fix style

Signed-off-by: Kyle Sayers <kylesayrs@gmail.com>

Author: kylesayrs

src/compressed_tensors/quantization/lifecycle/initialize.py

@@ -27,7 +27,6 @@
     FP8_E4M3_DATA,
     ActivationOrdering,
     DynamicType,
-    KVCacheScaleType,
     QuantizationArgs,
     QuantizationMetadata,
     QuantizationScheme,

src/compressed_tensors/quantization/quant_config.py

@@ -171,7 +171,7 @@ def from_pretrained(
         :param model: model to calculate quantization scheme of
         :return: filled out QuantizationScheme for the input model
         """
-        from compressed_tensors.quantization.lifecycle.initialize import (  # circular import, move to utils later
+        from compressed_tensors.quantization.lifecycle.initialize import (
             is_attention_module,
         )
 

src/compressed_tensors/quantization/utils/helpers.py

@@ -14,7 +14,7 @@
 
 import logging
 import math
-from typing import Generator, List, Optional, Tuple
+from typing import Generator, Optional, Tuple
 
 import torch
 from compressed_tensors.quantization.quant_args import (

tests/test_modeling/test_attention_and_cache.py

@@ -49,7 +49,7 @@ def test_attention_cache():
     assert torch.equal(outputs.logits, true_outputs.logits)
     assert all(k_called) and all(v_called)
 
-    ## apply attention quantization after kv cache quantization ##
+    """ apply attention quantization after kv cache quantization """
 
     # check if hooks work
     q_called = [False for _ in range(len(layers))]

tests/test_quantization/lifecycle/test_apply.py

@@ -367,6 +367,7 @@ def test_multi_apply_quantization_config():
                 and weight_zero_point.shape == torch.Size([1])
             )
 
+
 @requires_accelerate()
 def test_apply_kv_cache():
     from accelerate import init_empty_weights
@@ -376,7 +377,7 @@ def test_apply_kv_cache():
 
     args = QuantizationArgs(num_bits=8, type="float", strategy="tensor")
     config = QuantizationConfig(config_groups={}, kv_cache_scheme=args)
-    
+
     apply_quantization_config(model, config)
 
     for layer in model.model.layers:
@@ -397,11 +398,11 @@ def test_apply_attention():
         input_activations=QuantizationArgs(num_bits=8, type="float", strategy="tensor"),
     )
     config = QuantizationConfig(config_groups={"attention": scheme})
-    
+
     apply_quantization_config(model, config)
 
     for layer in model.model.layers:
         assert getattr(layer.self_attn, "quantization_scheme") == scheme
         assert hasattr(layer.self_attn, "q_scale")
         assert hasattr(layer.self_attn, "k_scale")
-        assert hasattr(layer.self_attn, "v_scale")
+        assert hasattr(layer.self_attn, "v_scale")