Instructions to use tsavage68/UTI2_M2_300steps_1e8rate_03beta_CSFTDPO with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use tsavage68/UTI2_M2_300steps_1e8rate_03beta_CSFTDPO with Transformers:

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

pipe = pipeline("text-generation", model="tsavage68/UTI2_M2_300steps_1e8rate_03beta_CSFTDPO")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("tsavage68/UTI2_M2_300steps_1e8rate_03beta_CSFTDPO")
model = AutoModelForCausalLM.from_pretrained("tsavage68/UTI2_M2_300steps_1e8rate_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/UTI2_M2_300steps_1e8rate_03beta_CSFTDPO with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "tsavage68/UTI2_M2_300steps_1e8rate_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/UTI2_M2_300steps_1e8rate_03beta_CSFTDPO",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/tsavage68/UTI2_M2_300steps_1e8rate_03beta_CSFTDPO

SGLang

How to use tsavage68/UTI2_M2_300steps_1e8rate_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/UTI2_M2_300steps_1e8rate_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/UTI2_M2_300steps_1e8rate_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/UTI2_M2_300steps_1e8rate_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/UTI2_M2_300steps_1e8rate_03beta_CSFTDPO",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use tsavage68/UTI2_M2_300steps_1e8rate_03beta_CSFTDPO with Docker Model Runner:
```
docker model run hf.co/tsavage68/UTI2_M2_300steps_1e8rate_03beta_CSFTDPO
```

UTI2_M2_300steps_1e8rate_03beta_CSFTDPO

This model is a fine-tuned version of tsavage68/UTI_M2_1000steps_1e7rate_SFT on an unknown dataset. It achieves the following results on the evaluation set:

Loss: 0.6666
Rewards/chosen: 0.0060
Rewards/rejected: -0.0495
Rewards/accuracies: 0.7100
Rewards/margins: 0.0555
Logps/rejected: -39.5212
Logps/chosen: -19.9016
Logits/rejected: -2.6824
Logits/chosen: -2.6798

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: 300

Training results

Training Loss	Epoch	Step	Validation Loss	Rewards/chosen	Rewards/rejected	Rewards/accuracies	Rewards/margins	Logps/rejected	Logps/chosen	Logits/rejected	Logits/chosen
0.6931	0.3333	25	0.6924	0.0019	0.0000	0.1900	0.0018	-39.3560	-19.9153	-2.6832	-2.6806
0.6873	0.6667	50	0.6859	0.0055	-0.0103	0.5400	0.0158	-39.3904	-19.9033	-2.6823	-2.6798
0.6944	1.0	75	0.6937	-0.0057	-0.0058	0.4500	0.0001	-39.3756	-19.9405	-2.6824	-2.6798
0.6899	1.3333	100	0.6855	-0.0113	-0.0281	0.5600	0.0168	-39.4498	-19.9591	-2.6820	-2.6794
0.6858	1.6667	125	0.6752	0.0071	-0.0308	0.6400	0.0379	-39.4588	-19.8979	-2.6822	-2.6796
0.6767	2.0	150	0.6734	0.0063	-0.0351	0.6600	0.0415	-39.4733	-19.9004	-2.6827	-2.6802
0.6625	2.3333	175	0.6598	0.0094	-0.0599	0.7600	0.0693	-39.5558	-19.8902	-2.6809	-2.6783
0.6658	2.6667	200	0.6644	0.0077	-0.0525	0.6900	0.0602	-39.5310	-19.8958	-2.6823	-2.6797
0.6793	3.0	225	0.6654	0.0092	-0.0489	0.7200	0.0581	-39.5192	-19.8907	-2.6819	-2.6793
0.6836	3.3333	250	0.6662	0.0062	-0.0499	0.7300	0.0561	-39.5225	-19.9009	-2.6824	-2.6798
0.6704	3.6667	275	0.6666	0.0060	-0.0495	0.7100	0.0555	-39.5212	-19.9016	-2.6824	-2.6798
0.6726	4.0	300	0.6666	0.0060	-0.0495	0.7100	0.0555	-39.5212	-19.9016	-2.6824	-2.6798

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/UTI2_M2_300steps_1e8rate_03beta_CSFTDPO

Base model

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

Finetuned

tsavage68/UTI_M2_1000steps_1e7rate_SFT

Finetuned

(32)

this model