[WIP] Disentangle auto modules from other modeling files (#13023)

* Initial work

* All auto models

* All tf auto models

* All flax auto models

* Tokenizers

* Add feature extractors

* Fix typos

* Fix other typo

* Use the right config

* Remove old mapping names and update logic in AutoTokenizer

* Update check_table

* Fix copies and check_repo script

* Fix last test

* Add back name

* clean up

* Update template

* Update template

* Forgot a )

* Use alternative to fixup

* Fix TF model template

* Address review comments

* Address review comments

* Style

templates/adding_a_new_model/cookiecutter-template-{{cookiecutter.modelname}}/modeling_tf_{{cookiecutter.lowercase_modelname}}.py

@@ -17,7 +17,7 @@
 {% if cookiecutter.is_encoder_decoder_model == "False" %}
 
 import math
-from typing import Any, Dict, Optional, Tuple, Union
+from typing import Dict, Optional, Tuple, Union
 
 import numpy as np
 import tensorflow as tf
@@ -1484,7 +1484,7 @@ from ...file_utils import (
 )
 from ...modeling_tf_outputs import (
     TFBaseModelOutput,
-    TFBaseModelOutputWithPast,
+    TFBaseModelOutputWithPastAndCrossAttentions,
     TFSeq2SeqLMOutput,
     TFSeq2SeqModelOutput,
 )
@@ -2162,7 +2162,7 @@ class TF{{cookiecutter.camelcase_modelname}}Encoder(tf.keras.layers.Layer):
             )
 
         # encoder layers
-        for encoder_layer in self.layers:
+        for idx, encoder_layer in enumerate(self.layers):
 
             if inputs["output_hidden_states"]:
                 encoder_states = encoder_states + (hidden_states,)

templates/adding_a_new_model/cookiecutter-template-{{cookiecutter.modelname}}/to_replace_{{cookiecutter.lowercase_modelname}}.py

@@ -172,17 +172,12 @@
 # To replace in: "src/transformers/models/auto/configuration_auto.py"
 # Below: "# Add configs here"
 # Replace with:
-        ("{{cookiecutter.lowercase_modelname}}", {{cookiecutter.camelcase_modelname}}Config),
+        ("{{cookiecutter.lowercase_modelname}}", "{{cookiecutter.camelcase_modelname}}Config"),
 # End.
 
 # Below: "# Add archive maps here"
 # Replace with:
-        {{cookiecutter.uppercase_modelname}}_PRETRAINED_CONFIG_ARCHIVE_MAP,
-# End.
-
-# Below: "from ..albert.configuration_albert import ALBERT_PRETRAINED_CONFIG_ARCHIVE_MAP, AlbertConfig",
-# Replace with:
-from ..{{cookiecutter.lowercase_modelname}}.configuration_{{cookiecutter.lowercase_modelname}} import {{cookiecutter.uppercase_modelname}}_PRETRAINED_CONFIG_ARCHIVE_MAP, {{cookiecutter.camelcase_modelname}}Config
+        ("{{cookiecutter.lowercase_modelname}}", "{{cookiecutter.uppercase_modelname}}_PRETRAINED_CONFIG_ARCHIVE_MAP"),
 # End.
 
 # Below: "# Add full (and cased) model names here"
@@ -193,75 +188,47 @@ from ..{{cookiecutter.lowercase_modelname}}.configuration_{{cookiecutter.lowerca
 
 
 # To replace in: "src/transformers/models/auto/modeling_auto.py" if generating PyTorch
-# Below: "from .configuration_auto import ("
-# Replace with:
-    {{cookiecutter.camelcase_modelname}}Config,
-# End.
-
-# Below: "# Add modeling imports here"
-# Replace with:
-{% if cookiecutter.is_encoder_decoder_model == "False" -%}
-from ..{{cookiecutter.lowercase_modelname}}.modeling_{{cookiecutter.lowercase_modelname}} import (
-    {{cookiecutter.camelcase_modelname}}ForMaskedLM,
-    {{cookiecutter.camelcase_modelname}}ForCausalLM,
-    {{cookiecutter.camelcase_modelname}}ForMultipleChoice,
-    {{cookiecutter.camelcase_modelname}}ForQuestionAnswering,
-    {{cookiecutter.camelcase_modelname}}ForSequenceClassification,
-    {{cookiecutter.camelcase_modelname}}ForTokenClassification,
-    {{cookiecutter.camelcase_modelname}}Model,
-)
-{% else -%}
-from ..{{cookiecutter.lowercase_modelname}}.modeling_{{cookiecutter.lowercase_modelname}} import (
-    {{cookiecutter.camelcase_modelname}}ForConditionalGeneration,
-    {{cookiecutter.camelcase_modelname}}ForCausalLM,
-    {{cookiecutter.camelcase_modelname}}ForQuestionAnswering,
-    {{cookiecutter.camelcase_modelname}}ForSequenceClassification,
-    {{cookiecutter.camelcase_modelname}}Model,
-)
-{% endif -%}
-# End.
-
 # Below: "# Base model mapping"
 # Replace with:
-        ({{cookiecutter.camelcase_modelname}}Config, {{cookiecutter.camelcase_modelname}}Model),
+        ("{{cookiecutter.lowercase_modelname}}", "{{cookiecutter.camelcase_modelname}}Model"),
 # End.
 
 # Below: "# Model with LM heads mapping"
 # Replace with:
 {% if cookiecutter.is_encoder_decoder_model == "False" -%}
-        ({{cookiecutter.camelcase_modelname}}Config, {{cookiecutter.camelcase_modelname}}ForMaskedLM),
+        ("{{cookiecutter.lowercase_modelname}}", "{{cookiecutter.camelcase_modelname}}ForMaskedLM"),
 {% else %}
-        ({{cookiecutter.camelcase_modelname}}Config, {{cookiecutter.camelcase_modelname}}ForConditionalGeneration),
+        ("{{cookiecutter.lowercase_modelname}}", "{{cookiecutter.camelcase_modelname}}ForConditionalGeneration"),
 {% endif -%}
 # End.
 
 # Below: "# Model for Causal LM mapping"
 # Replace with:
-        ({{cookiecutter.camelcase_modelname}}Config, {{cookiecutter.camelcase_modelname}}ForCausalLM),
+        ("{{cookiecutter.lowercase_modelname}}", "{{cookiecutter.camelcase_modelname}}ForCausalLM"),
 # End.
 
 # Below: "# Model for Masked LM mapping"
 # Replace with:
 {% if cookiecutter.is_encoder_decoder_model == "False" -%}
-        ({{cookiecutter.camelcase_modelname}}Config, {{cookiecutter.camelcase_modelname}}ForMaskedLM),
+        ("{{cookiecutter.lowercase_modelname}}", "{{cookiecutter.camelcase_modelname}}ForMaskedLM"),
 {% else -%}
 {% endif -%}
 # End.
 
 # Below: "# Model for Sequence Classification mapping"
 # Replace with:
-        ({{cookiecutter.camelcase_modelname}}Config, {{cookiecutter.camelcase_modelname}}ForSequenceClassification),
+        ("{{cookiecutter.lowercase_modelname}}", "{{cookiecutter.camelcase_modelname}}ForSequenceClassification"),
 # End.
 
 # Below: "# Model for Question Answering mapping"
 # Replace with:
-        ({{cookiecutter.camelcase_modelname}}Config, {{cookiecutter.camelcase_modelname}}ForQuestionAnswering),
+        ("{{cookiecutter.lowercase_modelname}}", "{{cookiecutter.camelcase_modelname}}ForQuestionAnswering"),
 # End.
 
 # Below: "# Model for Token Classification mapping"
 # Replace with:
 {% if cookiecutter.is_encoder_decoder_model == "False" -%}
-        ({{cookiecutter.camelcase_modelname}}Config, {{cookiecutter.camelcase_modelname}}ForTokenClassification),
+        ("{{cookiecutter.lowercase_modelname}}", "{{cookiecutter.camelcase_modelname}}ForTokenClassification"),
 {% else -%}
 {% endif -%}
 # End.
@@ -269,7 +236,7 @@ from ..{{cookiecutter.lowercase_modelname}}.modeling_{{cookiecutter.lowercase_mo
 # Below: "# Model for Multiple Choice mapping"
 # Replace with:
 {% if cookiecutter.is_encoder_decoder_model == "False" -%}
-        ({{cookiecutter.camelcase_modelname}}Config, {{cookiecutter.camelcase_modelname}}ForMultipleChoice),
+        ("{{cookiecutter.lowercase_modelname}}", "{{cookiecutter.camelcase_modelname}}ForMultipleChoice"),
 {% else -%}
 {% endif -%}
 # End.
@@ -278,54 +245,29 @@ from ..{{cookiecutter.lowercase_modelname}}.modeling_{{cookiecutter.lowercase_mo
 # Replace with:
 {% if cookiecutter.is_encoder_decoder_model == "False" -%}
 {% else %}
-        ({{cookiecutter.camelcase_modelname}}Config, {{cookiecutter.camelcase_modelname}}ForConditionalGeneration),
+        ("{{cookiecutter.lowercase_modelname}}", "{{cookiecutter.camelcase_modelname}}ForConditionalGeneration"),
 {% endif -%}
 # End.
 
 # To replace in: "src/transformers/models/auto/modeling_tf_auto.py" if generating TensorFlow
-# Below: "from .configuration_auto import ("
-# Replace with:
-    {{cookiecutter.camelcase_modelname}}Config,
-# End.
-
-# Below: "# Add modeling imports here"
-# Replace with:
-{% if cookiecutter.is_encoder_decoder_model == "False" -%}
-from ..{{cookiecutter.lowercase_modelname}}.modeling_tf_{{cookiecutter.lowercase_modelname}} import (
-    TF{{cookiecutter.camelcase_modelname}}ForMaskedLM,
-    TF{{cookiecutter.camelcase_modelname}}ForCausalLM,
-    TF{{cookiecutter.camelcase_modelname}}ForMultipleChoice,
-    TF{{cookiecutter.camelcase_modelname}}ForQuestionAnswering,
-    TF{{cookiecutter.camelcase_modelname}}ForSequenceClassification,
-    TF{{cookiecutter.camelcase_modelname}}ForTokenClassification,
-    TF{{cookiecutter.camelcase_modelname}}Model,
-)
-{% else -%}
-from ..{{cookiecutter.lowercase_modelname}}.modeling_tf_{{cookiecutter.lowercase_modelname}} import (
-    TF{{cookiecutter.camelcase_modelname}}ForConditionalGeneration,
-    TF{{cookiecutter.camelcase_modelname}}Model,
-)
-{% endif -%}
-# End.
-
 # Below: "# Base model mapping"
 # Replace with:
-        ({{cookiecutter.camelcase_modelname}}Config, TF{{cookiecutter.camelcase_modelname}}Model),
+        ("{{cookiecutter.lowercase_modelname}}", "TF{{cookiecutter.camelcase_modelname}}Model"),
 # End.
 
 # Below: "# Model with LM heads mapping"
 # Replace with:
 {% if cookiecutter.is_encoder_decoder_model == "False" -%}
-        ({{cookiecutter.camelcase_modelname}}Config, TF{{cookiecutter.camelcase_modelname}}ForMaskedLM),
+        ("{{cookiecutter.lowercase_modelname}}", "TF{{cookiecutter.camelcase_modelname}}ForMaskedLM"),
 {% else %}
-        ({{cookiecutter.camelcase_modelname}}Config, TF{{cookiecutter.camelcase_modelname}}ForConditionalGeneration),
+        ("{{cookiecutter.lowercase_modelname}}", "TF{{cookiecutter.camelcase_modelname}}ForConditionalGeneration"),
 {% endif -%}
 # End.
 
 # Below: "# Model for Causal LM mapping"
 # Replace with:
 {% if cookiecutter.is_encoder_decoder_model == "False" -%}
-        ({{cookiecutter.camelcase_modelname}}Config, TF{{cookiecutter.camelcase_modelname}}ForCausalLM),
+        ("{{cookiecutter.lowercase_modelname}}", "TF{{cookiecutter.camelcase_modelname}}ForCausalLM"),
 {% else -%}
 {% endif -%}
 # End.
@@ -333,7 +275,7 @@ from ..{{cookiecutter.lowercase_modelname}}.modeling_tf_{{cookiecutter.lowercase
 # Below: "# Model for Masked LM mapping"
 # Replace with:
 {% if cookiecutter.is_encoder_decoder_model == "False" -%}
-        ({{cookiecutter.camelcase_modelname}}Config, TF{{cookiecutter.camelcase_modelname}}ForMaskedLM),
+        ("{{cookiecutter.lowercase_modelname}}", "TF{{cookiecutter.camelcase_modelname}}ForMaskedLM"),
 {% else -%}
 {% endif -%}
 # End.
@@ -341,7 +283,7 @@ from ..{{cookiecutter.lowercase_modelname}}.modeling_tf_{{cookiecutter.lowercase
 # Below: "# Model for Sequence Classification mapping"
 # Replace with:
 {% if cookiecutter.is_encoder_decoder_model == "False" -%}
-        ({{cookiecutter.camelcase_modelname}}Config, TF{{cookiecutter.camelcase_modelname}}ForSequenceClassification),
+        ("{{cookiecutter.lowercase_modelname}}", "TF{{cookiecutter.camelcase_modelname}}ForSequenceClassification"),
 {% else -%}
 {% endif -%}
 # End.
@@ -349,7 +291,7 @@ from ..{{cookiecutter.lowercase_modelname}}.modeling_tf_{{cookiecutter.lowercase
 # Below: "# Model for Question Answering mapping"
 # Replace with:
 {% if cookiecutter.is_encoder_decoder_model == "False" -%}
-        ({{cookiecutter.camelcase_modelname}}Config, TF{{cookiecutter.camelcase_modelname}}ForQuestionAnswering),
+        ("{{cookiecutter.lowercase_modelname}}", "TF{{cookiecutter.camelcase_modelname}}ForQuestionAnswering"),
 {% else -%}
 {% endif -%}
 # End.
@@ -357,7 +299,7 @@ from ..{{cookiecutter.lowercase_modelname}}.modeling_tf_{{cookiecutter.lowercase
 # Below: "# Model for Token Classification mapping"
 # Replace with:
 {% if cookiecutter.is_encoder_decoder_model == "False" -%}
-        ({{cookiecutter.camelcase_modelname}}Config, TF{{cookiecutter.camelcase_modelname}}ForTokenClassification),
+        ("{{cookiecutter.lowercase_modelname}}", "TF{{cookiecutter.camelcase_modelname}}ForTokenClassification"),
 {% else -%}
 {% endif -%}
 # End.
@@ -365,7 +307,7 @@ from ..{{cookiecutter.lowercase_modelname}}.modeling_tf_{{cookiecutter.lowercase
 # Below: "# Model for Multiple Choice mapping"
 # Replace with:
 {% if cookiecutter.is_encoder_decoder_model == "False" -%}
-        ({{cookiecutter.camelcase_modelname}}Config, TF{{cookiecutter.camelcase_modelname}}ForMultipleChoice),
+        ("{{cookiecutter.lowercase_modelname}}", "TF{{cookiecutter.camelcase_modelname}}ForMultipleChoice"),
 {% else -%}
 {% endif -%}
 # End.
@@ -374,7 +316,7 @@ from ..{{cookiecutter.lowercase_modelname}}.modeling_tf_{{cookiecutter.lowercase
 # Replace with:
 {% if cookiecutter.is_encoder_decoder_model == "False" -%}
 {% else %}
-        ({{cookiecutter.camelcase_modelname}}Config, TF{{cookiecutter.camelcase_modelname}}ForConditionalGeneration),
+        ("{{cookiecutter.lowercase_modelname}}", "TF{{cookiecutter.camelcase_modelname}}ForConditionalGeneration"),
 {% endif -%}
 # End.