Update README.md (#1)

Update README.md

Update README.md

Update README.md

Create README.md

Delete README.md

--- ## Model Card for `gizadatateam/Arabic-ModernBERT` ### Summary `Arabic-ModernBERT` is a ModernBERT‐architecture Masked Language Model tailored for Arabic. It supports very long contexts (up to 8 192 tokens), rotary position embeddings, local‐global alternating attention, and GeGLU activations. ### Files & Versions - **.gitattributes** (1.63 kB) - **config.json** (1.41 kB) citeturn2view0turn4view0 - **model.safetensors** (845 MB; LFS) - **special_tokens_map.json** (866 Bytes) - **tokenizer.json** (18.6 MB; LFS) - **tokenizer_config.json** (14.2 kB; LFS) - **training_state.pt** (1.65 GB; LFS) citeturn2view0 ### Configuration (from `config.json`) ```jsonc { "_attn_implementation_autoset": true, "_name_or_path": "./checkpoint_step_20000/", "architectures": ["ModernBertForMaskedLM"], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 50281, "classifier_activation": "gelu", "classifier_bias": false, "classifier_dropout": 0.0, "classifier_pooling": "mean", "cls_token_id": 50281, "decoder_bias": true, "deterministic_flash_attn": false, "embedding_dropout": 0.0, "eos_token_id": 50282, "global_attn_every_n_layers": 3, "global_rope_theta": 160000.0, "gradient_checkpointing": false, "hidden_activation": "gelu", "hidden_size": 768, "initializer_cutoff_factor": 2.0, "initializer_range": 0.02, "intermediate_size": 1152, "layer_norm_eps": 1e-05, "local_attention": 128, "local_rope_theta": 10000.0, "max_position_embeddings": 8192, "mlp_bias": false, "mlp_dropout": 0.0, "model_type": "modernbert", "norm_bias": false, "norm_eps": 1e-05, "num_attention_heads": 12, "num_hidden_layers": 22, "pad_token_id": 50283, "position_embedding_type": "absolute", "reference_compile": false, "repad_logits_with_grad": false, "sep_token_id": 50282, "sparse_pred_ignore_index": -100, "sparse_prediction": false, "torch_dtype": "float32", "transformers_version": "4.48.3", "use_cache": false, "vocab_size": 130368 } ``` All fields above are verbatim from the repository’s `config.json`. citeturn4view0 ### Model Statistics - **Model size**: 211 M parameters - **Tensor type**: `float32` - **Downloads (last month)**: 0 - **Inference Providers**: none (not deployed) citeturn3view0 ### Usage Example ```python from transformers import AutoTokenizer, AutoModelForMaskedLM, pipeline model_id = "gizadatateam/Arabic-ModernBERT" tokenizer = AutoTokenizer.from_pretrained(model_id) model = AutoModelForMaskedLM.from_pretrained(model_id) fill_mask = pipeline("fill-mask", model=model, tokenizer=tokenizer) print(fill_mask("مرحبا أيها [MASK]!")) ``` ### Citation If you use this model, please cite the underlying ModernBERT paper: ```bibtex @misc {modernbert, title={Smarter, Better, Faster, Longer: A Modern Bidirectional Encoder for Fast, Memory Efficient, and Long Context Finetuning and Inference}, author={Warner et al.}, year={2024}, eprint={2412.13663}, archivePrefix={arXiv}, primaryClass={cs.CL}, url={https://arxiv.org/abs/2412.13663}, } ``` ---

Add tokenizer files

Add model files

initial commit