Instructions to use TinyLlama/TinyLlama-1.1B-step-50K-105b with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use TinyLlama/TinyLlama-1.1B-step-50K-105b with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="TinyLlama/TinyLlama-1.1B-step-50K-105b")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("TinyLlama/TinyLlama-1.1B-step-50K-105b")
model = AutoModelForCausalLM.from_pretrained("TinyLlama/TinyLlama-1.1B-step-50K-105b", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use TinyLlama/TinyLlama-1.1B-step-50K-105b with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "TinyLlama/TinyLlama-1.1B-step-50K-105b"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "TinyLlama/TinyLlama-1.1B-step-50K-105b",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/TinyLlama/TinyLlama-1.1B-step-50K-105b

SGLang

How to use TinyLlama/TinyLlama-1.1B-step-50K-105b 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 "TinyLlama/TinyLlama-1.1B-step-50K-105b" \
    --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": "TinyLlama/TinyLlama-1.1B-step-50K-105b",
		"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 "TinyLlama/TinyLlama-1.1B-step-50K-105b" \
        --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": "TinyLlama/TinyLlama-1.1B-step-50K-105b",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use TinyLlama/TinyLlama-1.1B-step-50K-105b with Docker Model Runner:
```
docker model run hf.co/TinyLlama/TinyLlama-1.1B-step-50K-105b
```

Should this work with llama2.c from karpathy?

by JulianX4 - opened Sep 4, 2023

Discussion

JulianX4

Sep 4, 2023

See title. Thank you!

PY007

TinyLlama org Sep 5, 2023

•

edited Sep 6, 2023

It is compatible with llama.cpp. I have not played with llama2.c yet but my best guess is Yes. We use the same arc. & tokenizer with Llama 2. Maybe you need to modify some config files a bit to fit our model shape.

deleted

Sep 6, 2023

This comment has been hidden

JulianX4

Sep 6, 2023

I tried it before and got “segmentation fault”.

I should have some time later tonight for tinkering, if I get a better result I’ll write an update.

deleted

Sep 9, 2023

This comment has been hidden

JulianX4

Sep 9, 2023

I tried it before and got “segmentation fault”.

I should have some time later tonight for tinkering, if I get a better result I’ll write an update.

I'm hitting the following error when running export.py:
RuntimeError: shape '[32, 2, 32, 2048]' is invalid for input of size 524288
May I know any update from your side? It will be great if you can share your findings so far, I think it will benefit everyone here.

Thank you.

Unfortunately that's as far as I got as well.

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment