Enable line-ending and other hygiene lints (#1006)

examples/int8_inference_huggingface.py

@@ -22,6 +22,3 @@ model = AutoModelForCausalLM.from_pretrained(
 )
 generated_ids = model.generate(input_ids, max_length=MAX_NEW_TOKENS)
 print(tokenizer.decode(generated_ids[0], skip_special_tokens=True))
-
-
-

tests/test_autograd.py

@@ -519,4 +519,3 @@ def test_matmul_fp8( dim1, dim2, dim3, dim4, funcs, dtype, req_grad, transpose):
                     torch.testing.assert_close(
                         gradB1, gradB2, atol=0.18, rtol=0.3
                     )
-

tests/test_cuda_setup_evaluator.py

@@ -19,11 +19,3 @@ def test_manual_override(requires_cuda):
     import bitsandbytes as bnb
     loaded_lib = bnb.cuda_setup.main.CUDASetup.get_instance().binary_name
     #assert loaded_lib == 'libbitsandbytes_cuda122.so'
-
-
-
-
-
-
-
-

tests/test_functional.py

@@ -2345,5 +2345,3 @@ def test_gemv_eye_4bit(storage_type, dtype, double_quant):
         torch.testing.assert_close(A, C2)
         #torch.testing.assert_close(A, C1, rtol=1e-5, atol=0.00001)
         #torch.testing.assert_close(A, C2, rtol=1e-5, atol=0.080)
-
-

tests/test_generation.py

@@ -120,6 +120,3 @@ def test_pi(requires_cuda, model_and_tokenizer, inference_kernel, DQ, dtype):
         for out in outputs:
             print(out)
         raise ValueError(f'Failure count: {failure_count}/{n_cases}')
-
-
-

tests/test_modules.py

@@ -637,6 +637,3 @@ def test_4bit_warnings():
         net(inp)
 
     assert len(record) == 2
-
-
-

tests/test_triton.py

@@ -58,4 +58,3 @@ def test_switchback(vector_wise_quantization):
 
             print('GX1', err_sb, err_baseline)
             assert err_sb < 2 * err_baseline
-