Tarjeta del modelo en español (reemplaza plantilla por defecto)

README.md

@@ -1,199 +1,167 @@
 ---
+language: es
+license: cc-by-4.0
 library_name: transformers
-tags: []
+pipeline_tag: token-classification
+base_model: dccuchile/bert-base-spanish-wwm-cased
+tags:
+  - token-classification
+  - ner
+  - named-entity-recognition
+  - spanish
+  - español
+  - biomedical
+  - clinical
+  - medical
+  - oncology
+  - prostate-cancer
+  - cancer-de-prostata
+  - beto
+metrics:
+  - f1
+  - precision
+  - recall
+widget:
+  - text: "Paciente masculino de 72 años con adenocarcinoma de próstata, Gleason 3+3 y PSA de 9.9 ng/dL. Se inicia tratamiento con radioterapia y bicalutamida 50 mg diarios."
+    example_title: "Caso clínico de próstata"
+  - text: "Varón de 68 años con cáncer de próstata estadio T2N0M0, PSA 12.4. Prostatectomía radical en marzo de 2023."
+    example_title: "Estadificación y cirugía"
+model-index:
+  - name: beto-prostata-ner
+    results:
+      - task:
+          type: token-classification
+          name: Named Entity Recognition
+        metrics:
+          - type: f1
+            value: 0.9757
+            name: F1 (micro, seqeval)
+          - type: precision
+            value: 0.9732
+            name: Precision
+          - type: recall
+            value: 0.9781
+            name: Recall
 ---
 
-# Model Card for Model ID
+# BETO fine-tuned para NER de cáncer de próstata (español)
 
-<!-- Provide a quick summary of what the model is/does. -->
+Modelo de **Reconocimiento de Entidades Nombradas (NER)** que extrae información clínica
+de textos en **español** sobre **cáncer de próstata**. Es un fine-tuning de
+[**BETO**](https://huggingface.co/dccuchile/bert-base-spanish-wwm-cased) (BERT entrenado
+desde cero en español) para clasificación de tokens en formato **BIO** con 21 etiquetas.
 
+> **F1 en test: 0.976** (seqeval, micro) · Precision 0.973 · Recall 0.978 · Accuracy 0.995
 
+## 🎯 Para qué sirve
 
-## Model Details
+Detecta y clasifica, palabra por palabra, **10 tipos de entidades clínicas** en informes y
+notas médicas de oncología prostática:
 
-### Model Description
+| Entidad | Qué captura | Ejemplo |
+|---|---|---|
+| `EDAD` | Edad del paciente | *72 años* |
+| `CANCER` | Tipo/diagnóstico de cáncer | *adenocarcinoma de próstata* |
+| `GLEASON` | Escala de Gleason | *Gleason 3+3* |
+| `BIOMARCADOR` | Biomarcadores (PSA, etc.) | *PSA de 9.9 ng/dL* |
+| `TNM` | Estadificación TNM | *T2N0M0* |
+| `TRATAMIENTO` | Tratamientos | *radioterapia* |
+| `MEDICAMENTO` | Fármacos | *bicalutamida* |
+| `DOSIS` | Dosis | *50 mg diarios* |
+| `CIRUGIA` | Procedimientos quirúrgicos | *prostatectomía radical* |
+| `FECHA` | Fechas | *marzo de 2023* |
 
-<!-- Provide a longer summary of what this model is. -->
+Cada entidad se etiqueta en esquema **BIO** (`B-` inicio, `I-` continuación, `O` fuera),
+para 21 etiquetas en total.
 
-This is the model card of a 🤗 transformers model that has been pushed on the Hub. This model card has been automatically generated.
+## 🚀 Uso rápido
 
-- **Developed by:** [More Information Needed]
-- **Funded by [optional]:** [More Information Needed]
-- **Shared by [optional]:** [More Information Needed]
-- **Model type:** [More Information Needed]
-- **Language(s) (NLP):** [More Information Needed]
-- **License:** [More Information Needed]
-- **Finetuned from model [optional]:** [More Information Needed]
+```python
+from transformers import pipeline
 
-### Model Sources [optional]
+ner = pipeline("token-classification", model="ralzate/beto-prostata-ner",
+               aggregation_strategy="simple")
 
-<!-- Provide the basic links for the model. -->
+texto = ("Paciente masculino de 72 años con adenocarcinoma de próstata, "
+         "Gleason 3+3 y PSA de 9.9 ng/dL. Se inicia radioterapia y bicalutamida 50 mg.")
 
-- **Repository:** [More Information Needed]
-- **Paper [optional]:** [More Information Needed]
-- **Demo [optional]:** [More Information Needed]
+for e in ner(texto):
+    print(f"{e['entity_group']:14s} {e['score']*100:5.1f}%  {e['word']}")
+```
 
-## Uses
+Salida esperada (entidades como `EDAD`, `CANCER`, `GLEASON`, `BIOMARCADOR`,
+`TRATAMIENTO`, `MEDICAMENTO`, `DOSIS`).
 
-<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+## 📊 Resultados (conjunto de test, seqeval)
 
-### Direct Use
+Barrido de *batch size* con 6 épocas y *early stopping*:
 
-<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+| batch size | Precision | Recall | F1 |
+|---|---|---|---|
+| **8** (recomendado) | 0.9732 | 0.9781 | **0.9757** |
+| 16 | 0.9627 | 0.9705 | 0.9666 |
+| 32 | 0.9488 | 0.9603 | 0.9545 |
 
-[More Information Needed]
+Este repositorio publica la **mejor configuración (batch size 8)**. El F1 baja levemente al
+aumentar el lote porque, con 6 épocas, los lotes pequeños dan más pasos de optimización.
 
-### Downstream Use [optional]
+### Comparación con XLM-RoBERTa
 
-<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+En el mismo dataset, **BETO supera a XLM-RoBERTa-base** (multilingüe) en todo el barrido
+(F1 0.955–0.976 vs 0.850–0.936). En NER clínico en español, un encoder **nativo en español**
+y compacto rinde mejor que un multilingüe más grande: la especialización lingüística pesa
+más que el tamaño.
 
-[More Information Needed]
+## 🧠 Detalles de entrenamiento
 
-### Out-of-Scope Use
+| | |
+|---|---|
+| **Modelo base** | `dccuchile/bert-base-spanish-wwm-cased` (BETO, ~110M parámetros) |
+| **Tarea** | Token Classification (NER), 21 etiquetas BIO |
+| **Dominio** | Texto clínico de cáncer de próstata en español |
+| **Épocas** | 6 (con *early stopping*, paciencia 2, métrica F1) |
+| **Learning rate** | 2e-5 |
+| **Batch size** | 8 (este checkpoint) |
+| **Longitud máxima** | 128 subtokens |
+| **Optimizador** | AdamW, weight decay 0.01, warmup 0.1 |
+| **Métrica** | seqeval (precision/recall/F1 a nivel de entidad) |
+| **Hardware** | Apple M1 Pro (MPS), entrenamiento local |
+| **Alineación** | etiqueta en el primer subtoken de cada palabra; resto a `-100` |
 
-<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+## ⚠️ Limitaciones
 
-[More Information Needed]
+- Entrenado en un **dominio específico** (cáncer de próstata en español); fuera de ese
+  dominio el rendimiento puede caer.
+- El conjunto de datos es de tamaño moderado; entidades poco frecuentes pueden tener menor
+  recall.
+- No sustituye el criterio clínico profesional. Pensado como apoyo a la extracción de
+  información, no para decisiones médicas autónomas.
+- Hereda los posibles sesgos de BETO y del corpus clínico de entrenamiento.
 
-## Bias, Risks, and Limitations
+## 🏷️ Etiquetas
 
-<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+```
+O
+B-EDAD, I-EDAD, B-CANCER, I-CANCER, B-GLEASON, I-GLEASON,
+B-BIOMARCADOR, I-BIOMARCADOR, B-TNM, I-TNM, B-TRATAMIENTO, I-TRATAMIENTO,
+B-MEDICAMENTO, I-MEDICAMENTO, B-DOSIS, I-DOSIS, B-CIRUGIA, I-CIRUGIA, B-FECHA, I-FECHA
+```
+
+## 📚 Cita
+
+Si usas este modelo, cita también a BETO:
 
-[More Information Needed]
+```bibtex
+@inproceedings{canete2020beto,
+  title={Spanish Pre-Trained BERT Model and Evaluation Data},
+  author={Cañete, José and Chaperon, Gabriel and Fuentes, Rodrigo and Ho, Jou-Hui and Kang, Hojin and Pérez, Jorge},
+  booktitle={PML4DC at ICLR 2020},
+  year={2020}
+}
+```
 
-### Recommendations
-
-<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
-
-Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
-
-## How to Get Started with the Model
-
-Use the code below to get started with the model.
-
-[More Information Needed]
-
-## Training Details
-
-### Training Data
-
-<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
-
-[More Information Needed]
-
-### Training Procedure
-
-<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
-
-#### Preprocessing [optional]
-
-[More Information Needed]
-
-
-#### Training Hyperparameters
-
-- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
-
-#### Speeds, Sizes, Times [optional]
-
-<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
-
-[More Information Needed]
-
-## Evaluation
-
-<!-- This section describes the evaluation protocols and provides the results. -->
-
-### Testing Data, Factors & Metrics
-
-#### Testing Data
-
-<!-- This should link to a Dataset Card if possible. -->
-
-[More Information Needed]
-
-#### Factors
-
-<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
-
-[More Information Needed]
-
-#### Metrics
-
-<!-- These are the evaluation metrics being used, ideally with a description of why. -->
-
-[More Information Needed]
-
-### Results
-
-[More Information Needed]
-
-#### Summary
-
-
-
-## Model Examination [optional]
-
-<!-- Relevant interpretability work for the model goes here -->
-
-[More Information Needed]
-
-## Environmental Impact
-
-<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
-
-Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
-
-- **Hardware Type:** [More Information Needed]
-- **Hours used:** [More Information Needed]
-- **Cloud Provider:** [More Information Needed]
-- **Compute Region:** [More Information Needed]
-- **Carbon Emitted:** [More Information Needed]
-
-## Technical Specifications [optional]
-
-### Model Architecture and Objective
-
-[More Information Needed]
-
-### Compute Infrastructure
-
-[More Information Needed]
-
-#### Hardware
-
-[More Information Needed]
-
-#### Software
-
-[More Information Needed]
-
-## Citation [optional]
-
-<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
-
-**BibTeX:**
-
-[More Information Needed]
-
-**APA:**
-
-[More Information Needed]
-
-## Glossary [optional]
-
-<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
-
-[More Information Needed]
-
-## More Information [optional]
-
-[More Information Needed]
-
-## Model Card Authors [optional]
-
-[More Information Needed]
-
-## Model Card Contact
+---
 
-[More Information Needed]
+*Modelo afinado y publicado como parte de un proyecto de PLN en español (clasificación y NER
+con Transformers). Reproducible en hardware modesto (Apple Silicon) mediante fine-tuning
+completo para BETO y LoRA para modelos grandes.*