Instructions to use tsavage68/UTI_L3_1000steps_1e5rate_01beta_CSFTDPO with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use tsavage68/UTI_L3_1000steps_1e5rate_01beta_CSFTDPO with Transformers:

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

pipe = pipeline("text-generation", model="tsavage68/UTI_L3_1000steps_1e5rate_01beta_CSFTDPO")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("tsavage68/UTI_L3_1000steps_1e5rate_01beta_CSFTDPO")
model = AutoModelForCausalLM.from_pretrained("tsavage68/UTI_L3_1000steps_1e5rate_01beta_CSFTDPO", 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/UTI_L3_1000steps_1e5rate_01beta_CSFTDPO with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "tsavage68/UTI_L3_1000steps_1e5rate_01beta_CSFTDPO"
# 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/UTI_L3_1000steps_1e5rate_01beta_CSFTDPO",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/tsavage68/UTI_L3_1000steps_1e5rate_01beta_CSFTDPO

SGLang

How to use tsavage68/UTI_L3_1000steps_1e5rate_01beta_CSFTDPO 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/UTI_L3_1000steps_1e5rate_01beta_CSFTDPO" \
    --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/UTI_L3_1000steps_1e5rate_01beta_CSFTDPO",
		"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/UTI_L3_1000steps_1e5rate_01beta_CSFTDPO" \
        --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/UTI_L3_1000steps_1e5rate_01beta_CSFTDPO",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use tsavage68/UTI_L3_1000steps_1e5rate_01beta_CSFTDPO with Docker Model Runner:
```
docker model run hf.co/tsavage68/UTI_L3_1000steps_1e5rate_01beta_CSFTDPO
```

UTI_L3_1000steps_1e5rate_01beta_CSFTDPO

This model is a fine-tuned version of tsavage68/UTI_L3_1000steps_1e5rate_SFT on an unknown dataset. It achieves the following results on the evaluation set:

Loss: 0.0069
Rewards/chosen: -2.8206
Rewards/rejected: -21.2510
Rewards/accuracies: 0.9900
Rewards/margins: 18.4304
Logps/rejected: -275.7048
Logps/chosen: -60.6847
Logits/rejected: -2.0149
Logits/chosen: -1.9810

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-05
train_batch_size: 2
eval_batch_size: 1
seed: 42
gradient_accumulation_steps: 2
total_train_batch_size: 4
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
0.0002	0.6667	50	0.0072	0.7766	-13.0912	0.9900	13.8678	-194.1069	-24.7133	-1.5529	-1.4952
0.0173	1.3333	100	0.0086	-1.2523	-14.7999	0.9900	13.5476	-211.1941	-45.0020	-1.7169	-1.6585
0.0371	2.0	150	0.0069	-2.9050	-20.9463	0.9900	18.0414	-272.6581	-61.5287	-2.0084	-1.9771
0.0	2.6667	200	0.0069	-2.8291	-21.1059	0.9900	18.2768	-274.2534	-60.7697	-2.0121	-1.9789
0.0173	3.3333	250	0.0069	-2.8268	-21.1156	0.9900	18.2889	-274.3510	-60.7466	-2.0124	-1.9791
0.0347	4.0	300	0.0069	-2.8254	-21.1309	0.9900	18.3055	-274.5038	-60.7333	-2.0126	-1.9792
0.0173	4.6667	350	0.0069	-2.8156	-21.1516	0.9900	18.3360	-274.7103	-60.6348	-2.0131	-1.9796
0.0173	5.3333	400	0.0069	-2.8155	-21.1665	0.9900	18.3511	-274.8600	-60.6336	-2.0133	-1.9797
0.0173	6.0	450	0.0069	-2.8146	-21.1758	0.9900	18.3612	-274.9522	-60.6250	-2.0136	-1.9799
0.0347	6.6667	500	0.0069	-2.8128	-21.1899	0.9900	18.3771	-275.0935	-60.6071	-2.0140	-1.9802
0.0	7.3333	550	0.0069	-2.8143	-21.2087	0.9900	18.3944	-275.2815	-60.6221	-2.0143	-1.9804
0.0347	8.0	600	0.0069	-2.8161	-21.2215	0.9900	18.4054	-275.4096	-60.6400	-2.0144	-1.9805
0.0	8.6667	650	0.0069	-2.8197	-21.2301	0.9900	18.4104	-275.4954	-60.6758	-2.0147	-1.9807
0.0173	9.3333	700	0.0069	-2.8217	-21.2410	0.9900	18.4193	-275.6051	-60.6962	-2.0148	-1.9809
0.0	10.0	750	0.0069	-2.8204	-21.2414	0.9900	18.4210	-275.6092	-60.6834	-2.0148	-1.9809
0.0173	10.6667	800	0.0069	-2.8221	-21.2513	0.9900	18.4292	-275.7073	-60.7001	-2.0148	-1.9808
0.0	11.3333	850	0.0069	-2.8219	-21.2497	0.9900	18.4278	-275.6921	-60.6985	-2.0148	-1.9808
0.0	12.0	900	0.0069	-2.8223	-21.2528	0.9900	18.4305	-275.7229	-60.7022	-2.0151	-1.9811
0.0173	12.6667	950	0.0069	-2.8218	-21.2512	0.9900	18.4295	-275.7072	-60.6970	-2.0149	-1.9810
0.0	13.3333	1000	0.0069	-2.8206	-21.2510	0.9900	18.4304	-275.7048	-60.6847	-2.0149	-1.9810

Framework versions

Transformers 4.41.2
Pytorch 2.0.0+cu117
Datasets 2.19.2
Tokenizers 0.19.1

Downloads last month: 6

Safetensors

Model size

8B params

Tensor type

F16

Model tree for tsavage68/UTI_L3_1000steps_1e5rate_01beta_CSFTDPO

Base model

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

Finetuned

tsavage68/UTI_L3_1000steps_1e5rate_SFT

Finetuned

(39)

this model