DELLA-Merging: Reducing Interference in Model Merging through Magnitude-Based Sampling
Paper โข 2406.11617 โข Published โข 12
How to use h-iida/gemma-2-9b-merge-della-2 with Transformers:
# Use a pipeline as a high-level helper
from transformers import pipeline
pipe = pipeline("text-generation", model="h-iida/gemma-2-9b-merge-della-2") # pip install -U transformers accelerate
# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM
tokenizer = AutoTokenizer.from_pretrained("h-iida/gemma-2-9b-merge-della-2")
model = AutoModelForCausalLM.from_pretrained("h-iida/gemma-2-9b-merge-della-2", device_map="auto")How to use h-iida/gemma-2-9b-merge-della-2 with vLLM:
# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "h-iida/gemma-2-9b-merge-della-2"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
-H "Content-Type: application/json" \
--data '{
"model": "h-iida/gemma-2-9b-merge-della-2",
"prompt": "Once upon a time,",
"max_tokens": 512,
"temperature": 0.5
}'docker model run hf.co/h-iida/gemma-2-9b-merge-della-2
How to use h-iida/gemma-2-9b-merge-della-2 with SGLang:
# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
--model-path "h-iida/gemma-2-9b-merge-della-2" \
--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": "h-iida/gemma-2-9b-merge-della-2",
"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 "h-iida/gemma-2-9b-merge-della-2" \
--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": "h-iida/gemma-2-9b-merge-della-2",
"prompt": "Once upon a time,",
"max_tokens": 512,
"temperature": 0.5
}'How to use h-iida/gemma-2-9b-merge-della-2 with Docker Model Runner:
docker model run hf.co/h-iida/gemma-2-9b-merge-della-2
This is a merge of pre-trained language models created using mergekit.
This model was merged using the DELLA merge method using google/gemma-2-9b as a base.
The following models were included in the merge:
The following YAML configuration was used to produce this model:
models:
- model: google/gemma-2-9b
- model: google/gemma-2-9b-it
parameters:
density: 1.0
weight: 0.6
- model: wzhouad/gemma-2-9b-it-WPO-HB
parameters:
density: 0.55
weight: 0.6
- model: princeton-nlp/gemma-2-9b-it-SimPO
parameters:
density: 0.35
weight: 0.6
- model: UCLA-AGI/Gemma-2-9B-It-SPPO-Iter3
parameters:
density: 0.25
weight: 0.4
merge_method: della
base_model: google/gemma-2-9b
parameters:
normalize: true
int8_mask: true
lambda: 1.0
epsilon: 0.1
dtype: float16