Text Generation
Transformers
Safetensors
English
tinybuddy
tiny-lm
tinystories
educational
built-with-llama
custom_code
Instructions to use Eeppa/TinyBuddy-30M with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use Eeppa/TinyBuddy-30M with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="Eeppa/TinyBuddy-30M", trust_remote_code=True)# Load model directly from transformers import AutoModelForCausalLM model = AutoModelForCausalLM.from_pretrained("Eeppa/TinyBuddy-30M", trust_remote_code=True, dtype="auto") - Notebooks
- Google Colab
- Kaggle
- Local Apps
- vLLM
How to use Eeppa/TinyBuddy-30M with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "Eeppa/TinyBuddy-30M" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "Eeppa/TinyBuddy-30M", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }'Use Docker
docker model run hf.co/Eeppa/TinyBuddy-30M
- SGLang
How to use Eeppa/TinyBuddy-30M 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 "Eeppa/TinyBuddy-30M" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "Eeppa/TinyBuddy-30M", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }'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 "Eeppa/TinyBuddy-30M" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "Eeppa/TinyBuddy-30M", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }' - Docker Model Runner
How to use Eeppa/TinyBuddy-30M with Docker Model Runner:
docker model run hf.co/Eeppa/TinyBuddy-30M
Create README.txt
Browse files- README.txt +28 -0
README.txt
ADDED
|
@@ -0,0 +1,28 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# TinyBuddy-30M
|
| 2 |
+
|
| 3 |
+
A 30 million parameter GPT-style transformer trained on TinyStories.
|
| 4 |
+
|
| 5 |
+
## Architecture
|
| 6 |
+
- 6 layers, 8 attention heads, 256 embedding dim
|
| 7 |
+
- 50,000 vocabulary size (untied weights)
|
| 8 |
+
- 512 context length (trained on 128 for speed)
|
| 9 |
+
|
| 10 |
+
## Training
|
| 11 |
+
- Dataset: TinyStories (5,000 stories)
|
| 12 |
+
- Steps: 1,500
|
| 13 |
+
- Hardware: CPU only
|
| 14 |
+
- Loss: ~5.5 (coherent but not good)
|
| 15 |
+
|
| 16 |
+
## What It Can Do
|
| 17 |
+
- Generate 2-3 word fragments that resemble story patterns
|
| 18 |
+
- Sometimes repeat words from the prompt
|
| 19 |
+
- Produce gibberish that's trying to be English
|
| 20 |
+
|
| 21 |
+
## What It Cannot Do
|
| 22 |
+
- Tell a coherent story
|
| 23 |
+
- Answer questions
|
| 24 |
+
- Anything useful
|
| 25 |
+
|
| 26 |
+
## Why It Exists
|
| 27 |
+
To demonstrate that even a tiny transformer learns *patterns*, not rules.
|
| 28 |
+
This is a real AI, just a very small one.
|