Revert "Parallel test with pytest-xdist" (#526)

This reverts commit cbaff6c.

Author: msaroufim

.github/workflows/regression_test.yml

@@ -68,5 +68,4 @@ jobs:
         pip install ${{ matrix.torch-spec }}
         pip install -r dev-requirements.txt
         pip install .
-        pytest test --verbose -s -m "not multi_gpu" --dist load --tx popen//env:CUDA_VISIBLE_DEVICES=0 --tx popen//env:CUDA_VISIBLE_DEVICES=1 --tx popen//env:CUDA_VISIBLE_DEVICES=2 --tx popen//env:CUDA_VISIBLE_DEVICES=3
-        pytest test --verbose -s -m "multi_gpu"
+        pytest test --verbose -s

dev-requirements.txt

@@ -7,7 +7,6 @@ transformers
 hypothesis # Avoid test derandomization warning
 sentencepiece # for gpt-fast tokenizer
 expecttest
-pytest-xdist
 
 # For prototype features and benchmarks
 bitsandbytes #needed for testing triton quant / dequant ops for 8-bit optimizers

pytest.ini

@@ -486,7 +486,6 @@ class TestQLoRA(FSDPTest):
     def world_size(self) -> int:
         return 2
 
-    @pytest.mark.multi_gpu
     @pytest.mark.skipif(
         version.parse(torch.__version__).base_version < "2.4.0",
         reason="torch >= 2.4 required",

test/dtypes/test_nf4.py

@@ -985,10 +985,7 @@ def forward(self, x):
         # save quantized state_dict
         api(model)
 
-        # unique filename to avoid collision in parallel tests
-        ckpt_name = f"{api.__name__}_{test_device}_{test_dtype}_test.pth"
-
-        torch.save(model.state_dict(), ckpt_name)
+        torch.save(model.state_dict(), "test.pth")
         # get quantized reference
         model_qc = torch.compile(model, mode="max-autotune")
         ref_q = model_qc(x).detach()
@@ -1001,8 +998,8 @@ def forward(self, x):
         api(model)
 
         # load quantized state_dict
-        state_dict = torch.load(ckpt_name, mmap=True)
-        os.remove(ckpt_name)
+        state_dict = torch.load("test.pth", mmap=True)
+        os.remove("test.pth")
 
         model.load_state_dict(state_dict, assign=True)
         model = model.to(device=test_device, dtype=test_dtype).eval()

test/integration/test_integration.py

@@ -163,7 +163,6 @@ class TestFSDP2(FSDPTest):
     def world_size(self) -> int:
         return 2
 
-    @pytest.mark.multi_gpu
     @pytest.mark.skipif(not TORCH_VERSION_AFTER_2_4, reason="torch >= 2.4 required")
     @skip_if_lt_x_gpu(2)
     def test_fsdp2(self):