Model soups: averaging weights of multiple fine-tuned models improves accuracy without increasing inference time
Paper • 2203.05482 • Published • 8
How to use mstyslavity/5tiny with Transformers:
# Use a pipeline as a high-level helper
from transformers import pipeline
pipe = pipeline("text-generation", model="mstyslavity/5tiny") # Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM
tokenizer = AutoTokenizer.from_pretrained("mstyslavity/5tiny")
model = AutoModelForCausalLM.from_pretrained("mstyslavity/5tiny")How to use mstyslavity/5tiny with vLLM:
# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "mstyslavity/5tiny"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
-H "Content-Type: application/json" \
--data '{
"model": "mstyslavity/5tiny",
"prompt": "Once upon a time,",
"max_tokens": 512,
"temperature": 0.5
}'docker model run hf.co/mstyslavity/5tiny
How to use mstyslavity/5tiny with SGLang:
# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
--model-path "mstyslavity/5tiny" \
--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": "mstyslavity/5tiny",
"prompt": "Once upon a time,",
"max_tokens": 512,
"temperature": 0.5
}'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 "mstyslavity/5tiny" \
--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": "mstyslavity/5tiny",
"prompt": "Once upon a time,",
"max_tokens": 512,
"temperature": 0.5
}'How to use mstyslavity/5tiny with Docker Model Runner:
docker model run hf.co/mstyslavity/5tiny
This is a merge of pre-trained language models created using mergekit.
This model was merged using the Linear merge method.
The following models were included in the merge:
The following YAML configuration was used to produce this model:
models:
- model: Jiayi-Pan/Tiny-Vicuna-1B
parameters:
weight: 0.2
- model: TinyLlama/TinyLlama-1.1B-Chat-v1.0
parameters:
weight: 0.2
- model: yihanwang617/tinyllama-sft-vicuna-random-100k
parameters:
weight: 0.2
- model: NickyNicky/cognitivecomputations_TinyDolphin-2.8-1.1b
parameters:
weight: 0.2
- model: QuixiAI/TinyDolphin-2.8.2-1.1b
parameters:
weight: 0.2
merge_method: linear
dtype: float32