Instructions to use Raghav-Singhal/dpo-tulu3-lr5e-7-tulu3sft-100B-no-bad-data-off-policy-if with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Raghav-Singhal/dpo-tulu3-lr5e-7-tulu3sft-100B-no-bad-data-off-policy-if with Transformers:

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

pipe = pipeline("text-generation", model="Raghav-Singhal/dpo-tulu3-lr5e-7-tulu3sft-100B-no-bad-data-off-policy-if")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Raghav-Singhal/dpo-tulu3-lr5e-7-tulu3sft-100B-no-bad-data-off-policy-if")
model = AutoModelForCausalLM.from_pretrained("Raghav-Singhal/dpo-tulu3-lr5e-7-tulu3sft-100B-no-bad-data-off-policy-if", 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 Raghav-Singhal/dpo-tulu3-lr5e-7-tulu3sft-100B-no-bad-data-off-policy-if with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Raghav-Singhal/dpo-tulu3-lr5e-7-tulu3sft-100B-no-bad-data-off-policy-if"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Raghav-Singhal/dpo-tulu3-lr5e-7-tulu3sft-100B-no-bad-data-off-policy-if",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Raghav-Singhal/dpo-tulu3-lr5e-7-tulu3sft-100B-no-bad-data-off-policy-if

SGLang

How to use Raghav-Singhal/dpo-tulu3-lr5e-7-tulu3sft-100B-no-bad-data-off-policy-if 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 "Raghav-Singhal/dpo-tulu3-lr5e-7-tulu3sft-100B-no-bad-data-off-policy-if" \
    --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": "Raghav-Singhal/dpo-tulu3-lr5e-7-tulu3sft-100B-no-bad-data-off-policy-if",
		"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 "Raghav-Singhal/dpo-tulu3-lr5e-7-tulu3sft-100B-no-bad-data-off-policy-if" \
        --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": "Raghav-Singhal/dpo-tulu3-lr5e-7-tulu3sft-100B-no-bad-data-off-policy-if",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Raghav-Singhal/dpo-tulu3-lr5e-7-tulu3sft-100B-no-bad-data-off-policy-if with Docker Model Runner:
```
docker model run hf.co/Raghav-Singhal/dpo-tulu3-lr5e-7-tulu3sft-100B-no-bad-data-off-policy-if
```

Model Card for dpo-tulu3-lr5e-7-tulu3sft-100B-no-bad-data-off-policy-if

This repository contains a DPO fine-tune of the local SFT checkpoint tulu3sft-normal-smollm-1p7b-100B-20n-2048sl-960gbsz-no-bad-data.

The final model weights are stored at the repository root. Intermediate training checkpoints are also included under checkpoint-500, checkpoint-1000, and checkpoint-1270.

Quick start

from transformers import pipeline

question = "If you had a time machine, but could only go to the past or the future once and never return, which would you choose and why?"
generator = pipeline(
    "text-generation",
    model="Raghav-Singhal/dpo-tulu3-lr5e-7-tulu3sft-100B-no-bad-data-off-policy-if",
    device="cuda",
)
output = generator(
    [{"role": "user", "content": question}],
    max_new_tokens=128,
    return_full_text=False,
)[0]
print(output["generated_text"])

Training procedure

This model was trained with DPO, a method introduced in Direct Preference Optimization: Your Language Model is Secretly a Reward Model.

Framework versions

TRL: 1.0.0
Transformers: 4.57.6
Pytorch: 2.10.0a0+b4e4ee81d3.nv25.12
Datasets: 4.8.4
Tokenizers: 0.22.1

Downloads last month: 10

Safetensors

Model size

2B params

Tensor type

BF16

Paper for Raghav-Singhal/dpo-tulu3-lr5e-7-tulu3sft-100B-no-bad-data-off-policy-if

Direct Preference Optimization: Your Language Model is Secretly a Reward Model

Paper • 2305.18290 • Published May 29, 2023 • 71