Instructions to use Korzadi/Atomic-Code-1.1 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use Korzadi/Atomic-Code-1.1 with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="Korzadi/Atomic-Code-1.1") messages = [ {"role": "user", "content": "Who are you?"}, ] pipe(messages)# pip install -U transformers accelerate # Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("Korzadi/Atomic-Code-1.1", device_map="auto") - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- llama.cpp
How to use Korzadi/Atomic-Code-1.1 with llama.cpp:
Install (macOS, Linux)
curl -LsSf https://llama.app/install.sh | sh # Start a local OpenAI-compatible server with a web UI: llama serve -hf Korzadi/Atomic-Code-1.1:Q4_K_M # Run inference directly in the terminal: llama cli -hf Korzadi/Atomic-Code-1.1:Q4_K_M
Install from WinGet (Windows)
winget install llama.cpp # Start a local OpenAI-compatible server with a web UI: llama serve -hf Korzadi/Atomic-Code-1.1:Q4_K_M # Run inference directly in the terminal: llama cli -hf Korzadi/Atomic-Code-1.1:Q4_K_M
Build from source code
git clone https://github.com/ggml-org/llama.cpp.git cd llama.cpp cmake -B build cmake --build build -j --target llama-app # Start a local OpenAI-compatible server with a web UI: ./build/bin/llama serve -hf Korzadi/Atomic-Code-1.1:Q4_K_M # Run inference directly in the terminal: ./build/bin/llama cli -hf Korzadi/Atomic-Code-1.1:Q4_K_M
Use Docker
docker model run hf.co/Korzadi/Atomic-Code-1.1:Q4_K_M
- Llama new
- LM Studio
- Jan
- vLLM
How to use Korzadi/Atomic-Code-1.1 with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "Korzadi/Atomic-Code-1.1" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "Korzadi/Atomic-Code-1.1", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker
docker model run hf.co/Korzadi/Atomic-Code-1.1:Q4_K_M
- SGLang
How to use Korzadi/Atomic-Code-1.1 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 "Korzadi/Atomic-Code-1.1" \ --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": "Korzadi/Atomic-Code-1.1", "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 "Korzadi/Atomic-Code-1.1" \ --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": "Korzadi/Atomic-Code-1.1", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }' - Ollama
How to use Korzadi/Atomic-Code-1.1 with Ollama:
ollama run hf.co/Korzadi/Atomic-Code-1.1:Q4_K_M
- Unsloth Desktop
- Pi
How to use Korzadi/Atomic-Code-1.1 with Pi:
Start the llama.cpp server
# Install llama.cpp: brew install llama.cpp # Start a local OpenAI-compatible server: llama serve -hf Korzadi/Atomic-Code-1.1:Q4_K_M
Configure the model in Pi
# Install Pi: npm install -g @earendil-works/pi-coding-agent # Add to ~/.pi/agent/models.json: { "providers": { "llama-cpp": { "baseUrl": "http://localhost:8080/v1", "api": "openai-completions", "apiKey": "none", "models": [ { "id": "Korzadi/Atomic-Code-1.1:Q4_K_M" } ] } } }Run Pi
# Start Pi in your project directory: pi
- Docker Model Runner
How to use Korzadi/Atomic-Code-1.1 with Docker Model Runner:
docker model run hf.co/Korzadi/Atomic-Code-1.1:Q4_K_M
- Lemonade
How to use Korzadi/Atomic-Code-1.1 with Lemonade:
Pull the model
# Download Lemonade from https://lemonade-server.ai/ lemonade pull Korzadi/Atomic-Code-1.1:Q4_K_M
Run and chat with the model
lemonade run user.Atomic-Code-1.1-Q4_K_M
List all available models
lemonade list
- Hermes Agent
How to use Korzadi/Atomic-Code-1.1 with Hermes Agent:
Start the llama.cpp server
# Install llama.cpp: brew install llama.cpp # Start a local OpenAI-compatible server: llama serve -hf Korzadi/Atomic-Code-1.1:Q4_K_M
Configure Hermes
# Install Hermes: curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash hermes setup # Point Hermes at the local server: hermes config set model.provider custom hermes config set model.base_url http://127.0.0.1:8080/v1 hermes config set model.default Korzadi/Atomic-Code-1.1:Q4_K_M
Run Hermes
hermes
- Atomic Chat
- OpenClaw
How to use Korzadi/Atomic-Code-1.1 with OpenClaw:
Start the llama.cpp server
# Install llama.cpp: brew install llama.cpp # Start a local OpenAI-compatible server: llama serve -hf Korzadi/Atomic-Code-1.1:Q4_K_M
Configure OpenClaw
# Install OpenClaw: npm install -g openclaw@latest # Register the local server and set it as the default model: openclaw onboard --non-interactive --mode local \ --auth-choice custom-api-key \ --custom-base-url http://127.0.0.1:8080/v1 \ --custom-model-id "Korzadi/Atomic-Code-1.1:Q4_K_M" \ --custom-provider-id llama-cpp \ --custom-compatibility openai \ --custom-text-input \ --accept-risk \ --skip-health
Run OpenClaw
openclaw agent --local --agent main --message "Hello from Hugging Face"
Atomic Code 1.1
Atomic Code 1.1 es la distribuciรณn oficial, completa e independiente de la familia de modelos Atomic, entrenada por Korzadi. Este repositorio unifica la experiencia de despliegue ofreciendo los pesos completos fusionados de alta precisiรณn, el adaptador LoRA original para desarrollo flexible y versiones optimizadas en formato GGUF.
๐ช Estructura de la Distribuciรณn
Para garantizar la mรกxima flexibilidad sin colisiones en el ecosistema Hugging Face, el repositorio se organiza de la siguiente manera:
- Pesos Consolidados (Modelo Completo): Ubicados bajo la carpeta
consolidated/en precisiรณnbfloat16/float16nativa, listos para cargarse directamente en entornos de producciรณn contransformers. - Adaptador LoRA Original: Ubicado en la raรญz (
adapter_model.safetensorsyadapter_config.json) para cargas dinรกmicas usando PEFT con Unsloth sobreQwen2.5-Coder-7B-Instruct. - Cuantizaciones GGUF: Disponibles en la carpeta
gguf/para inferencia local rรกpida en CPU/Edge usandollama.cppuOllama.
๐ Ficha Tรฉcnica y Especificaciones
| Parรกmetro | Valor de Especificaciรณn |
|---|---|
| Creador / Autor | Korzadi |
| Arquitectura Base | Qwen2 (Qwen2ForCausalLM) |
| Modelo Base de Origen | Qwen2.5-Coder-7B-Instruct |
| Tamaรฑo de Parรกmetros | ~7.61 Billones |
| Dimensiรณn Oculta (Hidden Size) | 3584 |
| Tamaรฑo de Vocabulario | 152064 |
| Longitud de Contexto | 32,768 tokens (ampliable) |
| Formato de Pesos | Safetensors |
| Dtype de Inferencia Recomendado | bfloat16 o float16 |
๐ ๏ธ Guรญa de Uso e Inferencia (Transformers)
1. Cargar el Modelo Completo Consolidado
Para ejecutar la versiรณn de pesos consolidados e independientes directamente desde la subcarpeta dedicada:
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch
model_id = "Korzadi/Atomic-Code-1.1"
# Carga del tokenizador y el modelo consolidado
tokenizer = AutoTokenizer.from_pretrained(model_id, subfolder="consolidated")
model = AutoModelForCausalLM.from_pretrained(
model_id,
subfolder="consolidated",
torch_dtype=torch.bfloat16,
device_map="auto"
)
# Prompt de ejemplo
prompt = "Escribe una funciรณn en Python para validar una direcciรณn IP v4 y v6."
messages = [
{"role": "system", "content": "Eres Atomic Code, un asistente de programaciรณn experto de nivel Senior."},
{"role": "user", "content": prompt}
]
text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
model_inputs = tokenizer([text], return_tensors="pt").to(model.device)
generated_ids = model.generate(model_inputs.input_ids, max_new_tokens=512)
response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
print(response)
2. Cargar como Adaptador PEFT / LoRA
Si deseas utilizar el adaptador ligero cargรกndolo dinรกmicamente sobre la base original:
from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftModel
import torch
base_model_id = "Qwen/Qwen2.5-Coder-7B-Instruct"
adapter_id = "Korzadi/Atomic-Code-1.1"
tokenizer = AutoTokenizer.from_pretrained(base_model_id)
base_model = AutoModelForCausalLM.from_pretrained(base_model_id, torch_dtype=torch.bfloat16, device_map="auto")
model = PeftModel.from_pretrained(base_model, adapter_id)
๐๏ธ Configuraciรณn para Ollama y Llama.cpp
Para ejecutar el modelo localmente utilizando el archivo unificado atomic-code-Q4_K_M.gguf provisto en el repositorio, crea un archivo Modelfile:
FROM ./gguf/atomic-code-Q4_K_M.gguf
# Configurar parรกmetros del sistema
PARAMETER temperature 0.2
PARAMETER top_p 0.95
PARAMETER stop <|im_start|>
PARAMETER stop <|im_end|>
TEMPLATE """<|im_start|>system
Eres Atomic Code, un asistente de programaciรณn altamente calificado creado por Korzadi.<|im_end|>
<|im_start|>user
{{ .Prompt }}<|im_end|>
<|im_start|>assistant
"""
Ejecuta en tu terminal:
ollama create atomic-code-1.1 -f Modelfile
ollama run atomic-code-1.1
๐ Historial de Versiones e Integridad
- v1.0 (Free): Publicaciรณn inicial de distribuciรณn.
- v1.1 (Active Adapter): Ajuste fino de precisiรณn entrenado mediante QLoRA y Unsloth enfocado en el dominio de desarrollo de software.
- v1.2 (Consolidated): Fusiรณn completa de pesos consolidados en topologรญa FP16/BF16 para garantizar compatibilidad nativa sin capas de adaptaciรณn adicionales.
๐ Licencia
Este modelo se distribuye bajo la licencia Apache 2.0.
Ficha oficial generada para el ecosistema Korzadi / Atomic.
๐ Cรณmo usar Atomic Code 1.1
Opciรณn 1: Carga del Checkpoint Completo Consolidado (BF16)
Para cargar la versiรณn consolidada y ajustada del modelo completo (sin necesidad de cargar adaptadores por separado), debes apuntar explรญcitamente a la subcarpeta consolidated:
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch
model_id = "Korzadi/Atomic-Code-1.1"
# Carga de Tokenizer y Modelo especificando subfolder
tokenizer = AutoTokenizer.from_pretrained(model_id, subfolder="consolidated")
model = AutoModelForCausalLM.from_pretrained(
model_id,
subfolder="consolidated",
torch_dtype=torch.bfloat16,
device_map="auto"
)
# Inferencia
messages = [
{"role": "system", "content": "You are Atomic Code, an advanced coding assistant."},
{"role": "user", "content": "Escribe una funciรณn de ordenamiento rรกpido en Python."}
]
formatted_chat = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
inputs = tokenizer(formatted_chat, return_tensors="pt").to(model.device)
outputs = model.generate(**inputs, max_new_tokens=512, do_sample=True, temperature=0.7)
print(tokenizer.decode(outputs[0][inputs.input_ids.shape[1]:], skip_special_tokens=True))
Opciรณn 2: Carga mediante Adaptador LoRA (PEFT)
Si prefieres utilizar el adaptador LoRA sobre el modelo base original de Qwen:
from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftModel
import torch
base_model_id = "Qwen/Qwen2.5-Coder-7B-Instruct"
adapter_id = "Korzadi/Atomic-Code-1.1"
tokenizer = AutoTokenizer.from_pretrained(base_model_id)
base_model = AutoModelForCausalLM.from_pretrained(
base_model_id,
torch_dtype=torch.bfloat16,
device_map="auto"
)
model = PeftModel.from_pretrained(base_model, adapter_id)
Opciรณn 3: Inferencia local con GGUF (llama.cpp)
El modelo de cuantizaciรณn intermedia recomendado estรก disponible en:
gguf/atomic-code-Q4_K_M.gguf
- Downloads last month
- 34
4-bit