Instructions to use fiel1986/Andromeda-Code-0.5B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use fiel1986/Andromeda-Code-0.5B with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="fiel1986/Andromeda-Code-0.5B")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("fiel1986/Andromeda-Code-0.5B")
model = AutoModelForCausalLM.from_pretrained("fiel1986/Andromeda-Code-0.5B", device_map="auto")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=256)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use fiel1986/Andromeda-Code-0.5B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "fiel1986/Andromeda-Code-0.5B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "fiel1986/Andromeda-Code-0.5B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/fiel1986/Andromeda-Code-0.5B

SGLang

How to use fiel1986/Andromeda-Code-0.5B with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "fiel1986/Andromeda-Code-0.5B" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "fiel1986/Andromeda-Code-0.5B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "fiel1986/Andromeda-Code-0.5B" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "fiel1986/Andromeda-Code-0.5B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use fiel1986/Andromeda-Code-0.5B with Docker Model Runner:
```
docker model run hf.co/fiel1986/Andromeda-Code-0.5B
```

Andromeda-Code-0.5B

File size: 2,157 Bytes

---
license: apache-2.0
base_model: Qwen/Qwen2.5-0.5B-Instruct
language:
  - es
pipeline_tag: text-generation
tags:
  - codigo
  - code
  - destilacion
  - response-distillation
  - deepseek
  - qwen
  - espanol
library_name: transformers
---

![Andromeda](https://cdn-uploads.huggingface.co/production/uploads/69dd3cdd659f1094458104e3/J2THbjZIFL_TTNpOXu6Wg.jpeg)

# Andromeda-Code-0.5B 💻

Modelo de lenguaje de **0.5B parámetros** especializado en **programación**,
destilado desde `deepseek-ai/deepseek-coder-1.3b-instruct` hacia
`Qwen/Qwen2.5-0.5B-Instruct`.

> DeepSeek y Qwen usan tokenizers distintos, así que la destilación por logits
> no aplica directamente. Acá se usó **response distillation**: el teacher genera
> respuestas de código, se filtran y limpian, y el student aprende con SFT sobre
> ese dataset sintético.

## 🎯 Método de destilación

- **Teacher:** `deepseek-ai/deepseek-coder-1.3b-instruct`
- **Student:** `Qwen/Qwen2.5-0.5B-Instruct`
- **Técnica:** Response Distillation (Answer-Guided) + SFT
  1. Generación de respuestas de código con el teacher
  2. Limpieza: extracción de bloques ```python, remoción de texto introductorio
  3. Conversión al chat template de Qwen (formato JSONL)
  4. Fine-tuning del student (3 epochs, lr=2e-5, bf16)

## ⚠️ Estado experimental

Este modelo fue entrenado con un **dataset de prueba de solo 20 ejemplos**.
Sirve para validar el pipeline de destilación cross-tokenizer, pero su calidad
es limitada.

## 🚀 Uso rápido

```python
from transformers import AutoModelForCausalLM, AutoTokenizer

model_id = "fiel1986/Andromeda-Code-0.5B"
tok = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id, device_map="auto")

messages = [
    {"role": "system", "content": "Sos un asistente experto en programacion."},
    {"role": "user", "content": "Escribe una funcion en Python que verifique si un numero es primo"}
]
inputs = tok.apply_chat_template(messages, return_tensors="pt").to(model.device)
out = model.generate(inputs, max_new_tokens=256, do_sample=True, temperature=0.6)
print(tok.decode(out[0], skip_special_tokens=True))