Instructions to use tsavage68/chat_1000STEPS_1e7_03beta_DPO with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use tsavage68/chat_1000STEPS_1e7_03beta_DPO with Transformers:

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

pipe = pipeline("text-generation", model="tsavage68/chat_1000STEPS_1e7_03beta_DPO")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("tsavage68/chat_1000STEPS_1e7_03beta_DPO")
model = AutoModelForCausalLM.from_pretrained("tsavage68/chat_1000STEPS_1e7_03beta_DPO", 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 tsavage68/chat_1000STEPS_1e7_03beta_DPO with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "tsavage68/chat_1000STEPS_1e7_03beta_DPO"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "tsavage68/chat_1000STEPS_1e7_03beta_DPO",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/tsavage68/chat_1000STEPS_1e7_03beta_DPO

SGLang

How to use tsavage68/chat_1000STEPS_1e7_03beta_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 "tsavage68/chat_1000STEPS_1e7_03beta_DPO" \
    --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": "tsavage68/chat_1000STEPS_1e7_03beta_DPO",
		"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 "tsavage68/chat_1000STEPS_1e7_03beta_DPO" \
        --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": "tsavage68/chat_1000STEPS_1e7_03beta_DPO",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use tsavage68/chat_1000STEPS_1e7_03beta_DPO with Docker Model Runner:
```
docker model run hf.co/tsavage68/chat_1000STEPS_1e7_03beta_DPO
```

chat_1000STEPS_1e7_03beta_DPO

This model is a fine-tuned version of meta-llama/Llama-2-7b-chat-hf on an unknown dataset. It achieves the following results on the evaluation set:

Loss: 0.6902
Rewards/chosen: -0.0000
Rewards/rejected: -0.0069
Rewards/accuracies: 0.4681
Rewards/margins: 0.0069
Logps/rejected: -18.8144
Logps/chosen: -16.7447
Logits/rejected: -0.5973
Logits/chosen: -0.5972

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: 1e-07
train_batch_size: 4
eval_batch_size: 1
seed: 42
gradient_accumulation_steps: 2
total_train_batch_size: 8
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: cosine
lr_scheduler_warmup_steps: 100
training_steps: 1000

Training results

Training Loss	Epoch	Step	Validation Loss	Rewards/chosen	Rewards/rejected	Rewards/accuracies	Rewards/margins	Logps/rejected	Logps/chosen	Logits/rejected	Logits/chosen
0.6935	0.2	100	0.6925	-0.0035	-0.0055	0.4286	0.0019	-18.8094	-16.7564	-0.5969	-0.5967
0.6934	0.39	200	0.6911	0.0022	-0.0027	0.4615	0.0049	-18.8003	-16.7374	-0.5979	-0.5977
0.6882	0.59	300	0.6929	-0.0047	-0.0060	0.4330	0.0013	-18.8112	-16.7601	-0.5973	-0.5972
0.6896	0.78	400	0.6907	-0.0013	-0.0070	0.4615	0.0057	-18.8147	-16.7490	-0.5982	-0.5981
0.6877	0.98	500	0.6904	0.0012	-0.0051	0.4923	0.0063	-18.8082	-16.7405	-0.5972	-0.5971
0.6829	1.17	600	0.6903	-0.0020	-0.0085	0.4703	0.0066	-18.8198	-16.7511	-0.5976	-0.5975
0.6832	1.37	700	0.6904	-0.0032	-0.0097	0.4593	0.0064	-18.8236	-16.7554	-0.5971	-0.5970
0.6802	1.56	800	0.6889	-0.0010	-0.0105	0.4923	0.0096	-18.8263	-16.7478	-0.5979	-0.5978
0.6826	1.76	900	0.6897	-0.0009	-0.0088	0.4769	0.0079	-18.8206	-16.7475	-0.5972	-0.5971
0.6761	1.95	1000	0.6902	-0.0000	-0.0069	0.4681	0.0069	-18.8144	-16.7447	-0.5973	-0.5972

Framework versions

Transformers 4.37.2
Pytorch 2.0.0+cu117
Datasets 2.17.0
Tokenizers 0.15.2

Downloads last month: 6

Safetensors

Model size

7B params

Tensor type

F16

Model tree for tsavage68/chat_1000STEPS_1e7_03beta_DPO

Base model

meta-llama/Llama-2-7b-chat-hf

Finetuned

(552)

this model