Instructions to use vtava/Tiny-LLM-PDelta3-GDN2-InputRoute with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use vtava/Tiny-LLM-PDelta3-GDN2-InputRoute with Transformers:

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

pipe = pipeline("text-generation", model="vtava/Tiny-LLM-PDelta3-GDN2-InputRoute")

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("vtava/Tiny-LLM-PDelta3-GDN2-InputRoute", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use vtava/Tiny-LLM-PDelta3-GDN2-InputRoute with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "vtava/Tiny-LLM-PDelta3-GDN2-InputRoute"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "vtava/Tiny-LLM-PDelta3-GDN2-InputRoute",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/vtava/Tiny-LLM-PDelta3-GDN2-InputRoute

SGLang

How to use vtava/Tiny-LLM-PDelta3-GDN2-InputRoute 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 "vtava/Tiny-LLM-PDelta3-GDN2-InputRoute" \
    --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": "vtava/Tiny-LLM-PDelta3-GDN2-InputRoute",
		"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 "vtava/Tiny-LLM-PDelta3-GDN2-InputRoute" \
        --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": "vtava/Tiny-LLM-PDelta3-GDN2-InputRoute",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use vtava/Tiny-LLM-PDelta3-GDN2-InputRoute with Docker Model Runner:
```
docker model run hf.co/vtava/Tiny-LLM-PDelta3-GDN2-InputRoute
```

Tiny-LLM-PDelta3-GDN2-InputRoute / selection.json

vtava

Upload conv4_gdn2_inputroute_f96 from TinyCeNN Tiny-LLM compatibility benchmark

c21da77 verified about 1 month ago

Raw History Blame Contribute Delete

626 Bytes

	{
	"quality_winner": "conv4_gdn2_inputroute_f96",
	"efficient_winner": "conv4_gdn2_f96",
	"quality_criterion": "weighted validation delta NLL: 0.20@256 + 0.30@512 + 0.50@1024",
	"validation_composite_delta_nll": {
	"conv4_pdelta_f96": 0.22705587267875638,
	"conv4_channel_decay_f96": 0.2314825379848476,
	"conv4_gdn2_f96": 0.2087166726589203,
	"conv4_gdn2_inputroute_f96": 0.20554421544075002
	},
	"test_not_used_for_selection": true,
	"tiny_llm_has_one_layer": true,
	"true_clvr_available": false,
	"route_proxy": "conv4_gdn2_inputroute_f96 routes current pre-convolution V; it is not cross-layer CLVR"
	}