Instructions to use shebxyd/Kaze-4.7B-A530M-MIT with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use shebxyd/Kaze-4.7B-A530M-MIT with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="shebxyd/Kaze-4.7B-A530M-MIT") 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("shebxyd/Kaze-4.7B-A530M-MIT") model = AutoModelForCausalLM.from_pretrained("shebxyd/Kaze-4.7B-A530M-MIT", 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
- llama.cpp
How to use shebxyd/Kaze-4.7B-A530M-MIT 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 shebxyd/Kaze-4.7B-A530M-MIT:Q4_K_M # Run inference directly in the terminal: llama cli -hf shebxyd/Kaze-4.7B-A530M-MIT:Q4_K_M
Install from WinGet (Windows)
winget install llama.cpp # Start a local OpenAI-compatible server with a web UI: llama serve -hf shebxyd/Kaze-4.7B-A530M-MIT:Q4_K_M # Run inference directly in the terminal: llama cli -hf shebxyd/Kaze-4.7B-A530M-MIT: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 shebxyd/Kaze-4.7B-A530M-MIT:Q4_K_M # Run inference directly in the terminal: ./build/bin/llama cli -hf shebxyd/Kaze-4.7B-A530M-MIT:Q4_K_M
Use Docker
docker model run hf.co/shebxyd/Kaze-4.7B-A530M-MIT:Q4_K_M
- Llama new
- LM Studio
- Jan
- vLLM
How to use shebxyd/Kaze-4.7B-A530M-MIT with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "shebxyd/Kaze-4.7B-A530M-MIT" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "shebxyd/Kaze-4.7B-A530M-MIT", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker
docker model run hf.co/shebxyd/Kaze-4.7B-A530M-MIT:Q4_K_M
- SGLang
How to use shebxyd/Kaze-4.7B-A530M-MIT 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 "shebxyd/Kaze-4.7B-A530M-MIT" \ --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": "shebxyd/Kaze-4.7B-A530M-MIT", "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 "shebxyd/Kaze-4.7B-A530M-MIT" \ --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": "shebxyd/Kaze-4.7B-A530M-MIT", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }' - Ollama
How to use shebxyd/Kaze-4.7B-A530M-MIT with Ollama:
ollama run hf.co/shebxyd/Kaze-4.7B-A530M-MIT:Q4_K_M
- Unsloth Desktop
- Docker Model Runner
How to use shebxyd/Kaze-4.7B-A530M-MIT with Docker Model Runner:
docker model run hf.co/shebxyd/Kaze-4.7B-A530M-MIT:Q4_K_M
- Lemonade
How to use shebxyd/Kaze-4.7B-A530M-MIT with Lemonade:
Pull the model
# Download Lemonade from https://lemonade-server.ai/ lemonade pull shebxyd/Kaze-4.7B-A530M-MIT:Q4_K_M
Run and chat with the model
lemonade run user.Kaze-4.7B-A530M-MIT-Q4_K_M
List all available models
lemonade list
- Atomic Chat
Kaze-4.7B-A530M-MIT
Kaze-4.7B-A530M-MIT es un modelo fundacional de lenguaje Sparse Mixture-of-Experts (MoE) entrenado desde cero por Kineyo (@kineyo.io / Dev_Kofu) bajo licencia MIT, optimizado para inferencia local y privada en dispositivos edge (CPU, Linux y Android).
Arquitectura
- Parámetros totales: ~4.72B | Activos por token: ~528M
- Capas: 23 | d_model: 1024 | GQA: 16Q / 4KV (
head_dim = 64) - MoE: 1 Shared Expert + 31 Routed Experts (
top_k = 2,d_ff = 2048SwiGLU) - Contexto: 4096 (
RoPE,theta = 100000.0) | Vocabulario: 32,768 (Byte-Level BPE propio) - Formato HF:
Qwen2MoeForCausalLM(conversión lossless,expert_bias = 0, sintrust_remote_code)
Archivos Incluidos
model-00001-of-00003.safetensors..model-00003-of-00003.safetensors+model.safetensors.index.json(BF16crudo dividido)kaze-4.7b-a530m-mit-q4_k_m.gguf(cuantizaciónQ4_K_Mlista parallama.cppyKofu_Mobile)
Uso Rápido (transformers)
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer
repo_id = "shebxyd/Kaze-4.7B-A530M-MIT"
tok = AutoTokenizer.from_pretrained(repo_id)
model = AutoModelForCausalLM.from_pretrained(repo_id, dtype=torch.bfloat16, device_map="auto")
prompt = "<|user|>\nHola, ¿quién eres y para qué sirves?\n<|assistant|>\n"
inputs = tok(prompt, return_tensors="pt").to(model.device)
out = model.generate(**inputs, max_new_tokens=128, temperature=0.65, top_p=0.9, repetition_penalty=1.18, do_sample=True)
print(tok.decode(out[0][inputs.input_ids.shape[1]:], skip_special_tokens=True))
Limitaciones
Modelo experimental entrenado desde cero con corpus bajo licencia MIT. Orientado a investigación de arquitecturas Sparse MoE en edge y tareas acotadas en español/código dentro del ecosistema Proyecto Kofu.
- Downloads last month
- 69