|
|
--- |
|
|
title: OpenLLM Inference Space |
|
|
emoji: 🚀 |
|
|
colorFrom: blue |
|
|
colorTo: purple |
|
|
sdk: gradio |
|
|
sdk_version: 4.44.1 |
|
|
app_file: app.py |
|
|
pinned: false |
|
|
license: gpl-3.0 |
|
|
--- |
|
|
|
|
|
# 🚀 OpenLLM Inference Space |
|
|
|
|
|
Welcome to the OpenLLM Inference Space! This is a comprehensive interface for running inference on our trained OpenLLM models with customizable parameters. |
|
|
|
|
|
## 🎯 Available Models |
|
|
|
|
|
We provide **7 different models** trained for varying numbers of steps: |
|
|
|
|
|
| Model | Training Steps | Description | Best Loss | |
|
|
|-------|---------------|-------------|-----------| |
|
|
| **4k Model** | 4,000 | Early training stage, basic language patterns | ~6.2 | |
|
|
| **6k Model** | 6,000 | Improved coherence, better vocabulary usage | ~5.8 | |
|
|
| **7k Model** | 7,000 | Enhanced text generation quality | ~5.5 | |
|
|
| **8k Model** | 8,000 | More sophisticated language understanding | ~5.3 | |
|
|
| **9k Model** | 9,000 | Best performing model (latest training) | ~5.2 | |
|
|
| **10k Model** | 10,000 | Latest extended training, maximum performance | ~5.22 | |
|
|
| **10k Improved** | 10,000 | Improved training process, proper checkpoint format | ~5.1774 | |
|
|
|
|
|
## 🎮 How to Use |
|
|
|
|
|
1. **Select a Model** from the dropdown menu |
|
|
2. **Load the Model** to see its information |
|
|
3. **Enter Your Prompt** in the text box |
|
|
4. **Adjust Parameters** (temperature, max length, etc.) |
|
|
5. **Generate Text** and see the results! |
|
|
|
|
|
## ⚙️ Parameters |
|
|
|
|
|
- **Temperature**: Controls randomness (0.1-2.0) |
|
|
- **Max Length**: Number of tokens to generate (10-500) |
|
|
- **Top-K**: Limits to top-k most likely tokens (1-100) |
|
|
- **Top-P**: Nucleus sampling threshold (0.1-1.0) |
|
|
|
|
|
## 🧠 Model Architecture |
|
|
|
|
|
- **Model Size**: Small (35.8M parameters) |
|
|
- **Layers**: 6 transformer layers |
|
|
- **Embedding**: 512 dimensions |
|
|
- **Vocabulary**: 32,000 tokens (SentencePiece) |
|
|
- **Context Length**: 1,024 tokens |
|
|
|
|
|
--- |
|
|
|
|
|
**OpenLLM Inference Space** - Experience the power of open-source language models! 🚀 |
|
|
|
