Remove deprecated

examples/seq2seq/test_finetune_trainer.py

@@ -138,7 +138,7 @@ def _compute_metrics(pred):
             per_device_train_batch_size=batch_size,
             per_device_eval_batch_size=batch_size,
             predict_with_generate=True,
-            evaluate_during_training=True,
+            evaluation_strategy="steps",
             do_train=True,
             do_eval=True,
             warmup_steps=0,
@@ -179,7 +179,7 @@ def run_trainer(self, eval_steps: int, max_len: str, model_name: str, num_train_
             --per_device_eval_batch_size 4
             --learning_rate 3e-3
             --warmup_steps 8
-            --evaluate_during_training
+            --evaluation_strategy steps
             --predict_with_generate
             --logging_steps 0
             --save_steps {str(eval_steps)}

examples/token-classification/run_ner_old.py

@@ -254,7 +254,7 @@ def compute_metrics(p: EvalPrediction) -> Dict:
         trainer.save_model()
         # For convenience, we also re-save the tokenizer to the same directory,
         # so that you can share your model easily on huggingface.co/models =)
-        if trainer.is_world_master():
+        if trainer.is_world_process_zero():
             tokenizer.save_pretrained(training_args.output_dir)
 
     # Evaluation
@@ -265,7 +265,7 @@ def compute_metrics(p: EvalPrediction) -> Dict:
         result = trainer.evaluate()
 
         output_eval_file = os.path.join(training_args.output_dir, "eval_results.txt")
-        if trainer.is_world_master():
+        if trainer.is_world_process_zero():
             with open(output_eval_file, "w") as writer:
                 logger.info("***** Eval results *****")
                 for key, value in result.items():

src/transformers/data/processors/squad.py

@@ -145,11 +145,11 @@ def squad_convert_example_to_features(
     # in the way they compute mask of added tokens.
     tokenizer_type = type(tokenizer).__name__.replace("Tokenizer", "").lower()
     sequence_added_tokens = (
-        tokenizer.max_len - tokenizer.max_len_single_sentence + 1
+        tokenizer.model_max_length - tokenizer.max_len_single_sentence + 1
         if tokenizer_type in MULTI_SEP_TOKENS_TOKENIZERS_SET
-        else tokenizer.max_len - tokenizer.max_len_single_sentence
+        else tokenizer.model_max_length - tokenizer.max_len_single_sentence
     )
-    sequence_pair_added_tokens = tokenizer.max_len - tokenizer.max_len_sentences_pair
+    sequence_pair_added_tokens = tokenizer.model_max_length - tokenizer.max_len_sentences_pair
 
     span_doc_tokens = all_doc_tokens
     while len(spans) * doc_stride < len(all_doc_tokens):

src/transformers/models/albert/modeling_albert.py

@@ -16,7 +16,6 @@
 
 import math
 import os
-import warnings
 from dataclasses import dataclass
 from typing import Optional, Tuple
 
@@ -742,7 +741,6 @@ def forward(
         output_attentions=None,
         output_hidden_states=None,
         return_dict=None,
-        **kwargs,
     ):
         r"""
         labels (``torch.LongTensor`` of shape ``(batch_size, sequence_length)``, `optional`):
@@ -753,8 +751,6 @@ def forward(
             Labels for computing the next sequence prediction (classification) loss. Input should be a sequence pair
             (see :obj:`input_ids` docstring) Indices should be in ``[0, 1]``. ``0`` indicates original order (sequence
             A, then sequence B), ``1`` indicates switched order (sequence B, then sequence A).
-        kwargs (:obj:`Dict[str, any]`, optional, defaults to `{}`):
-            Used to hide legacy arguments that have been deprecated.
 
         Returns:
 
@@ -773,14 +769,6 @@ def forward(
             >>> sop_logits = outputs.sop_logits
 
         """
-
-        if "masked_lm_labels" in kwargs:
-            warnings.warn(
-                "The `masked_lm_labels` argument is deprecated and will be removed in a future version, use `labels` instead.",
-                FutureWarning,
-            )
-            labels = kwargs.pop("masked_lm_labels")
-        assert kwargs == {}, f"Unexpected keyword arguments: {list(kwargs.keys())}."
         return_dict = return_dict if return_dict is not None else self.config.use_return_dict
 
         outputs = self.albert(
@@ -898,23 +886,13 @@ def forward(
         output_attentions=None,
         output_hidden_states=None,
         return_dict=None,
-        **kwargs
     ):
         r"""
         labels (:obj:`torch.LongTensor` of shape :obj:`(batch_size, sequence_length)`, `optional`):
             Labels for computing the masked language modeling loss. Indices should be in ``[-100, 0, ...,
             config.vocab_size]`` (see ``input_ids`` docstring) Tokens with indices set to ``-100`` are ignored
             (masked), the loss is only computed for the tokens with labels in ``[0, ..., config.vocab_size]``
-        kwargs (:obj:`Dict[str, any]`, optional, defaults to `{}`):
-            Used to hide legacy arguments that have been deprecated.
         """
-        if "masked_lm_labels" in kwargs:
-            warnings.warn(
-                "The `masked_lm_labels` argument is deprecated and will be removed in a future version, use `labels` instead.",
-                FutureWarning,
-            )
-            labels = kwargs.pop("masked_lm_labels")
-        assert kwargs == {}, f"Unexpected keyword arguments: {list(kwargs.keys())}."
         return_dict = return_dict if return_dict is not None else self.config.use_return_dict
 
         outputs = self.albert(

src/transformers/models/bart/modeling_bart.py

@@ -15,7 +15,6 @@
 """PyTorch BART model, ported from the fairseq repo."""
 import math
 import random
-import warnings
 from typing import Dict, List, Optional, Tuple
 
 import numpy as np
@@ -529,7 +528,6 @@ def forward(
         output_attentions=False,
         output_hidden_states=False,
         return_dict=True,
-        **unused,
     ):
         """
         Includes several features from "Jointly Learning to Align and Translate with Transformer Models" (Garg et al.,
@@ -551,18 +549,6 @@ def forward(
                 - hidden states
                 - attentions
         """
-        if "decoder_cached_states" in unused:
-            warnings.warn(
-                "The `decoder_cached_states` argument is deprecated and will be removed in a future version, use `past_key_values` instead.",
-                FutureWarning,
-            )
-            past_key_values = unused.pop("decoder_cached_states")
-        if "decoder_past_key_values" in unused:
-            warnings.warn(
-                "The `decoder_past_key_values` argument is deprecated and will be removed in a future version, use `past_key_values` instead.",
-                FutureWarning,
-            )
-            past_key_values = unused.pop("decoder_past_key_values")
 
         # check attention mask and invert
         if encoder_padding_mask is not None:
@@ -873,14 +859,7 @@ def forward(
         output_attentions=None,
         output_hidden_states=None,
         return_dict=None,
-        **kwargs,
     ):
-        if "decoder_past_key_values" in kwargs:
-            warnings.warn(
-                "The `decoder_past_key_values` argument is deprecated and will be removed in a future version, use `past_key_values` instead.",
-                FutureWarning,
-            )
-            past_key_values = kwargs.pop("decoder_past_key_values")
 
         if decoder_input_ids is None:
             use_cache = False
@@ -1006,7 +985,6 @@ def forward(
         output_attentions=None,
         output_hidden_states=None,
         return_dict=None,
-        **unused,
     ):
         r"""
         labels (:obj:`torch.LongTensor` of shape :obj:`(batch_size, sequence_length)`, `optional`):
@@ -1034,24 +1012,6 @@ def forward(
             >>> tokenizer.decode(predictions).split()
             >>> # ['good', 'great', 'all', 'really', 'very']
         """
-        if "lm_labels" in unused:
-            warnings.warn(
-                "The `lm_labels` argument is deprecated and will be removed in a future version, use `labels` instead.",
-                FutureWarning,
-            )
-            labels = unused.pop("lm_labels")
-        if "decoder_cached_states" in unused:
-            warnings.warn(
-                "The `decoder_cached_states` argument is deprecated and will be removed in a future version, use `past_key_values` instead.",
-                FutureWarning,
-            )
-            past_key_values = unused.pop("decoder_cached_states")
-        if "decoder_past_key_values" in unused:
-            warnings.warn(
-                "The `decoder_past_key_values` argument is deprecated and will be removed in a future version, use `past_key_values` instead.",
-                FutureWarning,
-            )
-            past_key_values = unused.pop("decoder_past_key_values")
         return_dict = return_dict if return_dict is not None else self.config.use_return_dict
 
         if labels is not None:

src/transformers/models/bert/modeling_bert.py

@@ -896,7 +896,6 @@ def forward(
         output_attentions=None,
         output_hidden_states=None,
         return_dict=None,
-        **kwargs
     ):
         r"""
         labels (:obj:`torch.LongTensor` of shape ``(batch_size, sequence_length)``, `optional`):
@@ -928,13 +927,6 @@ def forward(
             >>> prediction_logits = outputs.prediction_logits
             >>> seq_relationship_logits = outputs.seq_relationship_logits
         """
-        if "masked_lm_labels" in kwargs:
-            warnings.warn(
-                "The `masked_lm_labels` argument is deprecated and will be removed in a future version, use `labels` instead.",
-                FutureWarning,
-            )
-            labels = kwargs.pop("masked_lm_labels")
-        assert kwargs == {}, f"Unexpected keyword arguments: {list(kwargs.keys())}."
         return_dict = return_dict if return_dict is not None else self.config.use_return_dict
 
         outputs = self.bert(
@@ -1136,24 +1128,13 @@ def forward(
         output_attentions=None,
         output_hidden_states=None,
         return_dict=None,
-        **kwargs
     ):
         r"""
         labels (:obj:`torch.LongTensor` of shape :obj:`(batch_size, sequence_length)`, `optional`):
             Labels for computing the masked language modeling loss. Indices should be in ``[-100, 0, ...,
             config.vocab_size]`` (see ``input_ids`` docstring) Tokens with indices set to ``-100`` are ignored
             (masked), the loss is only computed for the tokens with labels in ``[0, ..., config.vocab_size]``
-        kwargs (:obj:`Dict[str, any]`, optional, defaults to `{}`):
-            Used to hide legacy arguments that have been deprecated.
         """
-        if "masked_lm_labels" in kwargs:
-            warnings.warn(
-                "The `masked_lm_labels` argument is deprecated and will be removed in a future version, use `labels` instead.",
-                FutureWarning,
-            )
-            labels = kwargs.pop("masked_lm_labels")
-        assert "lm_labels" not in kwargs, "Use `BertWithLMHead` for autoregressive language modeling task."
-        assert kwargs == {}, f"Unexpected keyword arguments: {list(kwargs.keys())}."
 
         return_dict = return_dict if return_dict is not None else self.config.use_return_dict
 

src/transformers/models/ctrl/modeling_ctrl.py

@@ -15,9 +15,6 @@
 # limitations under the License.
 """ PyTorch CTRL model."""
 
-
-import warnings
-
 import numpy as np
 import torch
 import torch.nn as nn
@@ -369,15 +366,7 @@ def forward(
         output_attentions=None,
         output_hidden_states=None,
         return_dict=None,
-        **kwargs,
     ):
-        if "past" in kwargs:
-            warnings.warn(
-                "The `past` argument is deprecated and will be removed in a future version, use `past_key_values` instead.",
-                FutureWarning,
-            )
-            past_key_values = kwargs.pop("past")
-        assert kwargs == {}, f"Unexpected keyword arguments: {list(kwargs.keys())}."
 
         output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
         use_cache = use_cache if use_cache is not None else self.config.use_cache
@@ -542,21 +531,13 @@ def forward(
         output_attentions=None,
         output_hidden_states=None,
         return_dict=None,
-        **kwargs,
     ):
         r"""
         labels (:obj:`torch.LongTensor` of shape :obj:`(batch_size, sequence_length)`, `optional`):
             Labels for language modeling. Note that the labels **are shifted** inside the model, i.e. you can set
             ``labels = input_ids`` Indices are selected in ``[-100, 0, ..., config.vocab_size]`` All labels set to
             ``-100`` are ignored (masked), the loss is only computed for labels in ``[0, ..., config.vocab_size]``
         """
-        if "past" in kwargs:
-            warnings.warn(
-                "The `past` argument is deprecated and will be removed in a future version, use `past_key_values` instead.",
-                FutureWarning,
-            )
-            past_key_values = kwargs.pop("past")
-        assert kwargs == {}, f"Unexpected keyword arguments: {list(kwargs.keys())}."
         return_dict = return_dict if return_dict is not None else self.config.use_return_dict
 
         transformer_outputs = self.transformer(

src/transformers/models/distilbert/modeling_distilbert.py

@@ -20,7 +20,6 @@
 
 import copy
 import math
-import warnings
 
 import numpy as np
 import torch
@@ -526,23 +525,13 @@ def forward(
         output_attentions=None,
         output_hidden_states=None,
         return_dict=None,
-        **kwargs
     ):
         r"""
         labels (:obj:`torch.LongTensor` of shape :obj:`(batch_size, sequence_length)`, `optional`):
             Labels for computing the masked language modeling loss. Indices should be in ``[-100, 0, ...,
             config.vocab_size]`` (see ``input_ids`` docstring) Tokens with indices set to ``-100`` are ignored
             (masked), the loss is only computed for the tokens with labels in ``[0, ..., config.vocab_size]``.
-        kwargs (:obj:`Dict[str, any]`, optional, defaults to `{}`):
-            Used to hide legacy arguments that have been deprecated.
         """
-        if "masked_lm_labels" in kwargs:
-            warnings.warn(
-                "The `masked_lm_labels` argument is deprecated and will be removed in a future version, use `labels` instead.",
-                FutureWarning,
-            )
-            labels = kwargs.pop("masked_lm_labels")
-        assert kwargs == {}, f"Unexpected keyword arguments: {list(kwargs.keys())}."
         return_dict = return_dict if return_dict is not None else self.config.use_return_dict
 
         dlbrt_output = self.distilbert(

src/transformers/models/electra/modeling_electra.py

@@ -16,7 +16,6 @@
 
 import math
 import os
-import warnings
 from dataclasses import dataclass
 from typing import Optional, Tuple
 
@@ -1000,23 +999,13 @@ def forward(
         output_attentions=None,
         output_hidden_states=None,
         return_dict=None,
-        **kwargs
     ):
         r"""
         labels (:obj:`torch.LongTensor` of shape :obj:`(batch_size, sequence_length)`, `optional`):
             Labels for computing the masked language modeling loss. Indices should be in ``[-100, 0, ...,
             config.vocab_size]`` (see ``input_ids`` docstring) Tokens with indices set to ``-100`` are ignored
             (masked), the loss is only computed for the tokens with labels in ``[0, ..., config.vocab_size]``
-        kwargs (:obj:`Dict[str, any]`, optional, defaults to `{}`):
-            Used to hide legacy arguments that have been deprecated.
         """
-        if "masked_lm_labels" in kwargs:
-            warnings.warn(
-                "The `masked_lm_labels` argument is deprecated and will be removed in a future version, use `labels` instead.",
-                FutureWarning,
-            )
-            labels = kwargs.pop("masked_lm_labels")
-        assert kwargs == {}, f"Unexpected keyword arguments: {list(kwargs.keys())}."
         return_dict = return_dict if return_dict is not None else self.config.use_return_dict
 
         generator_hidden_states = self.electra(