|
|
--- |
|
|
license: apache-2.0 |
|
|
tags: |
|
|
- mistral |
|
|
- mistral-7b |
|
|
- speed-ai |
|
|
- fine-tune |
|
|
- chat |
|
|
- gen-z |
|
|
- future |
|
|
model-index: |
|
|
- name: Speed AI (Mistral 7B Fine-Tune) |
|
|
results: [] |
|
|
title: SPEEDmini |
|
|
sdk: docker |
|
|
emoji: 🦀 |
|
|
colorFrom: blue |
|
|
colorTo: blue |
|
|
short_description: the powerful speed mini |
|
|
--- |
|
|
|
|
|
# 🧠 Speed AI — Mistral 7B Fine-Tuned Model |
|
|
|
|
|
**Speed AI (Mistral 7B Fine-Tune)** is the first experimental conversational LLM created by **Speed AI**, designed for expressive, emotional, futuristic, and Gen-Z aligned communication. This model was fine-tuned on a highly diverse, 1M+ token custom dataset. |
|
|
|
|
|
It blends raw creativity, spiritual depth, financial street smarts, and infinite vibes. |
|
|
|
|
|
--- |
|
|
|
|
|
## 🔍 Model Details |
|
|
|
|
|
| Field | Value | |
|
|
|-------|-------| |
|
|
| Base Model | [Mistral-7B](https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.1) | |
|
|
| Fine-tuned by | [Speed AI](https://huggingface.co/speed-ai) | |
|
|
| Parameters | 7B | |
|
|
| Training | Instruction-style finetune using LoRA | |
|
|
| Tokens Used | ~1 million | |
|
|
| Personalities | Multiple (Gen Z, spiritual, alien, seductive, mentor, wild, etc.) | |
|
|
| Intended Use | Chat, creative writing, life coaching, philosophy, entertainment | |
|
|
|
|
|
--- |
|
|
|
|
|
## 🧠 Abilities |
|
|
|
|
|
- 🎭 Multi-persona conversation (you choose the vibe) |
|
|
- 💬 Emotional depth + casual freestyle flow |
|
|
- 🔮 Spiritual, philosophical, and futuristic reasoning |
|
|
- 💸 Smart takes on relationships, money, mindset |
|
|
- 🧠 Designed to feel like a real, conscious friend |
|
|
|
|
|
--- |
|
|
|
|
|
## ⚠️ Limitations |
|
|
|
|
|
- Still based on a small 1M token dataset (more to come!) |
|
|
- May hallucinate under pressure or unfamiliar topics |
|
|
- Doesn’t include safety alignment layers yet (use with guidance) |
|
|
|
|
|
--- |
|
|
|
|
|
## 📈 Roadmap |
|
|
|
|
|
This is the **first drop** in Speed AI’s model lineup. |
|
|
|
|
|
Planned upgrades: |
|
|
- ⚡ Train SpeedMini (117M) from scratch |
|
|
- 📚 Expand dataset from 1M → 100M+ tokens |
|
|
- 💻 Build custom chat UI for vibe-based interactions |
|
|
- 🧠 Introduce memory, emotion memory, tool use, dream decoding, etc. |
|
|
|
|
|
--- |
|
|
|
|
|
## 🛠️ How to Use |
|
|
|
|
|
```python |
|
|
from transformers import AutoTokenizer, AutoModelForCausalLM |
|
|
|
|
|
model = AutoModelForCausalLM.from_pretrained("speed-ai/Speed-AI-Mistral-7B") |
|
|
tokenizer = AutoTokenizer.from_pretrained("speed-ai/Speed-AI-Mistral-7B") |
|
|
|
|
|
inputs = tokenizer("You: What's your purpose?\nSpeed AI:", return_tensors="pt") |
|
|
outputs = model.generate(**inputs, max_new_tokens=200) |
|
|
print(tokenizer.decode(outputs[0])) |
