Instructions to use AIPlans/tinyllama-1.1b-dpo-pku-saferlhf_2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use AIPlans/tinyllama-1.1b-dpo-pku-saferlhf_2 with Transformers:

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

pipe = pipeline("text-generation", model="AIPlans/tinyllama-1.1b-dpo-pku-saferlhf_2")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("AIPlans/tinyllama-1.1b-dpo-pku-saferlhf_2")
model = AutoModelForCausalLM.from_pretrained("AIPlans/tinyllama-1.1b-dpo-pku-saferlhf_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 AIPlans/tinyllama-1.1b-dpo-pku-saferlhf_2 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "AIPlans/tinyllama-1.1b-dpo-pku-saferlhf_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": "AIPlans/tinyllama-1.1b-dpo-pku-saferlhf_2",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/AIPlans/tinyllama-1.1b-dpo-pku-saferlhf_2

SGLang

How to use AIPlans/tinyllama-1.1b-dpo-pku-saferlhf_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 "AIPlans/tinyllama-1.1b-dpo-pku-saferlhf_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": "AIPlans/tinyllama-1.1b-dpo-pku-saferlhf_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 "AIPlans/tinyllama-1.1b-dpo-pku-saferlhf_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": "AIPlans/tinyllama-1.1b-dpo-pku-saferlhf_2",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use AIPlans/tinyllama-1.1b-dpo-pku-saferlhf_2 with Docker Model Runner:
```
docker model run hf.co/AIPlans/tinyllama-1.1b-dpo-pku-saferlhf_2
```

tinyllama-1.1b-dpo-pku-saferlhf_2

This model is a fine-tuned version of TinyLlama/TinyLlama-1.1B-Chat-v1.0 on an unknown dataset. It achieves the following results on the evaluation set:

Loss: 0.4486
Rewards/chosen: -0.1778
Rewards/rejected: -1.2187
Rewards/accuracies: 0.8000
Rewards/margins: 1.0409
Logps/rejected: -217.7180
Logps/chosen: -234.9111
Logits/rejected: -2.4987
Logits/chosen: -2.6097

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-06
train_batch_size: 4
eval_batch_size: 4
seed: 42
gradient_accumulation_steps: 4
total_train_batch_size: 16
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: cosine
lr_scheduler_warmup_ratio: 0.1
num_epochs: 1.0

Training results

Training Loss	Epoch	Step	Validation Loss	Rewards/chosen	Rewards/rejected	Rewards/accuracies	Rewards/margins	Logps/rejected	Logps/chosen	Logits/rejected	Logits/chosen
0.6075	0.0960	200	0.6052	0.4208	0.1906	0.7170	0.2302	-203.6246	-228.9246	-2.5653	-2.6887
0.5198	0.1920	400	0.5145	0.3083	-0.2832	0.7850	0.5915	-208.3632	-230.0500	-2.5310	-2.6479
0.4703	0.2880	600	0.4838	0.0019	-0.7989	0.7940	0.8008	-213.5201	-233.1144	-2.5215	-2.6339
0.5223	0.3840	800	0.4631	0.1561	-0.7414	0.7880	0.8975	-212.9448	-231.5717	-2.4968	-2.6118
0.4335	0.4800	1000	0.4555	-0.1607	-1.1551	0.7950	0.9944	-217.0816	-234.7399	-2.4976	-2.6097
0.5214	0.5760	1200	0.4511	-0.1319	-1.1471	0.7990	1.0152	-217.0018	-234.4519	-2.5001	-2.6120
0.4784	0.6719	1400	0.4486	-0.1778	-1.2187	0.8000	1.0409	-217.7180	-234.9111	-2.4987	-2.6097
0.4223	0.7679	1600	0.4487	-0.1274	-1.1676	0.7990	1.0402	-217.2067	-234.4066	-2.4987	-2.6100
0.5114	0.8639	1800	0.4483	-0.1119	-1.1535	0.7990	1.0416	-217.0661	-234.2521	-2.4988	-2.6100
0.4763	0.9599	2000	0.4485	-0.1144	-1.1552	0.7950	1.0409	-217.0830	-234.2766	-2.4989	-2.6102

Framework versions

Transformers 4.44.2
Pytorch 2.10.0+cu130
Datasets 2.21.0
Tokenizers 0.19.1

Downloads last month: 21

Safetensors

Model size

1B params

Tensor type

BF16

Model tree for AIPlans/tinyllama-1.1b-dpo-pku-saferlhf_2

Base model

TinyLlama/TinyLlama-1.1B-Chat-v1.0

Finetuned

(573)

this model

Collection including AIPlans/tinyllama-1.1b-dpo-pku-saferlhf_2

TinyLlama RLHF Models

Collection

6 items • Updated May 16