Instructions to use tsavage68/chat_1000STEPS_1e6_03beta_DPO with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use tsavage68/chat_1000STEPS_1e6_03beta_DPO with Transformers:

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

pipe = pipeline("text-generation", model="tsavage68/chat_1000STEPS_1e6_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_1e6_03beta_DPO")
model = AutoModelForCausalLM.from_pretrained("tsavage68/chat_1000STEPS_1e6_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_1e6_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_1e6_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_1e6_03beta_DPO",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/tsavage68/chat_1000STEPS_1e6_03beta_DPO

SGLang

How to use tsavage68/chat_1000STEPS_1e6_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_1e6_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_1e6_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_1e6_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_1e6_03beta_DPO",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use tsavage68/chat_1000STEPS_1e6_03beta_DPO with Docker Model Runner:
```
docker model run hf.co/tsavage68/chat_1000STEPS_1e6_03beta_DPO
```

chat_1000STEPS_1e6_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: 0.6804
Rewards/chosen: -0.5183
Rewards/rejected: -0.7327
Rewards/accuracies: 0.5363
Rewards/margins: 0.2144
Logps/rejected: -21.2336
Logps/chosen: -18.4723
Logits/rejected: -0.6767
Logits/chosen: -0.6766

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-06
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
0.6885	0.2	100	0.6933	-0.2467	-0.2660	0.4637	0.0193	-19.6779	-17.5670	-0.6067	-0.6066
0.683	0.39	200	0.6859	0.0215	-0.0664	0.4923	0.0879	-19.0127	-16.6730	-0.6150	-0.6148
0.6033	0.59	300	0.6999	-0.1969	-0.2977	0.4791	0.1009	-19.7837	-17.4008	-0.6311	-0.6309
0.6812	0.78	400	0.6942	-0.0785	-0.2126	0.4813	0.1340	-19.4998	-17.0064	-0.6041	-0.6039
0.6633	0.98	500	0.6789	-0.1266	-0.2799	0.5077	0.1533	-19.7242	-17.1665	-0.5557	-0.5555
0.2615	1.17	600	0.6788	-0.4082	-0.6084	0.5253	0.2002	-20.8192	-18.1052	-0.6281	-0.6279
0.3175	1.37	700	0.6809	-0.4980	-0.7087	0.5297	0.2107	-21.1536	-18.4046	-0.6655	-0.6653
0.2805	1.56	800	0.6794	-0.5125	-0.7293	0.5341	0.2169	-21.2224	-18.4529	-0.6754	-0.6753
0.3255	1.76	900	0.6807	-0.5148	-0.7297	0.5385	0.2149	-21.2235	-18.4605	-0.6768	-0.6766
0.2966	1.95	1000	0.6804	-0.5183	-0.7327	0.5363	0.2144	-21.2336	-18.4723	-0.6767	-0.6766

Framework versions

Transformers 4.37.2
Pytorch 2.0.0+cu117
Datasets 2.17.0
Tokenizers 0.15.2

Downloads last month: 5

Safetensors

Model size

7B params

Tensor type

F16

Model tree for tsavage68/chat_1000STEPS_1e6_03beta_DPO

Base model

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

Finetuned

(552)

this model