Instructions to use NoManDeRY/DPO-Shift-Llama-3-8B-Ultrafeedback-decrease_linear-1.0to0.95 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use NoManDeRY/DPO-Shift-Llama-3-8B-Ultrafeedback-decrease_linear-1.0to0.95 with Transformers:

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

pipe = pipeline("text-generation", model="NoManDeRY/DPO-Shift-Llama-3-8B-Ultrafeedback-decrease_linear-1.0to0.95")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("NoManDeRY/DPO-Shift-Llama-3-8B-Ultrafeedback-decrease_linear-1.0to0.95")
model = AutoModelForCausalLM.from_pretrained("NoManDeRY/DPO-Shift-Llama-3-8B-Ultrafeedback-decrease_linear-1.0to0.95")
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

vLLM

How to use NoManDeRY/DPO-Shift-Llama-3-8B-Ultrafeedback-decrease_linear-1.0to0.95 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "NoManDeRY/DPO-Shift-Llama-3-8B-Ultrafeedback-decrease_linear-1.0to0.95"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "NoManDeRY/DPO-Shift-Llama-3-8B-Ultrafeedback-decrease_linear-1.0to0.95",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/NoManDeRY/DPO-Shift-Llama-3-8B-Ultrafeedback-decrease_linear-1.0to0.95

SGLang

How to use NoManDeRY/DPO-Shift-Llama-3-8B-Ultrafeedback-decrease_linear-1.0to0.95 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 "NoManDeRY/DPO-Shift-Llama-3-8B-Ultrafeedback-decrease_linear-1.0to0.95" \
    --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": "NoManDeRY/DPO-Shift-Llama-3-8B-Ultrafeedback-decrease_linear-1.0to0.95",
		"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 "NoManDeRY/DPO-Shift-Llama-3-8B-Ultrafeedback-decrease_linear-1.0to0.95" \
        --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": "NoManDeRY/DPO-Shift-Llama-3-8B-Ultrafeedback-decrease_linear-1.0to0.95",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use NoManDeRY/DPO-Shift-Llama-3-8B-Ultrafeedback-decrease_linear-1.0to0.95 with Docker Model Runner:
```
docker model run hf.co/NoManDeRY/DPO-Shift-Llama-3-8B-Ultrafeedback-decrease_linear-1.0to0.95
```

llama-3-8b-dpo-ultrafeedback-decrease_linear-1.0to0.95

This is a model released from the preprint: DPO-Shift: Shifting the Distribution of Direct Preference Optimization. Please refer to our repository for more details.

This model is a fine-tuned version of princeton-nlp/Llama-3-Base-8B-SFT on the HuggingFaceH4/ultrafeedback_binarized dataset. It achieves the following results on the evaluation set:

Loss: 0.5619
Rewards/chosen: -0.3784
Rewards/rejected: -0.8957
Dpo Lambda: 0.9528
Rewards/accuracies: 0.7310
Rewards/margins: 0.5173
Logps/rejected: -360.6006
Logps/chosen: -338.4835
Logits/rejected: -1.0030
Logits/chosen: -0.9672

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 5e-07
train_batch_size: 2
eval_batch_size: 2
seed: 42
distributed_type: multi-GPU
num_devices: 2
gradient_accumulation_steps: 32
total_train_batch_size: 128
total_eval_batch_size: 4
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: cosine
lr_scheduler_warmup_ratio: 0.1
num_epochs: 1

Training results

Training Loss	Epoch	Step	Validation Loss	Rewards/chosen	Rewards/rejected	Dpo Lambda	Rewards/accuracies	Rewards/margins	Logps/rejected	Logps/chosen	Logits/rejected	Logits/chosen
0.6826	0.1047	50	0.6803	0.0669	0.0399	0.9948	0.6690	0.0270	-267.0431	-293.9557	-0.9094	-0.8412
0.5951	0.2094	100	0.6223	-0.0861	-0.2745	0.9895	0.7130	0.1884	-298.4850	-309.2591	-0.9195	-0.8667
0.6296	0.3141	150	0.5972	-0.2312	-0.5289	0.9843	0.7100	0.2977	-323.9177	-323.7625	-0.9008	-0.8554
0.6219	0.4187	200	0.5784	-0.4096	-0.8051	0.9790	0.7310	0.3955	-351.5381	-341.6022	-0.9313	-0.8927
0.5738	0.5234	250	0.5685	-0.4338	-0.8864	0.9738	0.7260	0.4526	-359.6707	-344.0276	-0.9691	-0.9333
0.5598	0.6281	300	0.5695	-0.4246	-0.9086	0.9686	0.7220	0.4840	-361.8922	-343.1057	-1.0002	-0.9608
0.566	0.7328	350	0.5613	-0.3470	-0.8404	0.9633	0.7260	0.4934	-355.0737	-335.3493	-0.9958	-0.9592
0.5423	0.8375	400	0.5613	-0.3837	-0.8996	0.9581	0.7290	0.5159	-360.9908	-339.0213	-1.0033	-0.9665
0.5357	0.9422	450	0.5619	-0.3784	-0.8957	0.9528	0.7310	0.5173	-360.6006	-338.4835	-1.0030	-0.9672

Framework versions

Transformers 4.44.2
Pytorch 2.4.0+cu121
Datasets 2.21.0
Tokenizers 0.19.1

Downloads last month: 5

Safetensors

Model size

8B params

Tensor type

BF16

Model tree for NoManDeRY/DPO-Shift-Llama-3-8B-Ultrafeedback-decrease_linear-1.0to0.95

Base model

princeton-nlp/Llama-3-Base-8B-SFT

Finetuned

(49)

this model

Dataset used to train NoManDeRY/DPO-Shift-Llama-3-8B-Ultrafeedback-decrease_linear-1.0to0.95

Paper for NoManDeRY/DPO-Shift-Llama-3-8B-Ultrafeedback-decrease_linear-1.0to0.95

DPO-Shift: Shifting the Distribution of Direct Preference Optimization

Paper • 2502.07599 • Published Feb 11, 2025 • 15