Upload SpikingLM checkpoint and code

README.md

@@ -0,0 +1,69 @@
+---
+library_name: transformers
+pipeline_tag: fill-mask
+tags:
+  - bert
+  - spiking-neural-network
+  - masked-language-modeling
+  - pytorch
+license: apache-2.0
+---
+
+# SpikingLM
+
+SpikingLM is a BERT-base style masked-language model with spiking attention blocks.
+
+This checkpoint uses:
+
+- temporal Spiking BERT with `T=4`
+- learnable Q/K/V scaling parameters initialized from `7`
+- LIF nodes for projection, Q, K, V, attention output, and MLP blocks
+- `FP16OptimizedExp2Softmax` through `self.learnmax(attention_scores)` for attention normalization
+
+## Files
+
+```text
+config.json
+model.safetensors
+tokenizer.json
+tokenizer_config.json
+special_tokens_map.json
+vocab.txt
+spiking_bert/modeling_spiking_bert.py
+scripts/finetune_glue.py
+requirements.txt
+```
+
+## Usage
+
+```python
+from safetensors.torch import load_file
+from transformers import AutoConfig, AutoTokenizer
+from spiking_bert import BertForMaskedLM
+
+repo_or_path = "YOUR_USERNAME/SpikingLM"
+
+config = AutoConfig.from_pretrained(repo_or_path)
+config.T = 4
+config._attn_implementation = "eager"
+
+tokenizer = AutoTokenizer.from_pretrained(repo_or_path)
+model = BertForMaskedLM(config)
+
+state = load_file("model.safetensors")
+model.load_state_dict(state)
+```
+
+If you clone the repository locally, replace `repo_or_path` with the local clone path and load `model.safetensors` from that directory.
+
+## Results
+
+Masked-language-model evaluation stored with the checkpoint:
+
+```json
+{"perplexity": 57.81198691730261}
+```
+
+## Notes
+
+Large binary weights are stored as `model.safetensors`. The model code requires `torch`, `transformers`, `safetensors`, and `spikingjelly`.

all_results.json

@@ -0,0 +1 @@
+{"perplexity": 57.81198691730261}

config.json

@@ -0,0 +1,26 @@
+{
+  "T": 4,
+  "architectures": [
+    "BertForMaskedLM"
+  ],
+  "attention_probs_dropout_prob": 0.1,
+  "classifier_dropout": null,
+  "dtype": "float32",
+  "gradient_checkpointing": false,
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 768,
+  "initializer_range": 0.02,
+  "intermediate_size": 3072,
+  "layer_norm_eps": 1e-12,
+  "max_position_embeddings": 512,
+  "model_type": "bert",
+  "num_attention_heads": 12,
+  "num_hidden_layers": 12,
+  "pad_token_id": 0,
+  "position_embedding_type": "absolute",
+  "transformers_version": "4.57.0",
+  "type_vocab_size": 2,
+  "use_cache": true,
+  "vocab_size": 30522
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5a591db6049da28c2356ca56b61ffb6ee5bd5a8fe5ce3d4d2f3b3722a81373b7
+size 438197096

requirements.txt

@@ -0,0 +1,11 @@
+accelerate>=0.28.0
+datasets>=2.14.0
+evaluate>=0.4.0
+huggingface-hub>=0.20.0
+numpy
+safetensors
+scikit-learn
+spikingjelly
+torch
+tqdm
+transformers>=4.57.0

scripts/finetune_glue.py

@@ -0,0 +1,689 @@
+#!/usr/bin/env python
+# 添加从预训练checkpoint加载的功能
+
+import argparse
+import json
+import logging
+import math
+import os
+import random
+import sys
+from pathlib import Path
+
+sys.path.insert(0, str(Path(__file__).resolve().parents[1]))
+
+import datasets
+import evaluate
+import torch
+from accelerate import Accelerator
+from accelerate.logging import get_logger
+from accelerate.utils import set_seed
+from datasets import load_dataset
+from huggingface_hub import HfApi
+from torch.utils.data import DataLoader
+from tqdm.auto import tqdm
+from safetensors.torch import load_file  # 添加这个导入
+
+import transformers
+from transformers import (
+    AutoConfig,
+    AutoModelForSequenceClassification,
+    AutoTokenizer,
+    DataCollatorWithPadding,
+    PretrainedConfig,
+    SchedulerType,
+    default_data_collator,
+    get_scheduler,
+)
+from transformers.utils import check_min_version
+
+from spikingjelly.activation_based import functional
+check_min_version("4.57.0")
+logger = get_logger(__name__)
+
+task_to_keys = {
+    "cola": ("sentence", None),
+    "mnli": ("premise", "hypothesis"),
+    "mrpc": ("sentence1", "sentence2"),
+    "qnli": ("question", "sentence"),
+    "qqp": ("question1", "question2"),
+    "rte": ("sentence1", "sentence2"),
+    "sst2": ("sentence", None),
+    "stsb": ("sentence1", "sentence2"),
+    "wnli": ("sentence1", "sentence2"),
+}
+
+
+def parse_args():
+    parser = argparse.ArgumentParser(description="Finetune a transformers model on a text classification task")
+    parser.add_argument(
+        "--task_name",
+        type=str,
+        default=None,
+        help="The name of the glue task to train on.",
+        choices=list(task_to_keys.keys()),
+    )
+    parser.add_argument(
+        "--train_file", type=str, default=None, help="A csv or a json file containing the training data."
+    )
+    parser.add_argument(
+        "--validation_file", type=str, default=None, help="A csv or a json file containing the validation data."
+    )
+    parser.add_argument(
+        "--max_length",
+        type=int,
+        default=128,
+        help="The maximum total input sequence length after tokenization.",
+    )
+    parser.add_argument(
+        "--pad_to_max_length",
+        action="store_true",
+        help="If passed, pad all samples to `max_length`. Otherwise, dynamic padding is used.",
+    )
+    parser.add_argument(
+        "--model_name_or_path",
+        type=str,
+        help="Path to pretrained model or model identifier from huggingface.co/models.",
+        required=True,
+    )
+    parser.add_argument(
+        "--pretrained_checkpoint",
+        type=str,
+        default=None,
+        help="Path to pretrained checkpoint directory containing model.safetensors (e.g., step_10000)",
+    )
+    parser.add_argument(
+        "--dataset_cache_dir",
+        type=str,
+        default=None,
+        help="Optional Hugging Face datasets cache directory.",
+    )
+    parser.add_argument(
+        "--use_slow_tokenizer",
+        action="store_true",
+        help="If passed, will use a slow tokenizer.",
+    )
+    parser.add_argument(
+        "--per_device_train_batch_size",
+        type=int,
+        default=8,
+        help="Batch size (per device) for the training dataloader.",
+    )
+    parser.add_argument(
+        "--per_device_eval_batch_size",
+        type=int,
+        default=8,
+        help="Batch size (per device) for the evaluation dataloader.",
+    )
+    parser.add_argument(
+        "--learning_rate",
+        type=float,
+        default=5e-5,
+        help="Initial learning rate to use.",
+    )
+    parser.add_argument("--weight_decay", type=float, default=0.0, help="Weight decay to use.")
+    parser.add_argument("--num_train_epochs", type=int, default=3, help="Total number of training epochs to perform.")
+    parser.add_argument(
+        "--max_train_steps",
+        type=int,
+        default=None,
+        help="Total number of training steps to perform.",
+    )
+    parser.add_argument(
+        "--gradient_accumulation_steps",
+        type=int,
+        default=1,
+        help="Number of updates steps to accumulate before performing a backward/update pass.",
+    )
+    parser.add_argument(
+        "--lr_scheduler_type",
+        type=SchedulerType,
+        default="linear",
+        help="The scheduler type to use.",
+        choices=["linear", "cosine", "cosine_with_restarts", "polynomial", "constant", "constant_with_warmup"],
+    )
+    parser.add_argument(
+        "--num_warmup_steps", type=int, default=0, help="Number of steps for the warmup in the lr scheduler."
+    )
+    parser.add_argument("--output_dir", type=str, default=None, help="Where to store the final model.")
+    parser.add_argument("--seed", type=int, default=None, help="A seed for reproducible training.")
+    parser.add_argument("--push_to_hub", action="store_true", help="Whether or not to push the model to the Hub.")
+    parser.add_argument(
+        "--hub_model_id", type=str, help="The name of the repository to keep in sync with the local `output_dir`."
+    )
+    parser.add_argument("--hub_token", type=str, help="The token to use to push to the Model Hub.")
+    parser.add_argument(
+        "--trust_remote_code",
+        type=bool,
+        default=False,
+        help="Whether or not to allow for custom models defined on the Hub.",
+    )
+    parser.add_argument(
+        "--checkpointing_steps",
+        type=str,
+        default=None,
+        help="Whether the various states should be saved at the end of every n steps, or 'epoch' for each epoch.",
+    )
+    parser.add_argument(
+        "--resume_from_checkpoint",
+        type=str,
+        default=None,
+        help="If the training should continue from a checkpoint folder.",
+    )
+    parser.add_argument(
+        "--with_tracking",
+        action="store_true",
+        help="Whether to enable experiment trackers for logging.",
+    )
+    parser.add_argument(
+        "--report_to",
+        type=str,
+        default="all",
+        help='The integration to report the results and logs to.',
+    )
+    parser.add_argument(
+        "--ignore_mismatched_sizes",
+        action="store_true",
+        help="Whether or not to enable to load a pretrained model whose head dimensions are different.",
+    )
+    parser.add_argument(
+        "--T",
+        type=int,
+        default=4,
+    )
+    args = parser.parse_args()
+
+    # Sanity checks
+    if args.task_name is None and args.train_file is None and args.validation_file is None:
+        raise ValueError("Need either a task name or a training/validation file.")
+    else:
+        if args.train_file is not None:
+            extension = args.train_file.split(".")[-1]
+            assert extension in ["csv", "json"], "`train_file` should be a csv or a json file."
+        if args.validation_file is not None:
+            extension = args.validation_file.split(".")[-1]
+            assert extension in ["csv", "json"], "`validation_file` should be a csv or a json file."
+
+    if args.push_to_hub:
+        assert args.output_dir is not None, "Need an `output_dir` to create a repo when `--push_to_hub` is passed."
+
+    return args
+
+
+def load_model_from_checkpoint(checkpoint_path, config, num_labels):
+    """
+    从包含model.safetensors的checkpoint目录加载模型
+    
+    Args:
+        checkpoint_path: checkpoint目录路径（如 step_10000）
+        config: 模型配置
+        num_labels: 分类任务的标签数量
+    
+    Returns:
+        加载了预训练权重的模型
+    """
+    from spiking_bert.modeling_spiking_bert import BertForSequenceClassification
+    
+    # 检查safetensors文件是否存在
+    safetensors_path = os.path.join(checkpoint_path, "model.safetensors")
+    if not os.path.exists(safetensors_path):
+        raise FileNotFoundError(f"Cannot find model.safetensors in {checkpoint_path}")
+    
+    logger.info(f"Loading pretrained weights from {safetensors_path}")
+    
+    # 修改config以适应分类任务
+    config.num_labels = num_labels
+    
+    # 创建分类模型（会初始化新的分类头）
+    model = BertForSequenceClassification(config)
+    
+    # 加载预训练的权重
+    pretrained_state_dict = load_file(safetensors_path)
+    
+    # 获取当前模型的state_dict
+    model_state_dict = model.state_dict()
+    
+    # 只加载匹配的权重（跳过分类头）
+    matched_keys = []
+    mismatched_keys = []
+    
+    for key in pretrained_state_dict.keys():
+        if key in model_state_dict:
+            if pretrained_state_dict[key].shape == model_state_dict[key].shape:
+                model_state_dict[key] = pretrained_state_dict[key]
+                matched_keys.append(key)
+            else:
+                mismatched_keys.append(key)
+                logger.warning(f"Shape mismatch for {key}: pretrained {pretrained_state_dict[key].shape} vs model {model_state_dict[key].shape}")
+        else:
+            logger.info(f"Key {key} not found in model, skipping...")
+    
+    # 加载权重
+    model.load_state_dict(model_state_dict)
+    
+    logger.info(f"Loaded {len(matched_keys)} matching weights from checkpoint")
+    logger.info(f"Skipped {len(mismatched_keys)} mismatched weights")
+    logger.info(f"Classifier head will be trained from scratch")
+    
+    return model
+
+
+def main():
+    args = parse_args()
+
+    # Initialize the accelerator
+    accelerator = (
+        Accelerator(log_with=args.report_to, project_dir=args.output_dir) if args.with_tracking else Accelerator()
+    )
+    
+    # Make one log on every process with the configuration for debugging.
+    logging.basicConfig(
+        format="%(asctime)s - %(levelname)s - %(name)s - %(message)s",
+        datefmt="%m/%d/%Y %H:%M:%S",
+        level=logging.INFO,
+    )
+    logger.info(accelerator.state, main_process_only=False)
+    if accelerator.is_local_main_process:
+        datasets.utils.logging.set_verbosity_warning()
+        transformers.utils.logging.set_verbosity_info()
+    else:
+        datasets.utils.logging.set_verbosity_error()
+        transformers.utils.logging.set_verbosity_error()
+
+    # Set seed
+    if args.seed is not None:
+        set_seed(args.seed)
+
+    # Handle the repository creation
+    if accelerator.is_main_process:
+        if args.push_to_hub:
+            repo_name = args.hub_model_id
+            if repo_name is None:
+                repo_name = Path(args.output_dir).absolute().name
+            api = HfApi()
+            repo_id = api.create_repo(repo_name, exist_ok=True, token=args.hub_token).repo_id
+
+            with open(os.path.join(args.output_dir, ".gitignore"), "w+") as gitignore:
+                if "step_*" not in gitignore:
+                    gitignore.write("step_*\n")
+                if "epoch_*" not in gitignore:
+                    gitignore.write("epoch_*\n")
+        elif args.output_dir is not None:
+            os.makedirs(args.output_dir, exist_ok=True)
+    accelerator.wait_for_everyone()
+
+    # Load datasets
+    if args.task_name is not None:
+        raw_datasets = load_dataset("nyu-mll/glue", args.task_name, cache_dir=args.dataset_cache_dir)
+    else:
+        data_files = {}
+        if args.train_file is not None:
+            data_files["train"] = args.train_file
+        if args.validation_file is not None:
+            data_files["validation"] = args.validation_file
+        extension = (args.train_file if args.train_file is not None else args.validation_file).split(".")[-1]
+        raw_datasets = load_dataset(extension, data_files=data_files)
+
+    # Labels
+    if args.task_name is not None:
+        is_regression = args.task_name == "stsb"
+        if not is_regression:
+            label_list = raw_datasets["train"].features["label"].names
+            num_labels = len(label_list)
+        else:
+            num_labels = 1
+    else:
+        is_regression = raw_datasets["train"].features["label"].dtype in ["float32", "float64"]
+        if is_regression:
+            num_labels = 1
+        else:
+            label_list = raw_datasets["train"].unique("label")
+            label_list.sort()
+            num_labels = len(label_list)
+
+    # Load config and tokenizer
+    config = AutoConfig.from_pretrained(
+        args.model_name_or_path,
+        num_labels=num_labels,
+        finetuning_task=args.task_name,
+        trust_remote_code=args.trust_remote_code,
+    )
+    tokenizer = AutoTokenizer.from_pretrained(
+        args.model_name_or_path, use_fast=not args.use_slow_tokenizer, trust_remote_code=args.trust_remote_code
+    )
+    if tokenizer.pad_token is None:
+        tokenizer.pad_token = tokenizer.eos_token
+    config.pad_token_id = tokenizer.pad_token_id
+    config.T = args.T
+    config._attn_implementation = 'eager'
+    
+    if args.pretrained_checkpoint is not None:
+        logger.info(f"Loading model from pretrained checkpoint: {args.pretrained_checkpoint}")
+        model = load_model_from_checkpoint(args.pretrained_checkpoint, config, num_labels)
+    else:
+        logger.info(f"Loading model from: {args.model_name_or_path}")
+        model = AutoModelForSequenceClassification.from_pretrained(
+            args.model_name_or_path,
+            from_tf=bool(".ckpt" in args.model_name_or_path),
+            config=config,
+            ignore_mismatched_sizes=args.ignore_mismatched_sizes,
+            trust_remote_code=args.trust_remote_code,
+        )
+    
+    logger.info(model)
+
+    # Preprocessing the datasets
+    if args.task_name is not None:
+        sentence1_key, sentence2_key = task_to_keys[args.task_name]
+    else:
+        non_label_column_names = [name for name in raw_datasets["train"].column_names if name != "label"]
+        if "sentence1" in non_label_column_names and "sentence2" in non_label_column_names:
+            sentence1_key, sentence2_key = "sentence1", "sentence2"
+        else:
+            if len(non_label_column_names) >= 2:
+                sentence1_key, sentence2_key = non_label_column_names[:2]
+            else:
+                sentence1_key, sentence2_key = non_label_column_names[0], None
+
+    # Label mapping
+    label_to_id = None
+    if (
+        model.config.label2id != PretrainedConfig(num_labels=num_labels).label2id
+        and args.task_name is not None
+        and not is_regression
+    ):
+        label_name_to_id = {k.lower(): v for k, v in model.config.label2id.items()}
+        if sorted(label_name_to_id.keys()) == sorted(label_list):
+            logger.info(f"Using label correspondence: {label_name_to_id}")
+            label_to_id = {i: label_name_to_id[label_list[i]] for i in range(num_labels)}
+        else:
+            logger.warning("Model labels don't match dataset labels, ignoring model labels.")
+    elif args.task_name is None and not is_regression:
+        label_to_id = {v: i for i, v in enumerate(label_list)}
+
+    if label_to_id is not None:
+        model.config.label2id = label_to_id
+        model.config.id2label = {id: label for label, id in config.label2id.items()}
+    elif args.task_name is not None and not is_regression:
+        model.config.label2id = {l: i for i, l in enumerate(label_list)}
+        model.config.id2label = {id: label for label, id in config.label2id.items()}
+
+    padding = "max_length" if args.pad_to_max_length else False
+
+    def preprocess_function(examples):
+        texts = (
+            (examples[sentence1_key],) if sentence2_key is None else (examples[sentence1_key], examples[sentence2_key])
+        )
+        result = tokenizer(*texts, padding=padding, max_length=args.max_length, truncation=True)
+
+        if "label" in examples:
+            if label_to_id is not None:
+                result["labels"] = [label_to_id[l] for l in examples["label"]]
+            else:
+                result["labels"] = examples["label"]
+        return result
+
+    with accelerator.main_process_first():
+        processed_datasets = raw_datasets.map(
+            preprocess_function,
+            batched=True,
+            remove_columns=raw_datasets["train"].column_names,
+            desc="Running tokenizer on dataset",
+        )
+
+    train_dataset = processed_datasets["train"]
+    eval_dataset = processed_datasets["validation_matched" if args.task_name == "mnli" else "validation"]
+
+    # Log a few random samples
+    for index in random.sample(range(len(train_dataset)), 3):
+        logger.info(f"Sample {index} of the training set: {train_dataset[index]}.")
+
+    # DataLoaders creation
+    if args.pad_to_max_length:
+        data_collator = default_data_collator
+    else:
+        if accelerator.mixed_precision == "fp8":
+            pad_to_multiple_of = 16
+        elif accelerator.mixed_precision != "no":
+            pad_to_multiple_of = 8
+        else:
+            pad_to_multiple_of = None
+        data_collator = DataCollatorWithPadding(tokenizer, pad_to_multiple_of=pad_to_multiple_of)
+
+    train_dataloader = DataLoader(
+        train_dataset, shuffle=True, collate_fn=data_collator, batch_size=args.per_device_train_batch_size
+    )
+    eval_dataloader = DataLoader(eval_dataset, collate_fn=data_collator, batch_size=args.per_device_eval_batch_size)
+
+    # Optimizer
+    no_decay = ["bias", "LayerNorm.weight"]
+    optimizer_grouped_parameters = [
+        {
+            "params": [p for n, p in model.named_parameters() if not any(nd in n for nd in no_decay)],
+            "weight_decay": args.weight_decay,
+        },
+        {
+            "params": [p for n, p in model.named_parameters() if any(nd in n for nd in no_decay)],
+            "weight_decay": 0.0,
+        },
+    ]
+    optimizer = torch.optim.AdamW(optimizer_grouped_parameters, lr=args.learning_rate)
+
+    # Scheduler
+    overrode_max_train_steps = False
+    num_update_steps_per_epoch = math.ceil(len(train_dataloader) / args.gradient_accumulation_steps)
+    if args.max_train_steps is None:
+        args.max_train_steps = args.num_train_epochs * num_update_steps_per_epoch
+        overrode_max_train_steps = True
+
+    lr_scheduler = get_scheduler(
+        name=args.lr_scheduler_type,
+        optimizer=optimizer,
+        num_warmup_steps=args.num_warmup_steps,
+        num_training_steps=args.max_train_steps,
+    )
+
+    # Prepare with accelerator
+    model, optimizer, train_dataloader, eval_dataloader, lr_scheduler = accelerator.prepare(
+        model, optimizer, train_dataloader, eval_dataloader, lr_scheduler
+    )
+
+    # Recalculate training steps
+    num_update_steps_per_epoch = math.ceil(len(train_dataloader) / args.gradient_accumulation_steps)
+    if overrode_max_train_steps:
+        args.max_train_steps = args.num_train_epochs * num_update_steps_per_epoch
+    args.num_train_epochs = math.ceil(args.max_train_steps / num_update_steps_per_epoch)
+
+    checkpointing_steps = args.checkpointing_steps
+    if checkpointing_steps is not None and checkpointing_steps.isdigit():
+        checkpointing_steps = int(checkpointing_steps)
+
+    # Initialize trackers
+    if args.with_tracking:
+        experiment_config = vars(args)
+        experiment_config["lr_scheduler_type"] = experiment_config["lr_scheduler_type"].value
+        accelerator.init_trackers("glue_no_trainer", experiment_config)
+    # Get metric
+    if args.task_name is not None:
+        metric = evaluate.load("glue", args.task_name)
+    else:
+        metric = evaluate.load("accuracy")
+
+    # Train!
+    total_batch_size = args.per_device_train_batch_size * accelerator.num_processes * args.gradient_accumulation_steps
+
+    logger.info("***** Running training *****")
+    logger.info(f"  Num examples = {len(train_dataset)}")
+    logger.info(f"  Num Epochs = {args.num_train_epochs}")
+    logger.info(f"  Instantaneous batch size per device = {args.per_device_train_batch_size}")
+    logger.info(f"  Total train batch size (w. parallel, distributed & accumulation) = {total_batch_size}")
+    logger.info(f"  Gradient Accumulation steps = {args.gradient_accumulation_steps}")
+    logger.info(f"  Total optimization steps = {args.max_train_steps}")
+    
+    progress_bar = tqdm(range(args.max_train_steps), disable=not accelerator.is_local_main_process)
+    completed_steps = 0
+    starting_epoch = 0
+    
+    # Resume from checkpoint if specified
+    if args.resume_from_checkpoint:
+        if args.resume_from_checkpoint is not None or args.resume_from_checkpoint != "":
+            checkpoint_path = args.resume_from_checkpoint
+            path = os.path.basename(args.resume_from_checkpoint)
+        else:
+            dirs = [f.name for f in os.scandir(os.getcwd()) if f.is_dir()]
+            dirs.sort(key=os.path.getctime)
+            path = dirs[-1]
+            checkpoint_path = path
+            path = os.path.basename(checkpoint_path)
+
+        accelerator.print(f"Resumed from checkpoint: {checkpoint_path}")
+        accelerator.load_state(checkpoint_path)
+        training_difference = os.path.splitext(path)[0]
+
+        if "epoch" in training_difference:
+            starting_epoch = int(training_difference.replace("epoch_", "")) + 1
+            resume_step = None
+            completed_steps = starting_epoch * num_update_steps_per_epoch
+        else:
+            resume_step = int(training_difference.replace("step_", "")) * args.gradient_accumulation_steps
+            starting_epoch = resume_step // len(train_dataloader)
+            completed_steps = resume_step // args.gradient_accumulation_steps
+            resume_step -= starting_epoch * len(train_dataloader)
+
+    progress_bar.update(completed_steps)
+
+    for epoch in range(starting_epoch, args.num_train_epochs):
+        model.train()
+        if args.with_tracking:
+            total_loss = 0
+        if args.resume_from_checkpoint and epoch == starting_epoch and resume_step is not None:
+            active_dataloader = accelerator.skip_first_batches(train_dataloader, resume_step)
+        else:
+            active_dataloader = train_dataloader
+            
+        for step, batch in enumerate(active_dataloader):
+            outputs = model(**batch)
+            loss = outputs.loss
+            if args.with_tracking:
+                total_loss += loss.detach().float()
+            loss = loss / args.gradient_accumulation_steps
+            accelerator.backward(loss)
+            if step % args.gradient_accumulation_steps == 0 or step == len(train_dataloader) - 1:
+                optimizer.step()
+                lr_scheduler.step()
+                optimizer.zero_grad()
+                progress_bar.update(1)
+                completed_steps += 1
+                functional.reset_net(model)
+            if isinstance(checkpointing_steps, int):
+                if completed_steps % checkpointing_steps == 0 and accelerator.sync_gradients:
+                    output_dir = f"step_{completed_steps}"
+                    if args.output_dir is not None:
+                        output_dir = os.path.join(args.output_dir, output_dir)
+                    accelerator.save_state(output_dir)
+
+            if completed_steps >= args.max_train_steps:
+                break
+
+        model.eval()
+        samples_seen = 0
+        for step, batch in enumerate(eval_dataloader):
+            with torch.no_grad():
+                outputs = model(**batch)
+            predictions = outputs.logits.argmax(dim=-1) if not is_regression else outputs.logits.squeeze()
+            predictions, references = accelerator.gather((predictions, batch["labels"]))
+            if accelerator.num_processes > 1:
+                if step == len(eval_dataloader) - 1:
+                    predictions = predictions[: len(eval_dataloader.dataset) - samples_seen]
+                    references = references[: len(eval_dataloader.dataset) - samples_seen]
+                else:
+                    samples_seen += references.shape[0]
+            metric.add_batch(
+                predictions=predictions,
+                references=references,
+            )
+            functional.reset_net(model)
+
+        eval_metric = metric.compute()
+        logger.info(f"epoch {epoch}: {eval_metric}")
+
+        if args.with_tracking:
+            accelerator.log(
+                {
+                    "accuracy" if args.task_name is not None else "glue": eval_metric,
+                    "train_loss": total_loss.item() / len(train_dataloader),
+                    "epoch": epoch,
+                    "step": completed_steps,
+                },
+                step=completed_steps,
+            )
+
+        if args.push_to_hub and epoch < args.num_train_epochs - 1:
+            accelerator.wait_for_everyone()
+            unwrapped_model = accelerator.unwrap_model(model)
+            unwrapped_model.save_pretrained(
+                args.output_dir, is_main_process=accelerator.is_main_process, save_function=accelerator.save
+            )
+            if accelerator.is_main_process:
+                tokenizer.save_pretrained(args.output_dir)
+                api.upload_folder(
+                    commit_message=f"Training in progress epoch {epoch}",
+                    folder_path=args.output_dir,
+                    repo_id=repo_id,
+                    repo_type="model",
+                    token=args.hub_token,
+                )
+
+        if args.checkpointing_steps == "epoch":
+            output_dir = f"epoch_{epoch}"
+            if args.output_dir is not None:
+                output_dir = os.path.join(args.output_dir, output_dir)
+            accelerator.save_state(output_dir)
+
+    if args.output_dir is not None:
+        accelerator.wait_for_everyone()
+        unwrapped_model = accelerator.unwrap_model(model)
+        unwrapped_model.save_pretrained(
+            args.output_dir, is_main_process=accelerator.is_main_process, save_function=accelerator.save
+        )
+        if accelerator.is_main_process:
+            tokenizer.save_pretrained(args.output_dir)
+            if args.push_to_hub:
+                api.upload_folder(
+                    commit_message="End of training",
+                    folder_path=args.output_dir,
+                    repo_id=repo_id,
+                    repo_type="model",
+                    token=args.hub_token,
+                )
+
+    if args.task_name == "mnli":
+        eval_dataset = processed_datasets["validation_mismatched"]
+        eval_dataloader = DataLoader(
+            eval_dataset, collate_fn=data_collator, batch_size=args.per_device_eval_batch_size
+        )
+        eval_dataloader = accelerator.prepare(eval_dataloader)
+
+        model.eval()
+        for step, batch in enumerate(eval_dataloader):
+            outputs = model(**batch)
+            predictions = outputs.logits.argmax(dim=-1)
+            metric.add_batch(
+                predictions=accelerator.gather(predictions),
+                references=accelerator.gather(batch["labels"]),
+            )
+            functional.reset_net(model)
+        eval_metric = metric.compute()
+        logger.info(f"mnli-mm: {eval_metric}")
+
+    if args.output_dir is not None:
+        all_results = {f"eval_{k}": v for k, v in eval_metric.items()}
+        with open(os.path.join(args.output_dir, "all_results.json"), "w") as f:
+            json.dump(all_results, f)
+
+    accelerator.wait_for_everyone()
+    accelerator.end_training()
+
+
+if __name__ == "__main__":
+    main()

special_tokens_map.json

@@ -0,0 +1,7 @@
+{
+  "cls_token": "[CLS]",
+  "mask_token": "[MASK]",
+  "pad_token": "[PAD]",
+  "sep_token": "[SEP]",
+  "unk_token": "[UNK]"
+}

spiking_bert/__init__.py

@@ -0,0 +1,13 @@
+from .modeling_spiking_bert import (
+    BertForMaskedLM,
+    BertForSequenceClassification,
+    BertModel,
+    BertPreTrainedModel,
+)
+
+__all__ = [
+    "BertForMaskedLM",
+    "BertForSequenceClassification",
+    "BertModel",
+    "BertPreTrainedModel",
+]

spiking_bert/modeling_spiking_bert.py

@@ -0,0 +1,1882 @@
+# coding=utf-8
+# Copyright 2018 The Google AI Language Team Authors and The HuggingFace Inc. team.
+# Copyright (c) 2018, NVIDIA CORPORATION.  All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# 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.
+"""PyTorch BERT model."""
+
+import math
+import os
+import warnings
+from dataclasses import dataclass
+from typing import Optional, Union
+
+import torch
+from torch import nn
+from torch.nn import BCEWithLogitsLoss, CrossEntropyLoss, MSELoss
+
+from transformers.activations import ACT2FN
+from transformers.cache_utils import Cache, DynamicCache, EncoderDecoderCache
+from transformers.generation import GenerationMixin
+from transformers.modeling_attn_mask_utils import _prepare_4d_attention_mask_for_sdpa, _prepare_4d_causal_attention_mask_for_sdpa
+from transformers.modeling_layers import GradientCheckpointingLayer
+from transformers.modeling_outputs import (
+    BaseModelOutputWithPastAndCrossAttentions,
+    BaseModelOutputWithPoolingAndCrossAttentions,
+    CausalLMOutputWithCrossAttentions,
+    MaskedLMOutput,
+    MultipleChoiceModelOutput,
+    NextSentencePredictorOutput,
+    QuestionAnsweringModelOutput,
+    SequenceClassifierOutput,
+    TokenClassifierOutput,
+)
+from transformers.modeling_utils import PreTrainedModel
+from transformers.pytorch_utils import apply_chunking_to_forward, find_pruneable_heads_and_indices, prune_linear_layer
+from transformers.utils import ModelOutput, auto_docstring, logging
+from transformers.utils.deprecation import deprecate_kwarg
+from transformers.models.bert.configuration_bert import BertConfig
+from spikingjelly.activation_based import neuron
+
+
+logger = logging.get_logger(__name__)
+
+
+def load_tf_weights_in_bert(model, config, tf_checkpoint_path):
+    """Load tf checkpoints in a pytorch model."""
+    try:
+        import re
+
+        import numpy as np
+        import tensorflow as tf
+    except ImportError:
+        logger.error(
+            "Loading a TensorFlow model in PyTorch, requires TensorFlow to be installed. Please see "
+            "https://www.tensorflow.org/install/ for installation instructions."
+        )
+        raise
+    tf_path = os.path.abspath(tf_checkpoint_path)
+    logger.info(f"Converting TensorFlow checkpoint from {tf_path}")
+    # Load weights from TF model
+    init_vars = tf.train.list_variables(tf_path)
+    names = []
+    arrays = []
+    for name, shape in init_vars:
+        logger.info(f"Loading TF weight {name} with shape {shape}")
+        array = tf.train.load_variable(tf_path, name)
+        names.append(name)
+        arrays.append(array)
+
+    for name, array in zip(names, arrays):
+        name = name.split("/")
+        # adam_v and adam_m are variables used in AdamWeightDecayOptimizer to calculated m and v
+        # which are not required for using pretrained model
+        if any(
+            n in ["adam_v", "adam_m", "AdamWeightDecayOptimizer", "AdamWeightDecayOptimizer_1", "global_step"]
+            for n in name
+        ):
+            logger.info(f"Skipping {'/'.join(name)}")
+            continue
+        pointer = model
+        for m_name in name:
+            if re.fullmatch(r"[A-Za-z]+_\d+", m_name):
+                scope_names = re.split(r"_(\d+)", m_name)
+            else:
+                scope_names = [m_name]
+            if scope_names[0] == "kernel" or scope_names[0] == "gamma":
+                pointer = getattr(pointer, "weight")
+            elif scope_names[0] == "output_bias" or scope_names[0] == "beta":
+                pointer = getattr(pointer, "bias")
+            elif scope_names[0] == "output_weights":
+                pointer = getattr(pointer, "weight")
+            elif scope_names[0] == "squad":
+                pointer = getattr(pointer, "classifier")
+            else:
+                try:
+                    pointer = getattr(pointer, scope_names[0])
+                except AttributeError:
+                    logger.info(f"Skipping {'/'.join(name)}")
+                    continue
+            if len(scope_names) >= 2:
+                num = int(scope_names[1])
+                pointer = pointer[num]
+        if m_name[-11:] == "_embeddings":
+            pointer = getattr(pointer, "weight")
+        elif m_name == "kernel":
+            array = np.transpose(array)
+        try:
+            if pointer.shape != array.shape:
+                raise ValueError(f"Pointer shape {pointer.shape} and array shape {array.shape} mismatched")
+        except ValueError as e:
+            e.args += (pointer.shape, array.shape)
+            raise
+        logger.info(f"Initialize PyTorch weight {name}")
+        pointer.data = torch.from_numpy(array)
+    return model
+
+
+class BertEmbeddings(nn.Module):
+    """Construct the embeddings from word, position and token_type embeddings."""
+
+    def __init__(self, config):
+        super().__init__()
+        self.word_embeddings = nn.Embedding(config.vocab_size, config.hidden_size, padding_idx=config.pad_token_id)
+        self.position_embeddings = nn.Embedding(config.max_position_embeddings, config.hidden_size)
+        self.token_type_embeddings = nn.Embedding(config.type_vocab_size, config.hidden_size)
+
+        # self.LayerNorm is not snake-cased to stick with TensorFlow model variable name and be able to load
+        # any TensorFlow checkpoint file
+        self.LayerNorm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps)
+        self.dropout = nn.Dropout(config.hidden_dropout_prob)
+        # position_ids (1, len position emb) is contiguous in memory and exported when serialized
+        self.position_embedding_type = getattr(config, "position_embedding_type", "absolute")
+        self.register_buffer(
+            "position_ids", torch.arange(config.max_position_embeddings).expand((1, -1)), persistent=False
+        )
+        self.register_buffer(
+            "token_type_ids", torch.zeros(self.position_ids.size(), dtype=torch.long), persistent=False
+        )
+
+    def forward(
+        self,
+        input_ids: Optional[torch.LongTensor] = None,
+        token_type_ids: Optional[torch.LongTensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        inputs_embeds: Optional[torch.FloatTensor] = None,
+        past_key_values_length: int = 0,
+    ) -> torch.Tensor:
+        if input_ids is not None:
+            input_shape = input_ids.size()
+        else:
+            input_shape = inputs_embeds.size()[:-1]
+
+        seq_length = input_shape[1]
+
+        if position_ids is None:
+            position_ids = self.position_ids[:, past_key_values_length : seq_length + past_key_values_length]
+
+        # Setting the token_type_ids to the registered buffer in constructor where it is all zeros, which usually occurs
+        # when its auto-generated, registered buffer helps users when tracing the model without passing token_type_ids, solves
+        # issue #5664
+        if token_type_ids is None:
+            if hasattr(self, "token_type_ids"):
+                buffered_token_type_ids = self.token_type_ids[:, :seq_length]
+                buffered_token_type_ids_expanded = buffered_token_type_ids.expand(input_shape[0], seq_length)
+                token_type_ids = buffered_token_type_ids_expanded
+            else:
+                token_type_ids = torch.zeros(input_shape, dtype=torch.long, device=self.position_ids.device)
+
+        if inputs_embeds is None:
+            inputs_embeds = self.word_embeddings(input_ids)
+        token_type_embeddings = self.token_type_embeddings(token_type_ids)
+
+        embeddings = inputs_embeds + token_type_embeddings
+        if self.position_embedding_type == "absolute":
+            position_embeddings = self.position_embeddings(position_ids)
+            embeddings += position_embeddings
+        embeddings = self.LayerNorm(embeddings)
+        embeddings = self.dropout(embeddings)
+        return embeddings
+
+class FP16OptimizedExp2Softmax(nn.Module):
+    def __init__(self, T, init_value=1.0):
+        """
+        Args:
+            T: Temporal dimension size
+            head_size: Number of attention heads
+            init_value: Initial value for scale parameter
+        """
+        super().__init__()
+        # 注册为可学习参数，维度 [T, head_size]
+        self.scale = nn.Parameter(torch.full((T,), init_value))
+    
+    def forward(self, tensor, k=1.0):
+        """
+        Args:
+            tensor: Input tensor of shape [T, B, head_size, seq_len, seq_len]
+            k: Temperature parameter
+        Returns:
+            Scaled softmax output
+        """
+        tensor = tensor - tensor.max(dim=-1, keepdim=True)[0] - 1
+        tensor.clamp_(min=-30.0, max=30.0)
+        tensor = torch.exp2(tensor)
+        # tensor = tensor / (tensor.sum(dim=-1, keepdim=True) + 1e-6)
+        
+        # scale: [T, head_size] -> [T, 1, head_size, 1, 1] for broadcasting
+        scale = self.scale[:, None, None, None, None]
+        tensor = tensor * scale
+        return tensor
+
+class BertSelfAttention(nn.Module):
+    def __init__(self, config, position_embedding_type=None, layer_idx=None):
+        super().__init__()
+        if config.hidden_size % config.num_attention_heads != 0 and not hasattr(config, "embedding_size"):
+            raise ValueError(
+                f"The hidden size ({config.hidden_size}) is not a multiple of the number of attention "
+                f"heads ({config.num_attention_heads})"
+            )
+
+        self.num_attention_heads = config.num_attention_heads
+        self.attention_head_size = int(config.hidden_size / config.num_attention_heads)
+        self.all_head_size = self.num_attention_heads * self.attention_head_size
+
+        self.proj_lif = neuron.LIFNode(tau=2.0, detach_reset=True, backend='cupy', step_mode='m')
+        self.q_lif = neuron.LIFNode(tau=2.0, detach_reset=True, backend='cupy', step_mode='m')
+        self.k_lif = neuron.LIFNode(tau=2.0, detach_reset=True, backend='cupy', step_mode='m')
+        self.v_lif = neuron.LIFNode(tau=2.0, detach_reset=True, backend='cupy', step_mode='m')
+        
+        self.scale = 0.125
+
+        self.query = nn.Linear(config.hidden_size, self.all_head_size)
+        self.q_lam = nn.Parameter(torch.ones(self.all_head_size)*7)
+        # self.q_norm = nn.RMSNorm(config.hidden_size)
+        self.key = nn.Linear(config.hidden_size, self.all_head_size)
+        self.k_lam = nn.Parameter(torch.ones(self.all_head_size)*7)
+        # self.k_norm = nn.RMSNorm(config.hidden_size)
+        self.value = nn.Linear(config.hidden_size, self.all_head_size)
+        self.v_lam = nn.Parameter(torch.ones(self.all_head_size)*7)
+        # self.v_norm = nn.RMSNorm(config.hidden_size)
+        self.learnmax = FP16OptimizedExp2Softmax(T=config.T)
+        self.dropout = nn.Dropout(config.attention_probs_dropout_prob)
+        self.position_embedding_type = position_embedding_type or getattr(
+            config, "position_embedding_type", "absolute"
+        )
+        if self.position_embedding_type == "relative_key" or self.position_embedding_type == "relative_key_query":
+            self.max_position_embeddings = config.max_position_embeddings
+            self.distance_embedding = nn.Embedding(2 * config.max_position_embeddings - 1, self.attention_head_size)
+
+        self.is_decoder = config.is_decoder
+        self.layer_idx = layer_idx
+        # True 时不做 q_lam/k_lam/v_lam 缩放（等价于恒等，便于与启用 lamb 的实验对比）
+        self.disable_qkv_lamb = bool(getattr(config, "disable_qkv_lamb", False))
+
+    @deprecate_kwarg("past_key_value", new_name="past_key_values", version="4.58")
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        attention_mask: Optional[torch.FloatTensor] = None,
+        head_mask: Optional[torch.FloatTensor] = None,
+        encoder_hidden_states: Optional[torch.FloatTensor] = None,
+        past_key_values: Optional[Cache] = None,
+        output_attentions: Optional[bool] = False,
+        cache_position: Optional[torch.Tensor] = None,
+    ) -> tuple[torch.Tensor]:
+        T, batch_size, seq_length, _ = hidden_states.shape
+        hidden_states = self.proj_lif(hidden_states)
+        query_layer = self.query(hidden_states) 
+        # query_layer = self.q_norm(query_layer.reshape(T*batch_size, seq_length, -1)).reshape(T, batch_size, seq_length, -1)
+        if not self.disable_qkv_lamb:
+            query_layer = self.q_lam * query_layer
+        query_layer = self.q_lif(query_layer)
+        query_layer = query_layer.view(T, batch_size, -1, self.num_attention_heads, self.attention_head_size).transpose(
+            2, 3
+        )
+
+        is_updated = False
+        is_cross_attention = encoder_hidden_states is not None
+        if past_key_values is not None:
+            if isinstance(past_key_values, EncoderDecoderCache):
+                is_updated = past_key_values.is_updated.get(self.layer_idx)
+                if is_cross_attention:
+                    # after the first generated id, we can subsequently re-use all key/value_layer from cache
+                    curr_past_key_value = past_key_values.cross_attention_cache
+                else:
+                    curr_past_key_value = past_key_values.self_attention_cache
+            else:
+                curr_past_key_value = past_key_values
+
+        current_states = encoder_hidden_states if is_cross_attention else hidden_states
+        if is_cross_attention and past_key_values is not None and is_updated:
+            # reuse k,v, cross_attentions
+            key_layer = curr_past_key_value.layers[self.layer_idx].keys
+            value_layer = curr_past_key_value.layers[self.layer_idx].values
+        else:
+            key_layer = self.key(current_states) 
+            # key_layer = self.k_norm(key_layer.reshape(T*batch_size, seq_length, -1)).reshape(T, batch_size, seq_length, -1)
+            if not self.disable_qkv_lamb:
+                key_layer = self.k_lam * key_layer
+            key_layer = self.k_lif(key_layer)
+            key_layer = key_layer.view(T, batch_size, -1, self.num_attention_heads, self.attention_head_size).transpose(
+                2, 3
+            )
+            value_layer = self.value(current_states) 
+            # value_layer = self.v_norm(value_layer.reshape(T*batch_size, seq_length, -1)).reshape(T, batch_size, seq_length, -1)
+            if not self.disable_qkv_lamb:
+                value_layer = self.v_lam * value_layer
+            value_layer = self.v_lif(value_layer)
+            value_layer = value_layer.view(
+                T, batch_size, -1, self.num_attention_heads, self.attention_head_size
+            ).transpose(2, 3)
+
+            if past_key_values is not None:
+                # save all key/value_layer to cache to be re-used for fast auto-regressive generation
+                cache_position = cache_position if not is_cross_attention else None
+                key_layer, value_layer = curr_past_key_value.update(
+                    key_layer, value_layer, self.layer_idx, {"cache_position": cache_position}
+                )
+                # set flag that curr layer for cross-attn is already updated so we can re-use in subsequent calls
+                if is_cross_attention and isinstance(past_key_values, EncoderDecoderCache):
+                    past_key_values.is_updated[self.layer_idx] = True
+
+        # Take the dot product between "query" and "key" to get the raw attention scores.
+        attention_scores = torch.matmul(query_layer, key_layer.transpose(-1, -2))
+
+        if self.position_embedding_type == "relative_key" or self.position_embedding_type == "relative_key_query":
+            query_length, key_length = query_layer.shape[2], key_layer.shape[2]
+            if past_key_values is not None:
+                position_ids_l = torch.tensor(key_length - 1, dtype=torch.long, device=hidden_states.device).view(
+                    -1, 1
+                )
+            else:
+                position_ids_l = torch.arange(query_length, dtype=torch.long, device=hidden_states.device).view(-1, 1)
+            position_ids_r = torch.arange(key_length, dtype=torch.long, device=hidden_states.device).view(1, -1)
+            distance = position_ids_l - position_ids_r
+
+            positional_embedding = self.distance_embedding(distance + self.max_position_embeddings - 1)
+            positional_embedding = positional_embedding.to(dtype=query_layer.dtype)  # fp16 compatibility
+
+            if self.position_embedding_type == "relative_key":
+                relative_position_scores = torch.einsum("bhld,lrd->bhlr", query_layer, positional_embedding)
+                attention_scores = attention_scores + relative_position_scores.unsqueeze(0)
+            elif self.position_embedding_type == "relative_key_query":
+                relative_position_scores_query = torch.einsum("bhld,lrd->bhlr", query_layer, positional_embedding)
+                relative_position_scores_key = torch.einsum("bhrd,lrd->bhlr", key_layer, positional_embedding)
+                attention_scores = attention_scores + relative_position_scores_query.unsqueeze(0) + relative_position_scores_key.unsqueeze(0)
+
+        # attention_scores = attention_scores / math.sqrt(self.attention_head_size)
+        # print(attention_mask)
+        if attention_mask is not None:
+            # Apply the attention mask is (precomputed for all layers in BertModel forward() function)
+            attention_scores = attention_scores + attention_mask
+
+        # # # Normalize the attention scores to probabilities.
+        attention_probs = self.learnmax(attention_scores)
+        # attention_probs = nn.functional.softmax(attention_scores, dim=-1)
+
+        # # This is actually dropping out entire tokens to attend to, which might
+        # # seem a bit unusual, but is taken from the original Transformer paper.
+        # attention_probs = self.dropout(attention_probs)
+        # if attention_mask is not None:
+        #     attention_scores = attention_scores * attention_mask.unsqueeze(0)
+        
+        # attention_probs = attention_scores * self.scale
+
+        # Mask heads if we want to
+        # if head_mask is not None:
+        #     attention_probs = attention_probs * head_mask
+
+        context_layer = torch.matmul(attention_probs, value_layer)  # t,b,h,l,d -> t,b,l,h,d
+
+        context_layer = context_layer.permute(0, 1, 3, 2, 4).contiguous()
+        new_context_layer_shape = context_layer.size()[:-2] + (self.all_head_size,)
+        context_layer = context_layer.view(new_context_layer_shape)
+
+        return context_layer, attention_probs
+
+
+class BertSdpaSelfAttention(BertSelfAttention):
+    def __init__(self, config, position_embedding_type=None, layer_idx=None):
+        super().__init__(config, position_embedding_type=position_embedding_type, layer_idx=layer_idx)
+        self.dropout_prob = config.attention_probs_dropout_prob
+
+    # Adapted from BertSelfAttention
+    @deprecate_kwarg("past_key_value", new_name="past_key_values", version="4.58")
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        attention_mask: Optional[torch.Tensor] = None,
+        head_mask: Optional[torch.FloatTensor] = None,
+        encoder_hidden_states: Optional[torch.FloatTensor] = None,
+        past_key_values: Optional[Cache] = None,
+        output_attentions: Optional[bool] = False,
+        cache_position: Optional[torch.Tensor] = None,
+    ) -> tuple[torch.Tensor]:
+        if self.position_embedding_type != "absolute" or output_attentions or head_mask is not None:
+            # TODO: Improve this warning with e.g. `model.config._attn_implementation = "manual"` once implemented.
+            logger.warning_once(
+                "BertSdpaSelfAttention is used but `torch.nn.functional.scaled_dot_product_attention` does not support "
+                "non-absolute `position_embedding_type` or `output_attentions=True` or `head_mask`. Falling back to "
+                "the manual attention implementation, but specifying the manual implementation will be required from "
+                "Transformers version v5.0.0 onwards. This warning can be removed using the argument "
+                '`attn_implementation="eager"` when loading the model.'
+            )
+            return super().forward(
+                hidden_states,
+                attention_mask,
+                head_mask,
+                encoder_hidden_states,
+                past_key_values,
+                output_attentions,
+                cache_position,
+            )
+
+        T, bsz, tgt_len, _ = hidden_states.size()
+
+        query_layer = (
+            self.query(hidden_states).view(bsz, -1, self.num_attention_heads, self.attention_head_size).transpose(1, 2)
+        )
+
+        is_updated = False
+        is_cross_attention = encoder_hidden_states is not None
+        current_states = encoder_hidden_states if is_cross_attention else hidden_states
+        if past_key_values is not None:
+            if isinstance(past_key_values, EncoderDecoderCache):
+                is_updated = past_key_values.is_updated.get(self.layer_idx)
+                if is_cross_attention:
+                    # after the first generated id, we can subsequently re-use all key/value_states from cache
+                    curr_past_key_value = past_key_values.cross_attention_cache
+                else:
+                    curr_past_key_value = past_key_values.self_attention_cache
+            else:
+                curr_past_key_value = past_key_values
+
+        current_states = encoder_hidden_states if is_cross_attention else hidden_states
+        if is_cross_attention and past_key_values is not None and is_updated:
+            # reuse k,v, cross_attentions
+            key_layer = curr_past_key_value.layers[self.layer_idx].keys
+            value_layer = curr_past_key_value.layers[self.layer_idx].values
+        else:
+            key_layer = (
+                self.key(current_states)
+                .view(bsz, -1, self.num_attention_heads, self.attention_head_size)
+                .transpose(1, 2)
+            )
+            value_layer = (
+                self.value(current_states)
+                .view(bsz, -1, self.num_attention_heads, self.attention_head_size)
+                .transpose(1, 2)
+            )
+
+            if past_key_values is not None:
+                # save all key/value_layer to cache to be re-used for fast auto-regressive generation
+                cache_position = cache_position if not is_cross_attention else None
+                key_layer, value_layer = curr_past_key_value.update(
+                    key_layer, value_layer, self.layer_idx, {"cache_position": cache_position}
+                )
+                # set flag that curr layer for cross-attn is already updated so we can re-use in subsequent calls
+                if is_cross_attention and isinstance(past_key_values, EncoderDecoderCache):
+                    past_key_values.is_updated[self.layer_idx] = True
+
+        # We dispatch to SDPA's Flash Attention or Efficient kernels via this `is_causal` if statement instead of an inline conditional assignment
+        # in SDPA to support both torch.compile's dynamic shapes and full graph options. An inline conditional prevents dynamic shapes from compiling.
+        # The tgt_len > 1 is necessary to match with AttentionMaskConverter.to_causal_4d that does not create
+        # a causal mask in case tgt_len == 1.
+        is_causal = self.is_decoder and not is_cross_attention and attention_mask is None and tgt_len > 1
+
+        attn_output = torch.nn.functional.scaled_dot_product_attention(
+            query_layer,
+            key_layer,
+            value_layer,
+            attn_mask=attention_mask,
+            dropout_p=self.dropout_prob if self.training else 0.0,
+            is_causal=is_causal,
+        )
+
+        attn_output = attn_output.transpose(1, 2)
+        attn_output = attn_output.reshape(bsz, tgt_len, self.all_head_size)
+
+        return attn_output, None
+
+
+class BertSelfOutput(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        self.dense = nn.Linear(config.hidden_size, config.hidden_size)
+        self.LayerNorm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps)
+        self.dropout = nn.Dropout(config.hidden_dropout_prob)
+
+        self.attn_lif = neuron.LIFNode(tau=2.0, v_threshold=0.5, detach_reset=True, backend='cupy', step_mode='m')
+
+    def forward(self, hidden_states: torch.Tensor, input_tensor: torch.Tensor) -> torch.Tensor:
+        hidden_states = self.attn_lif(hidden_states)
+        hidden_states = self.dense(hidden_states)
+        hidden_states = self.dropout(hidden_states)
+        hidden_states = self.LayerNorm(hidden_states + input_tensor)
+        return hidden_states
+
+
+BERT_SELF_ATTENTION_CLASSES = {
+    "eager": BertSelfAttention,
+    "sdpa": BertSdpaSelfAttention,
+}
+
+
+class BertAttention(nn.Module):
+    def __init__(self, config, position_embedding_type=None, layer_idx=None):
+        super().__init__()
+        self.self = BERT_SELF_ATTENTION_CLASSES[config._attn_implementation](
+            config,
+            position_embedding_type=position_embedding_type,
+            layer_idx=layer_idx,
+        )
+        self.output = BertSelfOutput(config)
+        self.pruned_heads = set()
+
+    def prune_heads(self, heads):
+        if len(heads) == 0:
+            return
+        heads, index = find_pruneable_heads_and_indices(
+            heads, self.self.num_attention_heads, self.self.attention_head_size, self.pruned_heads
+        )
+
+        # Prune linear layers
+        self.self.query = prune_linear_layer(self.self.query, index)
+        self.self.key = prune_linear_layer(self.self.key, index)
+        self.self.value = prune_linear_layer(self.self.value, index)
+        self.output.dense = prune_linear_layer(self.output.dense, index, dim=1)
+
+        # Update hyper params and store pruned heads
+        self.self.num_attention_heads = self.self.num_attention_heads - len(heads)
+        self.self.all_head_size = self.self.attention_head_size * self.self.num_attention_heads
+        self.pruned_heads = self.pruned_heads.union(heads)
+
+    @deprecate_kwarg("past_key_value", new_name="past_key_values", version="4.58")
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        attention_mask: Optional[torch.FloatTensor] = None,
+        head_mask: Optional[torch.FloatTensor] = None,
+        encoder_hidden_states: Optional[torch.FloatTensor] = None,
+        past_key_values: Optional[Cache] = None,
+        output_attentions: Optional[bool] = False,
+        cache_position: Optional[torch.Tensor] = None,
+    ) -> tuple[torch.Tensor]:
+        self_outputs = self.self(
+            hidden_states,
+            attention_mask=attention_mask,
+            head_mask=head_mask,
+            encoder_hidden_states=encoder_hidden_states,
+            past_key_values=past_key_values,
+            output_attentions=output_attentions,
+            cache_position=cache_position,
+        )
+        attention_output = self.output(self_outputs[0], hidden_states)
+        outputs = (attention_output,) + self_outputs[1:]  # add attentions if we output them
+        return outputs
+
+
+class BertIntermediate(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        self.dense = nn.Linear(config.hidden_size, config.intermediate_size)
+        if isinstance(config.hidden_act, str):
+            self.intermediate_act_fn = ACT2FN[config.hidden_act]
+        else:
+            self.intermediate_act_fn = config.hidden_act
+        
+        self.mlp1_lif = neuron.LIFNode(tau=2.0, detach_reset=True, backend='cupy', step_mode='m')
+
+    def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
+        hidden_states = self.mlp1_lif(hidden_states)
+        hidden_states = self.dense(hidden_states)
+        # hidden_states = self.intermediate_act_fn(hidden_states)
+        return hidden_states
+
+
+class BertOutput(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        self.dense = nn.Linear(config.intermediate_size, config.hidden_size)
+        self.LayerNorm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps)
+        self.dropout = nn.Dropout(config.hidden_dropout_prob)
+
+        self.mlp2_lif = neuron.LIFNode(tau=2.0, detach_reset=True, backend='cupy', step_mode='m')
+
+    def forward(self, hidden_states: torch.Tensor, input_tensor: torch.Tensor) -> torch.Tensor:
+        hidden_states = self.mlp2_lif(hidden_states)
+        hidden_states = self.dense(hidden_states)
+        hidden_states = self.dropout(hidden_states)
+        hidden_states = self.LayerNorm(hidden_states + input_tensor)
+        return hidden_states
+
+
+class BertLayer(GradientCheckpointingLayer):
+    def __init__(self, config, layer_idx=None):
+        super().__init__()
+        self.chunk_size_feed_forward = config.chunk_size_feed_forward
+        self.seq_len_dim = 1
+        self.attention = BertAttention(config, layer_idx=layer_idx)
+        self.is_decoder = config.is_decoder
+        self.add_cross_attention = config.add_cross_attention
+        if self.add_cross_attention:
+            if not self.is_decoder:
+                raise ValueError(f"{self} should be used as a decoder model if cross attention is added")
+            self.crossattention = BertAttention(config, position_embedding_type="absolute", layer_idx=layer_idx)
+        self.intermediate = BertIntermediate(config)
+        self.output = BertOutput(config)
+
+    @deprecate_kwarg("past_key_value", new_name="past_key_values", version="4.58")
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        attention_mask: Optional[torch.FloatTensor] = None,
+        head_mask: Optional[torch.FloatTensor] = None,
+        encoder_hidden_states: Optional[torch.FloatTensor] = None,
+        encoder_attention_mask: Optional[torch.FloatTensor] = None,
+        past_key_values: Optional[Cache] = None,
+        output_attentions: Optional[bool] = False,
+        cache_position: Optional[torch.Tensor] = None,
+    ) -> tuple[torch.Tensor]:
+        self_attention_outputs = self.attention(
+            hidden_states,
+            attention_mask=attention_mask,
+            head_mask=head_mask,
+            output_attentions=output_attentions,
+            past_key_values=past_key_values,
+            cache_position=cache_position,
+        )
+        attention_output = self_attention_outputs[0]
+        outputs = self_attention_outputs[1:]  # add self attentions if we output attention weights
+
+        if self.is_decoder and encoder_hidden_states is not None:
+            if not hasattr(self, "crossattention"):
+                raise ValueError(
+                    f"If `encoder_hidden_states` are passed, {self} has to be instantiated with cross-attention layers"
+                    " by setting `config.add_cross_attention=True`"
+                )
+
+            cross_attention_outputs = self.crossattention(
+                attention_output,
+                attention_mask=encoder_attention_mask,
+                head_mask=head_mask,
+                encoder_hidden_states=encoder_hidden_states,
+                past_key_values=past_key_values,
+                output_attentions=output_attentions,
+                cache_position=cache_position,
+            )
+            attention_output = cross_attention_outputs[0]
+            outputs = outputs + cross_attention_outputs[1:]  # add cross attentions if we output attention weights
+
+        layer_output = apply_chunking_to_forward(
+            self.feed_forward_chunk, self.chunk_size_feed_forward, self.seq_len_dim, attention_output
+        )
+        outputs = (layer_output,) + outputs
+
+        return outputs
+
+    def feed_forward_chunk(self, attention_output):
+        intermediate_output = self.intermediate(attention_output)
+        layer_output = self.output(intermediate_output, attention_output)
+        return layer_output
+
+
+class BertEncoder(nn.Module):
+    def __init__(self, config, layer_idx=None):
+        super().__init__()
+        self.config = config
+        self.layer = nn.ModuleList([BertLayer(config, layer_idx=i) for i in range(config.num_hidden_layers)])
+        self.gradient_checkpointing = False
+
+        self.T = config.T
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        attention_mask: Optional[torch.FloatTensor] = None,
+        head_mask: Optional[torch.FloatTensor] = None,
+        encoder_hidden_states: Optional[torch.FloatTensor] = None,
+        encoder_attention_mask: Optional[torch.FloatTensor] = None,
+        past_key_values: Optional[Cache] = None,
+        use_cache: Optional[bool] = None,
+        output_attentions: Optional[bool] = False,
+        output_hidden_states: Optional[bool] = False,
+        return_dict: Optional[bool] = True,
+        cache_position: Optional[torch.Tensor] = None,
+    ) -> Union[tuple[torch.Tensor], BaseModelOutputWithPastAndCrossAttentions]:
+        all_hidden_states = () if output_hidden_states else None
+        all_self_attentions = () if output_attentions else None
+        all_cross_attentions = () if output_attentions and self.config.add_cross_attention else None
+
+        if self.gradient_checkpointing and self.training:
+            if use_cache:
+                logger.warning_once(
+                    "`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`..."
+                )
+                use_cache = False
+
+        if use_cache and self.config.is_decoder and past_key_values is None:
+            past_key_values = EncoderDecoderCache(DynamicCache(config=self.config), DynamicCache(config=self.config))
+
+        if use_cache and self.config.is_decoder and isinstance(past_key_values, tuple):
+            logger.warning_once(
+                "Passing a tuple of `past_key_values` is deprecated and will be removed in Transformers v4.58.0. "
+                "You should pass an instance of `EncoderDecoderCache` instead, e.g. "
+                "`past_key_values=EncoderDecoderCache.from_legacy_cache(past_key_values)`."
+            )
+            past_key_values = EncoderDecoderCache.from_legacy_cache(past_key_values)
+
+        hidden_states = hidden_states.repeat(tuple([self.T] + torch.ones(len(hidden_states.size()), dtype=int).tolist())) # T B L D
+
+        for i, layer_module in enumerate(self.layer):
+            if output_hidden_states:
+                all_hidden_states = all_hidden_states + (hidden_states,)
+
+            layer_head_mask = head_mask[i] if head_mask is not None else None
+
+            layer_outputs = layer_module(
+                hidden_states,
+                attention_mask,
+                layer_head_mask,
+                encoder_hidden_states,  # as a positional argument for gradient checkpointing
+                encoder_attention_mask=encoder_attention_mask,
+                past_key_values=past_key_values,
+                output_attentions=output_attentions,
+                cache_position=cache_position,
+            )
+
+            hidden_states = layer_outputs[0]
+            if output_attentions:
+                all_self_attentions = all_self_attentions + (layer_outputs[1],)
+                if self.config.add_cross_attention:
+                    all_cross_attentions = all_cross_attentions + (layer_outputs[2],)
+
+        if output_hidden_states:
+            all_hidden_states = all_hidden_states + (hidden_states.mean(0),)
+
+        hidden_states = hidden_states.mean(0)
+
+        if not return_dict:
+            return tuple(
+                v
+                for v in [
+                    hidden_states,
+                    past_key_values,
+                    all_hidden_states,
+                    all_self_attentions,
+                    all_cross_attentions,
+                ]
+                if v is not None
+            )
+        return BaseModelOutputWithPastAndCrossAttentions(
+            last_hidden_state=hidden_states,
+            past_key_values=past_key_values,
+            hidden_states=all_hidden_states,
+            attentions=all_self_attentions,
+            cross_attentions=all_cross_attentions,
+        )
+
+
+class BertPooler(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        self.dense = nn.Linear(config.hidden_size, config.hidden_size)
+        self.activation = nn.Tanh()
+
+    def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
+        # We "pool" the model by simply taking the hidden state corresponding
+        # to the first token.
+        first_token_tensor = hidden_states[:, 0]
+        pooled_output = self.dense(first_token_tensor)
+        pooled_output = self.activation(pooled_output)
+        return pooled_output
+
+
+class BertPredictionHeadTransform(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        self.dense = nn.Linear(config.hidden_size, config.hidden_size)
+        if isinstance(config.hidden_act, str):
+            self.transform_act_fn = ACT2FN[config.hidden_act]
+        else:
+            self.transform_act_fn = config.hidden_act
+        self.LayerNorm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps)
+
+    def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
+        hidden_states = self.dense(hidden_states)
+        hidden_states = self.transform_act_fn(hidden_states)
+        hidden_states = self.LayerNorm(hidden_states)
+        return hidden_states
+
+
+class BertLMPredictionHead(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        self.transform = BertPredictionHeadTransform(config)
+
+        # The output weights are the same as the input embeddings, but there is
+        # an output-only bias for each token.
+        self.decoder = nn.Linear(config.hidden_size, config.vocab_size, bias=False)
+
+        self.bias = nn.Parameter(torch.zeros(config.vocab_size))
+
+        # Need a link between the two variables so that the bias is correctly resized with `resize_token_embeddings`
+        self.decoder.bias = self.bias
+
+    def _tie_weights(self):
+        self.decoder.bias = self.bias
+
+    def forward(self, hidden_states):
+        hidden_states = self.transform(hidden_states)
+        hidden_states = self.decoder(hidden_states)
+        return hidden_states
+
+
+class BertOnlyMLMHead(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        self.predictions = BertLMPredictionHead(config)
+
+    def forward(self, sequence_output: torch.Tensor) -> torch.Tensor:
+        prediction_scores = self.predictions(sequence_output)
+        return prediction_scores
+
+
+class BertOnlyNSPHead(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        self.seq_relationship = nn.Linear(config.hidden_size, 2)
+
+    def forward(self, pooled_output):
+        seq_relationship_score = self.seq_relationship(pooled_output)
+        return seq_relationship_score
+
+
+class BertPreTrainingHeads(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        self.predictions = BertLMPredictionHead(config)
+        self.seq_relationship = nn.Linear(config.hidden_size, 2)
+
+    def forward(self, sequence_output, pooled_output):
+        prediction_scores = self.predictions(sequence_output)
+        seq_relationship_score = self.seq_relationship(pooled_output)
+        return prediction_scores, seq_relationship_score
+
+
+@auto_docstring
+class BertPreTrainedModel(PreTrainedModel):
+    config: BertConfig
+    load_tf_weights = load_tf_weights_in_bert
+    base_model_prefix = "bert"
+    supports_gradient_checkpointing = True
+    _supports_sdpa = True
+
+    def _init_weights(self, module):
+        """Initialize the weights"""
+        if isinstance(module, nn.Linear):
+            # Slightly different from the TF version which uses truncated_normal for initialization
+            # cf https://github.com/pytorch/pytorch/pull/5617
+            module.weight.data.normal_(mean=0.0, std=self.config.initializer_range)
+            if module.bias is not None:
+                module.bias.data.zero_()
+        elif isinstance(module, nn.Embedding):
+            module.weight.data.normal_(mean=0.0, std=self.config.initializer_range)
+            if module.padding_idx is not None:
+                module.weight.data[module.padding_idx].zero_()
+        elif isinstance(module, nn.LayerNorm):
+            module.bias.data.zero_()
+            module.weight.data.fill_(1.0)
+        elif isinstance(module, BertLMPredictionHead):
+            module.bias.data.zero_()
+
+
+@dataclass
+@auto_docstring(
+    custom_intro="""
+    Output type of [`BertForPreTraining`].
+    """
+)
+class BertForPreTrainingOutput(ModelOutput):
+    r"""
+    loss (*optional*, returned when `labels` is provided, `torch.FloatTensor` of shape `(1,)`):
+        Total loss as the sum of the masked language modeling loss and the next sequence prediction
+        (classification) loss.
+    prediction_logits (`torch.FloatTensor` of shape `(batch_size, sequence_length, config.vocab_size)`):
+        Prediction scores of the language modeling head (scores for each vocabulary token before SoftMax).
+    seq_relationship_logits (`torch.FloatTensor` of shape `(batch_size, 2)`):
+        Prediction scores of the next sequence prediction (classification) head (scores of True/False continuation
+        before SoftMax).
+    """
+
+    loss: Optional[torch.FloatTensor] = None
+    prediction_logits: Optional[torch.FloatTensor] = None
+    seq_relationship_logits: Optional[torch.FloatTensor] = None
+    hidden_states: Optional[tuple[torch.FloatTensor]] = None
+    attentions: Optional[tuple[torch.FloatTensor]] = None
+
+
+@auto_docstring(
+    custom_intro="""
+    The model can behave as an encoder (with only self-attention) as well as a decoder, in which case a layer of
+    cross-attention is added between the self-attention layers, following the architecture described in [Attention is
+    all you need](https://huggingface.co/papers/1706.03762) by Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit,
+    Llion Jones, Aidan N. Gomez, Lukasz Kaiser and Illia Polosukhin.
+
+    To behave as an decoder the model needs to be initialized with the `is_decoder` argument of the configuration set
+    to `True`. To be used in a Seq2Seq model, the model needs to initialized with both `is_decoder` argument and
+    `add_cross_attention` set to `True`; an `encoder_hidden_states` is then expected as an input to the forward pass.
+    """
+)
+class BertModel(BertPreTrainedModel):
+    _no_split_modules = ["BertEmbeddings", "BertLayer"]
+
+    def __init__(self, config, add_pooling_layer=True):
+        r"""
+        add_pooling_layer (bool, *optional*, defaults to `True`):
+            Whether to add a pooling layer
+        """
+        super().__init__(config)
+        self.config = config
+
+        self.embeddings = BertEmbeddings(config)
+        self.encoder = BertEncoder(config)
+
+        self.pooler = BertPooler(config) if add_pooling_layer else None
+
+        self.attn_implementation = config._attn_implementation
+        self.position_embedding_type = config.position_embedding_type
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    def get_input_embeddings(self):
+        return self.embeddings.word_embeddings
+
+    def set_input_embeddings(self, value):
+        self.embeddings.word_embeddings = value
+
+    def _prune_heads(self, heads_to_prune):
+        """
+        Prunes heads of the model. heads_to_prune: dict of {layer_num: list of heads to prune in this layer} See base
+        class PreTrainedModel
+        """
+        for layer, heads in heads_to_prune.items():
+            self.encoder.layer[layer].attention.prune_heads(heads)
+
+    @auto_docstring
+    def forward(
+        self,
+        input_ids: Optional[torch.Tensor] = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        token_type_ids: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.Tensor] = None,
+        head_mask: Optional[torch.Tensor] = None,
+        inputs_embeds: Optional[torch.Tensor] = None,
+        encoder_hidden_states: Optional[torch.Tensor] = None,
+        encoder_attention_mask: Optional[torch.Tensor] = None,
+        past_key_values: Optional[Cache] = None,
+        use_cache: Optional[bool] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+        cache_position: Optional[torch.Tensor] = None,
+    ) -> Union[tuple[torch.Tensor], BaseModelOutputWithPoolingAndCrossAttentions]:
+        output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
+        output_hidden_states = (
+            output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
+        )
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        if self.config.is_decoder:
+            use_cache = use_cache if use_cache is not None else self.config.use_cache
+        else:
+            use_cache = False
+
+        if input_ids is not None and inputs_embeds is not None:
+            raise ValueError("You cannot specify both input_ids and inputs_embeds at the same time")
+        elif input_ids is not None:
+            self.warn_if_padding_and_no_attention_mask(input_ids, attention_mask)
+            input_shape = input_ids.size()
+        elif inputs_embeds is not None:
+            input_shape = inputs_embeds.size()[:-1]
+        else:
+            raise ValueError("You have to specify either input_ids or inputs_embeds")
+
+        batch_size, seq_length = input_shape
+        device = input_ids.device if input_ids is not None else inputs_embeds.device
+
+        past_key_values_length = 0
+        if past_key_values is not None:
+            past_key_values_length = (
+                past_key_values[0][0].shape[-2]
+                if not isinstance(past_key_values, Cache)
+                else past_key_values.get_seq_length()
+            )
+
+        if token_type_ids is None:
+            if hasattr(self.embeddings, "token_type_ids"):
+                buffered_token_type_ids = self.embeddings.token_type_ids[:, :seq_length]
+                buffered_token_type_ids_expanded = buffered_token_type_ids.expand(batch_size, seq_length)
+                token_type_ids = buffered_token_type_ids_expanded
+            else:
+                token_type_ids = torch.zeros(input_shape, dtype=torch.long, device=device)
+
+        embedding_output = self.embeddings(
+            input_ids=input_ids,
+            position_ids=position_ids,
+            token_type_ids=token_type_ids,
+            inputs_embeds=inputs_embeds,
+            past_key_values_length=past_key_values_length,
+        )
+
+        if attention_mask is None:
+            attention_mask = torch.ones((batch_size, seq_length + past_key_values_length), device=device)
+
+        use_sdpa_attention_masks = (
+            self.attn_implementation == "sdpa"
+            and self.position_embedding_type == "absolute"
+            and head_mask is None
+            and not output_attentions
+        )
+
+        # Expand the attention mask
+        if use_sdpa_attention_masks and attention_mask.dim() == 2:
+            # Expand the attention mask for SDPA.
+            # [bsz, seq_len] -> [bsz, 1, seq_len, seq_len]
+            if self.config.is_decoder:
+                extended_attention_mask = _prepare_4d_causal_attention_mask_for_sdpa(
+                    attention_mask,
+                    input_shape,
+                    embedding_output,
+                    past_key_values_length,
+                )
+            else:
+                extended_attention_mask = _prepare_4d_attention_mask_for_sdpa(
+                    attention_mask, embedding_output.dtype, tgt_len=seq_length
+                )
+        else:
+            # We can provide a self-attention mask of dimensions [batch_size, from_seq_length, to_seq_length]
+            # ourselves in which case we just need to make it broadcastable to all heads.
+            extended_attention_mask = self.get_extended_attention_mask(attention_mask, input_shape)
+
+        # If a 2D or 3D attention mask is provided for the cross-attention
+        # we need to make broadcastable to [batch_size, num_heads, seq_length, seq_length]
+        if self.config.is_decoder and encoder_hidden_states is not None:
+            encoder_batch_size, encoder_sequence_length, _ = encoder_hidden_states.size()
+            encoder_hidden_shape = (encoder_batch_size, encoder_sequence_length)
+            if encoder_attention_mask is None:
+                encoder_attention_mask = torch.ones(encoder_hidden_shape, device=device)
+
+            if use_sdpa_attention_masks and encoder_attention_mask.dim() == 2:
+                # Expand the attention mask for SDPA.
+                # [bsz, seq_len] -> [bsz, 1, seq_len, seq_len]
+                encoder_extended_attention_mask = _prepare_4d_attention_mask_for_sdpa(
+                    encoder_attention_mask, embedding_output.dtype, tgt_len=seq_length
+                )
+            else:
+                encoder_extended_attention_mask = self.invert_attention_mask(encoder_attention_mask)
+        else:
+            encoder_extended_attention_mask = None
+
+        # Prepare head mask if needed
+        # 1.0 in head_mask indicate we keep the head
+        # attention_probs has shape bsz x n_heads x N x N
+        # input head_mask has shape [num_heads] or [num_hidden_layers x num_heads]
+        # and head_mask is converted to shape [num_hidden_layers x batch x num_heads x seq_length x seq_length]
+
+        ### update extended_attention_mask during training here!!!
+        # extended_attention_mask = attention_mask.unsqueeze(1).unsqueeze(1)
+        head_mask = self.get_head_mask(head_mask, self.config.num_hidden_layers)
+        encoder_outputs = self.encoder(
+            embedding_output,
+            attention_mask=extended_attention_mask,
+            head_mask=head_mask,
+            encoder_hidden_states=encoder_hidden_states,
+            encoder_attention_mask=encoder_extended_attention_mask,
+            past_key_values=past_key_values,
+            use_cache=use_cache,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+            cache_position=cache_position,
+        )
+        sequence_output = encoder_outputs[0]
+        pooled_output = self.pooler(sequence_output) if self.pooler is not None else None
+
+        if not return_dict:
+            return (sequence_output, pooled_output) + encoder_outputs[1:]
+
+        return BaseModelOutputWithPoolingAndCrossAttentions(
+            last_hidden_state=sequence_output,
+            pooler_output=pooled_output,
+            past_key_values=encoder_outputs.past_key_values,
+            hidden_states=encoder_outputs.hidden_states,
+            attentions=encoder_outputs.attentions,
+            cross_attentions=encoder_outputs.cross_attentions,
+        )
+
+
+@auto_docstring(
+    custom_intro="""
+    Bert Model with two heads on top as done during the pretraining: a `masked language modeling` head and a `next
+    sentence prediction (classification)` head.
+    """
+)
+class BertForPreTraining(BertPreTrainedModel):
+    _tied_weights_keys = ["predictions.decoder.bias", "cls.predictions.decoder.weight"]
+
+    def __init__(self, config):
+        super().__init__(config)
+
+        self.bert = BertModel(config)
+        self.cls = BertPreTrainingHeads(config)
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    def get_output_embeddings(self):
+        return self.cls.predictions.decoder
+
+    def set_output_embeddings(self, new_embeddings):
+        self.cls.predictions.decoder = new_embeddings
+        self.cls.predictions.bias = new_embeddings.bias
+
+    @auto_docstring
+    def forward(
+        self,
+        input_ids: Optional[torch.Tensor] = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        token_type_ids: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.Tensor] = None,
+        head_mask: Optional[torch.Tensor] = None,
+        inputs_embeds: Optional[torch.Tensor] = None,
+        labels: Optional[torch.Tensor] = None,
+        next_sentence_label: Optional[torch.Tensor] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+    ) -> Union[tuple[torch.Tensor], BertForPreTrainingOutput]:
+        r"""
+        labels (`torch.LongTensor` of shape `(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]`
+        next_sentence_label (`torch.LongTensor` of shape `(batch_size,)`, *optional*):
+            Labels for computing the next sequence prediction (classification) loss. Input should be a sequence
+            pair (see `input_ids` docstring) Indices should be in `[0, 1]`:
+
+            - 0 indicates sequence B is a continuation of sequence A,
+            - 1 indicates sequence B is a random sequence.
+
+        Example:
+
+        ```python
+        >>> from transformers import AutoTokenizer, BertForPreTraining
+        >>> import torch
+
+        >>> tokenizer = AutoTokenizer.from_pretrained("google-bert/bert-base-uncased")
+        >>> model = BertForPreTraining.from_pretrained("google-bert/bert-base-uncased")
+
+        >>> inputs = tokenizer("Hello, my dog is cute", return_tensors="pt")
+        >>> outputs = model(**inputs)
+
+        >>> prediction_logits = outputs.prediction_logits
+        >>> seq_relationship_logits = outputs.seq_relationship_logits
+        ```
+        """
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        outputs = self.bert(
+            input_ids,
+            attention_mask=attention_mask,
+            token_type_ids=token_type_ids,
+            position_ids=position_ids,
+            head_mask=head_mask,
+            inputs_embeds=inputs_embeds,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+        )
+
+        sequence_output, pooled_output = outputs[:2]
+        prediction_scores, seq_relationship_score = self.cls(sequence_output, pooled_output)
+
+        total_loss = None
+        if labels is not None and next_sentence_label is not None:
+            loss_fct = CrossEntropyLoss()
+            masked_lm_loss = loss_fct(prediction_scores.view(-1, self.config.vocab_size), labels.view(-1))
+            next_sentence_loss = loss_fct(seq_relationship_score.view(-1, 2), next_sentence_label.view(-1))
+            total_loss = masked_lm_loss + next_sentence_loss
+
+        if not return_dict:
+            output = (prediction_scores, seq_relationship_score) + outputs[2:]
+            return ((total_loss,) + output) if total_loss is not None else output
+
+        return BertForPreTrainingOutput(
+            loss=total_loss,
+            prediction_logits=prediction_scores,
+            seq_relationship_logits=seq_relationship_score,
+            hidden_states=outputs.hidden_states,
+            attentions=outputs.attentions,
+        )
+
+
+@auto_docstring(
+    custom_intro="""
+    Bert Model with a `language modeling` head on top for CLM fine-tuning.
+    """
+)
+class BertLMHeadModel(BertPreTrainedModel, GenerationMixin):
+    _tied_weights_keys = ["cls.predictions.decoder.bias", "cls.predictions.decoder.weight"]
+
+    def __init__(self, config):
+        super().__init__(config)
+
+        if not config.is_decoder:
+            logger.warning("If you want to use `BertLMHeadModel` as a standalone, add `is_decoder=True.`")
+
+        self.bert = BertModel(config, add_pooling_layer=False)
+        self.cls = BertOnlyMLMHead(config)
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    def get_output_embeddings(self):
+        return self.cls.predictions.decoder
+
+    def set_output_embeddings(self, new_embeddings):
+        self.cls.predictions.decoder = new_embeddings
+        self.cls.predictions.bias = new_embeddings.bias
+
+    @auto_docstring
+    def forward(
+        self,
+        input_ids: Optional[torch.Tensor] = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        token_type_ids: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.Tensor] = None,
+        head_mask: Optional[torch.Tensor] = None,
+        inputs_embeds: Optional[torch.Tensor] = None,
+        encoder_hidden_states: Optional[torch.Tensor] = None,
+        encoder_attention_mask: Optional[torch.Tensor] = None,
+        labels: Optional[torch.Tensor] = None,
+        past_key_values: Optional[Cache] = None,
+        use_cache: Optional[bool] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+        cache_position: Optional[torch.Tensor] = None,
+        **loss_kwargs,
+    ) -> Union[tuple[torch.Tensor], CausalLMOutputWithCrossAttentions]:
+        r"""
+        labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
+            Labels for computing the left-to-right language modeling loss (next word prediction). 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 n `[0, ..., config.vocab_size]`
+        """
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+        if labels is not None:
+            use_cache = False
+
+        outputs = self.bert(
+            input_ids,
+            attention_mask=attention_mask,
+            token_type_ids=token_type_ids,
+            position_ids=position_ids,
+            head_mask=head_mask,
+            inputs_embeds=inputs_embeds,
+            encoder_hidden_states=encoder_hidden_states,
+            encoder_attention_mask=encoder_attention_mask,
+            past_key_values=past_key_values,
+            use_cache=use_cache,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+            cache_position=cache_position,
+        )
+
+        sequence_output = outputs[0]
+        prediction_scores = self.cls(sequence_output)
+
+        lm_loss = None
+        if labels is not None:
+            lm_loss = self.loss_function(prediction_scores, labels, self.config.vocab_size, **loss_kwargs)
+
+        if not return_dict:
+            output = (prediction_scores,) + outputs[2:]
+            return ((lm_loss,) + output) if lm_loss is not None else output
+
+        return CausalLMOutputWithCrossAttentions(
+            loss=lm_loss,
+            logits=prediction_scores,
+            past_key_values=outputs.past_key_values,
+            hidden_states=outputs.hidden_states,
+            attentions=outputs.attentions,
+            cross_attentions=outputs.cross_attentions,
+        )
+
+
+@auto_docstring
+class BertForMaskedLM(BertPreTrainedModel):
+    _tied_weights_keys = ["predictions.decoder.bias", "cls.predictions.decoder.weight"]
+
+    def __init__(self, config):
+        super().__init__(config)
+
+        if config.is_decoder:
+            logger.warning(
+                "If you want to use `BertForMaskedLM` make sure `config.is_decoder=False` for "
+                "bi-directional self-attention."
+            )
+
+        self.bert = BertModel(config, add_pooling_layer=False)
+        self.cls = BertOnlyMLMHead(config)
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    def get_output_embeddings(self):
+        return self.cls.predictions.decoder
+
+    def set_output_embeddings(self, new_embeddings):
+        self.cls.predictions.decoder = new_embeddings
+        self.cls.predictions.bias = new_embeddings.bias
+
+    @auto_docstring
+    def forward(
+        self,
+        input_ids: Optional[torch.Tensor] = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        token_type_ids: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.Tensor] = None,
+        head_mask: Optional[torch.Tensor] = None,
+        inputs_embeds: Optional[torch.Tensor] = None,
+        encoder_hidden_states: Optional[torch.Tensor] = None,
+        encoder_attention_mask: Optional[torch.Tensor] = None,
+        labels: Optional[torch.Tensor] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+    ) -> Union[tuple[torch.Tensor], MaskedLMOutput]:
+        r"""
+        labels (`torch.LongTensor` of shape `(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]`
+        """
+
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        outputs = self.bert(
+            input_ids,
+            attention_mask=attention_mask,
+            token_type_ids=token_type_ids,
+            position_ids=position_ids,
+            head_mask=head_mask,
+            inputs_embeds=inputs_embeds,
+            encoder_hidden_states=encoder_hidden_states,
+            encoder_attention_mask=encoder_attention_mask,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+        )
+
+        sequence_output = outputs[0]
+        prediction_scores = self.cls(sequence_output)
+
+        masked_lm_loss = None
+        if labels is not None:
+            loss_fct = CrossEntropyLoss()  # -100 index = padding token
+            masked_lm_loss = loss_fct(prediction_scores.view(-1, self.config.vocab_size), labels.view(-1))
+
+        if not return_dict:
+            output = (prediction_scores,) + outputs[2:]
+            return ((masked_lm_loss,) + output) if masked_lm_loss is not None else output
+
+        return MaskedLMOutput(
+            loss=masked_lm_loss,
+            logits=prediction_scores,
+            hidden_states=outputs.hidden_states,
+            attentions=outputs.attentions,
+        )
+
+    def prepare_inputs_for_generation(self, input_ids, attention_mask=None, **model_kwargs):
+        input_shape = input_ids.shape
+        effective_batch_size = input_shape[0]
+
+        #  add a dummy token
+        if self.config.pad_token_id is None:
+            raise ValueError("The PAD token should be defined for generation")
+
+        attention_mask = torch.cat([attention_mask, attention_mask.new_zeros((attention_mask.shape[0], 1))], dim=-1)
+        dummy_token = torch.full(
+            (effective_batch_size, 1), self.config.pad_token_id, dtype=torch.long, device=input_ids.device
+        )
+        input_ids = torch.cat([input_ids, dummy_token], dim=1)
+
+        return {"input_ids": input_ids, "attention_mask": attention_mask}
+
+    @classmethod
+    def can_generate(cls) -> bool:
+        """
+        Legacy correction: BertForMaskedLM can't call `generate()` from `GenerationMixin`, even though it has a
+        `prepare_inputs_for_generation` method.
+        """
+        return False
+
+
+@auto_docstring(
+    custom_intro="""
+    Bert Model with a `next sentence prediction (classification)` head on top.
+    """
+)
+class BertForNextSentencePrediction(BertPreTrainedModel):
+    def __init__(self, config):
+        super().__init__(config)
+
+        self.bert = BertModel(config)
+        self.cls = BertOnlyNSPHead(config)
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    @auto_docstring
+    def forward(
+        self,
+        input_ids: Optional[torch.Tensor] = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        token_type_ids: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.Tensor] = None,
+        head_mask: Optional[torch.Tensor] = None,
+        inputs_embeds: Optional[torch.Tensor] = None,
+        labels: Optional[torch.Tensor] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+        **kwargs,
+    ) -> Union[tuple[torch.Tensor], NextSentencePredictorOutput]:
+        r"""
+        labels (`torch.LongTensor` of shape `(batch_size,)`, *optional*):
+            Labels for computing the next sequence prediction (classification) loss. Input should be a sequence pair
+            (see `input_ids` docstring). Indices should be in `[0, 1]`:
+
+            - 0 indicates sequence B is a continuation of sequence A,
+            - 1 indicates sequence B is a random sequence.
+
+        Example:
+
+        ```python
+        >>> from transformers import AutoTokenizer, BertForNextSentencePrediction
+        >>> import torch
+
+        >>> tokenizer = AutoTokenizer.from_pretrained("google-bert/bert-base-uncased")
+        >>> model = BertForNextSentencePrediction.from_pretrained("google-bert/bert-base-uncased")
+
+        >>> prompt = "In Italy, pizza served in formal settings, such as at a restaurant, is presented unsliced."
+        >>> next_sentence = "The sky is blue due to the shorter wavelength of blue light."
+        >>> encoding = tokenizer(prompt, next_sentence, return_tensors="pt")
+
+        >>> outputs = model(**encoding, labels=torch.LongTensor([1]))
+        >>> logits = outputs.logits
+        >>> assert logits[0, 0] < logits[0, 1]  # next sentence was random
+        ```
+        """
+
+        if "next_sentence_label" in kwargs:
+            warnings.warn(
+                "The `next_sentence_label` argument is deprecated and will be removed in a future version, use"
+                " `labels` instead.",
+                FutureWarning,
+            )
+            labels = kwargs.pop("next_sentence_label")
+
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        outputs = self.bert(
+            input_ids,
+            attention_mask=attention_mask,
+            token_type_ids=token_type_ids,
+            position_ids=position_ids,
+            head_mask=head_mask,
+            inputs_embeds=inputs_embeds,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+        )
+
+        pooled_output = outputs[1]
+
+        seq_relationship_scores = self.cls(pooled_output)
+
+        next_sentence_loss = None
+        if labels is not None:
+            loss_fct = CrossEntropyLoss()
+            next_sentence_loss = loss_fct(seq_relationship_scores.view(-1, 2), labels.view(-1))
+
+        if not return_dict:
+            output = (seq_relationship_scores,) + outputs[2:]
+            return ((next_sentence_loss,) + output) if next_sentence_loss is not None else output
+
+        return NextSentencePredictorOutput(
+            loss=next_sentence_loss,
+            logits=seq_relationship_scores,
+            hidden_states=outputs.hidden_states,
+            attentions=outputs.attentions,
+        )
+
+
+@auto_docstring(
+    custom_intro="""
+    Bert Model transformer with a sequence classification/regression head on top (a linear layer on top of the pooled
+    output) e.g. for GLUE tasks.
+    """
+)
+class BertForSequenceClassification(BertPreTrainedModel):
+    def __init__(self, config):
+        super().__init__(config)
+        self.num_labels = config.num_labels
+        self.config = config
+
+        self.bert = BertModel(config)
+        classifier_dropout = (
+            config.classifier_dropout if config.classifier_dropout is not None else config.hidden_dropout_prob
+        )
+        self.dropout = nn.Dropout(classifier_dropout)
+        self.classifier = nn.Linear(config.hidden_size, config.num_labels)
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    @auto_docstring
+    def forward(
+        self,
+        input_ids: Optional[torch.Tensor] = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        token_type_ids: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.Tensor] = None,
+        head_mask: Optional[torch.Tensor] = None,
+        inputs_embeds: Optional[torch.Tensor] = None,
+        labels: Optional[torch.Tensor] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+    ) -> Union[tuple[torch.Tensor], SequenceClassifierOutput]:
+        r"""
+        labels (`torch.LongTensor` of shape `(batch_size,)`, *optional*):
+            Labels for computing the sequence classification/regression loss. Indices should be in `[0, ...,
+            config.num_labels - 1]`. If `config.num_labels == 1` a regression loss is computed (Mean-Square loss), If
+            `config.num_labels > 1` a classification loss is computed (Cross-Entropy).
+        """
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        outputs = self.bert(
+            input_ids,
+            attention_mask=attention_mask,
+            token_type_ids=token_type_ids,
+            position_ids=position_ids,
+            head_mask=head_mask,
+            inputs_embeds=inputs_embeds,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+        )
+
+        pooled_output = outputs[1]
+
+        pooled_output = self.dropout(pooled_output)
+        logits = self.classifier(pooled_output)
+
+        loss = None
+        if labels is not None:
+            if self.config.problem_type is None:
+                if self.num_labels == 1:
+                    self.config.problem_type = "regression"
+                elif self.num_labels > 1 and (labels.dtype == torch.long or labels.dtype == torch.int):
+                    self.config.problem_type = "single_label_classification"
+                else:
+                    self.config.problem_type = "multi_label_classification"
+
+            if self.config.problem_type == "regression":
+                loss_fct = MSELoss()
+                if self.num_labels == 1:
+                    loss = loss_fct(logits.squeeze(), labels.squeeze())
+                else:
+                    loss = loss_fct(logits, labels)
+            elif self.config.problem_type == "single_label_classification":
+                loss_fct = CrossEntropyLoss()
+                loss = loss_fct(logits.view(-1, self.num_labels), labels.view(-1))
+            elif self.config.problem_type == "multi_label_classification":
+                loss_fct = BCEWithLogitsLoss()
+                loss = loss_fct(logits, labels)
+        if not return_dict:
+            output = (logits,) + outputs[2:]
+            return ((loss,) + output) if loss is not None else output
+
+        return SequenceClassifierOutput(
+            loss=loss,
+            logits=logits,
+            hidden_states=outputs.hidden_states,
+            attentions=outputs.attentions,
+        )
+
+
+@auto_docstring
+class BertForMultipleChoice(BertPreTrainedModel):
+    def __init__(self, config):
+        super().__init__(config)
+
+        self.bert = BertModel(config)
+        classifier_dropout = (
+            config.classifier_dropout if config.classifier_dropout is not None else config.hidden_dropout_prob
+        )
+        self.dropout = nn.Dropout(classifier_dropout)
+        self.classifier = nn.Linear(config.hidden_size, 1)
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    @auto_docstring
+    def forward(
+        self,
+        input_ids: Optional[torch.Tensor] = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        token_type_ids: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.Tensor] = None,
+        head_mask: Optional[torch.Tensor] = None,
+        inputs_embeds: Optional[torch.Tensor] = None,
+        labels: Optional[torch.Tensor] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+    ) -> Union[tuple[torch.Tensor], MultipleChoiceModelOutput]:
+        r"""
+        input_ids (`torch.LongTensor` of shape `(batch_size, num_choices, sequence_length)`):
+            Indices of input sequence tokens in the vocabulary.
+
+            Indices can be obtained using [`AutoTokenizer`]. See [`PreTrainedTokenizer.encode`] and
+            [`PreTrainedTokenizer.__call__`] for details.
+
+            [What are input IDs?](../glossary#input-ids)
+        token_type_ids (`torch.LongTensor` of shape `(batch_size, num_choices, sequence_length)`, *optional*):
+            Segment token indices to indicate first and second portions of the inputs. Indices are selected in `[0,
+            1]`:
+
+            - 0 corresponds to a *sentence A* token,
+            - 1 corresponds to a *sentence B* token.
+
+            [What are token type IDs?](../glossary#token-type-ids)
+        position_ids (`torch.LongTensor` of shape `(batch_size, num_choices, sequence_length)`, *optional*):
+            Indices of positions of each input sequence tokens in the position embeddings. Selected in the range `[0,
+            config.max_position_embeddings - 1]`.
+
+            [What are position IDs?](../glossary#position-ids)
+        inputs_embeds (`torch.FloatTensor` of shape `(batch_size, num_choices, sequence_length, hidden_size)`, *optional*):
+            Optionally, instead of passing `input_ids` you can choose to directly pass an embedded representation. This
+            is useful if you want more control over how to convert `input_ids` indices into associated vectors than the
+            model's internal embedding lookup matrix.
+        labels (`torch.LongTensor` of shape `(batch_size,)`, *optional*):
+            Labels for computing the multiple choice classification loss. Indices should be in `[0, ...,
+            num_choices-1]` where `num_choices` is the size of the second dimension of the input tensors. (See
+            `input_ids` above)
+        """
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+        num_choices = input_ids.shape[1] if input_ids is not None else inputs_embeds.shape[1]
+
+        input_ids = input_ids.view(-1, input_ids.size(-1)) if input_ids is not None else None
+        attention_mask = attention_mask.view(-1, attention_mask.size(-1)) if attention_mask is not None else None
+        token_type_ids = token_type_ids.view(-1, token_type_ids.size(-1)) if token_type_ids is not None else None
+        position_ids = position_ids.view(-1, position_ids.size(-1)) if position_ids is not None else None
+        inputs_embeds = (
+            inputs_embeds.view(-1, inputs_embeds.size(-2), inputs_embeds.size(-1))
+            if inputs_embeds is not None
+            else None
+        )
+
+        outputs = self.bert(
+            input_ids,
+            attention_mask=attention_mask,
+            token_type_ids=token_type_ids,
+            position_ids=position_ids,
+            head_mask=head_mask,
+            inputs_embeds=inputs_embeds,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+        )
+
+        pooled_output = outputs[1]
+
+        pooled_output = self.dropout(pooled_output)
+        logits = self.classifier(pooled_output)
+        reshaped_logits = logits.view(-1, num_choices)
+
+        loss = None
+        if labels is not None:
+            loss_fct = CrossEntropyLoss()
+            loss = loss_fct(reshaped_logits, labels)
+
+        if not return_dict:
+            output = (reshaped_logits,) + outputs[2:]
+            return ((loss,) + output) if loss is not None else output
+
+        return MultipleChoiceModelOutput(
+            loss=loss,
+            logits=reshaped_logits,
+            hidden_states=outputs.hidden_states,
+            attentions=outputs.attentions,
+        )
+
+
+@auto_docstring
+class BertForTokenClassification(BertPreTrainedModel):
+    def __init__(self, config):
+        super().__init__(config)
+        self.num_labels = config.num_labels
+
+        self.bert = BertModel(config, add_pooling_layer=False)
+        classifier_dropout = (
+            config.classifier_dropout if config.classifier_dropout is not None else config.hidden_dropout_prob
+        )
+        self.dropout = nn.Dropout(classifier_dropout)
+        self.classifier = nn.Linear(config.hidden_size, config.num_labels)
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    @auto_docstring
+    def forward(
+        self,
+        input_ids: Optional[torch.Tensor] = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        token_type_ids: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.Tensor] = None,
+        head_mask: Optional[torch.Tensor] = None,
+        inputs_embeds: Optional[torch.Tensor] = None,
+        labels: Optional[torch.Tensor] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+    ) -> Union[tuple[torch.Tensor], TokenClassifierOutput]:
+        r"""
+        labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
+            Labels for computing the token classification loss. Indices should be in `[0, ..., config.num_labels - 1]`.
+        """
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        outputs = self.bert(
+            input_ids,
+            attention_mask=attention_mask,
+            token_type_ids=token_type_ids,
+            position_ids=position_ids,
+            head_mask=head_mask,
+            inputs_embeds=inputs_embeds,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+        )
+
+        sequence_output = outputs[0]
+
+        sequence_output = self.dropout(sequence_output)
+        logits = self.classifier(sequence_output)
+
+        loss = None
+        if labels is not None:
+            loss_fct = CrossEntropyLoss()
+            loss = loss_fct(logits.view(-1, self.num_labels), labels.view(-1))
+
+        if not return_dict:
+            output = (logits,) + outputs[2:]
+            return ((loss,) + output) if loss is not None else output
+
+        return TokenClassifierOutput(
+            loss=loss,
+            logits=logits,
+            hidden_states=outputs.hidden_states,
+            attentions=outputs.attentions,
+        )
+
+
+@auto_docstring
+class BertForQuestionAnswering(BertPreTrainedModel):
+    def __init__(self, config):
+        super().__init__(config)
+        self.num_labels = config.num_labels
+
+        self.bert = BertModel(config, add_pooling_layer=False)
+        self.qa_outputs = nn.Linear(config.hidden_size, config.num_labels)
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    @auto_docstring
+    def forward(
+        self,
+        input_ids: Optional[torch.Tensor] = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        token_type_ids: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.Tensor] = None,
+        head_mask: Optional[torch.Tensor] = None,
+        inputs_embeds: Optional[torch.Tensor] = None,
+        start_positions: Optional[torch.Tensor] = None,
+        end_positions: Optional[torch.Tensor] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+    ) -> Union[tuple[torch.Tensor], QuestionAnsweringModelOutput]:
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        outputs = self.bert(
+            input_ids,
+            attention_mask=attention_mask,
+            token_type_ids=token_type_ids,
+            position_ids=position_ids,
+            head_mask=head_mask,
+            inputs_embeds=inputs_embeds,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+        )
+
+        sequence_output = outputs[0]
+
+        logits = self.qa_outputs(sequence_output)
+        start_logits, end_logits = logits.split(1, dim=-1)
+        start_logits = start_logits.squeeze(-1).contiguous()
+        end_logits = end_logits.squeeze(-1).contiguous()
+
+        total_loss = None
+        if start_positions is not None and end_positions is not None:
+            # If we are on multi-GPU, split add a dimension
+            if len(start_positions.size()) > 1:
+                start_positions = start_positions.squeeze(-1)
+            if len(end_positions.size()) > 1:
+                end_positions = end_positions.squeeze(-1)
+            # sometimes the start/end positions are outside our model inputs, we ignore these terms
+            ignored_index = start_logits.size(1)
+            start_positions = start_positions.clamp(0, ignored_index)
+            end_positions = end_positions.clamp(0, ignored_index)
+
+            loss_fct = CrossEntropyLoss(ignore_index=ignored_index)
+            start_loss = loss_fct(start_logits, start_positions)
+            end_loss = loss_fct(end_logits, end_positions)
+            total_loss = (start_loss + end_loss) / 2
+
+        if not return_dict:
+            output = (start_logits, end_logits) + outputs[2:]
+            return ((total_loss,) + output) if total_loss is not None else output
+
+        return QuestionAnsweringModelOutput(
+            loss=total_loss,
+            start_logits=start_logits,
+            end_logits=end_logits,
+            hidden_states=outputs.hidden_states,
+            attentions=outputs.attentions,
+        )
+
+
+__all__ = [
+    "BertForMaskedLM",
+    "BertForMultipleChoice",
+    "BertForNextSentencePrediction",
+    "BertForPreTraining",
+    "BertForQuestionAnswering",
+    "BertForSequenceClassification",
+    "BertForTokenClassification",
+    "BertLayer",
+    "BertLMHeadModel",
+    "BertModel",
+    "BertPreTrainedModel",
+    "load_tf_weights_in_bert",
+]

tokenizer_config.json

@@ -0,0 +1,56 @@
+{
+  "added_tokens_decoder": {
+    "0": {
+      "content": "[PAD]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "100": {
+      "content": "[UNK]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "101": {
+      "content": "[CLS]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "102": {
+      "content": "[SEP]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "103": {
+      "content": "[MASK]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "clean_up_tokenization_spaces": false,
+  "cls_token": "[CLS]",
+  "do_lower_case": true,
+  "extra_special_tokens": {},
+  "mask_token": "[MASK]",
+  "model_max_length": 512,
+  "pad_token": "[PAD]",
+  "sep_token": "[SEP]",
+  "strip_accents": null,
+  "tokenize_chinese_chars": true,
+  "tokenizer_class": "BertTokenizer",
+  "unk_token": "[UNK]"
+}