Update model card with detailed information

README.md

@@ -1,70 +1,172 @@
 ---
-library_name: transformers
-language:
-- multilingual
 license: other
 base_model: DedalusHealthCare/tinybert-mlm-de
 tags:
-- generated_from_trainer
-datasets:
-- ner_demo_de
+- token-classification
+- ner
+- medical
+- demo
+- de
+- pytorch
+- transformers
+language:
+- de
+pipeline_tag: token-classification
+library_name: transformers
 model-index:
-- name: tinybert-demo-de
-  results: []
+- name: TinyBERT for Demo NER
+  results:
+  - task:
+      type: token-classification
+      name: Named Entity Recognition
+    dataset:
+      type: demo
+      name: Demo Dataset
+      config: de
+    metrics:
+    - type: f1
+      value: # Will be updated after evaluation
+      name: F1 Score
+    - type: precision
+      value: # Will be updated after evaluation
+      name: Precision
+    - type: recall
+      value: # Will be updated after evaluation
+      name: Recall
 ---
 
-<!-- This model card has been generated automatically according to the information the Trainer had access to. You
-should probably proofread and complete it, then remove this comment. -->
+# TinyBERT for Demo NER (DE)
+
+## Model Description
+
+This model is a fine-tuned TinyBERT model for Named Entity Recognition (NER) of DISORDER_FINDING entities in German medical texts.
+
+**Base Model**: DedalusHealthCare/tinybert-mlm-de
+
+**Language**: German (de)
+
+**Task**: Token Classification (NER)
+
+**Entities**: DISORDER_FINDING
+
+## Training Details
+
+### Training Dataset
+
+**Dataset**: `DedalusHealthCare/ner_demo_de@2025.10.16.13.40.41`
+
+The model was trained on a versioned dataset with timestamp-based versioning for reproducibility.
+
+### Training Configuration
+- **Training epochs**: 1
+- **Learning rate**: 5e-05
+- **Training batch size**: 32
+- **Evaluation batch size**: 32
+- **Max sequence length**: N/A
+- **Warmup steps**: 0
+- **Weight decay**: 0.01
+- **Gradient accumulation steps**: 2
+- **Mixed precision (FP16)**: False
+
+### Training Framework
+- **Framework**: PyTorch with HuggingFace Transformers
+- **Optimizer**: AdamW
+- **Scheduler**: Linear with warmup
+
+## Usage
+
+### Quick Start with Pipeline
+
+```python
+from transformers import pipeline
+
+# Initialize the NER pipeline
+ner_pipeline = pipeline(
+    "ner",
+    model="DedalusHealthCare/tinybert-demo-de",
+    tokenizer="DedalusHealthCare/tinybert-demo-de",
+    aggregation_strategy="simple"
+)
+
+# Example usage
+text = "Your medical text here"
+entities = ner_pipeline(text)
+print(entities)
+```
+
+### Advanced Usage
+
+```python
+from transformers import AutoTokenizer, AutoModelForTokenClassification
+import torch
+
+# Load model and tokenizer
+model_name = "DedalusHealthCare/tinybert-demo-de"
+tokenizer = AutoTokenizer.from_pretrained(model_name)
+model = AutoModelForTokenClassification.from_pretrained(model_name)
+
+# Set model to evaluation mode
+model.eval()
+
+# Tokenize text
+text = "Your medical text here"
+inputs = tokenizer(text, return_tensors="pt", truncation=True, padding=True)
+
+# Get predictions
+with torch.no_grad():
+    outputs = model(**inputs)
+    predictions = torch.nn.functional.softmax(outputs.logits, dim=-1)
 
-# tinybert-demo-de
+# Get predicted labels
+predicted_token_class_ids = predictions.argmax(-1)
+labels = [model.config.id2label[id.item()] for id in predicted_token_class_ids[0]]
+```
 
-This model is a fine-tuned version of [DedalusHealthCare/tinybert-mlm-de](https://huggingface.co/DedalusHealthCare/tinybert-mlm-de) on the ner_demo_de dataset.
-It achieves the following results on the evaluation set:
-- Loss: 0.4069
-- Disorder Finding Precision: 0.25
-- Disorder Finding Recall: 0.1818
-- Disorder Finding F1: 0.2105
-- Disorder Finding Number: 11
-- Overall Precision: 0.25
-- Overall Recall: 0.1818
-- Overall F1: 0.2105
-- Overall Accuracy: 0.9286
+## Model Performance
 
-## Model description
+Performance metrics will be updated after evaluation on the validation set.
 
-More information needed
+## Intended Use
 
-## Intended uses & limitations
+This model is specifically designed for:
+- Named Entity Recognition in German medical texts
+- Identification of DISORDER_FINDING entities
+- Medical document processing and analysis
+- Clinical NLP research and applications
 
-More information needed
+## Limitations
 
-## Training and evaluation data
+- Trained specifically for German medical texts
+- Performance may vary on different medical domains or institutions
+- May require domain adaptation for optimal performance on new datasets
+- Subject to biases present in the training data
 
-More information needed
+## Ethical Considerations
 
-## Training procedure
+- This model processes medical data and should be used responsibly
+- All predictions should be validated by qualified medical professionals
+- Patient privacy and data protection regulations must be followed
+- The model may exhibit biases from the training data
 
-### Training hyperparameters
+## Citation
 
-The following hyperparameters were used during training:
-- learning_rate: 5e-05
-- train_batch_size: 32
-- eval_batch_size: 32
-- seed: 33
-- gradient_accumulation_steps: 2
-- total_train_batch_size: 64
-- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
-- lr_scheduler_type: linear
-- lr_scheduler_warmup_ratio: 0.1
-- num_epochs: 1
+If you use this model, please cite:
 
-### Training results
+```bibtex
+@model{demo_de_ner_model,
+  title = {TinyBERT for Demo NER (DE)},
+  author = {DH Healthcare GmbH},
+  year = {2025},
+  publisher = {Hugging Face},
+  base_model = {DedalusHealthCare/tinybert-mlm-de},
+  url = {https://huggingface.co/DedalusHealthCare/tinybert-demo-de}
+}
+```
 
+## License
 
+This model is proprietary and owned by DH Healthcare GmbH. All rights reserved.
 
-### Framework versions
+## Contact
 
-- Transformers 4.45.1
-- Pytorch 2.6.0+cu124
-- Datasets 2.16.0
-- Tokenizers 0.20.3
+For questions or support regarding this model, please contact DH Healthcare GmbH.