Remove _supports_dynamic_cache_class attribute after rebase

src/transformers/generation/candidate_generator.py

@@ -120,7 +120,7 @@ def __init__(
  if "past_key_values" in assistant_kwargs.keys():
  if (
  isinstance(assistant_kwargs["past_key_values"], DynamicCache)
- and not self.assistant_model._supports_dynamic_cache_class
+ and not self.assistant_model._supports_cache_class
  ):
  # Cache is empty -> remove it from kwargs
  if len(assistant_kwargs["past_key_values"]) == 0:

src/transformers/generation/utils.py

@@ -1101,7 +1101,7 @@ def _validate_model_kwargs(self, model_kwargs: Dict[str, Any]):
  """Validates model kwargs for generation. Generate argument typos will also be caught here."""
  # If a `Cache` instance is passed, checks whether the model is compatible with it
  past = model_kwargs.get("past_key_values", None)
- if isinstance(past, DynamicCache) and not self._supports_dynamic_cache_class:
+ if isinstance(past, DynamicCache) and not self._supports_cache_class:
  raise ValueError(
  f"{self.__class__.__name__} does not support an instance of `DynamicCache` as `past_key_values`. Please "
  "check the model documentation for supported cache formats."
@@ -1639,7 +1639,7 @@ def generate(
  )
  # Use DynamicCache() instance by default. This will avoid back and forth from legacy format that
  # keeps copying the cache thus using much more memory
- elif generation_config.cache_implementation is None and self._supports_dynamic_cache_class:
+ elif generation_config.cache_implementation is None and self._supports_cache_class:
  past = model_kwargs.get("past_key_values", None)
  if past is None:
  model_kwargs["past_key_values"] = DynamicCache()

src/transformers/modeling_utils.py

@@ -1284,10 +1284,6 @@ class PreTrainedModel(nn.Module, ModuleUtilsMixin, GenerationMixin, PushToHubMix
  _supports_cache_class = False
  _supports_static_cache = False
 
- # Has support for a `DynamicCache` instance as `past_key_values`. Some models support it but not other caches
- # Using it is a big memory gain
- _supports_dynamic_cache_class = False
-
  @property
  def dummy_inputs(self) -> Dict[str, torch.Tensor]:
  """

src/transformers/models/cohere/modeling_cohere.py

@@ -713,7 +713,6 @@ class CoherePreTrainedModel(PreTrainedModel):
  _supports_sdpa = True
  _supports_cache_class = True
  _supports_static_cache = True
- _supports_dynamic_cache_class = True
 
  def _init_weights(self, module):
  std = self.config.initializer_range

src/transformers/models/dbrx/modeling_dbrx.py

@@ -938,7 +938,6 @@ class DbrxPreTrainedModel(PreTrainedModel):
  _supports_sdpa = True
  _supports_cache_class = True
  _supports_static_cache = True
- _supports_dynamic_cache_class = True
 
  def _init_weights(self, module: nn.Module):
  std = self.config.initializer_range

src/transformers/models/gemma/modeling_gemma.py

@@ -699,7 +699,6 @@ class GemmaPreTrainedModel(PreTrainedModel):
  _supports_sdpa = True
  _supports_cache_class = True
  _supports_static_cache = True
- _supports_dynamic_cache_class = True
 
  def _init_weights(self, module):
  std = self.config.initializer_range

src/transformers/models/idefics2/modeling_idefics2.py

@@ -1342,7 +1342,6 @@ class Idefics2PreTrainedModel(PreTrainedModel):
  _skip_keys_device_placement = "past_key_values"
  _supports_flash_attn_2 = True
  _supports_cache_class = True
- _supports_dynamic_cache_class = True
 
  def _init_weights(self, module):
  # important: this ported version of Idefics2 isn't meant for training from scratch - only

src/transformers/models/llama/modeling_llama.py

@@ -768,7 +768,6 @@ class LlamaPreTrainedModel(PreTrainedModel):
  _supports_sdpa = True
  _supports_cache_class = True
  _supports_static_cache = True
- _supports_dynamic_cache_class = True
 
  def _init_weights(self, module):
  std = self.config.initializer_range

src/transformers/models/mistral/modeling_mistral.py

@@ -771,7 +771,6 @@ class MistralPreTrainedModel(PreTrainedModel):
  _supports_sdpa = True
  _supports_cache_class = True
  _supports_static_cache = True
- _supports_dynamic_cache_class = True
 
  def _init_weights(self, module):
  std = self.config.initializer_range

src/transformers/models/mixtral/modeling_mixtral.py

@@ -973,7 +973,6 @@ class MixtralPreTrainedModel(PreTrainedModel):
  _supports_flash_attn_2 = True
  _supports_sdpa = True
  _supports_cache_class = True
- _supports_dynamic_cache_class = True
 
  def _init_weights(self, module):
  std = self.config.initializer_range

src/transformers/models/olmo/modeling_olmo.py

@@ -746,7 +746,6 @@ class OlmoPreTrainedModel(PreTrainedModel):
  _supports_sdpa = True
  _supports_cache_class = True
  _supports_static_cache = True
- _supports_dynamic_cache_class = True
 
  def _init_weights(self, module):
  std = self.config.initializer_range

src/transformers/models/persimmon/modeling_persimmon.py

@@ -463,7 +463,6 @@ class PersimmonPreTrainedModel(PreTrainedModel):
  _no_split_modules = ["PersimmonDecoderLayer"]
  _skip_keys_device_placement = "past_key_values"
  _supports_cache_class = True
- _supports_dynamic_cache_class = True
 
  def _init_weights(self, module):
  std = self.config.initializer_range

src/transformers/models/phi/modeling_phi.py

@@ -830,7 +830,6 @@ class PhiPreTrainedModel(PreTrainedModel):
  _supports_flash_attn_2 = True
  _supports_sdpa = True
  _supports_cache_class = True
- _supports_dynamic_cache_class = True
 
  def _init_weights(self, module):
  std = self.config.initializer_range

src/transformers/models/phi3/modeling_phi3.py

@@ -910,7 +910,6 @@ class Phi3PreTrainedModel(PreTrainedModel):
  _supports_flash_attn_2 = True
  _supports_sdpa = False
  _supports_cache_class = True
- _supports_dynamic_cache_class = True
 
  _version = "0.0.5"
 

src/transformers/models/qwen2/modeling_qwen2.py

@@ -811,7 +811,6 @@ class Qwen2PreTrainedModel(PreTrainedModel):
  _supports_flash_attn_2 = True
  _supports_sdpa = True
  _supports_cache_class = True
- _supports_dynamic_cache_class = True
 
  def _init_weights(self, module):
  std = self.config.initializer_range

src/transformers/models/qwen2_moe/modeling_qwen2_moe.py

@@ -965,7 +965,6 @@ class Qwen2MoePreTrainedModel(PreTrainedModel):
  _supports_flash_attn_2 = True
  _supports_sdpa = True
  _supports_cache_class = True
- _supports_dynamic_cache_class = True
 
  def _init_weights(self, module):
  std = self.config.initializer_range

src/transformers/models/stablelm/modeling_stablelm.py

@@ -831,7 +831,6 @@ class StableLmPreTrainedModel(PreTrainedModel):
  _supports_flash_attn_2 = True
  _supports_cache_class = True
  _supports_sdpa = True
- _supports_dynamic_cache_class = True
 
  def _init_weights(self, module):
  std = self.config.initializer_range

src/transformers/models/starcoder2/modeling_starcoder2.py

@@ -790,7 +790,6 @@ class Starcoder2PreTrainedModel(PreTrainedModel):
  _supports_flash_attn_2 = True
  _supports_sdpa = True
  _supports_cache_class = True
- _supports_dynamic_cache_class = True
 
  def _init_weights(self, module):
  std = self.config.initializer_range