Instructions to use Solshine/Llama-3-1-big-thoughtful-passthrough-merge-2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Solshine/Llama-3-1-big-thoughtful-passthrough-merge-2 with Transformers:

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

pipe = pipeline("text-generation", model="Solshine/Llama-3-1-big-thoughtful-passthrough-merge-2")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Solshine/Llama-3-1-big-thoughtful-passthrough-merge-2")
model = AutoModelForCausalLM.from_pretrained("Solshine/Llama-3-1-big-thoughtful-passthrough-merge-2", device_map="auto")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use Solshine/Llama-3-1-big-thoughtful-passthrough-merge-2 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Solshine/Llama-3-1-big-thoughtful-passthrough-merge-2"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Solshine/Llama-3-1-big-thoughtful-passthrough-merge-2",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Solshine/Llama-3-1-big-thoughtful-passthrough-merge-2

SGLang

How to use Solshine/Llama-3-1-big-thoughtful-passthrough-merge-2 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 "Solshine/Llama-3-1-big-thoughtful-passthrough-merge-2" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Solshine/Llama-3-1-big-thoughtful-passthrough-merge-2",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

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 "Solshine/Llama-3-1-big-thoughtful-passthrough-merge-2" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Solshine/Llama-3-1-big-thoughtful-passthrough-merge-2",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Solshine/Llama-3-1-big-thoughtful-passthrough-merge-2 with Docker Model Runner:
```
docker model run hf.co/Solshine/Llama-3-1-big-thoughtful-passthrough-merge-2
```

State of the art for size on Open LLM Leaderboard on acc_norm score

SOTA at size level as of acc_norm score on 9/30/2024, viewable at open-llm-leaderboard/Solshine__Llama-3-1-big-thoughtful-passthrough-merge-2-details

acc_norm of 31.5% according to open llm leaderboard test result dataset.

Due to the merged and minimally retrained nature of this model, this score may not reflect in human evaluated general performance in some domains.

merge

This is a merge of pre-trained language models created using mergekit.

Merge Details

Merge Method

This model was merged using the passthrough merge method.

Models Merged

The following models were included in the merge:

Configuration

The following YAML configuration was used to produce this model:

slices:
- sources:
  - layer_range: [0, 16]
    model: mlabonne/Hermes-3-Llama-3.1-8B-lorablated
- sources:
  - layer_range: [4, 20]
    model: Solshine/reflection-llama-3.1-8B
- sources:
  - layer_range: [8, 24]
    model: Solshine/Meta-Llama-3.1-8B-Instruct-Python-Coder
- sources:
  - layer_range: [12, 28]
    model: Solshine/reflection-llama-3.1-8B
- sources:
  - layer_range: [16, 32]
    model: mlabonne/Hermes-3-Llama-3.1-8B-lorablated
merge_method: passthrough
dtype: float16

Downloads last month: 4

Safetensors

Model size

18B params

Tensor type

F16

Model tree for Solshine/Llama-3-1-big-thoughtful-passthrough-merge-2

Solshine/Meta-Llama-3.1-8B-Instruct-Python-Coder

Solshine/reflection-llama-3.1-8B

mlabonne/Hermes-3-Llama-3.1-8B-lorablated

Merge model

this model

Evaluation results

strict accuracy on IFEval (0-Shot)
Open LLM Leaderboard

25.470
normalized accuracy on BBH (3-Shot)
Open LLM Leaderboard

5.010
exact match on MATH Lvl 5 (4-Shot)
Open LLM Leaderboard

0.150
acc_norm on GPQA (0-shot)
Open LLM Leaderboard

1.230
acc_norm on MuSR (0-shot)
Open LLM Leaderboard

6.750
accuracy on MMLU-PRO (5-shot)
test set Open LLM Leaderboard

2.060