Upload folder using huggingface_hub

1_Pooling/config.json

@@ -0,0 +1,5 @@
+{
+    "embedding_dimension": 768,
+    "pooling_mode": "lasttoken",
+    "include_prompt": true
+}

README.md

@@ -0,0 +1,265 @@
+---
+pipeline_tag: sentence-similarity
+tags:
+- gguf
+- embedding
+- eurobert
+- llama-cpp
+- jina-embeddings-v5
+- feature-extraction
+- mteb
+- vllm
+- sentence-transformers
+language:
+- multilingual
+base_model: jinaai/jina-embeddings-v5-text-nano
+base_model_relation: quantized
+inference: false
+license: cc-by-nc-4.0
+library_name: llama.cpp
+---
+<br><br>
+
+<p align="center">
+<img src="https://huggingface.co/datasets/jinaai/documentation-images/resolve/main/logo.webp" alt="Jina AI: Your Search Foundation, Supercharged!" width="150px">
+</p>
+
+### **jina-embeddings-v5-text**: Task-Targeted Embedding Distillation
+
+[Elastic Inference Service](https://www.elastic.co/docs/explore-analyze/elastic-inference/eis) | [ArXiv](https://arxiv.org/abs/2602.15547) | [Release Note](https://jina.ai/news/jina-embeddings-v5-text-distilling-4b-quality-into-sub-1b-multilingual-embeddings) | [Blog](https://www.elastic.co/search-labs/blog/jina-embeddings-v5-text)
+
+### Model Overview
+
+<p align="center">
+<img src="https://jina-ai-gmbh.ghost.io/content/images/2026/02/v5_architecture_1771470917.png" alt="jina-embeddings-v5-text Architecture" width="600px">
+</p>
+`jina-embeddings-v5-text-nano-retrieval` is a compact, high-performance text embedding model designed for information retrieval.
+
+It is part of the **jina-embeddings-v5-text** model family, which also includes [jina-embeddings-v5-text-small](https://huggingface.co/jinaai/jina-embeddings-v5-text-small), for better performance at a bigger size.
+
+Trained using a novel approach that combines distillation with task-specific contrastive losses, `jina-embeddings-v5-text-nano-retrieval` outperforms existing state-of-the-art models of similar size across diverse embedding benchmarks.
+| Feature | Value |
+| --- | --- |
+| Parameters | 239M |
+| Supported Tasks | `retrieval` |
+| Max Sequence Length | 8192 |
+| Embedding Dimension | 768 |
+| Matryoshka Dimensions | 32, 64, 128, 256, 512, 768 |
+| Pooling Strategy | Last-token pooling |
+| Base Model | jinaai/jina-embeddings-v5-text-nano |
+
+![image](https://cdn-uploads.huggingface.co/production/uploads/6476ff2699a5ce743ccea3fc/SJw9j09PkErQ0v9P052S9.png)
+
+
+### Training and Evaluation
+
+For training details and evaluation results, see our [technical report](https://arxiv.org/abs/2602.15547).
+
+### Usage
+
+<details>
+  <summary>Requirements</a></summary>
+  
+The following Python packages are required:
+
+- `transformers>=5.1.0`
+- `torch>=2.8.0`
+- `peft>=0.15.2`
+- `vllm==0.15.1`
+  
+### Optional / Recommended
+- **flash-attention**: Installing [flash-attention](https://github.com/Dao-AILab/flash-attention) is recommended for improved inference speed and efficiency, but not mandatory.
+- **sentence-transformers**: If you want to use the model via the `sentence-transformers` interface, install this package as well.
+
+</details>
+
+<details open>
+  <summary>via <a href="https://www.elastic.co/docs/explore-analyze/elastic-inference/eis">Elastic Inference Service</a></summary>
+
+The fastest way to use v5-text in production. Elastic Inference Service (EIS) provides managed embedding inference with built-in scaling, so you can generate embeddings directly within your Elastic deployment.
+
+```bash
+PUT _inference/text_embedding/jina-v5
+{
+  "service": "elastic",
+  "service_settings": {
+    "model_id": "jina-embeddings-v5-text-nano"
+  }
+}
+```
+
+See the [Elastic Inference Service documentation](https://www.elastic.co/docs/explore-analyze/elastic-inference/eis) for setup details.
+
+</details>
+
+
+<details>
+  <summary>via <a href="https://sbert.net/">sentence-transformers</a></summary>
+
+```python
+from sentence_transformers import SentenceTransformer
+import torch
+
+model = SentenceTransformer(
+    "jinaai/jina-embeddings-v5-text-nano-retrieval",
+    trust_remote_code=True,
+    model_kwargs={"dtype": torch.bfloat16},  # Recommended for GPUs
+    config_kwargs={"_attn_implementation": "flash_attention_2"},  # Recommended but optional
+)
+# Optional: set truncate_dim in encode() to control embedding size
+
+query = "Which planet is known as the Red Planet?"
+documents = [
+    "Venus is often called Earth's twin because of its similar size and proximity.",
+    "Mars, known for its reddish appearance, is often referred to as the Red Planet.",
+    "Jupiter, the largest planet in our solar system, has a prominent red spot.",
+    "Saturn, famous for its rings, is sometimes mistaken for the Red Planet.",
+]
+
+# Encode query and documents
+query_embeddings = model.encode(sentences=query, prompt_name="query")
+document_embeddings = model.encode(sentences=documents, prompt_name="document")
+print(query_embeddings.shape, document_embeddings.shape)
+# (768,) (4, 768)
+
+similarity = model.similarity(query_embeddings, document_embeddings)
+print(similarity)
+# tensor([[0.5013, 0.7914, 0.6133, 0.5736]])
+```
+
+</details>
+
+<details>
+  <summary>via <a href="https://github.com/vllm-project/vllm">vLLM</a></summary>
+
+```python
+from vllm import LLM
+from vllm.config.pooler import PoolerConfig
+
+# Initialize model
+name = "jinaai/jina-embeddings-v5-text-nano-retrieval"
+model = LLM(
+    model=name,
+    dtype="float16",
+    runner="pooling",
+    trust_remote_code=True,
+    pooler_config=PoolerConfig(seq_pooling_type="LAST", normalize=True),
+)
+
+# Create text prompts
+query = "Overview of climate change impacts on coastal cities"
+query_prompt = f"Query: {query}"
+
+document = "The impacts of climate change on coastal cities are significant.."
+document_prompt = f"Document: {document}"
+
+# Encode all prompts
+prompts = [query_prompt, document_prompt]
+outputs = model.encode(prompts, pooling_task="embed")
+
+```
+
+</details>
+
+<details>
+  <summary>via <a href="https://github.com/ggml-org/llama.cpp">llama.cpp (GGUF)</a></summary>
+
+Since our nano model is based on `jinaai/jina-embeddings-v5-text-nano`, which is not yet supported by llama.cpp, we provide our own branch of [llama.cpp](https://github.com/jina-ai/llama.cpp/tree/feat-jina-v5-text), which implements the necessary changes to support it for now.
+
+To start the OpenAI API compatible HTTP server, run with the respective model version:
+
+```
+llama-server \
+  -hf jinaai/jina-embeddings-v5-text-nano-retrieval:F16 \
+  --embedding \
+  --pooling last \
+  --batch-size 8192 \
+  --ubatch-size 8192 \
+  --ctx-size 8192
+```
+
+Client:
+
+```
+curl -X POST "http://127.0.0.1:8080/v1/embeddings" \
+  -H "Content-Type: application/json" \
+  -d '{
+    "input": [
+      "Query: A beautiful sunset over the beach",
+      "Query: Un beau coucher de soleil sur la plage",
+      "Document: 海滩上美丽的日落",
+      "Document: 浜辺に沈む美しい夕日",
+      "Document: Golden sunlight melts into the horizon, painting waves in warm amber and rose, while the sky whispers goodnight to the quiet, endless sea."
+    ]
+  }'
+```
+
+Note: For the retrieval variant, add `Query: ` or `Document: ` prefix in front of your input as shown above.
+
+</details>
+
+<details>
+  <summary> via <a href="https://huggingface.co/docs/optimum/index">Optimum (ONNX)</a></summary>
+
+You can run the ONNX-optimized version of the model locally using Hugging Face's `optimum` library. Make sure you have the required dependencies installed (e.g., `pip install optimum[onnxruntime] transformers torch`):
+
+```python
+from optimum.onnxruntime import ORTModelForFeatureExtraction
+from transformers import AutoTokenizer
+import torch
+
+model_id = "jinaai/jina-embeddings-v5-text-nano-retrieval"
+
+# 1. Load tokenizer and ONNX model
+# We specify the subfolder 'onnx' where the weights are located
+tokenizer = AutoTokenizer.from_pretrained(model_id, trust_remote_code=True)
+model = ORTModelForFeatureExtraction.from_pretrained(
+    model_id,
+    subfolder="onnx",
+    file_name="model.onnx",
+    provider="CPUExecutionProvider",  # Or "CUDAExecutionProvider" for GPU
+    trust_remote_code=True,
+)
+
+# 2. Prepare input
+texts = ["Query: How do I use Jina ONNX models?", "Document: Information about semantic matching."]
+inputs = tokenizer(texts, padding=True, truncation=True, return_tensors="pt")
+
+
+# 4. Inference
+with torch.no_grad():
+    outputs = model(**inputs)
+
+# 5. Pooling (Crucial for Jina-v5)
+# Jina-v5 uses LAST-TOKEN pooling.
+# We take the hidden state of the last non-padding token.
+last_hidden_state = outputs.last_hidden_state
+# Find the indices of the last token (usually the end of the sequence)
+sequence_lengths = inputs.attention_mask.sum(dim=1) - 1
+embeddings = last_hidden_state[torch.arange(last_hidden_state.size(0)), sequence_lengths]
+
+print('embeddings shape:', embeddings.shape)
+print('embeddings:', embeddings)
+```
+
+</details>
+
+### License
+
+The model is licensed under CC BY-NC 4.0. For commercial use, please [contact us](mailto:sales@jina.ai).
+
+### Citation
+
+If you find `jina-embeddings-v5-text-nano-retrieval` useful in your research, please cite the following paper:
+
+```
+@misc{akram2026jinaembeddingsv5texttasktargetedembeddingdistillation,
+      title={jina-embeddings-v5-text: Task-Targeted Embedding Distillation}, 
+      author={Mohammad Kalim Akram and Saba Sturua and Nastia Havriushenko and Quentin Herreros and Michael Günther and Maximilian Werk and Han Xiao},
+      year={2026},
+      eprint={2602.15547},
+      archivePrefix={arXiv},
+      primaryClass={cs.CL},
+      url={https://arxiv.org/abs/2602.15547}, 
+}
+```

config.json

@@ -0,0 +1,81 @@
+{
+  "architectures": [
+    "EuroBertModel"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "auto_map": {
+    "AutoConfig": "configuration_eurobert.EuroBertConfig",
+    "AutoModel": "modeling_eurobert.EuroBertModel",
+    "AutoModelForMaskedLM": "modeling_eurobert.EuroBertForMaskedLM"
+  },
+  "bos_token": "<|begin_of_text|>",
+  "bos_token_id": 33717,
+  "classifier_pooling": "late",
+  "clf_pooling": "late",
+  "dtype": "bfloat16",
+  "eos_token": "<|end_of_text|>",
+  "eos_token_id": 33718,
+  "head_dim": 64,
+  "hidden_act": "silu",
+  "hidden_dropout": 0.0,
+  "hidden_size": 768,
+  "initializer_range": 0.02,
+  "intermediate_size": 3072,
+  "is_decoder": false,
+  "mask_token": "<|mask|>",
+  "mask_token_id": 33719,
+  "max_position_embeddings": 8192,
+  "mlp_bias": false,
+  "model_type": "eurobert",
+  "num_attention_heads": 12,
+  "num_hidden_layers": 12,
+  "num_key_value_heads": 12,
+  "pad_token": "<|end_of_text|>",
+  "pad_token_id": 33718,
+  "pretraining_tp": 1,
+  "rms_norm_eps": 1e-05,
+  "rope_parameters": {
+    "factor": 1.0,
+    "rope_theta": 1000000,
+    "rope_type": "default"
+  },
+  "task_names": [
+    "retrieval",
+    "text-matching",
+    "clustering",
+    "classification"
+  ],
+  "tie_word_embeddings": false,
+  "transformers.js_config": {
+    "use_external_data_format": true
+  },
+  "transformers_version": "5.5.4",
+  "use_cache": false,
+  "vocab_size": 34048,
+  "vocabtrimmer": {
+    "mining_config": {
+      "dataset": [
+        "minhnguyent546/mc4-vi"
+      ],
+      "dataset_column": [
+        "text",
+        "pos"
+      ],
+      "dataset_split": "validation",
+      "min_frequency": 20,
+      "subset": null,
+      "target_vocab_size": null
+    },
+    "stats": {
+      "compression_rate_embedding": 26.546906187624753,
+      "compression_rate_full": 65.83411003963923,
+      "parameter_size_embedding/raw": 98500608,
+      "parameter_size_embedding/trimmed": 26148864,
+      "parameter_size_full/raw": 211766016,
+      "parameter_size_full/trimmed": 139414272,
+      "vocab_size/raw": 128256,
+      "vocab_size/trimmed": 34048
+    }
+  }
+}

config_sentence_transformers.json

@@ -0,0 +1,14 @@
+{
+  "__version__": {
+    "pytorch": "2.9.1+cu128",
+    "sentence_transformers": "5.4.1",
+    "transformers": "5.5.4"
+  },
+  "default_prompt_name": null,
+  "model_type": "SentenceTransformer",
+  "prompts": {
+    "document": "Document: ",
+    "query": "Query: "
+  },
+  "similarity_fn_name": "cosine"
+}

configuration_eurobert.py

@@ -0,0 +1,216 @@
+#                🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨
+#           This file was automatically generated from src/transformers/models/eurobert/modular_eurobert.py.
+#               Do NOT edit this file manually as any edits will be overwritten by the generation of
+#             the file from the modular. If any change should be done, please apply the change to the
+#                          modular_eurobert.py file directly. One of our CI enforces this.
+#                🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨
+# coding=utf-8
+# Copyright 2025 Nicolas Boizard, Duarte M. Alves, Hippolyte Gisserot-Boukhlef and the EuroBert team. 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.
+
+from transformers.utils import logging
+from transformers.models.llama import LlamaConfig
+
+
+logger = logging.get_logger(__name__)
+
+
+class EuroBertConfig(LlamaConfig):
+    r"""
+    This is the configuration class to store the configuration of a [`EuroBertModel`]. It is used to instantiate an EuroBert
+    model according to the specified arguments, defining the model architecture. Instantiating a configuration with the
+    defaults will yield a similar configuration to that of the EuroBERT-210m.
+
+    Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the
+    documentation from [`PretrainedConfig`] for more information.
+
+
+    Args:
+        vocab_size (`int`, *optional*, defaults to 128256):
+            Vocabulary size of the EuroBert model. Defines the number of different tokens that can be represented by the
+            `inputs_ids` passed when calling [`EuroBertModel`]
+        hidden_size (`int`, *optional*, defaults to 768):
+            Dimensionality of the encoder layers and the pooler layer.
+        intermediate_size (`int`, *optional*, defaults to 3072):
+            Dimensionality of the "intermediate" (often named feed-forward) layer in the Transformer encoder.
+        num_hidden_layers (`int`, *optional*, defaults to 12):
+            Number of hidden layers in the Transformer encoder.
+        num_attention_heads (`int`, *optional*, defaults to 12):
+            Number of attention heads for each attention layer in the Transformer encoder.
+        num_key_value_heads (`int`, *optional*):
+            This is the number of key_value heads that should be used to implement Grouped Query Attention. If
+            `num_key_value_heads=num_attention_heads`, the model will use Multi Head Attention (MHA), if
+            `num_key_value_heads=1` the model will use Multi Query Attention (MQA) otherwise GQA is used. When
+            converting a multi-head checkpoint to a GQA checkpoint, each group key and value head should be constructed
+            by meanpooling all the original heads within that group. For more details checkout [this
+            paper](https://arxiv.org/pdf/2305.13245.pdf). If it is not specified, will default to
+            `num_attention_heads`.
+        hidden_act (`str` or `function`, *optional*, defaults to `"silu"`):
+            The non-linear activation function (function or string) in the encoder and pooler.
+        max_position_embeddings (`int`, *optional*, defaults to 8192):
+            The maximum sequence length that this model might ever be used with. EuroBert supports up to 8192 tokens,
+            EuroBert-pretrained up to 2048.
+        initializer_range (`float`, *optional*, defaults to 0.02):
+            The standard deviation of the truncated_normal_initializer for initializing all weight matrices.
+        rms_norm_eps (`float`, *optional*, defaults to 1e-05):
+            The epsilon used by the rms normalization layers.
+        bos_token_id (`int`, *optional*, defaults to 128000):
+            Beginning of stream token id.
+        eos_token_id (`int`, *optional*, defaults to 128001):
+            End of stream token id.
+        pad_token_id (`int`, *optional*, defaults to 128001):
+            Padding token id.
+        mask_token_id (`int`, *optional*, defaults to 128002):
+            Mask token id.
+        pretraining_tp (`int`, *optional*, defaults to 1):
+            Experimental feature. Tensor parallelism rank used during pretraining. Please refer to [this
+            document](https://huggingface.co/docs/transformers/main/perf_train_gpu_many#tensor-parallelism) to
+            understand more about it. This value is necessary to ensure exact reproducibility of the pretraining
+            results. Please refer to [this issue](https://github.com/pytorch/pytorch/issues/76232).
+        tie_word_embeddings (`bool`, *optional*, defaults to `False`):
+            Whether to tie weight embeddings
+        rope_theta (`float`, *optional*, defaults to 250000.0):
+            The base period of the RoPE embeddings. EuroBert used base period of 250000.0,
+            EuroBert-pretrained 10000.0.
+        rope_scaling (`Dict`, *optional*):
+            Dictionary containing the scaling configuration for the RoPE embeddings. NOTE: if you apply new rope type
+            and you expect the model to work on longer `max_position_embeddings`, we recommend you to update this value
+            accordingly.
+            Expected contents:
+                `rope_type` (`str`):
+                    The sub-variant of RoPE to use. Can be one of ['default', 'linear', 'dynamic', 'yarn', 'longrope',
+                    'eurobert3'], with 'default' being the original RoPE implementation.
+                `factor` (`float`, *optional*):
+                    Used with all rope types except 'default'. The scaling factor to apply to the RoPE embeddings. In
+                    most scaling types, a `factor` of x will enable the model to handle sequences of length x *
+                    original maximum pre-trained length.
+                `original_max_position_embeddings` (`int`, *optional*):
+                    Used with 'dynamic', 'longrope' and 'eurobert3'. The original max position embeddings used during
+                    pretraining.
+                `attention_factor` (`float`, *optional*):
+                    Used with 'yarn' and 'longrope'. The scaling factor to be applied on the attention
+                    computation. If unspecified, it defaults to value recommended by the implementation, using the
+                    `factor` field to infer the suggested value.
+                `beta_fast` (`float`, *optional*):
+                    Only used with 'yarn'. Parameter to set the boundary for extrapolation (only) in the linear
+                    ramp function. If unspecified, it defaults to 32.
+                `beta_slow` (`float`, *optional*):
+                    Only used with 'yarn'. Parameter to set the boundary for interpolation (only) in the linear
+                    ramp function. If unspecified, it defaults to 1.
+                `short_factor` (`List[float]`, *optional*):
+                    Only used with 'longrope'. The scaling factor to be applied to short contexts (<
+                    `original_max_position_embeddings`). Must be a list of numbers with the same length as the hidden
+                    size divided by the number of attention heads divided by 2
+                `long_factor` (`List[float]`, *optional*):
+                    Only used with 'longrope'. The scaling factor to be applied to long contexts (<
+                    `original_max_position_embeddings`). Must be a list of numbers with the same length as the hidden
+                    size divided by the number of attention heads divided by 2
+                `low_freq_factor` (`float`, *optional*):
+                    Only used with 'eurobert3'. Scaling factor applied to low frequency components of the RoPE
+                `high_freq_factor` (`float`, *optional*):
+                    Only used with 'eurobert3'. Scaling factor applied to high frequency components of the RoPE
+        attention_bias (`bool`, *optional*, defaults to `False`):
+            Whether to use a bias in the query, key, value and output projection layers during self-attention.
+        attention_dropout (`float`, *optional*, defaults to 0.0):
+            The dropout ratio for the attention probabilities.
+        mlp_bias (`bool`, *optional*, defaults to `False`):
+            Whether to use a bias in up_proj, down_proj and gate_proj layers in the MLP layers.
+        head_dim (`int`, *optional*):
+            The attention head dimension. If None, it will default to hidden_size // num_attention_heads
+        classifier_pooling (`str`, *optional*, defaults to `"late"`):
+            The pooling strategy to use for the classifier. Can be one of ['bos', 'mean', 'late'].
+
+    ```python
+    >>> from transformers import EuroBertModel, EuroBertConfig
+
+    >>> # Initializing a EuroBert eurobert-base style configuration
+    >>> configuration = EuroBertConfig()
+
+    >>> # Initializing a model from the eurobert-base style configuration
+    >>> model = EuroBertModel(configuration)
+
+    >>> # Accessing the model configuration
+    >>> configuration = model.config
+    ```"""
+
+    model_type = "eurobert"
+
+    def __init__(
+        self,
+        vocab_size=128256,
+        hidden_size=768,
+        intermediate_size=3072,
+        num_hidden_layers=12,
+        num_attention_heads=12,
+        num_key_value_heads=None,
+        hidden_act="silu",
+        max_position_embeddings=8192,
+        initializer_range=0.02,
+        rms_norm_eps=1e-05,
+        bos_token_id=128000,
+        eos_token_id=128001,
+        pad_token_id=128001,
+        mask_token_id=128002,
+        pretraining_tp=1,
+        tie_word_embeddings=False,
+        rope_theta=250000.0,
+        rope_scaling=None,
+        attention_bias=False,
+        attention_dropout=0.0,
+        mlp_bias=False,
+        head_dim=None,
+        classifier_pooling="late",
+        **kwargs,
+    ):
+        # use_cache is specific to decoder models and should be set to False for encoder models
+        use_cache = kwargs.pop("use_cache", None)
+        if use_cache:
+            logger.warning_once(
+                "The `use_cache` argument to EuroBertConfig is set to `False`, as caching is never used for encoder models."
+            )
+
+        if num_key_value_heads is None:
+            num_key_value_heads = num_attention_heads
+
+        super().__init__(
+            vocab_size=vocab_size,
+            hidden_size=hidden_size,
+            intermediate_size=intermediate_size,
+            num_hidden_layers=num_hidden_layers,
+            num_attention_heads=num_attention_heads,
+            num_key_value_heads=num_key_value_heads,
+            hidden_act=hidden_act,
+            max_position_embeddings=max_position_embeddings,
+            initializer_range=initializer_range,
+            rms_norm_eps=rms_norm_eps,
+            use_cache=False,
+            bos_token_id=bos_token_id,
+            eos_token_id=eos_token_id,
+            pad_token_id=pad_token_id,
+            pretraining_tp=pretraining_tp,
+            tie_word_embeddings=tie_word_embeddings,
+            rope_theta=rope_theta,
+            rope_scaling=rope_scaling,
+            attention_bias=attention_bias,
+            attention_dropout=attention_dropout,
+            mlp_bias=mlp_bias,
+            head_dim=head_dim,
+            **kwargs,
+        )
+        self.mask_token_id = mask_token_id
+        self.clf_pooling = classifier_pooling
+
+
+__all__ = ["EuroBertConfig"]

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5eabe908e7c1933301f9e283dc009e9f65705c7149404b79386d3d4e8ef83abf
+size 278840176

modeling_eurobert.py

@@ -0,0 +1,1094 @@
+#                🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨
+#           This file was automatically generated from src/transformers/models/eurobert/modular_eurobert.py.
+#               Do NOT edit this file manually as any edits will be overwritten by the generation of
+#             the file from the modular. If any change should be done, please apply the change to the
+#                          modular_eurobert.py file directly. One of our CI enforces this.
+#                🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨
+# coding=utf-8
+# Copyright 2025 Nicolas Boizard, Duarte M. Alves, Hippolyte Gisserot-Boukhlef and the EuroBert team. 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.
+
+from typing import Callable, Optional, Tuple, Union
+
+import torch
+from torch import nn
+from torch.nn import BCEWithLogitsLoss, CrossEntropyLoss, MSELoss
+
+try:
+    from transformers import initialization as init
+except ImportError:  # transformers < v5
+    from transformers import modeling_utils as _modeling_utils
+    init = getattr(_modeling_utils, "init", torch.nn.init)
+
+from transformers.activations import ACT2FN
+from transformers.cache_utils import Cache, StaticCache
+from transformers.modeling_attn_mask_utils import AttentionMaskConverter
+from transformers.modeling_flash_attention_utils import FlashAttentionKwargs
+from transformers.modeling_outputs import (
+    BaseModelOutput,
+    BaseModelOutputWithPast,
+    MaskedLMOutput,
+    QuestionAnsweringModelOutput,
+    SequenceClassifierOutput,
+    TokenClassifierOutput,
+)
+from transformers.modeling_rope_utils import ROPE_INIT_FUNCTIONS
+from transformers.modeling_utils import ALL_ATTENTION_FUNCTIONS, PreTrainedModel
+from transformers.processing_utils import Unpack
+from transformers.utils import (
+    add_code_sample_docstrings,
+    add_start_docstrings,
+    add_start_docstrings_to_model_forward,
+    logging,
+)
+
+try:
+    from .configuration_eurobert import EuroBertConfig
+except ImportError:
+    from configuration_eurobert import EuroBertConfig
+
+
+logger = logging.get_logger(__name__)
+
+_CHECKPOINT_FOR_DOC = "EuroBERT/EuroBERT-210m"
+_CONFIG_FOR_DOC = "EuroBertConfig"
+
+
+class EuroBertRMSNorm(nn.Module):
+    def __init__(self, hidden_size, eps=1e-5):
+        """
+        EuroBertRMSNorm is equivalent to T5LayerNorm
+        """
+        super().__init__()
+        self.weight = nn.Parameter(torch.ones(hidden_size))
+        self.variance_epsilon = eps
+
+    def forward(self, hidden_states):
+        input_dtype = hidden_states.dtype
+        hidden_states = hidden_states.to(torch.float32)
+        variance = hidden_states.pow(2).mean(-1, keepdim=True)
+        hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)
+        return self.weight * hidden_states.to(input_dtype)
+
+    def extra_repr(self):
+        return f"{tuple(self.weight.shape)}, eps={self.variance_epsilon}"
+
+
+def rotate_half(x):
+    """Rotates half the hidden dims of the input."""
+    x1 = x[..., : x.shape[-1] // 2]
+    x2 = x[..., x.shape[-1] // 2 :]
+    return torch.cat((-x2, x1), dim=-1)
+
+
+def apply_rotary_pos_emb(q, k, cos, sin, position_ids=None, unsqueeze_dim=1):
+    """Applies Rotary Position Embedding to the query and key tensors.
+
+    Args:
+        q (`torch.Tensor`): The query tensor.
+        k (`torch.Tensor`): The key tensor.
+        cos (`torch.Tensor`): The cosine part of the rotary embedding.
+        sin (`torch.Tensor`): The sine part of the rotary embedding.
+        position_ids (`torch.Tensor`, *optional*):
+            Deprecated and unused.
+        unsqueeze_dim (`int`, *optional*, defaults to 1):
+            The 'unsqueeze_dim' argument specifies the dimension along which to unsqueeze cos[position_ids] and
+            sin[position_ids] so that they can be properly broadcasted to the dimensions of q and k. For example, note
+            that cos[position_ids] and sin[position_ids] have the shape [batch_size, seq_len, head_dim]. Then, if q and
+            k have the shape [batch_size, heads, seq_len, head_dim], then setting unsqueeze_dim=1 makes
+            cos[position_ids] and sin[position_ids] broadcastable to the shapes of q and k. Similarly, if q and k have
+            the shape [batch_size, seq_len, heads, head_dim], then set unsqueeze_dim=2.
+    Returns:
+        `tuple(torch.Tensor)` comprising of the query and key tensors rotated using the Rotary Position Embedding.
+    """
+    cos = cos.unsqueeze(unsqueeze_dim)
+    sin = sin.unsqueeze(unsqueeze_dim)
+    q_embed = (q * cos) + (rotate_half(q) * sin)
+    k_embed = (k * cos) + (rotate_half(k) * sin)
+    return q_embed, k_embed
+
+
+def repeat_kv(hidden_states: torch.Tensor, n_rep: int) -> torch.Tensor:
+    """
+    This is the equivalent of torch.repeat_interleave(x, dim=1, repeats=n_rep). The hidden states go from (batch,
+    num_key_value_heads, seqlen, head_dim) to (batch, num_attention_heads, seqlen, head_dim)
+    """
+    batch, num_key_value_heads, slen, head_dim = hidden_states.shape
+    if n_rep == 1:
+        return hidden_states
+    hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)
+    return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)
+
+
+def eager_attention_forward(
+    module: nn.Module,
+    query: torch.Tensor,
+    key: torch.Tensor,
+    value: torch.Tensor,
+    attention_mask: Optional[torch.Tensor],
+    scaling: float,
+    dropout: float = 0.0,
+    **kwargs,
+):
+    key_states = repeat_kv(key, module.num_key_value_groups)
+    value_states = repeat_kv(value, module.num_key_value_groups)
+
+    attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling
+    if attention_mask is not None:
+        causal_mask = attention_mask[:, :, :, : key_states.shape[-2]]
+        attn_weights = attn_weights + causal_mask
+
+    attn_weights = nn.functional.softmax(attn_weights, dim=-1, dtype=torch.float32).to(query.dtype)
+    attn_weights = nn.functional.dropout(attn_weights, p=dropout, training=module.training)
+    attn_output = torch.matmul(attn_weights, value_states)
+    attn_output = attn_output.transpose(1, 2).contiguous()
+
+    return attn_output, attn_weights
+
+
+class EuroBertAttention(nn.Module):
+    """Multi-headed attention from 'Attention Is All You Need' paper"""
+
+    def __init__(self, config: EuroBertConfig, layer_idx: int):
+        super().__init__()
+        self.config = config
+        self.layer_idx = layer_idx
+        self.head_dim = getattr(config, "head_dim", config.hidden_size // config.num_attention_heads)
+        self.num_key_value_groups = config.num_attention_heads // config.num_key_value_heads
+        self.scaling = self.head_dim**-0.5
+        self.attention_dropout = config.attention_dropout
+        self.is_causal = False
+
+        self.q_proj = nn.Linear(
+            config.hidden_size, config.num_attention_heads * self.head_dim, bias=config.attention_bias
+        )
+        self.k_proj = nn.Linear(
+            config.hidden_size, config.num_key_value_heads * self.head_dim, bias=config.attention_bias
+        )
+        self.v_proj = nn.Linear(
+            config.hidden_size, config.num_key_value_heads * self.head_dim, bias=config.attention_bias
+        )
+        self.o_proj = nn.Linear(
+            config.num_attention_heads * self.head_dim, config.hidden_size, bias=config.attention_bias
+        )
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        position_embeddings: Tuple[torch.Tensor, torch.Tensor],
+        attention_mask: Optional[torch.Tensor],
+        **kwargs: Unpack[FlashAttentionKwargs],
+    ) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
+        input_shape = hidden_states.shape[:-1]
+
+        hidden_shape = (*input_shape, -1, self.head_dim)
+
+        query_states = self.q_proj(hidden_states).view(hidden_shape).transpose(1, 2)
+        key_states = self.k_proj(hidden_states).view(hidden_shape).transpose(1, 2)
+        value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)
+
+        cos, sin = position_embeddings
+        query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)
+
+        attention_interface: Callable = eager_attention_forward
+        if self.config._attn_implementation != "eager":
+            if self.config._attn_implementation == "sdpa" and kwargs.get("output_attentions", False):
+                logger.warning_once(
+                    "`torch.nn.functional.scaled_dot_product_attention` does not support `output_attentions=True`. Falling back to "
+                    'eager attention. This warning can be removed using the argument `attn_implementation="eager"` when loading the model.'
+                )
+            else:
+                attention_interface = ALL_ATTENTION_FUNCTIONS[self.config._attn_implementation]
+        attn_output, attn_weights = attention_interface(
+            self,
+            query_states,
+            key_states,
+            value_states,
+            attention_mask,
+            dropout=0.0 if not self.training else self.attention_dropout,
+            scaling=self.scaling,
+            is_causal=False,
+            **kwargs,
+        )
+        attn_output = attn_output.reshape(*input_shape, -1).contiguous()
+        attn_output = self.o_proj(attn_output)
+        return attn_output, attn_weights
+
+
+EUROBERT_START_DOCSTRING = r"""
+    This model inherits from [`PreTrainedModel`]. Check the superclass documentation for the generic methods the
+    library implements for all its model (such as downloading or saving, resizing the input embeddings, pruning heads
+    etc.)
+
+    This model is also a PyTorch [torch.nn.Module](https://pytorch.org/docs/stable/nn.html#torch.nn.Module) subclass.
+    Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage
+    and behavior.
+
+    Parameters:
+        config ([`EuroBertConfig`]):
+            Model configuration class with all the parameters of the model. Initializing with a config file does not
+            load the weights associated with the model, only the configuration. Check out the
+            [`~PreTrainedModel.from_pretrained`] method to load the model weights.
+"""
+
+
+@add_start_docstrings(
+    "The bare EuroBERT Model outputting raw hidden-states without any specific head on top.",
+    EUROBERT_START_DOCSTRING,
+)
+class EuroBertPreTrainedModel(PreTrainedModel):
+    config_class = EuroBertConfig
+    base_model_prefix = "model"
+    supports_gradient_checkpointing = True
+    _no_split_modules = ["EuroBertDecoderLayer"]
+    _skip_keys_device_placement = ["past_key_values"]
+    _supports_flash_attn_2 = True
+    _supports_sdpa = True
+    _supports_flex_attn = True
+    _supports_cache_class = True
+    _supports_quantized_cache = True
+    _supports_static_cache = True
+    _supports_attention_backend = True
+
+    def _init_weights(self, module):
+        std = self.config.initializer_range
+        if isinstance(module, nn.Linear):
+            init.normal_(module.weight, mean=0.0, std=std)
+            if module.bias is not None:
+                init.zeros_(module.bias)
+        elif isinstance(module, nn.Embedding):
+            init.normal_(module.weight, mean=0.0, std=std)
+            if module.padding_idx is not None and not getattr(module.weight, "_is_hf_initialized", False):
+                init.zeros_(module.weight[module.padding_idx])
+        elif isinstance(module, EuroBertRotaryEmbedding):
+            rope_fn = ROPE_INIT_FUNCTIONS[module.rope_type]
+            buffer_value, _ = rope_fn(module.config, device=module.inv_freq.device)
+            if hasattr(init, "copy_"):
+                init.copy_(module.inv_freq, buffer_value)
+                init.copy_(module.original_inv_freq, buffer_value)
+            else:
+                module.inv_freq.copy_(buffer_value)
+                module.original_inv_freq.copy_(buffer_value)
+
+
+class EuroBertRotaryEmbedding(nn.Module):
+    def __init__(self, config: EuroBertConfig, device=None):
+        super().__init__()
+        # BC: "rope_type" was originally "type"
+        if hasattr(config, "rope_scaling") and config.rope_scaling is not None:
+            self.rope_type = config.rope_scaling.get("rope_type", config.rope_scaling.get("type"))
+        else:
+            self.rope_type = "default"
+        if self.rope_type == "default":
+            self.rope_type = "linear"
+            # Ensure rope_scaling is set up with factor=1.0 for linear (no scaling, equivalent to default)
+            if not hasattr(config, "rope_scaling") or config.rope_scaling is None:
+                config.rope_scaling = {"rope_type": "linear", "factor": 1.0}
+            elif "factor" not in config.rope_scaling:
+                config.rope_scaling["factor"] = 1.0
+        self.max_seq_len_cached = config.max_position_embeddings
+        self.original_max_seq_len = config.max_position_embeddings
+
+        self.config = config
+        self.rope_init_fn = ROPE_INIT_FUNCTIONS[self.rope_type]
+        inv_freq, self.attention_scaling = self.rope_init_fn(self.config, device)
+        self.register_buffer("inv_freq", inv_freq, persistent=False)
+        self.original_inv_freq = self.inv_freq
+
+    def _dynamic_frequency_update(self, position_ids, device):
+        """
+        dynamic RoPE layers should recompute `inv_freq` in the following situations:
+        1 - growing beyond the cached sequence length (allow scaling)
+        2 - the current sequence length is in the original scale (avoid losing precision with small sequences)
+        """
+        seq_len = torch.max(position_ids) + 1
+        if seq_len > self.max_seq_len_cached:  # growth
+            inv_freq, self.attention_scaling = self.rope_init_fn(self.config, device, seq_len=seq_len)
+            self.register_buffer("inv_freq", inv_freq, persistent=False)  # TODO joao: may break with compilation
+            self.max_seq_len_cached = seq_len
+
+        if seq_len < self.original_max_seq_len and self.max_seq_len_cached > self.original_max_seq_len:  # reset
+            # This .to() is needed if the model has been moved to a device after being initialized (because
+            # the buffer is automatically moved, but not the original copy)
+            self.original_inv_freq = self.original_inv_freq.to(device)
+            self.register_buffer("inv_freq", self.original_inv_freq, persistent=False)
+            self.max_seq_len_cached = self.original_max_seq_len
+
+    @torch.no_grad()
+    def forward(self, x, position_ids):
+        if "dynamic" in self.rope_type:
+            self._dynamic_frequency_update(position_ids, device=x.device)
+
+        # Core RoPE block
+        inv_freq_expanded = self.inv_freq[None, :, None].float().expand(position_ids.shape[0], -1, 1)
+        position_ids_expanded = position_ids[:, None, :].float()
+        # Force float32 (see https://github.com/huggingface/transformers/pull/29285)
+        device_type = x.device.type
+        device_type = device_type if isinstance(device_type, str) and device_type != "mps" else "cpu"
+        with torch.autocast(device_type=device_type, enabled=False):
+            freqs = (inv_freq_expanded.float() @ position_ids_expanded.float()).transpose(1, 2)
+            emb = torch.cat((freqs, freqs), dim=-1)
+            cos = emb.cos()
+            sin = emb.sin()
+
+        # Advanced RoPE types (e.g. yarn) apply a post-processing scaling factor, equivalent to scaling attention
+        cos = cos * self.attention_scaling
+        sin = sin * self.attention_scaling
+
+        return cos.to(dtype=x.dtype), sin.to(dtype=x.dtype)
+
+
+class EuroBertMLP(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        self.config = config
+        self.hidden_size = config.hidden_size
+        self.intermediate_size = config.intermediate_size
+        self.gate_proj = nn.Linear(self.hidden_size, self.intermediate_size, bias=config.mlp_bias)
+        self.up_proj = nn.Linear(self.hidden_size, self.intermediate_size, bias=config.mlp_bias)
+        self.down_proj = nn.Linear(self.intermediate_size, self.hidden_size, bias=config.mlp_bias)
+        self.act_fn = ACT2FN[config.hidden_act]
+
+    def forward(self, x):
+        down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))
+        return down_proj
+
+
+class EuroBertDecoderLayer(nn.Module):
+    def __init__(self, config: EuroBertConfig, layer_idx: int):
+        super().__init__()
+        self.hidden_size = config.hidden_size
+
+        self.self_attn = EuroBertAttention(config=config, layer_idx=layer_idx)
+
+        self.mlp = EuroBertMLP(config)
+        self.input_layernorm = EuroBertRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
+        self.post_attention_layernorm = EuroBertRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_value: Optional[Cache] = None,
+        output_attentions: Optional[bool] = False,
+        use_cache: Optional[bool] = False,
+        cache_position: Optional[torch.LongTensor] = None,
+        position_embeddings: Optional[Tuple[torch.Tensor, torch.Tensor]] = None,  # necessary, but kept here for BC
+        **kwargs: Unpack[FlashAttentionKwargs],
+    ) -> Tuple[torch.FloatTensor, Optional[Tuple[torch.FloatTensor, torch.FloatTensor]]]:
+        residual = hidden_states
+        
+        hidden_states = self.input_layernorm(hidden_states)
+
+        # Self Attention
+        hidden_states, self_attn_weights = self.self_attn(
+            hidden_states=hidden_states,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            past_key_value=past_key_value,
+            output_attentions=output_attentions,
+            use_cache=use_cache,
+            cache_position=cache_position,
+            position_embeddings=position_embeddings,
+            **kwargs,
+        )
+
+        hidden_states = residual + hidden_states
+
+        # Fully Connected
+        residual = hidden_states
+        hidden_states = self.post_attention_layernorm(hidden_states)
+        hidden_states = self.mlp(hidden_states)
+        hidden_states = residual + hidden_states
+
+        outputs = (hidden_states,)
+        if output_attentions:
+            outputs += (self_attn_weights,)
+
+        return outputs
+
+
+EUROBERT_INPUTS_DOCSTRING = r"""
+    Args:
+        input_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`):
+            Indices of input sequence tokens in the vocabulary. Padding will be ignored by default should you provide
+            it.
+
+            Indices can be obtained using [`AutoTokenizer`]. See [`PreTrainedTokenizer.encode`] and
+            [`PreTrainedTokenizer.__call__`] for details.
+
+            [What are input IDs?](../glossary#input-ids)
+        attention_mask (`torch.Tensor` of shape `(batch_size, sequence_length)`, *optional*):
+            Mask to avoid performing attention on padding token indices. Mask values selected in `[0, 1]`:
+
+            - 1 for tokens that are **not masked**,
+            - 0 for tokens that are **masked**.
+
+            [What are attention masks?](../glossary#attention-mask)
+
+            Indices can be obtained using [`AutoTokenizer`]. See [`PreTrainedTokenizer.encode`] and
+            [`PreTrainedTokenizer.__call__`] for details.
+
+            If `past_key_values` is used, optionally only the last `input_ids` have to be input (see
+            `past_key_values`).
+
+            If you want to change padding behavior, you should read [`modeling_opt._prepare_decoder_attention_mask`]
+            and modify to your needs. See diagram 1 in [the paper](https://arxiv.org/abs/1910.13461) for more
+            information on the default strategy.
+
+            - 1 indicates the head is **not masked**,
+            - 0 indicates the head is **masked**.
+        position_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
+            Indices of positions of each input sequence tokens in the position embeddings. Selected in the range `[0,
+            config.n_positions - 1]`.
+
+            [What are position IDs?](../glossary#position-ids)
+        past_key_values (`Cache` or `tuple(tuple(torch.FloatTensor))`, *optional*):
+            Pre-computed hidden-states (key and values in the self-attention blocks and in the cross-attention
+            blocks) that can be used to speed up sequential decoding. This typically consists in the `past_key_values`
+            returned by the model at a previous stage of decoding, when `use_cache=True` or `config.use_cache=True`.
+
+            Two formats are allowed:
+            - a [`~cache_utils.Cache`] instance, see our
+            [kv cache guide](https://huggingface.co/docs/transformers/en/kv_cache);
+            - Tuple of `tuple(torch.FloatTensor)` of length `config.n_layers`, with each tuple having 2 tensors of
+            shape `(batch_size, num_heads, sequence_length, embed_size_per_head)`). This is also known as the legacy
+            cache format.
+
+            The model will output the same cache format that is fed as input. If no `past_key_values` are passed, the
+            legacy cache format will be returned.
+
+            If `past_key_values` are used, the user can optionally input only the last `input_ids` (those that don't
+            have their past key value states given to this model) of shape `(batch_size, 1)` instead of all `input_ids`
+            of shape `(batch_size, sequence_length)`.
+        inputs_embeds (`torch.FloatTensor` of shape `(batch_size, 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.
+        use_cache (`bool`, *optional*):
+            If set to `True`, `past_key_values` key value states are returned and can be used to speed up decoding (see
+            `past_key_values`).
+        output_attentions (`bool`, *optional*):
+            Whether or not to return the attentions tensors of all attention layers. See `attentions` under returned
+            tensors for more detail.
+        output_hidden_states (`bool`, *optional*):
+            Whether or not to return the hidden states of all layers. See `hidden_states` under returned tensors for
+            more detail.
+        return_dict (`bool`, *optional*):
+            Whether or not to return a [`~utils.ModelOutput`] instead of a plain tuple.
+        cache_position (`torch.LongTensor` of shape `(sequence_length)`, *optional*):
+            Indices depicting the position of the input sequence tokens in the sequence. Contrarily to `position_ids`,
+            this tensor is not affected by padding. It is used to update the cache in the correct position and to infer
+            the complete sequence length.
+"""
+
+
+@add_start_docstrings(
+    "The bare EuroBert Model outputting raw hidden-states without any specific head on top.",
+    EUROBERT_START_DOCSTRING,
+)
+class EuroBertModel(EuroBertPreTrainedModel):
+    """
+    Transformer encoder consisting of *config.num_hidden_layers* layers. Each layer is a [`EuroBertDecoderLayer`]
+
+    Args:
+        config: EuroBertConfig
+    """
+
+    def __init__(self, config: EuroBertConfig):
+        super().__init__(config)
+        self.padding_idx = config.pad_token_id
+        self.vocab_size = config.vocab_size
+
+        self.embed_tokens = nn.Embedding(config.vocab_size, config.hidden_size, self.padding_idx)
+        self.layers = nn.ModuleList(
+            [EuroBertDecoderLayer(config, layer_idx) for layer_idx in range(config.num_hidden_layers)]
+        )
+        self.norm = EuroBertRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
+        self.rotary_emb = EuroBertRotaryEmbedding(config=config)
+        self.gradient_checkpointing = False
+        self.mask_converter = AttentionMaskConverter(is_causal=False)
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    def get_input_embeddings(self):
+        return self.embed_tokens
+
+    def set_input_embeddings(self, value):
+        self.embed_tokens = value
+
+    @add_start_docstrings_to_model_forward(EUROBERT_INPUTS_DOCSTRING)
+    @add_code_sample_docstrings(
+        checkpoint=_CHECKPOINT_FOR_DOC,
+        output_type=BaseModelOutput,
+        config_class=_CONFIG_FOR_DOC,
+    )
+    def forward(
+        self,
+        input_ids: torch.LongTensor = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        inputs_embeds: Optional[torch.FloatTensor] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+        **flash_attn_kwargs: Unpack[FlashAttentionKwargs],
+    ) -> Union[Tuple, BaseModelOutputWithPast]:
+        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 (input_ids is None) ^ (inputs_embeds is not None):
+            raise ValueError("You must specify exactly one of input_ids or inputs_embeds")
+
+        if inputs_embeds is None:
+            inputs_embeds = self.embed_tokens(input_ids)
+
+        if attention_mask is not None and self.config._attn_implementation != "flash_attention_2":
+            mask = self.mask_converter.to_4d(attention_mask, attention_mask.shape[1], inputs_embeds.dtype)
+        else:
+            mask = attention_mask
+
+        hidden_states = inputs_embeds
+
+        # create position embeddings to be shared across the encoder layers
+        if position_ids is None:
+            position_ids = torch.arange(inputs_embeds.shape[1], device=inputs_embeds.device).unsqueeze(0)
+        position_embeddings = self.rotary_emb(hidden_states, position_ids)
+
+        # encoder layers
+        all_hidden_states = () if output_hidden_states else None
+        all_self_attns = () if output_attentions else None
+
+        for encoder_layer in self.layers[: self.config.num_hidden_layers]:
+            if output_hidden_states:
+                all_hidden_states += (hidden_states,)
+
+            if self.gradient_checkpointing and self.training:
+                layer_outputs = self._gradient_checkpointing_func(
+                    encoder_layer.__call__,
+                    hidden_states,
+                    mask,
+                    position_ids,
+                    None,
+                    output_attentions,
+                    False,
+                    None,
+                    position_embeddings,
+                )
+            else:
+                layer_outputs = encoder_layer(
+                    hidden_states,
+                    attention_mask=mask,
+                    position_ids=position_ids,
+                    output_attentions=output_attentions,
+                    position_embeddings=position_embeddings,
+                    **flash_attn_kwargs,
+                )
+
+            hidden_states = layer_outputs[0]
+
+            if output_attentions:
+                all_self_attns += (layer_outputs[1],)
+
+        hidden_states = self.norm(hidden_states)
+
+        # add hidden states from the last encoder layer
+        if output_hidden_states:
+            all_hidden_states += (hidden_states,)
+
+        output = BaseModelOutput(
+            last_hidden_state=hidden_states,
+            hidden_states=all_hidden_states,
+            attentions=all_self_attns,
+        )
+        return output if return_dict else output.to_tuple()
+
+    def _update_causal_mask(
+        self,
+        attention_mask: torch.Tensor,
+        input_tensor: torch.Tensor,
+        cache_position: torch.Tensor,
+        past_key_values: Cache,
+        output_attentions: bool,
+    ):
+        if self.config._attn_implementation == "flash_attention_2":
+            if attention_mask is not None and (attention_mask == 0.0).any():
+                return attention_mask
+            return None
+
+        # For SDPA, when possible, we will rely on its `is_causal` argument instead of its `attn_mask` argument, in
+        # order to dispatch on Flash Attention 2. This feature is not compatible with static cache, as SDPA will fail
+        # to infer the attention mask.
+        past_seen_tokens = past_key_values.get_seq_length() if past_key_values is not None else 0
+        using_static_cache = isinstance(past_key_values, StaticCache)
+
+        # When output attentions is True, sdpa implementation's forward method calls the eager implementation's forward
+        if self.config._attn_implementation == "sdpa" and not using_static_cache and not output_attentions:
+            if AttentionMaskConverter._ignore_causal_mask_sdpa(
+                attention_mask,
+                inputs_embeds=input_tensor,
+                past_key_values_length=past_seen_tokens,
+                is_training=self.training,
+            ):
+                return None
+
+        dtype, device = input_tensor.dtype, input_tensor.device
+        sequence_length = input_tensor.shape[1]
+        if using_static_cache:
+            target_length = past_key_values.get_max_cache_shape()
+        else:
+            target_length = (
+                attention_mask.shape[-1]
+                if isinstance(attention_mask, torch.Tensor)
+                else past_seen_tokens + sequence_length + 1
+            )
+
+        # In case the provided `attention` mask is 2D, we generate a causal mask here (4D).
+        causal_mask = self._prepare_4d_causal_attention_mask_with_cache_position(
+            attention_mask,
+            sequence_length=sequence_length,
+            target_length=target_length,
+            dtype=dtype,
+            device=device,
+            cache_position=cache_position,
+            batch_size=input_tensor.shape[0],
+        )
+
+        if (
+            self.config._attn_implementation == "sdpa"
+            and attention_mask is not None
+            and attention_mask.device.type in ["cuda", "xpu"]
+            and not output_attentions
+        ):
+            # Attend to all tokens in fully masked rows in the causal_mask, for example the relevant first rows when
+            # using left padding. This is required by F.scaled_dot_product_attention memory-efficient attention path.
+            # Details: https://github.com/pytorch/pytorch/issues/110213
+            min_dtype = torch.finfo(dtype).min
+            causal_mask = AttentionMaskConverter._unmask_unattended(causal_mask, min_dtype)
+
+        return causal_mask
+
+    @staticmethod
+    def _prepare_4d_causal_attention_mask_with_cache_position(
+        attention_mask: torch.Tensor,
+        sequence_length: int,
+        target_length: int,
+        dtype: torch.dtype,
+        device: torch.device,
+        cache_position: torch.Tensor,
+        batch_size: int,
+        **kwargs,
+    ):
+        """
+        Creates a causal 4D mask of shape `(batch_size, 1, query_length, key_value_length)` from a 2D mask of shape
+        `(batch_size, key_value_length)`, or if the input `attention_mask` is already 4D, do nothing.
+
+        Args:
+            attention_mask (`torch.Tensor`):
+                A 2D attention mask of shape `(batch_size, key_value_length)` or a 4D attention mask of shape
+                `(batch_size, 1, query_length, key_value_length)`.
+            sequence_length (`int`):
+                The sequence length being processed.
+            target_length (`int`):
+                The target length: when generating with static cache, the mask should be as long as the static cache,
+                to account for the 0 padding, the part of the cache that is not filled yet.
+            dtype (`torch.dtype`):
+                The dtype to use for the 4D attention mask.
+            device (`torch.device`):
+                The device to plcae the 4D attention mask on.
+            cache_position (`torch.Tensor`):
+                Indices depicting the position of the input sequence tokens in the sequence.
+            batch_size (`torch.Tensor`):
+                Batch size.
+        """
+        if attention_mask is not None and attention_mask.dim() == 4:
+            # In this case we assume that the mask comes already in inverted form and requires no inversion or slicing.
+            causal_mask = attention_mask
+        else:
+            min_dtype = torch.finfo(dtype).min
+            causal_mask = torch.full(
+                (sequence_length, target_length), fill_value=min_dtype, dtype=dtype, device=device
+            )
+            if sequence_length != 1:
+                causal_mask = torch.triu(causal_mask, diagonal=1)
+            causal_mask *= torch.arange(target_length, device=device) > cache_position.reshape(-1, 1)
+            causal_mask = causal_mask[None, None, :, :].expand(batch_size, 1, -1, -1)
+            if attention_mask is not None:
+                causal_mask = causal_mask.clone()  # copy to contiguous memory for in-place edit
+                mask_length = attention_mask.shape[-1]
+                padding_mask = causal_mask[:, :, :, :mask_length] + attention_mask[:, None, None, :].to(
+                    causal_mask.device
+                )
+                padding_mask = padding_mask == 0
+                causal_mask[:, :, :, :mask_length] = causal_mask[:, :, :, :mask_length].masked_fill(
+                    padding_mask, min_dtype
+                )
+
+        return causal_mask
+
+
+@add_start_docstrings(
+    "The EuroBert Model with a decoder head on top that is used for masked language modeling.",
+    EUROBERT_START_DOCSTRING,
+)
+class EuroBertForMaskedLM(EuroBertPreTrainedModel):
+    def __init__(self, config: EuroBertConfig):
+        super().__init__(config)
+        self.model = EuroBertModel(config)
+        self.lm_head = nn.Linear(config.hidden_size, config.vocab_size, config.mlp_bias)
+        self.post_init()
+
+    @add_start_docstrings_to_model_forward(EUROBERT_INPUTS_DOCSTRING)
+    @add_code_sample_docstrings(
+        checkpoint=_CHECKPOINT_FOR_DOC,
+        output_type=BaseModelOutput,
+        config_class=_CONFIG_FOR_DOC,
+    )
+    def forward(
+        self,
+        input_ids: Optional[torch.LongTensor] = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        inputs_embeds: Optional[torch.FloatTensor] = None,
+        labels: Optional[torch.LongTensor] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+    ) -> Union[Tuple[torch.Tensor], MaskedLMOutput]:
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        encoder_output = self.model(
+            input_ids,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            inputs_embeds=inputs_embeds,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+        )
+
+        prediction_scores = self.lm_head(encoder_output[0])
+        masked_lm_loss = None
+        if labels is not None:
+            labels = labels.to(prediction_scores.device)
+            masked_lm_loss = self.loss_function(prediction_scores, labels, vocab_size=self.config.vocab_size)
+
+        if not return_dict:
+            output = (prediction_scores,) + encoder_output[1:]
+            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=encoder_output.hidden_states,
+            attentions=encoder_output.attentions,
+        )
+
+
+@add_start_docstrings(
+    "The EuroBert Model with a sequence classification head on top that performs pooling.",
+    EUROBERT_START_DOCSTRING,
+)
+class EuroBertForSequenceClassification(EuroBertPreTrainedModel):
+    def __init__(self, config: EuroBertConfig):
+        super().__init__(config)
+        self.num_labels = config.num_labels
+        self.clf_pooling = config.clf_pooling
+
+        self.model = EuroBertModel(config)
+        self.dense = nn.Linear(config.hidden_size, config.hidden_size)
+        self.activation = nn.GELU()
+        self.classifier = nn.Linear(config.hidden_size, self.num_labels)
+        self.post_init()
+
+    @add_start_docstrings_to_model_forward(EUROBERT_INPUTS_DOCSTRING)
+    @add_code_sample_docstrings(
+        checkpoint=_CHECKPOINT_FOR_DOC,
+        output_type=BaseModelOutput,
+        config_class=_CONFIG_FOR_DOC,
+    )
+    def forward(
+        self,
+        input_ids: Optional[torch.LongTensor] = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        inputs_embeds: Optional[torch.FloatTensor] = None,
+        labels: Optional[torch.LongTensor] = 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
+
+        encoder_output = self.model(
+            input_ids,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            inputs_embeds=inputs_embeds,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+        )
+        last_hidden_state = encoder_output[0]
+
+        if self.clf_pooling in ["bos", "mean"]:
+            if self.clf_pooling == "bos":
+                pooled_output = last_hidden_state[:, 0]
+
+            elif self.clf_pooling == "mean":
+                if attention_mask is None:
+                    pooled_output = last_hidden_state.mean(dim=1)
+                else:
+                    pooled_output = (last_hidden_state * attention_mask.unsqueeze(-1)).sum(dim=1)
+                    pooled_output /= attention_mask.sum(dim=1, keepdim=True)
+
+            pooled_output = self.dense(pooled_output)
+            pooled_output = self.activation(pooled_output)
+            logits = self.classifier(pooled_output)
+
+        elif self.clf_pooling == "late":
+            x = self.dense(last_hidden_state)
+            x = self.activation(x)
+            logits = self.classifier(x)
+            if attention_mask is None:
+                logits = logits.mean(dim=1)
+            else:
+                logits = (logits * attention_mask.unsqueeze(-1)).sum(dim=1)
+                logits /= attention_mask.sum(dim=1, keepdim=True)
+
+        loss = None
+        if labels is not None:
+            labels = labels.to(logits.device)
+            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,) + encoder_output[1:]
+            return ((loss,) + output) if loss is not None else output
+
+        return SequenceClassifierOutput(
+            loss=loss,
+            logits=logits,
+            hidden_states=encoder_output.hidden_states,
+            attentions=encoder_output.attentions,
+        )
+
+
+@add_start_docstrings(
+    """
+    The EuroBert Model with a token classification head on top (a linear layer on top of the hidden-states
+    output) e.g. for Named-Entity-Recognition (NER) tasks."
+    """,
+    EUROBERT_START_DOCSTRING,
+)
+class EuroBertForTokenClassification(EuroBertPreTrainedModel):
+    def __init__(self, config: EuroBertConfig):
+        super().__init__(config)
+        self.num_labels = config.num_labels
+        self.model = EuroBertModel(config)
+
+        self.classifier = nn.Linear(config.hidden_size, config.num_labels)
+        self.post_init()
+
+    def get_input_embeddings(self):
+        return self.model.embed_tokens
+
+    def set_input_embeddings(self, value):
+        self.model.embed_tokens = value
+
+    @add_start_docstrings_to_model_forward(EUROBERT_INPUTS_DOCSTRING)
+    def forward(
+        self,
+        input_ids: Optional[torch.LongTensor] = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        inputs_embeds: Optional[torch.FloatTensor] = None,
+        labels: Optional[torch.LongTensor] = None,
+        use_cache: Optional[bool] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+    ) -> Union[Tuple, TokenClassifierOutput]:
+        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.model(
+            input_ids,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            inputs_embeds=inputs_embeds,
+            use_cache=use_cache,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+        )
+        sequence_output = outputs[0]
+        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,
+        )
+
+
+@add_start_docstrings(
+    """
+    The EuroBert Model with a span classification head on top for extractive question-answering tasks 
+    like SQuAD (a linear layers on top of the hidden-states output to compute span start logits 
+    and span end logits).
+    """,
+    EUROBERT_START_DOCSTRING,
+)
+class EuroBertForQuestionAnswering(EuroBertPreTrainedModel):
+    def __init__(self, config: EuroBertConfig):
+        super().__init__(config)
+        self.num_labels = config.num_labels
+        self.model = EuroBertModel(config)
+
+        self.qa_outputs = nn.Linear(config.hidden_size, config.num_labels)
+        self.post_init()
+
+    def get_input_embeddings(self):
+        return self.model.embed_tokens
+
+    def set_input_embeddings(self, value):
+        self.model.embed_tokens = value
+
+    @add_start_docstrings_to_model_forward(EUROBERT_INPUTS_DOCSTRING)
+    def forward(
+        self,
+        input_ids: Optional[torch.Tensor] = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.Tensor] = None,
+        inputs_embeds: Optional[torch.Tensor] = None,
+        use_cache: Optional[bool] = 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]:
+        r"""
+        start_positions (`torch.LongTensor` of shape `(batch_size,)`, *optional*):
+            Labels for position (index) of the start of the labelled span for computing the token classification loss.
+            Positions are clamped to the length of the sequence (`sequence_length`). Position outside of the sequence
+            are not taken into account for computing the loss.
+        end_positions (`torch.LongTensor` of shape `(batch_size,)`, *optional*):
+            Labels for position (index) of the end of the labelled span for computing the token classification loss.
+            Positions are clamped to the length of the sequence (`sequence_length`). Position outside of the sequence
+            are not taken into account for computing the loss.
+        """
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        outputs = self.model(
+            input_ids,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            inputs_embeds=inputs_embeds,
+            use_cache=use_cache,
+            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__ = [
+    "EuroBertPreTrainedModel",
+    "EuroBertModel",
+    "EuroBertForMaskedLM",
+    "EuroBertForSequenceClassification",
+    "EuroBertForTokenClassification",
+    "EuroBertForQuestionAnswering",
+]

modules.json

@@ -0,0 +1,20 @@
+[
+  {
+    "idx": 0,
+    "name": "0",
+    "path": "",
+    "type": "sentence_transformers.base.modules.transformer.Transformer"
+  },
+  {
+    "idx": 1,
+    "name": "1",
+    "path": "1_Pooling",
+    "type": "sentence_transformers.sentence_transformer.modules.pooling.Pooling"
+  },
+  {
+    "idx": 2,
+    "name": "2",
+    "path": "2_Normalize",
+    "type": "sentence_transformers.sentence_transformer.modules.normalize.Normalize"
+  }
+]

sentence_bert_config.json

@@ -0,0 +1,10 @@
+{
+    "transformer_task": "feature-extraction",
+    "modality_config": {
+        "text": {
+            "method": "forward",
+            "method_output_name": "last_hidden_state"
+        }
+    },
+    "module_output_name": "token_embeddings"
+}

tokenizer_config.json

@@ -0,0 +1,19 @@
+{
+  "backend": "tokenizers",
+  "bos_token": "<|begin_of_text|>",
+  "clean_up_tokenization_spaces": true,
+  "eos_token": "<|end_of_text|>",
+  "is_local": true,
+  "mask_token": "<|mask|>",
+  "max_length": null,
+  "model_input_names": [
+    "input_ids",
+    "attention_mask"
+  ],
+  "model_max_length": 1000000000000000019884624838656,
+  "pad_to_multiple_of": null,
+  "pad_token": "<|pad|>",
+  "pad_token_type_id": 0,
+  "padding_side": "right",
+  "tokenizer_class": "TokenizersBackend"
+}