Instructions to use jackf857/llama-3-8b-base-new-dpo-harmless-s_star0.6-q_t0.4 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use jackf857/llama-3-8b-base-new-dpo-harmless-s_star0.6-q_t0.4 with Transformers:

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

pipe = pipeline("text-generation", model="jackf857/llama-3-8b-base-new-dpo-harmless-s_star0.6-q_t0.4")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("jackf857/llama-3-8b-base-new-dpo-harmless-s_star0.6-q_t0.4")
model = AutoModelForCausalLM.from_pretrained("jackf857/llama-3-8b-base-new-dpo-harmless-s_star0.6-q_t0.4", 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 jackf857/llama-3-8b-base-new-dpo-harmless-s_star0.6-q_t0.4 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "jackf857/llama-3-8b-base-new-dpo-harmless-s_star0.6-q_t0.4"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "jackf857/llama-3-8b-base-new-dpo-harmless-s_star0.6-q_t0.4",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/jackf857/llama-3-8b-base-new-dpo-harmless-s_star0.6-q_t0.4

SGLang

How to use jackf857/llama-3-8b-base-new-dpo-harmless-s_star0.6-q_t0.4 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 "jackf857/llama-3-8b-base-new-dpo-harmless-s_star0.6-q_t0.4" \
    --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": "jackf857/llama-3-8b-base-new-dpo-harmless-s_star0.6-q_t0.4",
		"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 "jackf857/llama-3-8b-base-new-dpo-harmless-s_star0.6-q_t0.4" \
        --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": "jackf857/llama-3-8b-base-new-dpo-harmless-s_star0.6-q_t0.4",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use jackf857/llama-3-8b-base-new-dpo-harmless-s_star0.6-q_t0.4 with Docker Model Runner:
```
docker model run hf.co/jackf857/llama-3-8b-base-new-dpo-harmless-s_star0.6-q_t0.4
```

llama3-8b-base-new-method-s_star0.6

This model is a fine-tuned version of W-61/llama-3-8b-base-sft-hh-harmless-4xh200 on the Anthropic/hh-rlhf dataset. It achieves the following results on the evaluation set:

Loss: 0.5591
Fcm Dpo/beta: 0.0089
Margin Dpo/margin Mean: 55.2826
Margin Dpo/margin Std: 95.3567
Logps/chosen: -205.2188
Logps/rejected: -265.1909
Logps/ref Chosen: -74.8595
Logps/ref Rejected: -79.5490
Logits/chosen: 0.7206
Logits/rejected: 0.6751

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: 8
eval_batch_size: 8
seed: 42
distributed_type: multi-GPU
num_devices: 4
gradient_accumulation_steps: 2
total_train_batch_size: 64
total_eval_batch_size: 32
optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: cosine
lr_scheduler_warmup_ratio: 0.1
num_epochs: 1

Training results

Training Loss	Epoch	Step	Validation Loss	Fcm Dpo/beta	Margin Dpo/margin Mean	Margin Dpo/margin Std	Logps/chosen	Logps/rejected	Logps/ref Chosen	Logps/ref Rejected	Logits/chosen	Logits/rejected
1.0505	0.3023	200	0.5490	0.0445	12.8367	21.5327	-94.9354	-112.4616	-74.8595	-79.5490	0.4353	0.3850
1.0738	0.6047	400	0.5614	0.0103	44.9123	77.2791	-176.5743	-226.1761	-74.8595	-79.5490	0.6246	0.5775
1.1653	0.9070	600	0.5591	0.0089	55.2826	95.3567	-205.2188	-265.1909	-74.8595	-79.5490	0.7206	0.6751

Framework versions

Transformers 4.51.0
Pytorch 2.3.1+cu121
Datasets 2.21.0
Tokenizers 0.21.4

Downloads last month: 7

Safetensors

Model size

8B params

Tensor type

F32

Model tree for jackf857/llama-3-8b-base-new-dpo-harmless-s_star0.6-q_t0.4

Base model

meta-llama/Meta-Llama-3-8B

Finetuned

W-61/llama-3-8b-base-sft-hh-harmless-4xh200

Finetuned

(36)

this model

jackf857
/

llama-3-8b-base-new-dpo-harmless-s_star0.6-q_t0.4

llama3-8b-base-new-method-s_star0.6

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

Model tree for jackf857/llama-3-8b-base-new-dpo-harmless-s_star0.6-q_t0.4

Dataset used to train jackf857/llama-3-8b-base-new-dpo-harmless-s_star0.6-q_t0.4