Instructions to use Locutusque/Hyperion-3.0-Mistral-7B-DPO with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Locutusque/Hyperion-3.0-Mistral-7B-DPO with Transformers:

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

pipe = pipeline("text-generation", model="Locutusque/Hyperion-3.0-Mistral-7B-DPO")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Locutusque/Hyperion-3.0-Mistral-7B-DPO")
model = AutoModelForCausalLM.from_pretrained("Locutusque/Hyperion-3.0-Mistral-7B-DPO")

Inference
Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use Locutusque/Hyperion-3.0-Mistral-7B-DPO with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Locutusque/Hyperion-3.0-Mistral-7B-DPO"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Locutusque/Hyperion-3.0-Mistral-7B-DPO",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/Locutusque/Hyperion-3.0-Mistral-7B-DPO

SGLang

How to use Locutusque/Hyperion-3.0-Mistral-7B-DPO 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 "Locutusque/Hyperion-3.0-Mistral-7B-DPO" \
    --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": "Locutusque/Hyperion-3.0-Mistral-7B-DPO",
		"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 "Locutusque/Hyperion-3.0-Mistral-7B-DPO" \
        --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": "Locutusque/Hyperion-3.0-Mistral-7B-DPO",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use Locutusque/Hyperion-3.0-Mistral-7B-DPO with Docker Model Runner:
```
docker model run hf.co/Locutusque/Hyperion-3.0-Mistral-7B-DPO
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

Hyperion-3.0-Mistral-7B-DPO

Model Details

Model Name: Locutusque/Hyperion-3.0-Mistral-7B-DPO
Base Model: mistralai/Mistral-7B-v0.1
Publisher: Locutusque
Model Type: Question answering, conversational AI, code generation, medical text comprehension, mathematical reasoning, logical reasoning
Language: Multi-domain, English language
License: Apache-2.0

Model Description

Locutusque/Hyperion-3.0-Mistral-7B-DPO is an advanced language model fine-tuned with a dataset of 20,000 meticulously curated high-quality preference pairs using Direct Preference Optimization (DPO). The examples were generated by GPT-4 to ensure exceptional quality and relevance. This model is designed to provide superior performance across a wide range of complex tasks, including question answering, conversational AI, code generation, medical text comprehension, mathematical reasoning, and logical reasoning.

Intended Use

This model is intended for researchers, developers, and organizations seeking a highly capable and reliable language model for tackling challenging problems across various domains. Potential use cases include:

Intelligent tutoring systems and educational applications in science, medicine, mathematics, and computer science
Advanced conversational AI for technical support, customer service, and domain-specific chatbots
Code generation and analysis tools for software development and programming assistance
Medical text analysis and information retrieval for healthcare professionals and researchers
Mathematical problem-solving and logical reasoning applications for academia and industry

Training Data

The Locutusque/Hyperion-3.0-Mistral-7B-DPO model was fine-tuned on a carefully curated dataset of 20,000 preference pairs, where 4,000 examples were used to fine-tune. These examples were generated by GPT-4 to ensure the highest quality and relevance across various domains, including programming, medical texts, mathematical problems, and reasoning tasks. The training data was further optimized using Direct Preference Optimization (DPO) to align the model's outputs with human preferences and improve overall performance.

Quants

ExLlamaV2: https://huggingface.co/bartowski/Hyperion-3.0-Mistral-7B-DPO-exl2

GGUF: https://huggingface.co/bartowski/Hyperion-3.0-Mistral-7B-DPO-GGUF

Evaluation Results

mmlu flan cot 5-shot

Tasks	Version	Filter	Metric	Value		Stderr
mmlu_flan_cot_fewshot	N/A	get-answer	exact_match	0.5833	±	0.0118
- mmlu_flan_cot_fewshot_humanities	N/A	get-answer	exact_match	0.5039	±	0.0205
- mmlu_flan_cot_fewshot_formal_logic	0	get-answer	exact_match	0.2143	±	0.1138
- mmlu_flan_cot_fewshot_high_school_european_history	0	get-answer	exact_match	0.6667	±	0.1143
- mmlu_flan_cot_fewshot_high_school_us_history	0	get-answer	exact_match	0.7727	±	0.0914
- mmlu_flan_cot_fewshot_high_school_world_history	0	get-answer	exact_match	0.5385	±	0.0997
- mmlu_flan_cot_fewshot_international_law	0	get-answer	exact_match	0.9231	±	0.0769
- mmlu_flan_cot_fewshot_jurisprudence	0	get-answer	exact_match	0.5455	±	0.1575
- mmlu_flan_cot_fewshot_logical_fallacies	0	get-answer	exact_match	0.7778	±	0.1008
- mmlu_flan_cot_fewshot_moral_disputes	0	get-answer	exact_match	0.5526	±	0.0817
- mmlu_flan_cot_fewshot_moral_scenarios	0	get-answer	exact_match	0.4000	±	0.0492
- mmlu_flan_cot_fewshot_philosophy	0	get-answer	exact_match	0.7647	±	0.0738
- mmlu_flan_cot_fewshot_prehistory	0	get-answer	exact_match	0.6571	±	0.0814
- mmlu_flan_cot_fewshot_professional_law	0	get-answer	exact_match	0.3294	±	0.0362
- mmlu_flan_cot_fewshot_world_religions	0	get-answer	exact_match	0.8947	±	0.0723
- mmlu_flan_cot_fewshot_other	N/A	get-answer	exact_match	0.6833	±	0.0244
- mmlu_flan_cot_fewshot_business_ethics	0	get-answer	exact_match	0.9091	±	0.0909
- mmlu_flan_cot_fewshot_clinical_knowledge	0	get-answer	exact_match	0.5862	±	0.0931
- mmlu_flan_cot_fewshot_college_medicine	0	get-answer	exact_match	0.6364	±	0.1050
- mmlu_flan_cot_fewshot_global_facts	0	get-answer	exact_match	0.6000	±	0.1633
- mmlu_flan_cot_fewshot_human_aging	0	get-answer	exact_match	0.6087	±	0.1041
- mmlu_flan_cot_fewshot_management	0	get-answer	exact_match	0.9091	±	0.0909
- mmlu_flan_cot_fewshot_marketing	0	get-answer	exact_match	0.8000	±	0.0816
- mmlu_flan_cot_fewshot_medical_genetics	0	get-answer	exact_match	1.0000	±	0.0000
- mmlu_flan_cot_fewshot_miscellaneous	0	get-answer	exact_match	0.8023	±	0.0432
- mmlu_flan_cot_fewshot_nutrition	0	get-answer	exact_match	0.6667	±	0.0833
- mmlu_flan_cot_fewshot_professional_accounting	0	get-answer	exact_match	0.4839	±	0.0912
- mmlu_flan_cot_fewshot_professional_medicine	0	get-answer	exact_match	0.5806	±	0.0901
- mmlu_flan_cot_fewshot_virology	0	get-answer	exact_match	0.3889	±	0.1182
- mmlu_flan_cot_fewshot_social_sciences	N/A	get-answer	exact_match	0.7003	±	0.0239
- mmlu_flan_cot_fewshot_econometrics	0	get-answer	exact_match	0.4167	±	0.1486
- mmlu_flan_cot_fewshot_high_school_geography	0	get-answer	exact_match	0.9091	±	0.0627
- mmlu_flan_cot_fewshot_high_school_government_and_politics	0	get-answer	exact_match	0.8095	±	0.0878
- mmlu_flan_cot_fewshot_high_school_macroeconomics	0	get-answer	exact_match	0.6512	±	0.0735
- mmlu_flan_cot_fewshot_high_school_microeconomics	0	get-answer	exact_match	0.5769	±	0.0988
- mmlu_flan_cot_fewshot_high_school_psychology	0	get-answer	exact_match	0.9000	±	0.0391
- mmlu_flan_cot_fewshot_human_sexuality	0	get-answer	exact_match	0.6667	±	0.1421
- mmlu_flan_cot_fewshot_professional_psychology	0	get-answer	exact_match	0.6522	±	0.0578
- mmlu_flan_cot_fewshot_public_relations	0	get-answer	exact_match	0.5833	±	0.1486
- mmlu_flan_cot_fewshot_security_studies	0	get-answer	exact_match	0.4074	±	0.0964
- mmlu_flan_cot_fewshot_sociology	0	get-answer	exact_match	0.8182	±	0.0842
- mmlu_flan_cot_fewshot_us_foreign_policy	0	get-answer	exact_match	0.7273	±	0.1408
- mmlu_flan_cot_fewshot_stem	N/A	get-answer	exact_match	0.4866	±	0.0262
- mmlu_flan_cot_fewshot_abstract_algebra	0	get-answer	exact_match	0.0909	±	0.0909
- mmlu_flan_cot_fewshot_anatomy	0	get-answer	exact_match	0.4286	±	0.1373
- mmlu_flan_cot_fewshot_astronomy	0	get-answer	exact_match	0.5625	±	0.1281
- mmlu_flan_cot_fewshot_college_biology	0	get-answer	exact_match	0.5000	±	0.1291
- mmlu_flan_cot_fewshot_college_chemistry	0	get-answer	exact_match	0.5000	±	0.1890
- mmlu_flan_cot_fewshot_college_computer_science	0	get-answer	exact_match	0.2727	±	0.1408
- mmlu_flan_cot_fewshot_college_mathematics	0	get-answer	exact_match	0.3636	±	0.1521
- mmlu_flan_cot_fewshot_college_physics	0	get-answer	exact_match	0.3636	±	0.1521
- mmlu_flan_cot_fewshot_computer_security	0	get-answer	exact_match	0.7273	±	0.1408
- mmlu_flan_cot_fewshot_conceptual_physics	0	get-answer	exact_match	0.6538	±	0.0951
- mmlu_flan_cot_fewshot_electrical_engineering	0	get-answer	exact_match	0.7500	±	0.1118
- mmlu_flan_cot_fewshot_elementary_mathematics	0	get-answer	exact_match	0.7317	±	0.0701
- mmlu_flan_cot_fewshot_high_school_biology	0	get-answer	exact_match	0.5938	±	0.0882
- mmlu_flan_cot_fewshot_high_school_chemistry	0	get-answer	exact_match	0.3636	±	0.1050
- mmlu_flan_cot_fewshot_high_school_computer_science	0	get-answer	exact_match	0.5556	±	0.1757
- mmlu_flan_cot_fewshot_high_school_mathematics	0	get-answer	exact_match	0.3103	±	0.0874
- mmlu_flan_cot_fewshot_high_school_physics	0	get-answer	exact_match	0.2353	±	0.1060
- mmlu_flan_cot_fewshot_high_school_statistics	0	get-answer	exact_match	0.3043	±	0.0981
- mmlu_flan_cot_fewshot_machine_learning	0	get-answer	exact_match	0.4545	±	0.1575

Groups	Version	Filter	Metric	Value		Stderr
mmlu_flan_cot_fewshot	N/A	get-answer	exact_match	0.5833	±	0.0118
- mmlu_flan_cot_fewshot_humanities	N/A	get-answer	exact_match	0.5039	±	0.0205
- mmlu_flan_cot_fewshot_other	N/A	get-answer	exact_match	0.6833	±	0.0244
- mmlu_flan_cot_fewshot_social_sciences	N/A	get-answer	exact_match	0.7003	±	0.0239
- mmlu_flan_cot_fewshot_stem	N/A	get-answer	exact_match	0.4866	±	0.0262

How to Use

from transformers import AutoModelForCausalLM, AutoTokenizer

model_name = "Locutusque/Hyperion-3.0-Mistral-7B-DPO"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name)

# For a text generation task
input_text = "<|im_start|>user\nExplain the implications of quantum entanglement in layman's terms.<|im_end|>\n<|im_start|>assistant\n"
input_ids = tokenizer.encode(input_text, return_tensors="pt")

# Generate a response
outputs = model.generate(input_ids, max_length=200, do_sample=True, top_p=0.7, top_k=6) # These are the recommended sample settings.
print(tokenizer.decode(outputs[0], skip_special_tokens=True))

Known Limitations

While the training data has been carefully curated and optimized, there may still be some inconsistencies or biases present due to the inherent complexity and diversity of the source dataset. Users should be aware of potential limitations and carefully evaluate the model's outputs for their specific use case.

Additionally, this model is highly compliant and will attempt to respond to most requests. For enterprise-level deployment, it is strongly recommended to further fine-tune the model using DPO to align its behavior with specific requirements and constraints.

Licensing Information

This model is released under the Apache-2.0 license.

Downloads last month: 64

Safetensors

Model size

7B params

Tensor type

BF16

Model tree for Locutusque/Hyperion-3.0-Mistral-7B-DPO

Merges

2 models

Quantizations

4 models

Locutusque
/

Hyperion-3.0-Mistral-7B-DPO