Update model card (BERT-style tags + license)

README.md

@@ -1,29 +1,71 @@
----
-library_name: kerasformers
-license: apache-2.0
-pipeline_tag: fill-mask
-tags:
-  - keras
-  - modernbert
-base_model: answerdotai/ModernBERT-large
----
-
-# modernbert_large
-
-ModernBERT converted to [kerasformers](https://github.com/IMvision12/KerasFormers)
-(pure Keras 3, runnable on JAX / PyTorch / TensorFlow). Converted from [`answerdotai/ModernBERT-large`](https://huggingface.co/answerdotai/ModernBERT-large).
-
-```python
-from kerasformers.models.modernbert import ModernBertModel, ModernBertTokenizer
-
-model = ModernBertModel.from_weights("kerasformers/modernbert_large")
-tokenizer = ModernBertTokenizer.from_weights("kerasformers/modernbert_large")
-out = model(tokenizer("Hello, world."))["last_hidden_state"]
-```
-
-| layers | embed_dim | heads | mlp_dim |
-|---|---|---|---|
-| 28 | 1024 | 16 | 2624 |
-
-The single `model.weights.h5` is the full masked-LM checkpoint; the encoder,
-masked-LM, and task-head classes each load their own subset.
+---
+pipeline_tag: fill-mask
+license: apache-2.0
+base_model: answerdotai/ModernBERT-large
+library_name: kerasformers
+tags:
+- keras
+- kerasformers
+- modernbert
+- fill-mask
+- text-encoder
+- arxiv:2412.13663
+- pytorch
+- jax
+- tf
+---
+
+# Run ModernBERT with Keras 3: JAX, PyTorch, or TensorFlow
+
+[![GitHub](https://img.shields.io/badge/GitHub-KerasFormers-black?logo=github)](https://github.com/IMvision12/KerasFormers) [![Docs](https://img.shields.io/badge/Docs-ModernBERT-blue)](https://imvision12.github.io/KerasFormers/modernbert/)
+
+# kerasformers/modernbert_large
+
+Paper: [Smarter, Better, Faster, Longer: A Modern Bidirectional Encoder (arXiv:2412.13663)](https://arxiv.org/abs/2412.13663) · [HF Papers](https://huggingface.co/papers/2412.13663)
+
+ModernBERT is Answer.AI / LightOn's modernized bidirectional transformer text encoder: rotary position embeddings, attention that alternates between a global (full) layer and local sliding-window layers, GeGLU feed-forwards, and pre-LayerNorm, with an 8192-token context. Byte-level BPE tokenizer; mask token `[MASK]`. No token-type ids.
+
+For more details on the model, please go to the upstream [model card](https://huggingface.co/answerdotai/ModernBERT-large).
+
+Pure-**Keras 3** conversion of [`answerdotai/ModernBERT-large`](https://huggingface.co/answerdotai/ModernBERT-large) for [kerasformers](https://github.com/IMvision12/KerasFormers). One implementation runs unmodified on **TensorFlow / Torch / JAX**.
+
+This is a **fill-mask / encoder** checkpoint (`ModernBertMaskedLM`, modernbert_large). Task heads (sequence / token classify, QA, multiple choice) load via `hf:` fine-tunes.
+
+## ✨ Quick start (fill-mask)
+
+```python
+import os
+os.environ["KERAS_BACKEND"] = "torch"  # or "jax" / "tensorflow"
+
+from kerasformers.models.modernbert import ModernBertMaskedLM, ModernBertTokenizer
+
+mlm = ModernBertMaskedLM.from_weights("kerasformers/modernbert_large")
+tokenizer = ModernBertTokenizer.from_weights("kerasformers/modernbert_large")
+
+inputs = tokenizer("The capital of France is [MASK].")
+logits = mlm(inputs)  # (1, L, vocab_size)
+mask = int((inputs["input_ids"][0] == tokenizer.mask_token_id).argmax())
+print(tokenizer.decode([int(logits[0, mask].argmax())]))
+```
+
+Load any ModernBERT variant the same way with `from_weights("kerasformers/<variant>")`:
+
+| Variant | Hub | layers | embed_dim |
+|---|---|---|---|
+| `modernbert_base` | [`kerasformers/modernbert_base`](https://huggingface.co/kerasformers/modernbert_base) | 22 | 768 |
+| `modernbert_large` | [`kerasformers/modernbert_large`](https://huggingface.co/kerasformers/modernbert_large) | 28 | 1024 |
+
+## Tips
+
+- Set `KERAS_BACKEND` **before** importing Keras / kerasformers.
+- Prefer `ModernBertTokenizer.from_weights(...)` so tokenization matches.
+- Use `[MASK]` (not `<mask>`).
+- ModernBERT has no token-type ids; the tokenizer emits only `input_ids` / `attention_mask`.
+- See [ModernBERT docs](https://imvision12.github.io/KerasFormers/modernbert/) and [Loading Weights](https://imvision12.github.io/KerasFormers/loading_weights/).
+- Community / upstream safetensors still work via the `hf:` prefix, e.g. `ModernBertMaskedLM.from_weights("hf:answerdotai/ModernBERT-large")`.
+
+## Special Thanks
+
+A huge thank you to the Answer.AI and LightOn authors for creating and releasing ModernBERT.
+
+License: Apache 2.0.