Remove static pretrained maps from the library's internals

examples/research_projects/bertabs/modeling_bertabs.py

@@ -33,10 +33,6 @@
 
 MAX_SIZE = 5000
 
-BERTABS_FINETUNED_MODEL_ARCHIVE_LIST = [
-    "remi/bertabs-finetuned-cnndm-extractive-abstractive-summarization",
-]
-
 
 class BertAbsPreTrainedModel(PreTrainedModel):
     config_class = BertAbsConfig

src/transformers/__init__.py

@@ -726,7 +726,7 @@
         "ProphetNetTokenizer",
     ],
     "models.pvt": ["PVT_PRETRAINED_CONFIG_ARCHIVE_MAP", "PvtConfig"],
-    "models.pvt_v2": ["PVT_V2_PRETRAINED_CONFIG_ARCHIVE_MAP", "PvtV2Config"],
+    "models.pvt_v2": ["PvtV2Config"],
     "models.qdqbert": ["QDQBERT_PRETRAINED_CONFIG_ARCHIVE_MAP", "QDQBertConfig"],
     "models.qwen2": [
         "QWEN2_PRETRAINED_CONFIG_ARCHIVE_MAP",
@@ -3050,7 +3050,6 @@
     )
     _import_structure["models.pvt_v2"].extend(
         [
-            "PVT_V2_PRETRAINED_MODEL_ARCHIVE_LIST",
             "PvtV2Backbone",
             "PvtV2ForImageClassification",
             "PvtV2Model",
@@ -5602,7 +5601,7 @@
         ProphetNetTokenizer,
     )
     from .models.pvt import PVT_PRETRAINED_CONFIG_ARCHIVE_MAP, PvtConfig
-    from .models.pvt_v2 import PVT_V2_PRETRAINED_CONFIG_ARCHIVE_MAP, PvtV2Config
+    from .models.pvt_v2 import PvtV2Config
     from .models.qdqbert import QDQBERT_PRETRAINED_CONFIG_ARCHIVE_MAP, QDQBertConfig
     from .models.qwen2 import QWEN2_PRETRAINED_CONFIG_ARCHIVE_MAP, Qwen2Config, Qwen2Tokenizer
     from .models.rag import RagConfig, RagRetriever, RagTokenizer
@@ -7623,7 +7622,6 @@
             PvtPreTrainedModel,
         )
         from .models.pvt_v2 import (
-            PVT_V2_PRETRAINED_MODEL_ARCHIVE_LIST,
             PvtV2Backbone,
             PvtV2ForImageClassification,
             PvtV2Model,

src/transformers/commands/add_new_model_like.py

@@ -527,35 +527,6 @@ def duplicate_module(
     # Loop and treat all objects
     new_objects = []
     for obj in objects:
-        # Special cases
-        if "PRETRAINED_CONFIG_ARCHIVE_MAP = {" in obj:
-            # docstyle-ignore
-            obj = (
-                f"{new_model_patterns.model_upper_cased}_PRETRAINED_CONFIG_ARCHIVE_MAP = "
-                + "{"
-                + f"""
-    "{new_model_patterns.checkpoint}": "https://huggingface.co/{new_model_patterns.checkpoint}/resolve/main/config.json",
-"""
-                + "}\n"
-            )
-            new_objects.append(obj)
-            continue
-        elif "PRETRAINED_MODEL_ARCHIVE_LIST = [" in obj:
-            if obj.startswith("TF_"):
-                prefix = "TF_"
-            elif obj.startswith("FLAX_"):
-                prefix = "FLAX_"
-            else:
-                prefix = ""
-            # docstyle-ignore
-            obj = f"""{prefix}{new_model_patterns.model_upper_cased}_PRETRAINED_MODEL_ARCHIVE_LIST = [
-    "{new_model_patterns.checkpoint}",
-    # See all {new_model_patterns.model_name} models at https://huggingface.co/models?filter={new_model_patterns.model_type}
-]
-"""
-            new_objects.append(obj)
-            continue
-
         special_pattern = False
         for pattern, attr in SPECIAL_PATTERNS.items():
             if pattern in obj:
@@ -785,7 +756,6 @@ def retrieve_info_for_model(model_type, frameworks: Optional[List[str]] = None):
 
     model_name = auto_module.MODEL_NAMES_MAPPING[model_type]
     config_class = auto_module.configuration_auto.CONFIG_MAPPING_NAMES[model_type]
-    archive_map = auto_module.configuration_auto.CONFIG_ARCHIVE_MAP_MAPPING_NAMES.get(model_type, None)
     if model_type in auto_module.tokenization_auto.TOKENIZER_MAPPING_NAMES:
         tokenizer_classes = auto_module.tokenization_auto.TOKENIZER_MAPPING_NAMES[model_type]
         tokenizer_class = tokenizer_classes[0] if tokenizer_classes[0] is not None else tokenizer_classes[1]
@@ -814,19 +784,7 @@ def retrieve_info_for_model(model_type, frameworks: Optional[List[str]] = None):
 
     model_classes = retrieve_model_classes(model_type, frameworks=frameworks)
 
-    # Retrieve model upper-cased name from the constant name of the pretrained archive map.
-    if archive_map is None:
-        model_upper_cased = model_camel_cased.upper()
-    else:
-        parts = archive_map.split("_")
-        idx = 0
-        while idx < len(parts) and parts[idx] != "PRETRAINED":
-            idx += 1
-        if idx < len(parts):
-            model_upper_cased = "_".join(parts[:idx])
-        else:
-            model_upper_cased = model_camel_cased.upper()
-
+    model_upper_cased = model_camel_cased.upper()
     model_patterns = ModelPatterns(
         model_name,
         checkpoint=find_base_model_checkpoint(model_type, model_files=model_files),
@@ -1135,14 +1093,6 @@ def add_model_to_auto_classes(
             for attr in ["model_type", "model_name"]:
                 old_model_line = old_model_line.replace("{" + attr + "}", getattr(old_model_patterns, attr))
                 new_model_line = new_model_line.replace("{" + attr + "}", getattr(new_model_patterns, attr))
-            if "pretrained_archive_map" in pattern:
-                old_model_line = old_model_line.replace(
-                    "{pretrained_archive_map}", f"{old_model_patterns.model_upper_cased}_PRETRAINED_CONFIG_ARCHIVE_MAP"
-                )
-                new_model_line = new_model_line.replace(
-                    "{pretrained_archive_map}", f"{new_model_patterns.model_upper_cased}_PRETRAINED_CONFIG_ARCHIVE_MAP"
-                )
-
             new_model_line = new_model_line.replace(
                 old_model_patterns.model_camel_cased, new_model_patterns.model_camel_cased
             )

src/transformers/convert_pytorch_checkpoint_to_tf2.py

@@ -19,28 +19,6 @@
 import os
 
 from . import (
-    ALBERT_PRETRAINED_CONFIG_ARCHIVE_MAP,
-    BART_PRETRAINED_MODEL_ARCHIVE_LIST,
-    BERT_PRETRAINED_CONFIG_ARCHIVE_MAP,
-    CAMEMBERT_PRETRAINED_CONFIG_ARCHIVE_MAP,
-    CTRL_PRETRAINED_CONFIG_ARCHIVE_MAP,
-    DISTILBERT_PRETRAINED_CONFIG_ARCHIVE_MAP,
-    DPR_CONTEXT_ENCODER_PRETRAINED_MODEL_ARCHIVE_LIST,
-    DPR_QUESTION_ENCODER_PRETRAINED_MODEL_ARCHIVE_LIST,
-    DPR_READER_PRETRAINED_MODEL_ARCHIVE_LIST,
-    ELECTRA_PRETRAINED_CONFIG_ARCHIVE_MAP,
-    FLAUBERT_PRETRAINED_CONFIG_ARCHIVE_MAP,
-    GPT2_PRETRAINED_CONFIG_ARCHIVE_MAP,
-    LAYOUTLM_PRETRAINED_MODEL_ARCHIVE_LIST,
-    LXMERT_PRETRAINED_CONFIG_ARCHIVE_MAP,
-    OPENAI_GPT_PRETRAINED_CONFIG_ARCHIVE_MAP,
-    ROBERTA_PRETRAINED_CONFIG_ARCHIVE_MAP,
-    T5_PRETRAINED_CONFIG_ARCHIVE_MAP,
-    TRANSFO_XL_PRETRAINED_CONFIG_ARCHIVE_MAP,
-    WAV_2_VEC_2_PRETRAINED_CONFIG_ARCHIVE_MAP,
-    XLM_PRETRAINED_CONFIG_ARCHIVE_MAP,
-    XLM_ROBERTA_PRETRAINED_CONFIG_ARCHIVE_MAP,
-    XLNET_PRETRAINED_CONFIG_ARCHIVE_MAP,
     AlbertConfig,
     BartConfig,
     BertConfig,
@@ -140,31 +118,26 @@
         TFBartForConditionalGeneration,
         TFBartForSequenceClassification,
         BartForConditionalGeneration,
-        BART_PRETRAINED_MODEL_ARCHIVE_LIST,
     ),
     "bert": (
         BertConfig,
         TFBertForPreTraining,
         BertForPreTraining,
-        BERT_PRETRAINED_CONFIG_ARCHIVE_MAP,
     ),
     "google-bert/bert-large-uncased-whole-word-masking-finetuned-squad": (
         BertConfig,
         TFBertForQuestionAnswering,
         BertForQuestionAnswering,
-        BERT_PRETRAINED_CONFIG_ARCHIVE_MAP,
     ),
     "google-bert/bert-large-cased-whole-word-masking-finetuned-squad": (
         BertConfig,
         TFBertForQuestionAnswering,
         BertForQuestionAnswering,
-        BERT_PRETRAINED_CONFIG_ARCHIVE_MAP,
     ),
     "google-bert/bert-base-cased-finetuned-mrpc": (
         BertConfig,
         TFBertForSequenceClassification,
         BertForSequenceClassification,
-        BERT_PRETRAINED_CONFIG_ARCHIVE_MAP,
     ),
     "dpr": (
         DPRConfig,
@@ -174,130 +147,107 @@
         DPRQuestionEncoder,
         DPRContextEncoder,
         DPRReader,
-        DPR_CONTEXT_ENCODER_PRETRAINED_MODEL_ARCHIVE_LIST,
-        DPR_QUESTION_ENCODER_PRETRAINED_MODEL_ARCHIVE_LIST,
-        DPR_READER_PRETRAINED_MODEL_ARCHIVE_LIST,
     ),
     "openai-community/gpt2": (
         GPT2Config,
         TFGPT2LMHeadModel,
         GPT2LMHeadModel,
-        GPT2_PRETRAINED_CONFIG_ARCHIVE_MAP,
     ),
     "xlnet": (
         XLNetConfig,
         TFXLNetLMHeadModel,
         XLNetLMHeadModel,
-        XLNET_PRETRAINED_CONFIG_ARCHIVE_MAP,
     ),
     "xlm": (
         XLMConfig,
         TFXLMWithLMHeadModel,
         XLMWithLMHeadModel,
-        XLM_PRETRAINED_CONFIG_ARCHIVE_MAP,
     ),
     "xlm-roberta": (
         XLMRobertaConfig,
         TFXLMRobertaForMaskedLM,
         XLMRobertaForMaskedLM,
-        XLM_ROBERTA_PRETRAINED_CONFIG_ARCHIVE_MAP,
     ),
     "transfo-xl": (
         TransfoXLConfig,
         TFTransfoXLLMHeadModel,
         TransfoXLLMHeadModel,
-        TRANSFO_XL_PRETRAINED_CONFIG_ARCHIVE_MAP,
     ),
     "openai-community/openai-gpt": (
         OpenAIGPTConfig,
         TFOpenAIGPTLMHeadModel,
         OpenAIGPTLMHeadModel,
-        OPENAI_GPT_PRETRAINED_CONFIG_ARCHIVE_MAP,
     ),
     "roberta": (
         RobertaConfig,
         TFRobertaForCausalLM,
         TFRobertaForMaskedLM,
         RobertaForMaskedLM,
-        ROBERTA_PRETRAINED_CONFIG_ARCHIVE_MAP,
     ),
     "layoutlm": (
         LayoutLMConfig,
         TFLayoutLMForMaskedLM,
         LayoutLMForMaskedLM,
-        LAYOUTLM_PRETRAINED_MODEL_ARCHIVE_LIST,
     ),
     "FacebookAI/roberta-large-mnli": (
         RobertaConfig,
         TFRobertaForSequenceClassification,
         RobertaForSequenceClassification,
-        ROBERTA_PRETRAINED_CONFIG_ARCHIVE_MAP,
     ),
     "camembert": (
         CamembertConfig,
         TFCamembertForMaskedLM,
         CamembertForMaskedLM,
-        CAMEMBERT_PRETRAINED_CONFIG_ARCHIVE_MAP,
     ),
     "flaubert": (
         FlaubertConfig,
         TFFlaubertWithLMHeadModel,
         FlaubertWithLMHeadModel,
-        FLAUBERT_PRETRAINED_CONFIG_ARCHIVE_MAP,
     ),
     "distilbert": (
         DistilBertConfig,
         TFDistilBertForMaskedLM,
         DistilBertForMaskedLM,
-        DISTILBERT_PRETRAINED_CONFIG_ARCHIVE_MAP,
     ),
     "distilbert-base-distilled-squad": (
         DistilBertConfig,
         TFDistilBertForQuestionAnswering,
         DistilBertForQuestionAnswering,
-        DISTILBERT_PRETRAINED_CONFIG_ARCHIVE_MAP,
     ),
     "lxmert": (
         LxmertConfig,
         TFLxmertForPreTraining,
         LxmertForPreTraining,
-        LXMERT_PRETRAINED_CONFIG_ARCHIVE_MAP,
     ),
     "lxmert-visual-feature-encoder": (
         LxmertConfig,
         TFLxmertVisualFeatureEncoder,
         LxmertVisualFeatureEncoder,
-        LXMERT_PRETRAINED_CONFIG_ARCHIVE_MAP,
     ),
     "Salesforce/ctrl": (
         CTRLConfig,
         TFCTRLLMHeadModel,
         CTRLLMHeadModel,
-        CTRL_PRETRAINED_CONFIG_ARCHIVE_MAP,
     ),
     "albert": (
         AlbertConfig,
         TFAlbertForPreTraining,
         AlbertForPreTraining,
-        ALBERT_PRETRAINED_CONFIG_ARCHIVE_MAP,
     ),
     "t5": (
         T5Config,
         TFT5ForConditionalGeneration,
         T5ForConditionalGeneration,
-        T5_PRETRAINED_CONFIG_ARCHIVE_MAP,
     ),
     "electra": (
         ElectraConfig,
         TFElectraForPreTraining,
         ElectraForPreTraining,
-        ELECTRA_PRETRAINED_CONFIG_ARCHIVE_MAP,
     ),
     "wav2vec2": (
         Wav2Vec2Config,
         TFWav2Vec2Model,
         Wav2Vec2Model,
-        WAV_2_VEC_2_PRETRAINED_CONFIG_ARCHIVE_MAP,
     ),
 }
 

src/transformers/models/albert/configuration_albert.py

@@ -19,18 +19,7 @@
 
 from ...configuration_utils import PretrainedConfig
 from ...onnx import OnnxConfig
-
-
-ALBERT_PRETRAINED_CONFIG_ARCHIVE_MAP = {
-    "albert/albert-base-v1": "https://huggingface.co/albert/albert-base-v1/resolve/main/config.json",
-    "albert/albert-large-v1": "https://huggingface.co/albert/albert-large-v1/resolve/main/config.json",
-    "albert/albert-xlarge-v1": "https://huggingface.co/albert/albert-xlarge-v1/resolve/main/config.json",
-    "albert/albert-xxlarge-v1": "https://huggingface.co/albert/albert-xxlarge-v1/resolve/main/config.json",
-    "albert/albert-base-v2": "https://huggingface.co/albert/albert-base-v2/resolve/main/config.json",
-    "albert/albert-large-v2": "https://huggingface.co/albert/albert-large-v2/resolve/main/config.json",
-    "albert/albert-xlarge-v2": "https://huggingface.co/albert/albert-xlarge-v2/resolve/main/config.json",
-    "albert/albert-xxlarge-v2": "https://huggingface.co/albert/albert-xxlarge-v2/resolve/main/config.json",
-}
+from ..deprecated._archive_maps import ALBERT_PRETRAINED_CONFIG_ARCHIVE_MAP  # noqa: F401, E402
 
 
 class AlbertConfig(PretrainedConfig):

src/transformers/models/albert/modeling_albert.py

@@ -52,17 +52,7 @@
 _CONFIG_FOR_DOC = "AlbertConfig"
 
 
-ALBERT_PRETRAINED_MODEL_ARCHIVE_LIST = [
-    "albert/albert-base-v1",
-    "albert/albert-large-v1",
-    "albert/albert-xlarge-v1",
-    "albert/albert-xxlarge-v1",
-    "albert/albert-base-v2",
-    "albert/albert-large-v2",
-    "albert/albert-xlarge-v2",
-    "albert/albert-xxlarge-v2",
-    # See all ALBERT models at https://huggingface.co/models?filter=albert
-]
+from ..deprecated._archive_maps import ALBERT_PRETRAINED_MODEL_ARCHIVE_LIST  # noqa: F401, E402
 
 
 def load_tf_weights_in_albert(model, config, tf_checkpoint_path):

src/transformers/models/albert/modeling_tf_albert.py

@@ -65,17 +65,8 @@
 _CHECKPOINT_FOR_DOC = "albert/albert-base-v2"
 _CONFIG_FOR_DOC = "AlbertConfig"
 
-TF_ALBERT_PRETRAINED_MODEL_ARCHIVE_LIST = [
-    "albert/albert-base-v1",
-    "albert/albert-large-v1",
-    "albert/albert-xlarge-v1",
-    "albert/albert-xxlarge-v1",
-    "albert/albert-base-v2",
-    "albert/albert-large-v2",
-    "albert/albert-xlarge-v2",
-    "albert/albert-xxlarge-v2",
-    # See all ALBERT models at https://huggingface.co/models?filter=albert
-]
+
+from ..deprecated._archive_maps import TF_ALBERT_PRETRAINED_MODEL_ARCHIVE_LIST  # noqa: F401, E402
 
 
 class TFAlbertPreTrainingLoss: