Instructions to use tsavage68/UTI_M2_1000steps_1e8rate_05beta_CSFTDPO with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use tsavage68/UTI_M2_1000steps_1e8rate_05beta_CSFTDPO with Transformers:

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

pipe = pipeline("text-generation", model="tsavage68/UTI_M2_1000steps_1e8rate_05beta_CSFTDPO")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("tsavage68/UTI_M2_1000steps_1e8rate_05beta_CSFTDPO")
model = AutoModelForCausalLM.from_pretrained("tsavage68/UTI_M2_1000steps_1e8rate_05beta_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_M2_1000steps_1e8rate_05beta_CSFTDPO with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "tsavage68/UTI_M2_1000steps_1e8rate_05beta_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_M2_1000steps_1e8rate_05beta_CSFTDPO",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/tsavage68/UTI_M2_1000steps_1e8rate_05beta_CSFTDPO

SGLang

How to use tsavage68/UTI_M2_1000steps_1e8rate_05beta_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_M2_1000steps_1e8rate_05beta_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_M2_1000steps_1e8rate_05beta_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_M2_1000steps_1e8rate_05beta_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_M2_1000steps_1e8rate_05beta_CSFTDPO",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use tsavage68/UTI_M2_1000steps_1e8rate_05beta_CSFTDPO with Docker Model Runner:
```
docker model run hf.co/tsavage68/UTI_M2_1000steps_1e8rate_05beta_CSFTDPO
```

UTI_M2_1000steps_1e8rate_05beta_CSFTDPO

This model is a fine-tuned version of tsavage68/UTI_M2_1000steps_1e5rate_SFT on an unknown dataset. It achieves the following results on the evaluation set:

Loss: 0.6733
Rewards/chosen: 0.0076
Rewards/rejected: -0.0329
Rewards/accuracies: 0.8100
Rewards/margins: 0.0405
Logps/rejected: -44.2758
Logps/chosen: -20.2691
Logits/rejected: -3.8168
Logits/chosen: -3.7448

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-08
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
0.6936	0.3333	25	0.6942	0.0013	0.0032	0.4000	-0.0020	-44.1553	-20.2903	-3.8168	-3.7448
0.6921	0.6667	50	0.6936	0.0023	0.0030	0.5	-0.0007	-44.1562	-20.2868	-3.8168	-3.7448
0.6943	1.0	75	0.6955	0.0005	0.0049	0.4200	-0.0044	-44.1498	-20.2931	-3.8169	-3.7449
0.6933	1.3333	100	0.6933	0.0014	0.0014	0.4200	-0.0001	-44.1614	-20.2900	-3.8168	-3.7448
0.6886	1.6667	125	0.6920	0.0002	-0.0022	0.4800	0.0024	-44.1735	-20.2938	-3.8168	-3.7448
0.6896	2.0	150	0.6887	0.0040	-0.0053	0.5700	0.0092	-44.1837	-20.2814	-3.8168	-3.7448
0.6879	2.3333	175	0.6864	0.0033	-0.0104	0.6200	0.0138	-44.2009	-20.2835	-3.8168	-3.7447
0.683	2.6667	200	0.6824	0.0048	-0.0171	0.6700	0.0218	-44.2230	-20.2786	-3.8168	-3.7447
0.6815	3.0	225	0.6825	0.0042	-0.0174	0.7100	0.0217	-44.2242	-20.2805	-3.8169	-3.7449
0.6791	3.3333	250	0.6800	0.0039	-0.0229	0.7300	0.0268	-44.2426	-20.2815	-3.8168	-3.7448
0.6772	3.6667	275	0.6798	0.0062	-0.0210	0.6900	0.0273	-44.2362	-20.2738	-3.8167	-3.7447
0.6753	4.0	300	0.6784	0.0059	-0.0242	0.7400	0.0301	-44.2468	-20.2747	-3.8169	-3.7448
0.6821	4.3333	325	0.6767	0.0069	-0.0268	0.7700	0.0336	-44.2554	-20.2717	-3.8167	-3.7447
0.6744	4.6667	350	0.6770	0.0060	-0.0270	0.7100	0.0330	-44.2561	-20.2747	-3.8169	-3.7448
0.6741	5.0	375	0.6750	0.0088	-0.0281	0.7300	0.0370	-44.2598	-20.2651	-3.8168	-3.7448
0.6738	5.3333	400	0.6753	0.0084	-0.0281	0.7700	0.0365	-44.2599	-20.2667	-3.8168	-3.7447
0.6731	5.6667	425	0.6746	0.0074	-0.0306	0.75	0.0379	-44.2681	-20.2701	-3.8169	-3.7448
0.6756	6.0	450	0.6755	0.0071	-0.0289	0.7700	0.0360	-44.2625	-20.2710	-3.8168	-3.7448
0.6703	6.3333	475	0.6750	0.0093	-0.0279	0.75	0.0371	-44.2590	-20.2637	-3.8168	-3.7448
0.6796	6.6667	500	0.6744	0.0074	-0.0308	0.7800	0.0383	-44.2689	-20.2698	-3.8168	-3.7448
0.6676	7.0	525	0.6735	0.0091	-0.0311	0.7800	0.0402	-44.2699	-20.2642	-3.8168	-3.7447
0.6744	7.3333	550	0.6738	0.0067	-0.0330	0.7600	0.0397	-44.2760	-20.2721	-3.8168	-3.7448
0.6725	7.6667	575	0.6729	0.0083	-0.0330	0.8000	0.0413	-44.2761	-20.2668	-3.8168	-3.7447
0.6739	8.0	600	0.6732	0.0080	-0.0327	0.8100	0.0407	-44.2751	-20.2679	-3.8168	-3.7447
0.6675	8.3333	625	0.6748	0.0084	-0.0291	0.7800	0.0375	-44.2632	-20.2665	-3.8169	-3.7448
0.6706	8.6667	650	0.6732	0.0087	-0.0320	0.8100	0.0407	-44.2728	-20.2656	-3.8168	-3.7447
0.6718	9.0	675	0.6741	0.0086	-0.0303	0.7800	0.0389	-44.2671	-20.2658	-3.8168	-3.7448
0.6715	9.3333	700	0.6743	0.0085	-0.0300	0.8000	0.0385	-44.2662	-20.2662	-3.8168	-3.7447
0.6723	9.6667	725	0.6727	0.0066	-0.0352	0.7700	0.0417	-44.2834	-20.2727	-3.8168	-3.7448
0.6715	10.0	750	0.6729	0.0067	-0.0348	0.7700	0.0415	-44.2822	-20.2723	-3.8168	-3.7448
0.669	10.3333	775	0.6743	0.0074	-0.0310	0.7600	0.0384	-44.2694	-20.2698	-3.8168	-3.7447
0.6738	10.6667	800	0.6729	0.0079	-0.0336	0.8000	0.0415	-44.2780	-20.2682	-3.8168	-3.7448
0.6738	11.0	825	0.6735	0.0088	-0.0312	0.8100	0.0400	-44.2703	-20.2653	-3.8169	-3.7448
0.6682	11.3333	850	0.6736	0.0079	-0.0321	0.8100	0.0400	-44.2730	-20.2681	-3.8168	-3.7448
0.6787	11.6667	875	0.6733	0.0076	-0.0329	0.8100	0.0405	-44.2758	-20.2691	-3.8168	-3.7448
0.6771	12.0	900	0.6733	0.0076	-0.0329	0.8100	0.0405	-44.2758	-20.2691	-3.8168	-3.7448
0.6705	12.3333	925	0.6733	0.0076	-0.0329	0.8100	0.0405	-44.2758	-20.2691	-3.8168	-3.7448
0.6727	12.6667	950	0.6733	0.0076	-0.0329	0.8100	0.0405	-44.2758	-20.2691	-3.8168	-3.7448
0.6748	13.0	975	0.6733	0.0076	-0.0329	0.8100	0.0405	-44.2758	-20.2691	-3.8168	-3.7448
0.6809	13.3333	1000	0.6733	0.0076	-0.0329	0.8100	0.0405	-44.2758	-20.2691	-3.8168	-3.7448

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

7B params

Tensor type

F16

Model tree for tsavage68/UTI_M2_1000steps_1e8rate_05beta_CSFTDPO

Base model

mistralai/Mistral-7B-Instruct-v0.2

Finetuned

tsavage68/UTI_M2_1000steps_1e5rate_SFT

Finetuned

(12)

this model