[deprecations] Remove generate-related deprecations up to v4.56

setup.py

@@ -79,14 +79,12 @@
 stale_egg_info = Path(__file__).parent / "transformers.egg-info"
 if stale_egg_info.exists():
     print(
-        (
-            "Warning: {} exists.\n\n"
-            "If you recently updated transformers to 3.0 or later, this is expected,\n"
-            "but it may prevent transformers from installing in editable mode.\n\n"
-            "This directory is automatically generated by Python's packaging tools.\n"
-            "I will remove it now.\n\n"
-            "See https://github.com/pypa/pip/issues/5466 for details.\n"
-        ).format(stale_egg_info)
+        f"Warning: {stale_egg_info} exists.\n\n"
+        "If you recently updated transformers to 3.0 or later, this is expected,\n"
+        "but it may prevent transformers from installing in editable mode.\n\n"
+        "This directory is automatically generated by Python's packaging tools.\n"
+        "I will remove it now.\n\n"
+        "See https://github.com/pypa/pip/issues/5466 for details.\n"
     )
     shutil.rmtree(stale_egg_info)
 

src/transformers/generation/configuration_utils.py

@@ -564,12 +564,6 @@ def validate(self, strict=False):
             )
         # 1.4. Watermarking attributes
         if self.watermarking_config is not None:
-            if not (isinstance(self.watermarking_config, (WatermarkingConfig, SynthIDTextWatermarkingConfig))):
-                minor_issues["watermarking_config"] = (
-                    "`watermarking_config` as a dict is deprecated and will be removed in v4.54.0. Please construct "
-                    "`watermarking_config` object with `WatermarkingConfig` or `SynthIDTextWatermarkingConfig` class."
-                )
-                self.watermarking_config = WatermarkingConfig.from_dict(self.watermarking_config)
             self.watermarking_config.validate()
 
         # 2. Validation of attribute combinations

src/transformers/models/bloom/modeling_bloom.py

@@ -545,10 +545,6 @@ def forward(
         if inputs_embeds is None:
             inputs_embeds = self.word_embeddings(input_ids)
 
-        # TODO (joao): remove this exception in v4.56 -- it exists for users that try to pass a legacy cache
-        if not isinstance(past_key_values, (type(None), Cache)):
-            raise ValueError("The `past_key_values` should be either a `Cache` object or `None`.")
-
         if use_cache and past_key_values is None:
             past_key_values = DynamicCache(config=self.config)
 

src/transformers/models/codegen/modeling_codegen.py

@@ -376,10 +376,6 @@ def forward(
         if inputs_embeds is None:
             inputs_embeds = self.wte(input_ids)
 
-        # TODO (joao): remove this exception in v4.56 -- it exists for users that try to pass a legacy cache
-        if not isinstance(past_key_values, (type(None), Cache)):
-            raise ValueError("The `past_key_values` should be either a `Cache` object or `None`.")
-
         if use_cache and past_key_values is None:
             past_key_values = DynamicCache(config=self.config)
 

src/transformers/models/cohere2/configuration_cohere2.py

@@ -19,8 +19,6 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
-import warnings
-
 from ...configuration_utils import PretrainedConfig, layer_type_validation
 from ...modeling_rope_utils import rope_config_validation
 
@@ -230,17 +228,5 @@ def __init__(
             ]
         layer_type_validation(self.layer_types)
 
-    @property
-    def sliding_window_pattern(self):
-        warnings.warn(
-            "The `sliding_window_pattern` attribute is deprecated and will be removed in v4.55.0.",
-            FutureWarning,
-        )
-        return self._sliding_window_pattern
-
-    @sliding_window_pattern.setter
-    def sliding_window_pattern(self, value):
-        self._sliding_window_pattern = value
-
 
 __all__ = ["Cohere2Config"]

src/transformers/models/cohere2/modular_cohere2.py

@@ -13,7 +13,6 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
-import warnings
 from typing import Callable, Optional
 
 import torch
@@ -250,18 +249,6 @@ def __init__(
             ]
         layer_type_validation(self.layer_types)
 
-    @property
-    def sliding_window_pattern(self):
-        warnings.warn(
-            "The `sliding_window_pattern` attribute is deprecated and will be removed in v4.55.0.",
-            FutureWarning,
-        )
-        return self._sliding_window_pattern
-
-    @sliding_window_pattern.setter
-    def sliding_window_pattern(self, value):
-        self._sliding_window_pattern = value
-
 
 class Cohere2RotaryEmbedding(CohereRotaryEmbedding):
     pass

src/transformers/models/dbrx/modeling_dbrx.py

@@ -910,10 +910,6 @@ def forward(
 
         inputs_embeds = nn.functional.dropout(inputs_embeds, p=self.emb_pdrop, training=self.training)
 
-        # TODO (joao): remove this exception in v4.56 -- it exists for users that try to pass a legacy cache
-        if not isinstance(past_key_values, (type(None), Cache)):
-            raise ValueError("The `past_key_values` should be either a `Cache` object or `None`.")
-
         if use_cache and past_key_values is None:
             past_key_values = DynamicCache(config=self.config)
 

src/transformers/models/decision_transformer/modeling_decision_transformer.py

@@ -570,15 +570,6 @@ def forward(
         if use_cache:
             if past_key_values is None:
                 past_key_values = DynamicCache(config=self.config)
-            elif isinstance(past_key_values, tuple):
-                logger.warning_once(
-                    "Passing a tuple of `past_key_values` is deprecated and will be removed in Transformers v4.53.0. "
-                    "You should pass an instance of `Cache` instead, e.g. "
-                    "`past_key_values=DynamicCache.from_legacy_cache(past_key_values)`."
-                )
-                past_key_values = DynamicCache.from_legacy_cache(past_key_values)
-            elif past_key_values is None:
-                past_key_values = DynamicCache(config=self.config)
 
             if self.config.add_cross_attention and not isinstance(past_key_values, EncoderDecoderCache):
                 past_key_values = EncoderDecoderCache(past_key_values, DynamicCache(config=self.config))

src/transformers/models/falcon/modeling_falcon.py

@@ -760,10 +760,6 @@ def forward(
         if inputs_embeds is None:
             inputs_embeds = self.word_embeddings(input_ids)
 
-        # TODO (joao): remove this exception in v4.56 -- it exists for users that try to pass a legacy cache
-        if not isinstance(past_key_values, (type(None), Cache)):
-            raise ValueError("The `past_key_values` should be either a `Cache` object or `None`.")
-
         if use_cache and past_key_values is None:
             past_key_values = DynamicCache(config=self.config)
 

src/transformers/models/gemma3/configuration_gemma3.py

@@ -19,7 +19,6 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
-import warnings
 from typing import Any, Optional, Union
 
 from ...configuration_utils import PretrainedConfig, layer_type_validation
@@ -241,18 +240,6 @@ def __init__(
             ]
         layer_type_validation(self.layer_types)
 
-    @property
-    def sliding_window_pattern(self):
-        warnings.warn(
-            "The `sliding_window_pattern` attribute is deprecated and will be removed in v4.55.0.",
-            FutureWarning,
-        )
-        return self._sliding_window_pattern
-
-    @sliding_window_pattern.setter
-    def sliding_window_pattern(self, value):
-        self._sliding_window_pattern = value
-
 
 class Gemma3Config(PretrainedConfig):
     r"""

src/transformers/models/gemma3/modular_gemma3.py

@@ -14,7 +14,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 import copy
-import warnings
 from collections.abc import Callable
 from typing import Any, Optional, Union
 
@@ -252,18 +251,6 @@ def __init__(
             ]
         layer_type_validation(self.layer_types)
 
-    @property
-    def sliding_window_pattern(self):
-        warnings.warn(
-            "The `sliding_window_pattern` attribute is deprecated and will be removed in v4.55.0.",
-            FutureWarning,
-        )
-        return self._sliding_window_pattern
-
-    @sliding_window_pattern.setter
-    def sliding_window_pattern(self, value):
-        self._sliding_window_pattern = value
-
 
 class Gemma3Config(PretrainedConfig):
     r"""

src/transformers/models/git/modeling_git.py

@@ -405,10 +405,6 @@ def forward(
                 )
                 use_cache = False
 
-        # TODO (joao): remove this exception in v4.56 -- it exists for users that try to pass a legacy cache
-        if not isinstance(past_key_values, (type(None), Cache)):
-            raise ValueError("The `past_key_values` should be either a `Cache` object or `None`.")
-
         if use_cache and past_key_values is None:
             past_key_values = DynamicCache(config=self.config)
 

src/transformers/models/glm4v/modeling_glm4v.py

@@ -38,14 +38,11 @@
 from ...modeling_rope_utils import ROPE_INIT_FUNCTIONS, dynamic_rope_update
 from ...modeling_utils import ALL_ATTENTION_FUNCTIONS, PreTrainedModel
 from ...processing_utils import Unpack
-from ...utils import TransformersKwargs, auto_docstring, can_return_tuple, is_torchdynamo_compiling, logging
+from ...utils import TransformersKwargs, auto_docstring, can_return_tuple, is_torchdynamo_compiling
 from ...utils.generic import check_model_inputs
 from .configuration_glm4v import Glm4vConfig, Glm4vTextConfig, Glm4vVisionConfig
 
 
-logger = logging.get_logger(__name__)
-
-
 @use_kernel_forward_from_hub("RMSNorm")
 class Glm4vRMSNorm(nn.Module):
     def __init__(self, hidden_size, eps=1e-6):
@@ -304,18 +301,7 @@ def forward(
         query_states, key_states, value_states = (
             self.qkv(hidden_states).reshape(seq_length, 3, self.num_heads, -1).permute(1, 0, 2, 3).unbind(0)
         )
-        if position_embeddings is None:
-            logger.warning_once(
-                "The attention layers in this model are transitioning from computing the RoPE embeddings internally "
-                "through `rotary_pos_emb` (2D tensor of RoPE theta values), to using externally computed "
-                "`position_embeddings` (Tuple of tensors, containing cos and sin). In v4.54 `rotary_pos_emb` will be "
-                "removed and `position_embeddings` will be mandatory."
-            )
-            emb = torch.cat((rotary_pos_emb, rotary_pos_emb), dim=-1)
-            cos = emb.cos()
-            sin = emb.sin()
-        else:
-            cos, sin = position_embeddings
+        cos, sin = position_embeddings
         query_states, key_states = apply_rotary_pos_emb_vision(query_states, key_states, cos, sin)
 
         query_states = query_states.transpose(0, 1).unsqueeze(0)

src/transformers/models/glm4v_moe/modeling_glm4v_moe.py

@@ -38,15 +38,12 @@
 from ...modeling_rope_utils import ROPE_INIT_FUNCTIONS, dynamic_rope_update
 from ...modeling_utils import ALL_ATTENTION_FUNCTIONS, PreTrainedModel
 from ...processing_utils import Unpack
-from ...utils import TransformersKwargs, auto_docstring, can_return_tuple, is_torchdynamo_compiling, logging
+from ...utils import TransformersKwargs, auto_docstring, can_return_tuple, is_torchdynamo_compiling
 from ...utils.deprecation import deprecate_kwarg
 from ...utils.generic import check_model_inputs
 from .configuration_glm4v_moe import Glm4vMoeConfig, Glm4vMoeTextConfig, Glm4vMoeVisionConfig
 
 
-logger = logging.get_logger(__name__)
-
-
 @use_kernel_forward_from_hub("RMSNorm")
 class Glm4vMoeRMSNorm(nn.Module):
     def __init__(self, hidden_size, eps=1e-6):
@@ -648,18 +645,7 @@ def forward(
         query_states, key_states, value_states = (
             self.qkv(hidden_states).reshape(seq_length, 3, self.num_heads, -1).permute(1, 0, 2, 3).unbind(0)
         )
-        if position_embeddings is None:
-            logger.warning_once(
-                "The attention layers in this model are transitioning from computing the RoPE embeddings internally "
-                "through `rotary_pos_emb` (2D tensor of RoPE theta values), to using externally computed "
-                "`position_embeddings` (Tuple of tensors, containing cos and sin). In v4.54 `rotary_pos_emb` will be "
-                "removed and `position_embeddings` will be mandatory."
-            )
-            emb = torch.cat((rotary_pos_emb, rotary_pos_emb), dim=-1)
-            cos = emb.cos()
-            sin = emb.sin()
-        else:
-            cos, sin = position_embeddings
+        cos, sin = position_embeddings
         query_states, key_states = apply_rotary_pos_emb_vision(query_states, key_states, cos, sin)
 
         query_states = query_states.transpose(0, 1).unsqueeze(0)

src/transformers/models/gpt_neo/modeling_gpt_neo.py

@@ -571,10 +571,6 @@ def forward(
         if inputs_embeds is None:
             inputs_embeds = self.wte(input_ids)
 
-        # TODO (joao): remove this exception in v4.56 -- it exists for users that try to pass a legacy cache
-        if not isinstance(past_key_values, (type(None), Cache)):
-            raise ValueError("The `past_key_values` should be either a `Cache` object or `None`.")
-
         if use_cache and past_key_values is None:
             past_key_values = DynamicCache(config=self.config)
 

src/transformers/models/gpt_neox/modeling_gpt_neox.py

@@ -428,10 +428,6 @@ def forward(
         if inputs_embeds is None:
             inputs_embeds = self.embed_in(input_ids)
 
-        # TODO (joao): remove this exception in v4.56 -- it exists for users that try to pass a legacy cache
-        if not isinstance(past_key_values, (type(None), Cache)):
-            raise ValueError("The `past_key_values` should be either a `Cache` object or `None`.")
-
         if use_cache and past_key_values is None:
             past_key_values = DynamicCache(config=self.config)
 

src/transformers/models/gpt_neox/modular_gpt_neox.py

@@ -306,10 +306,6 @@ def forward(
         if inputs_embeds is None:
             inputs_embeds = self.embed_in(input_ids)
 
-        # TODO (joao): remove this exception in v4.56 -- it exists for users that try to pass a legacy cache
-        if not isinstance(past_key_values, (type(None), Cache)):
-            raise ValueError("The `past_key_values` should be either a `Cache` object or `None`.")
-
         if use_cache and past_key_values is None:
             past_key_values = DynamicCache(config=self.config)
 

src/transformers/models/gpt_neox_japanese/modeling_gpt_neox_japanese.py

@@ -450,10 +450,6 @@ def forward(
         if inputs_embeds is None:
             inputs_embeds = self.embed_in(input_ids)
 
-        # TODO (joao): remove this exception in v4.56 -- it exists for users that try to pass a legacy cache
-        if not isinstance(past_key_values, (type(None), Cache)):
-            raise ValueError("The `past_key_values` should be either a `Cache` object or `None`.")
-
         if use_cache and past_key_values is None:
             past_key_values = DynamicCache(config=self.config)
 

src/transformers/models/gptj/modeling_gptj.py

@@ -654,10 +654,6 @@ def forward(
         if inputs_embeds is None:
             inputs_embeds = self.wte(input_ids)
 
-        # TODO (joao): remove this exception in v4.56 -- it exists for users that try to pass a legacy cache
-        if not isinstance(past_key_values, (type(None), Cache)):
-            raise ValueError("The `past_key_values` should be either a `Cache` object or `None`.")
-
         if use_cache and past_key_values is None:
             past_key_values = DynamicCache(config=self.config)
 

src/transformers/models/granitemoe/modeling_granitemoe.py

@@ -673,10 +673,6 @@ def forward(
 
         inputs_embeds = inputs_embeds * self.embedding_multiplier
 
-        # TODO (joao): remove this exception in v4.56 -- it exists for users that try to pass a legacy cache
-        if not isinstance(past_key_values, (type(None), Cache)):
-            raise ValueError("The `past_key_values` should be either a `Cache` object or `None`.")
-
         if use_cache and past_key_values is None:
             past_key_values = DynamicCache(config=self.config)
 

src/transformers/models/granitemoeshared/modeling_granitemoeshared.py

@@ -649,10 +649,6 @@ def forward(
 
         inputs_embeds = inputs_embeds * self.embedding_multiplier
 
-        # TODO (joao): remove this exception in v4.56 -- it exists for users that try to pass a legacy cache
-        if not isinstance(past_key_values, (type(None), Cache)):
-            raise ValueError("The `past_key_values` should be either a `Cache` object or `None`.")
-
         if use_cache and past_key_values is None:
             past_key_values = DynamicCache(config=self.config)
 

src/transformers/models/idefics/modeling_idefics.py

@@ -1048,10 +1048,6 @@ def forward(
         if inputs_embeds is None:
             inputs_embeds = self.embed_tokens(input_ids)
 
-        # TODO (joao): remove this exception in v4.56 -- it exists for users that try to pass a legacy cache
-        if not isinstance(past_key_values, (type(None), Cache)):
-            raise ValueError("The `past_key_values` should be either a `Cache` object or `None`.")
-
         if use_cache and past_key_values is None:
             past_key_values = DynamicCache(config=self.config)
 

src/transformers/models/idefics2/modeling_idefics2.py

@@ -1053,10 +1053,6 @@ def forward(
         else:
             raise ValueError("You have to specify either input_ids or inputs_embeds")
 
-        # TODO (joao): remove this exception in v4.56 -- it exists for users that try to pass a legacy cache
-        if not isinstance(past_key_values, (type(None), Cache)):
-            raise ValueError("The `past_key_values` should be either a `Cache` object or `None`.")
-
         if use_cache and past_key_values is None:
             past_key_values = DynamicCache(config=self.config)
 

src/transformers/models/jetmoe/modeling_jetmoe.py

@@ -921,10 +921,6 @@ def forward(
         if inputs_embeds is None:
             inputs_embeds = self.embed_tokens(input_ids)
 
-        # TODO (joao): remove this exception in v4.56 -- it exists for users that try to pass a legacy cache
-        if not isinstance(past_key_values, (type(None), Cache)):
-            raise ValueError("The `past_key_values` should be either a `Cache` object or `None`.")
-
         if use_cache and past_key_values is None:
             past_key_values = DynamicCache(config=self.config)