跳到主要内容

Model Guide

AtomGit provides a rich set of model resources, which you can easily create, search, download, and use. This guide will help you understand how to perform model-related operations on the platform.

Model Creation​

Create a New Model​

  1. Log in to your AtomGit account
  2. Click "+" at the top right > "New Model"
  3. Fill in the basic information of the model:
    • Model ID
    • Model Name
    • Select a LICENSE template
    • Choose whether to make it public
  4. Select a LICENSE template:
    • PyTorch Creative Commons Attribution Non Commercial 3.0
    • Creative Commons Attribution Non Commercial No Derivatives 3.0
    • Creative Commons Attribution Non Commercial Share Alike 2.0
    • Creative Commons Attribution Non Commercial Share Alike 3.0
    • H Research License
    • Open Model, Data & Weights License Agreement
    • Unknown
  5. Click "Create Model" to complete

Model creation page screenshot Model creation page screenshot

Model Configuration File​

Each model requires a model-config.yaml configuration file, as shown in the example:

model-name: my-awesome-model
version: 1.0.0
framework: pytorch
task: image-classification
dependencies:
  - torch>=2.0.0
  - transformers>=4.30.0
  1. Enter a keyword in the search box
  2. Use filters to sort:
    • Task type
    • Framework
    • License
    • Download count
    • Update time

The following advanced search syntax is supported:

  • framework:pytorch - Search by framework
  • task:nlp - Search by task type
  • stars:>100 - Search by star count
  • language:python - Search by programming language

Model Download​

Download Using Web Interface​

  1. Go to the model details page
  2. Click the "Download" button
  3. Choose version and format

Download Using Command Line​

# Install GitCode CLI
pip install gitcode

# Download model
gitcode download username/model-name

# Download specific version
gitcode download username/model-name --version v1.0.0

Model Usage​

Python Code Example​

from gitcode_hub import load_model

# Load model
model = load_model("username/model-name")

# Use model for inference
result = model.predict(input_data)

API Call Example​

import requests

API_URL = "https://api.gitcode.com/v1/models/username/model-name"
headers = {"Authorization": f"Bearer {API_TOKEN}"}

def query(payload):
    response = requests.post(API_URL, headers=headers, json=payload)
    return response.json()

# Send inference request
output = query({
    "inputs": "Hello, World!" ,
})

Best Practices​

  1. Version Control

    • Use semantic version numbers
    • Maintain backward compatibility
    • Record version change logs

  2. Documentation

    • Provide detailed model descriptions
    • Include usage examples
    • Explain model limitations and notes

  3. Performance Optimization

    • Provide quantized model versions
    • Support batch inference
    • Optimize inference speed

  4. Security

    • Conduct model security testing
    • Provide model cards describing potential risks
    • Comply with data privacy requirements

Frequently Asked Questions​

Q: How do I update a published model? A: You can publish a new version through the version management feature, or update the files of the existing version.

Q: What formats does the model support? A: It supports mainstream deep learning framework formats, including PyTorch, TensorFlow, ONNX, etc.

Q: How to handle model dependencies? A: Declare dependencies in model-config.yaml, or provide a requirements.txt file.