv/device/conv2d_fprop_fixed_channels_f16nhwc_f16nhwc_f16nhwc_tensor_op_f32_sm80.cu
import pycutlass
from pycutlass.test import *
from pycutlass.utils.device import device_cc
import unittest
@unittest.skipIf(device_cc() < 80, "Device compute capability is insufficient for SM80 tests.")
def conv2d_fixed_channel_problemsizes(channels):
problem_sizes = [
cutlass.conv.Conv2dProblemSize(
tools/library/scripts/pycutlass/test/conv/conv2d_fprop_implicit_gemm_f16nhwc_f16nhwc_f16nhwc_tensor_op_f16_sm80.py
+3
View File
@@ -2,8 +2,11 @@
import pycutlass
from pycutlass import *
from pycutlass.test import *
from pycutlass.utils.device import device_cc
import unittest
@unittest.skipIf(device_cc() < 80, "Device compute capability is insufficient for SM80 tests.")
class Conv2dFpropImplicitGemmF16nhwcF16nhwcF16nhwcTensorOpF16SM80(unittest.TestCase):
def test_SM80_Device_Conv2d_Fprop_Analytic_ImplicitGemm_f16nhwc_f16nhwc_f16nhwc_tensor_op_f16(self):
math_inst = MathInstruction(
tools/library/scripts/pycutlass/test/conv/conv2d_fprop_implicit_gemm_f16nhwc_f16nhwc_f32nhwc_tensor_op_f32_sm80.py
+3
View File
@@ -2,8 +2,11 @@
import pycutlass
from pycutlass import *
from pycutlass.test import *
from pycutlass.utils.device import device_cc
import unittest
@unittest.skipIf(device_cc() < 80, "Device compute capability is insufficient for SM80 tests.")
class Conv2dFpropImplicitGemmF16nhwcF16nhwcF32nhwcTensorOpF32SM80(unittest.TestCase):
def test_SM80_Device_Conv2d_Fprop_Analytic_ImplicitGemm_f16nhwc_f16nhwc_f32nhwc_tensor_op_f32(self):
math_inst = MathInstruction(
tools/library/scripts/pycutlass/test/conv/conv2d_fprop_implicit_gemm_f32nhwc_f32nhwc_f32nhwc_simt_f32_sm80.py
+3
View File
@@ -3,8 +3,11 @@ import pycutlass
from pycutlass.conv2d_operation import *
from pycutlass import *
from pycutlass.test import *
from pycutlass.utils.device import device_cc
import unittest
@unittest.skipIf(device_cc() < 80, "Device compute capability is insufficient for SM80 tests.")
class Conv2dFpropImplicitGemmF32nhwcF32nhwcF32nhwcSimtF32SM80(unittest.TestCase):
def test_SM80_Device_Conv2d_Fprop_Analytic_ImplicitGemm_f32nhwc_f32nhwc_f32nhwc_simt_f32(self):
math_inst = MathInstruction(
tools/library/scripts/pycutlass/test/conv/conv2d_fprop_implicit_gemm_tf32nhwc_tf32nhwc_f32nhwc_tensor_op_f32_sm80.py
+3
View File
@@ -2,8 +2,11 @@
import pycutlass
from pycutlass import *
from pycutlass.test import *
from pycutlass.utils.device import device_cc
import unittest
@unittest.skipIf(device_cc() < 80, "Device compute capability is insufficient for SM80 tests.")
class Conv2dFpropImplicitGemmTF32nhwcTF32nhwcTF32nhwcTensorOpF32SM80(unittest.TestCase):
def test_SM80_Device_Conv2d_Fprop_Analytic_ImplicitGemm_tf32nhwc_tf32nhwc_f32nhwc_tensor_op_f32(self):
math_inst = MathInstruction(
tools/library/scripts/pycutlass/test/conv/conv2d_strided_dgrad_implicit_gemm_f16nhwc_f16nhwc_f32nhwc_tensor_op_f32_sm80.py
+3
View File
@@ -2,8 +2,11 @@
import pycutlass
from pycutlass import *
from pycutlass.test import *
from pycutlass.utils.device import device_cc
import unittest
@unittest.skipIf(device_cc() < 80, "Device compute capability is insufficient for SM80 tests.")
class Conv2dStridedDgradImplicitGemmF16NHWCF16NHWCF32NHWCTensorOpF32SM80(unittest.TestCase):
def test_SM80_Device_Conv2d_Strided_Dgrad_Analytic_ImplicitGemm_f16nhwc_f16nhwc_f32nhwc_tensor_op_f32_128x128_32x3_64x64x32(self):
math_inst = MathInstruction(
tools/library/scripts/pycutlass/test/conv/conv2d_wgrad_implicit_gemm_f16nhwc_f16nhwc_f16nhwc_tensor_op_f16_sm80.py
+3
View File
@@ -2,8 +2,11 @@
import pycutlass
from pycutlass import *
from pycutlass.test import *
from pycutlass.utils.device import device_cc
import unittest
@unittest.skipIf(device_cc() < 80, "Device compute capability is insufficient for SM80 tests.")
class Conv2dWgradImplicitGemmF16nhwcF16nhwcF16nhwcTensorOpF16SM80(unittest.TestCase):
def test_Device_Conv2d_Wgrad_Analytic_ImplicitGemm_f16nhwc_f16nhwc_f16nhwc_tensor_op_f16(self):
math_inst = MathInstruction(
tools/library/scripts/pycutlass/test/conv/conv2d_wgrad_implicit_gemm_f16nhwc_f16nhwc_f32nhwc_tensor_op_f32_sm80.py
+3
View File
@@ -2,8 +2,11 @@
import pycutlass
from pycutlass import *
from pycutlass.test import *
from pycutlass.utils.device import device_cc
import unittest
@unittest.skipIf(device_cc() < 80, "Device compute capability is insufficient for SM80 tests.")
class Conv2dWgradImplicitGemmF16nhwcF16nhwcF32nhwcTensorOpF32SM80(unittest.TestCase):
def test_Device_Conv2d_Wgrad_Analytic_ImplicitGemm_f16nhwc_f16nhwc_f32nhwc_tensor_op_f32(self):
math_inst = MathInstruction(
tools/library/scripts/pycutlass/test/conv/conv2d_wgrad_implicit_gemm_f32nhwc_f32nhwc_f32nhwc_simt_f32_sm80.py
+3
View File
@@ -3,8 +3,11 @@ import pycutlass
from pycutlass.conv2d_operation import *
from pycutlass import *
from pycutlass.test import *
from pycutlass.utils.device import device_cc
import unittest
@unittest.skipIf(device_cc() < 80, "Device compute capability is insufficient for SM80 tests.")
class Conv2dWgradImplicitGemmF32nhwcF32nhwcF32nhwcSimtF32SM80(unittest.TestCase):
def test_SM80_Device_Conv2d_Wgrad_Analytic_ImplicitGemm_f32nhwc_f32nhwc_f32nhwc_simt_f32(self):
math_inst = MathInstruction(
tools/library/scripts/pycutlass/test/conv/conv2d_wgrad_implicit_gemm_tf32nhwc_tf32nhwc_f32nhwc_tensor_op_f32_sm80.py
+3
View File
@@ -2,8 +2,11 @@
import pycutlass
from pycutlass import *
from pycutlass.test import *
from pycutlass.utils.device import device_cc
import unittest
@unittest.skipIf(device_cc() < 80, "Device compute capability is insufficient for SM80 tests.")
class Conv2dWgradImplicitGemmTF32nhwcTF32nhwcTF32nhwcTensorOpF32SM80(unittest.TestCase):
def test_SM80_Device_Conv2d_Wgrad_Optimized_ImplicitGemm_tf32nhwc_tf32nhwc_f32nhwc_tensor_op_f32(self):
math_inst = MathInstruction(