Instructions to use tsavage68/UTI_L3_1000steps_1e5rate_03beta_CSFTDPO with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use tsavage68/UTI_L3_1000steps_1e5rate_03beta_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_03beta_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_03beta_CSFTDPO")
model = AutoModelForCausalLM.from_pretrained("tsavage68/UTI_L3_1000steps_1e5rate_03beta_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_03beta_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_03beta_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_03beta_CSFTDPO",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/tsavage68/UTI_L3_1000steps_1e5rate_03beta_CSFTDPO

SGLang

How to use tsavage68/UTI_L3_1000steps_1e5rate_03beta_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_03beta_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_03beta_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_03beta_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_03beta_CSFTDPO",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use tsavage68/UTI_L3_1000steps_1e5rate_03beta_CSFTDPO with Docker Model Runner:
```
docker model run hf.co/tsavage68/UTI_L3_1000steps_1e5rate_03beta_CSFTDPO
```

UTI_L3_1000steps_1e5rate_03beta_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: 2.2757
Rewards/rejected: -15.6836
Rewards/accuracies: 0.9900
Rewards/margins: 17.9593
Logps/rejected: -115.4733
Logps/chosen: -24.8934
Logits/rejected: -1.4719
Logits/chosen: -1.4307

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
0.0	0.6667	50	0.0072	1.8402	-13.3590	0.9900	15.1992	-107.7247	-26.3451	-1.4305	-1.3941
0.0173	1.3333	100	0.0071	1.8455	-14.4051	0.9900	16.2506	-111.2116	-26.3273	-1.4331	-1.3960
0.0347	2.0	150	0.0069	2.3483	-14.9050	0.9900	17.2533	-112.8780	-24.6513	-1.4557	-1.4154
0.0	2.6667	200	0.0069	2.3179	-15.0160	0.9900	17.3339	-113.2480	-24.7526	-1.4584	-1.4180
0.0173	3.3333	250	0.0069	2.3120	-15.0851	0.9900	17.3971	-113.4783	-24.7723	-1.4616	-1.4212
0.0347	4.0	300	0.0069	2.3109	-15.1144	0.9900	17.4254	-113.5761	-24.7759	-1.4624	-1.4219
0.0173	4.6667	350	0.0069	2.3085	-15.1859	0.9900	17.4944	-113.8144	-24.7841	-1.4649	-1.4242
0.0173	5.3333	400	0.0069	2.2984	-15.2571	0.9900	17.5555	-114.0517	-24.8176	-1.4668	-1.4260
0.0173	6.0	450	0.0069	2.2945	-15.3467	0.9900	17.6412	-114.3504	-24.8307	-1.4680	-1.4272
0.0347	6.6667	500	0.0069	2.2859	-15.4295	0.9900	17.7154	-114.6264	-24.8593	-1.4694	-1.4284
0.0	7.3333	550	0.0069	2.2833	-15.5057	0.9900	17.7890	-114.8804	-24.8681	-1.4703	-1.4293
0.0347	8.0	600	0.0069	2.2775	-15.5762	0.9900	17.8538	-115.1155	-24.8872	-1.4709	-1.4298
0.0	8.6667	650	0.0069	2.2759	-15.6206	0.9900	17.8965	-115.2633	-24.8928	-1.4712	-1.4301
0.0173	9.3333	700	0.0069	2.2757	-15.6425	0.9900	17.9182	-115.3363	-24.8933	-1.4714	-1.4302
0.0	10.0	750	0.0069	2.2743	-15.6650	0.9900	17.9392	-115.4112	-24.8982	-1.4717	-1.4305
0.0173	10.6667	800	0.0069	2.2739	-15.6785	0.9900	17.9524	-115.4563	-24.8992	-1.4719	-1.4307
0.0	11.3333	850	0.0069	2.2703	-15.6667	0.9900	17.9370	-115.4169	-24.9113	-1.4717	-1.4306
0.0	12.0	900	0.0069	2.2749	-15.6771	0.9900	17.9520	-115.4516	-24.8959	-1.4719	-1.4307
0.0173	12.6667	950	0.0069	2.2732	-15.6753	0.9900	17.9485	-115.4458	-24.9018	-1.4719	-1.4307
0.0	13.3333	1000	0.0069	2.2757	-15.6836	0.9900	17.9593	-115.4733	-24.8934	-1.4719	-1.4307

Framework versions

Transformers 4.41.2
Pytorch 2.0.0+cu117
Datasets 2.19.2
Tokenizers 0.19.1

Downloads last month: 5

Safetensors

Model size

8B params

Tensor type

F16

Model tree for tsavage68/UTI_L3_1000steps_1e5rate_03beta_CSFTDPO

Base model

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

Finetuned

tsavage68/UTI_L3_1000steps_1e5rate_SFT

Finetuned

(39)

this model