Merge branch 'main' into dependabot-github_actions-pypa-gh-action-pyp…

…i-publish-1.12.2

thunder/tests/test_jit_general.py

@@ -680,6 +680,8 @@ def test_litgpt_variants(name, device):
 
     if device == "cuda" and not torch.cuda.is_available():
         pytest.skip("CUDA not available")
+    if device == "cuda" and name == "falcon-40b-like":
+        pytest.skip("NVFuser reenable when https://github.com/NVIDIA/Fuser/issues/3505 is fixed, Thunder issue #1504")
     if device == "cuda" and name == "falcon-7b-like":
         pytest.skip("NVFuser reenable when https://github.com/NVIDIA/Fuser/issues/3292 is fixed")
 
@@ -783,7 +785,7 @@ def sample(logits):
     ("cpu", "cuda"),
 )
 def test_tom_overrides_proxy(device):
-    from litgpt.config import Config
+    from thunder.tests.litgpt_model import Config
     from litgpt.model import GPT
 
     if device == "cuda" and not torch.cuda.is_available():

thunder/tests/test_torch_compile_executor.py

@@ -20,6 +20,7 @@ def test_supported_ops_are_in_pytorch_executor():
 # appropriate visual studio config.
 @pytest.mark.skipif(not is_inductor_supported() or platform.system() == "Windows", reason="inductor unsupported")
 def test_torch_compile_litgpt():
+    from thunder.tests.litgpt_model import Config
     from litgpt.model import GPT
 
     model = GPT.from_name("llama1-like", n_layer=1)
@@ -40,7 +41,7 @@ def test_torch_compile_litgpt():
 @requiresCUDA
 @pytest.mark.skipif(not device_supports_bf16(torch.device("cuda")), reason="bf16 is not supported")
 def test_torch_compile_cat_nvfuser_phi2_tanh():
-    from litgpt.config import Config
+    from thunder.tests.litgpt_model import Config
     from litgpt.model import GPT
 
     device = torch.device("cuda")

thunder/tests/test_transforms.py

@@ -74,7 +74,8 @@ def _test_equal_nvtx_push_and_pop(trc):
 @requiresCUDA
 def test_materialization():
     from thunder.transforms import MaterializationTransform
-    from litgpt.config import Config
+    from thunder.tests.litgpt_model import Config
+
     from litgpt.model import GPT
 
     config = Config.from_name("llama2-like")
@@ -121,7 +122,7 @@ def test_materialization():
 def test_quantization_on_meta():
     from thunder.transforms import MaterializationTransform
     from thunder.transforms.quantization import BitsAndBytesLinearQuant4bit, get_bitsandbytes_executor
-    from litgpt.config import Config
+    from thunder.tests.litgpt_model import Config
     from litgpt.model import GPT
 
     bitsandbytes_executor = get_bitsandbytes_executor()