diff --git a/.github/workflows/test.yml b/.github/workflows/test.yml
index 6fc8a5f59..fd166f5d6 100644
--- a/.github/workflows/test.yml
+++ b/.github/workflows/test.yml
@@ -411,13 +411,13 @@ jobs:
           DEBUG=5 PYTHONPATH=${{ github.workspace }} FORWARD_ONLY=1 python3 test/test_ops.py TestOps.test_add
       - name: Run pytest (not cuda)
         if: matrix.backend!='cuda' && matrix.backend!='ptx' && matrix.backend!='triton'
-        run: python -m pytest -n=auto test/ -k '${{matrix.backend=='llvm'&&'not (test_nn.py and test_conv_transpose2d)'||'test'}}' -m 'not exclude_${{matrix.backend}}' --durations=20
+        run: python -m pytest -n=auto test/ -m 'not exclude_${{matrix.backend}}' --durations=20
       - name: Run ONNX (only LLVM)
         if: matrix.backend == 'llvm'
         run: python -m pytest -n=auto test/external/external_test_onnx_backend.py --durations=20
       - name: Run pytest (cuda)
         if: matrix.backend=='cuda'||matrix.backend=='ptx'||matrix.backend=='triton'
-        run: python -m pytest -n=auto test/ -k 'not (half or test_efficientnet_safetensors) and not (test_conv2d and test_tensor.py)' -m 'not exclude_cuda' --ignore=test/external --ignore=test/models --durations=20
+        run: python -m pytest -n=auto test/ -k 'not (half or test_efficientnet_safetensors)' -m 'not exclude_cuda' --ignore=test/external --ignore=test/models --durations=20
 
   #testunicorn:
   #  name: ARM64 unicorn Test
diff --git a/pytest.ini b/pytest.ini
index 42214cbd5..f203b19cc 100644
--- a/pytest.ini
+++ b/pytest.ini
@@ -3,5 +3,4 @@ markers =
   exclude_cuda
   exclude_gpu
   exclude_clang
-  webgpu
   onnx_coverage
diff --git a/test/test_conv.py b/test/test_conv.py
index 1463eba48..6adde7af7 100644
--- a/test/test_conv.py
+++ b/test/test_conv.py
@@ -1,9 +1,6 @@
 import unittest
 import numpy as np
 from tinygrad.tensor import Tensor, Device
-import pytest
-
-pytestmark = [pytest.mark.exclude_cuda]
 
 class TestConv(unittest.TestCase):
   def test_simple(self):
diff --git a/test/test_custom_function.py b/test/test_custom_function.py
index 659a85e4d..a844bf2f2 100644
--- a/test/test_custom_function.py
+++ b/test/test_custom_function.py
@@ -11,9 +11,6 @@ from tinygrad.helpers import prod, dtypes
 from tinygrad.lazy import Buffer, create_lazybuffer
 from tinygrad.device import CompiledASTRunner, Device
 from tinygrad.shape.shapetracker import ShapeTracker
-import pytest
-
-pytestmark = pytest.mark.webgpu
 
 # we don't always have GPU support, so the type signature is the abstract CompiledBuffer instead of GPUBuffer
 def atan2_gpu(ret:Buffer, a:Buffer, b:Buffer):
diff --git a/test/test_randomness.py b/test/test_randomness.py
index b5653c978..d9b5a9079 100644
--- a/test/test_randomness.py
+++ b/test/test_randomness.py
@@ -4,12 +4,9 @@ import numpy as np
 import torch
 from tinygrad.tensor import Tensor
 import tinygrad.nn as nn
-import pytest
 from tinygrad.helpers import dtypes
 from functools import partial
 
-pytestmark = pytest.mark.webgpu
-
 # https://gist.github.com/devries/11405101
 def ksprob(a):
   fac, total, termbf = 2.0, 0.0, 0.0
diff --git a/test/test_specific_conv.py b/test/test_specific_conv.py
index 992f7a8ab..fa7f53db4 100644
--- a/test/test_specific_conv.py
+++ b/test/test_specific_conv.py
@@ -1,12 +1,10 @@
 import unittest
 from tinygrad.tensor import Tensor
-from tinygrad.helpers import dtypes
+from tinygrad.helpers import CI, dtypes
 from tinygrad import Device
-import pytest
 # similar to test/external/external_test_gpu_ast.py, but universal
 
-pytestmark = pytest.mark.exclude_cuda
-
+@unittest.skipIf(Device.DEFAULT == "CUDA" and CI, "slow on CUDA CI")
 class TestSpecific(unittest.TestCase):
   # from openpilot
 
@@ -22,7 +20,7 @@ class TestSpecific(unittest.TestCase):
     w = Tensor.randn(2048, 512)
     (x @ w).reshape(1, 128, 4).contiguous().realize()
 
-  @unittest.skipIf(Device.DEFAULT in ["LLVM", "WEBGPU"], "Broken on LLVM and webgpu")
+  @unittest.skipIf(Device.DEFAULT in ["LLVM", "WEBGPU", "CUDA"], "Broken on LLVM, WEBGPU and CUDA")
   def test_big_vec_mul(self):
     # from LLaMA
     #   0 buffer<4096, dtypes.float>                      [View((1024, 1, 1, 4), (4, 0, 0, 1), 0, None)]