Instructions to use tsavage68/chat_1000_STEPS_05beta_1e6rate_CDPOSFT with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use tsavage68/chat_1000_STEPS_05beta_1e6rate_CDPOSFT with Transformers:

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

pipe = pipeline("text-generation", model="tsavage68/chat_1000_STEPS_05beta_1e6rate_CDPOSFT")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("tsavage68/chat_1000_STEPS_05beta_1e6rate_CDPOSFT")
model = AutoModelForCausalLM.from_pretrained("tsavage68/chat_1000_STEPS_05beta_1e6rate_CDPOSFT", 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_1000_STEPS_05beta_1e6rate_CDPOSFT with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "tsavage68/chat_1000_STEPS_05beta_1e6rate_CDPOSFT"
# 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_1000_STEPS_05beta_1e6rate_CDPOSFT",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/tsavage68/chat_1000_STEPS_05beta_1e6rate_CDPOSFT

SGLang

How to use tsavage68/chat_1000_STEPS_05beta_1e6rate_CDPOSFT 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_1000_STEPS_05beta_1e6rate_CDPOSFT" \
    --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_1000_STEPS_05beta_1e6rate_CDPOSFT",
		"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_1000_STEPS_05beta_1e6rate_CDPOSFT" \
        --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_1000_STEPS_05beta_1e6rate_CDPOSFT",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use tsavage68/chat_1000_STEPS_05beta_1e6rate_CDPOSFT with Docker Model Runner:
```
docker model run hf.co/tsavage68/chat_1000_STEPS_05beta_1e6rate_CDPOSFT
```

chat_1000_STEPS_05beta_1e6rate_CDPOSFT

This model is a fine-tuned version of tsavage68/chat_600STEPS_1e8rate_SFT on an unknown dataset. It achieves the following results on the evaluation set:

Loss: 0.7027
Rewards/chosen: -0.5988
Rewards/rejected: -0.8712
Rewards/accuracies: 0.5099
Rewards/margins: 0.2724
Logps/rejected: -20.5443
Logps/chosen: -17.9521
Logits/rejected: -0.6530
Logits/chosen: -0.6528

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
0.6871	0.0977	50	0.6897	0.0517	0.0417	0.4352	0.0100	-18.7185	-16.6512	-0.6010	-0.6009
0.6399	0.1953	100	0.6728	-0.1560	-0.2548	0.5099	0.0989	-19.3116	-17.0666	-0.6090	-0.6089
0.7472	0.2930	150	0.6915	-0.1391	-0.2449	0.4725	0.1058	-19.2918	-17.0328	-0.6010	-0.6008
0.7204	0.3906	200	0.7024	-0.2454	-0.3692	0.4923	0.1239	-19.5404	-17.2454	-0.5899	-0.5897
0.7965	0.4883	250	0.7140	-0.1153	-0.2604	0.4835	0.1451	-19.3228	-16.9852	-0.5793	-0.5791
0.7139	0.5859	300	0.6956	-0.2819	-0.4872	0.5121	0.2052	-19.7763	-17.3185	-0.6031	-0.6029
0.6559	0.6836	350	0.7078	-0.3820	-0.5600	0.5099	0.1780	-19.9219	-17.5186	-0.5854	-0.5853
0.6696	0.7812	400	0.6942	-0.3456	-0.5617	0.5143	0.2160	-19.9254	-17.4459	-0.6059	-0.6057
0.7053	0.8789	450	0.7006	-0.1577	-0.3434	0.5033	0.1857	-19.4887	-17.0699	-0.5689	-0.5687
0.7143	0.9766	500	0.7009	-0.1658	-0.3724	0.5055	0.2066	-19.5467	-17.0862	-0.5925	-0.5923
0.2701	1.0742	550	0.6978	-0.2341	-0.4595	0.5121	0.2254	-19.7210	-17.2229	-0.6038	-0.6036
0.2867	1.1719	600	0.6987	-0.3718	-0.6187	0.5077	0.2469	-20.0393	-17.4982	-0.6186	-0.6184
0.3128	1.2695	650	0.7018	-0.4995	-0.7601	0.5055	0.2605	-20.3221	-17.7537	-0.6358	-0.6356
0.1953	1.3672	700	0.7004	-0.5365	-0.8019	0.5165	0.2653	-20.4057	-17.8277	-0.6439	-0.6437
0.1831	1.4648	750	0.7014	-0.5837	-0.8544	0.5143	0.2707	-20.5108	-17.9220	-0.6497	-0.6495
0.3253	1.5625	800	0.7019	-0.5862	-0.8575	0.5077	0.2713	-20.5169	-17.9270	-0.6514	-0.6512
0.2709	1.6602	850	0.7043	-0.6000	-0.8668	0.5077	0.2668	-20.5356	-17.9547	-0.6522	-0.6520
0.3225	1.7578	900	0.7035	-0.6017	-0.8716	0.5033	0.2699	-20.5452	-17.9580	-0.6530	-0.6528
0.2806	1.8555	950	0.7020	-0.5987	-0.8741	0.5121	0.2754	-20.5502	-17.9521	-0.6531	-0.6529
0.2262	1.9531	1000	0.7027	-0.5988	-0.8712	0.5099	0.2724	-20.5443	-17.9521	-0.6530	-0.6528

Framework versions

Transformers 4.40.1
Pytorch 2.0.0+cu117
Datasets 2.19.0
Tokenizers 0.19.1

Downloads last month: 6

Safetensors

Model size

7B params

Tensor type

F16

Model tree for tsavage68/chat_1000_STEPS_05beta_1e6rate_CDPOSFT

Base model

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

Finetuned

tsavage68/chat_600STEPS_1e8rate_SFT

Finetuned

(23)

this model