Instructions to use tsavage68/UTI2_M2_125steps_1e7rate_03beta_CSFTDPO with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use tsavage68/UTI2_M2_125steps_1e7rate_03beta_CSFTDPO with Transformers:

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

pipe = pipeline("text-generation", model="tsavage68/UTI2_M2_125steps_1e7rate_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_125steps_1e7rate_03beta_CSFTDPO")
model = AutoModelForCausalLM.from_pretrained("tsavage68/UTI2_M2_125steps_1e7rate_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_125steps_1e7rate_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_125steps_1e7rate_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_125steps_1e7rate_03beta_CSFTDPO",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/tsavage68/UTI2_M2_125steps_1e7rate_03beta_CSFTDPO

SGLang

How to use tsavage68/UTI2_M2_125steps_1e7rate_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_125steps_1e7rate_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_125steps_1e7rate_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_125steps_1e7rate_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_125steps_1e7rate_03beta_CSFTDPO",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use tsavage68/UTI2_M2_125steps_1e7rate_03beta_CSFTDPO with Docker Model Runner:
```
docker model run hf.co/tsavage68/UTI2_M2_125steps_1e7rate_03beta_CSFTDPO
```

UTI2_M2_125steps_1e7rate_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: 0.6514
Rewards/chosen: -0.0649
Rewards/rejected: -2.2854
Rewards/accuracies: 0.1900
Rewards/margins: 2.2205
Logps/rejected: -16.9921
Logps/chosen: -4.7589
Logits/rejected: -2.6309
Logits/chosen: -2.6305

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-07
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: 125

Training results

Training Loss	Epoch	Step	Validation Loss	Rewards/chosen	Rewards/rejected	Rewards/accuracies	Rewards/margins	Logps/rejected	Logps/chosen	Logits/rejected	Logits/chosen
0.6972	0.3333	25	0.6778	-0.0051	-0.0396	0.1700	0.0345	-9.5061	-4.5596	-2.7055	-2.7048
0.6009	0.6667	50	0.5879	0.0346	-0.3529	0.2100	0.3875	-10.5502	-4.4271	-2.6953	-2.6946
0.5832	1.0	75	0.5855	-0.0014	-1.5604	0.1900	1.5590	-14.5753	-4.5473	-2.6506	-2.6501
0.6192	1.3333	100	0.5646	-0.0198	-2.1667	0.1900	2.1469	-16.5963	-4.6084	-2.6315	-2.6312
0.5545	1.6667	125	0.6514	-0.0649	-2.2854	0.1900	2.2205	-16.9921	-4.7589	-2.6309	-2.6305

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_125steps_1e7rate_03beta_CSFTDPO

Base model

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

Finetuned

tsavage68/UTI_M2_1000steps_1e7rate_SFT

Finetuned

(32)

this model