Update README.md

README.md

@@ -7,7 +7,11 @@ base_model:
 - mistralai/Mistral-7B-Instruct-v0.3
 tags:
 - code
+- python
+- sql
+- data-science
 ---
+
 # Code Specialist 7B  
 
 <p align="left">
@@ -21,59 +25,67 @@ tags:
     <img src="https://img.shields.io/badge/Transformers-4.56+-purple?style=flat-square&logo=huggingface&logoColor=white" alt="Transformers"/>
   </a>
   <a href="https://github.com/Ricardouchub">
-    <img src="https://img.shields.io/badge/Autor-Ricardo_Urdaneta-000000?style=flat-square&logo=github&logoColor=white" alt="Author"/>
+    <img src="https://img.shields.io/badge/Author-Ricardo_Urdaneta-000000?style=flat-square&logo=github&logoColor=white" alt="Author"/>
   </a>
 </p>
 
-## Descripción  
+---
+
+## Description  
 
-**Code Specialist 7B** es un modelo de lenguaje basado en **Mistral-7B-Instruct-v0.3**, adaptado mediante **SFT (Supervised Fine-Tuning)** con datasets especializados en **Python y SQL**.  
-El entrenamiento fue realizado con el objetivo de mejorar la capacidad del modelo en **resolución de problemas de analisis de datos y data science**.
+**Code Specialist 7B** is a fine-tuned version of **Mistral-7B-Instruct-v0.3**, trained through **Supervised Fine-Tuning (SFT)** using datasets focused on **Python and SQL**.  
+The goal of this training was to enhance the model’s performance in **data analysis, programming problem-solving, and technical reasoning**.
 
-Este modelo mantiene la arquitectura original de **7B parámetros**, pero incorpora un ajuste fino orientado a código, lo que lo hace más robusto en generación de funciones, queries SQL y respuestas técnicas.  
+The model preserves the **7B parameter Transformer decoder-only** architecture while introducing a code-oriented fine-tuning, resulting in improved robustness for function generation, SQL queries, and technical answers.
 
 ---
 
-## Modelo base  
+## Base Model  
 
 - [Mistral-7B-Instruct-v0.3](https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.3)  
-- Arquitectura: Transformer Decoder-only  
-- Parámetros: ~7B  
+- Architecture: Transformer (decoder-only)  
+- Parameters: ~7B  
 
 ---
 
-## Dataset utilizado para SFT  
+## Datasets Used for SFT  
 
 - [CodeAlpaca-20k](https://huggingface.co/datasets/sahil2801/CodeAlpaca-20k)  
-- [code_instructions_122k (alpaca-style)](https://huggingface.co/datasets/TokenBender/code_instructions_122k_alpaca_style)  
+- [Code Instructions 122k (Alpaca-style)](https://huggingface.co/datasets/TokenBender/code_instructions_122k_alpaca_style)  
 
-Ambos datasets fueron **filtrados a ejemplos de Python y SQL**, con formato de prompts estilo **Alpaca/Mistral**.  
+Both datasets were **filtered to include only Python and SQL examples**, following **Alpaca/Mistral-style** instruction formatting.
 
-Ejemplo de formato aplicado:  
+Example prompt format:  
 
 ```
-[INST] Escribe una función en Python que sume dos números. [/INST]  
+[INST] Write a Python function that adds two numbers. [/INST]  
 def add(a, b):  
     return a + b  
 ```
 
 ---
 
-## Entrenamiento  
+## Training Details  
+
+| **Aspect**        | **Detail** |
+|--------------------|-------------|
+| **Method**         | QLoRA with final weight merge |
+| **Frameworks**     | `transformers`, `trl`, `peft`, `bitsandbytes` |
+| **Hardware**       | GPU with 12 GB VRAM (4-bit quantization for training) |
+
+### Main Hyperparameters
 
-- **Método:** QLoRA con Merge final de pesos
-- **Frameworks:** transformers, trl, peft, bitsandbytes  
-- **Hardware:** GPU 12GB VRAM (cuantización en 4-bit para entrenamiento)  
-- **Hiperparámetros principales:**  
-  - per_device_train_batch_size=2  
-  - gradient_accumulation_steps=4  
-  - learning_rate=2e-4  
-  - num_train_epochs=1  
-  - max_seq_length=1024  
+| **Parameter** | **Value** |
+|----------------|-----------|
+| `per_device_train_batch_size` | 2 |
+| `gradient_accumulation_steps` | 4 |
+| `learning_rate` | 2e-4 |
+| `num_train_epochs` | 1 |
+| `max_seq_length` | 1024 |
 
 ---
 
-## Uso  
+## Usage  
 
 ```python
 from transformers import AutoTokenizer, AutoModelForCausalLM  
@@ -82,7 +94,7 @@ model_id = "Ricardouchub/Code-Specialist-7B"
 tok = AutoTokenizer.from_pretrained(model_id)  
 mdl = AutoModelForCausalLM.from_pretrained(model_id, device_map="auto")  
 
-prompt = "[INST] Escribe una función en Python que calcule la media de una lista. [/INST]"  
+prompt = "[INST] Write a Python function that calculates the average of a list. [/INST]"  
 inputs = tok(prompt, return_tensors="pt").to(mdl.device)  
 
 out = mdl.generate(**inputs, max_new_tokens=256)  
@@ -91,14 +103,14 @@ print(tok.decode(out[0], skip_special_tokens=True))
 
 ---
 
-## Benchmarks iniciales  
+## Initial Benchmarks  
 
-- **Eval simple (Python tasks):** mejora en tareas básicas de programación para uso en **data analysis, SQL queries y snippets Python** comparado con el modelo base.  
-- Se recomienda evaluar en HumanEval / MBPP para métricas reproducibles.  
+- **Simple evaluation (Python tasks):** Improved results on small programming and data-related tasks, including **data analysis, SQL query generation, and Python snippets**, compared to the base model.  
+- Further evaluation on **HumanEval** or **MBPP** is recommended for reproducible metrics.
 
 ---
 
-## Autor  
+## Author  
 
 **Ricardo Urdaneta**  
 - [LinkedIn](https://www.linkedin.com/in/ricardourdanetacastro/)  
@@ -106,13 +118,13 @@ print(tok.decode(out[0], skip_special_tokens=True))
 
 ---
 
-## Limitaciones  
+## Limitations  
 
-- El modelo **no garantiza exactitud 100%** en código complejo.  
-- Puede generar respuestas incoherentes en prompts ambiguos.  
+- The model does **not guarantee 100% accuracy** on complex programming tasks.  
+- It may produce inconsistent results for ambiguous or incomplete prompts.  
 
 ---
 
-## Licencia  
+## License  
 
-El modelo se publica con la misma licencia que **Mistral-7B-Instruct-v0.3**: **MIT**
+This model is released under the same license as **Mistral-7B-Instruct-v0.3** — **MIT License**.