Instructions to use tsavage68/UTI_M2_1000steps_1e7rate_05beta_CSFTDPO with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use tsavage68/UTI_M2_1000steps_1e7rate_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_1e7rate_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_1e7rate_05beta_CSFTDPO")
model = AutoModelForCausalLM.from_pretrained("tsavage68/UTI_M2_1000steps_1e7rate_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_1e7rate_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_1e7rate_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_1e7rate_05beta_CSFTDPO",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/tsavage68/UTI_M2_1000steps_1e7rate_05beta_CSFTDPO

SGLang

How to use tsavage68/UTI_M2_1000steps_1e7rate_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_1e7rate_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_1e7rate_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_1e7rate_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_1e7rate_05beta_CSFTDPO",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use tsavage68/UTI_M2_1000steps_1e7rate_05beta_CSFTDPO with Docker Model Runner:
```
docker model run hf.co/tsavage68/UTI_M2_1000steps_1e7rate_05beta_CSFTDPO
```

UTI_M2_1000steps_1e7rate_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: 0.0694
Rewards/chosen: 1.9080
Rewards/rejected: -7.6159
Rewards/accuracies: 0.9000
Rewards/margins: 9.5239
Logps/rejected: -59.3979
Logps/chosen: -16.4785
Logits/rejected: -3.8250
Logits/chosen: -3.7455

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: 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
0.6901	0.3333	25	0.6692	0.0087	-0.0409	0.7200	0.0495	-44.2478	-20.2772	-3.8170	-3.7449
0.5385	0.6667	50	0.4603	0.1249	-0.5004	0.8900	0.6253	-45.1669	-20.0447	-3.8171	-3.7449
0.2906	1.0	75	0.2278	0.4625	-1.4619	0.9000	1.9243	-47.0899	-19.3696	-3.8191	-3.7461
0.0574	1.3333	100	0.1164	0.8924	-2.9525	0.9000	3.8450	-50.0712	-18.5097	-3.8224	-3.7481
0.0769	1.6667	125	0.0739	1.6307	-4.2975	0.9000	5.9282	-52.7612	-17.0332	-3.8223	-3.7453
0.1052	2.0	150	0.0702	2.1207	-5.3845	0.9000	7.5052	-54.9352	-16.0532	-3.8235	-3.7451
0.0523	2.3333	175	0.0699	2.1700	-5.7373	0.9000	7.9072	-55.6406	-15.9546	-3.8236	-3.7449
0.0525	2.6667	200	0.0697	2.1737	-6.1919	0.9000	8.3656	-56.5499	-15.9472	-3.8243	-3.7452
0.0348	3.0	225	0.0696	2.1037	-6.4097	0.9000	8.5134	-56.9855	-16.0871	-3.8246	-3.7455
0.0521	3.3333	250	0.0695	2.0825	-6.6031	0.9000	8.6856	-57.3723	-16.1296	-3.8248	-3.7457
0.0694	3.6667	275	0.0695	2.0856	-6.6814	0.9000	8.7670	-57.5290	-16.1234	-3.8249	-3.7457
0.0867	4.0	300	0.0695	2.0657	-6.7649	0.9000	8.8305	-57.6958	-16.1632	-3.8248	-3.7457
0.104	4.3333	325	0.0695	2.0705	-6.8399	0.9000	8.9103	-57.8458	-16.1536	-3.8248	-3.7456
0.0347	4.6667	350	0.0695	2.0425	-6.9434	0.9000	8.9860	-58.0530	-16.2095	-3.8251	-3.7458
0.0694	5.0	375	0.0695	2.0050	-7.0675	0.9000	9.0725	-58.3011	-16.2845	-3.8249	-3.7457
0.0867	5.3333	400	0.0695	1.9818	-7.1406	0.9000	9.1224	-58.4472	-16.3309	-3.8249	-3.7457
0.0521	5.6667	425	0.0694	1.9889	-7.2091	0.9000	9.1979	-58.5843	-16.3168	-3.8249	-3.7456
0.0867	6.0	450	0.0694	1.9778	-7.2729	0.9000	9.2507	-58.7119	-16.3389	-3.8250	-3.7456
0.0174	6.3333	475	0.0694	1.9885	-7.3057	0.9000	9.2942	-58.7776	-16.3176	-3.8250	-3.7457
0.1214	6.6667	500	0.0694	1.9533	-7.3744	0.9000	9.3277	-58.9150	-16.3880	-3.8250	-3.7457
0.0347	7.0	525	0.0694	1.9561	-7.4213	0.9000	9.3773	-59.0087	-16.3825	-3.8251	-3.7457
0.0867	7.3333	550	0.0694	1.9348	-7.4844	0.9000	9.4192	-59.1348	-16.4249	-3.8250	-3.7456
0.052	7.6667	575	0.0694	1.9383	-7.5106	0.9000	9.4489	-59.1874	-16.4180	-3.8251	-3.7457
0.104	8.0	600	0.0694	1.9329	-7.5416	0.9000	9.4745	-59.2494	-16.4287	-3.8250	-3.7456
0.1213	8.3333	625	0.0694	1.9293	-7.5550	0.9000	9.4842	-59.2761	-16.4360	-3.8251	-3.7457
0.0174	8.6667	650	0.0694	1.9201	-7.5740	0.9000	9.4942	-59.3142	-16.4543	-3.8250	-3.7456
0.052	9.0	675	0.0694	1.9221	-7.5851	0.9000	9.5072	-59.3363	-16.4504	-3.8251	-3.7457
0.052	9.3333	700	0.0694	1.9173	-7.5999	0.9000	9.5173	-59.3660	-16.4599	-3.8250	-3.7455
0.0867	9.6667	725	0.0694	1.9099	-7.6110	0.9000	9.5209	-59.3882	-16.4749	-3.8249	-3.7455
0.0867	10.0	750	0.0694	1.9157	-7.6160	0.9000	9.5316	-59.3980	-16.4633	-3.8249	-3.7455
0.0867	10.3333	775	0.0694	1.9136	-7.6107	0.9000	9.5243	-59.3875	-16.4673	-3.8249	-3.7455
0.0867	10.6667	800	0.0694	1.9097	-7.6200	0.9000	9.5297	-59.4061	-16.4752	-3.8250	-3.7456
0.0693	11.0	825	0.0694	1.9094	-7.6202	0.9000	9.5296	-59.4065	-16.4758	-3.8250	-3.7456
0.052	11.3333	850	0.0694	1.9095	-7.6258	0.9000	9.5353	-59.4176	-16.4756	-3.8250	-3.7456
0.0694	11.6667	875	0.0694	1.9107	-7.6193	0.9000	9.5300	-59.4047	-16.4731	-3.8250	-3.7456
0.0694	12.0	900	0.0694	1.9091	-7.6182	0.9000	9.5273	-59.4025	-16.4765	-3.8250	-3.7455
0.0347	12.3333	925	0.0694	1.9084	-7.6163	0.9000	9.5247	-59.3988	-16.4778	-3.8250	-3.7455
0.0694	12.6667	950	0.0694	1.9080	-7.6159	0.9000	9.5239	-59.3979	-16.4785	-3.8250	-3.7455
0.052	13.0	975	0.0694	1.9080	-7.6159	0.9000	9.5239	-59.3979	-16.4785	-3.8250	-3.7455
0.0867	13.3333	1000	0.0694	1.9080	-7.6159	0.9000	9.5239	-59.3979	-16.4785	-3.8250	-3.7455

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_1e7rate_05beta_CSFTDPO

Base model

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

Finetuned

tsavage68/UTI_M2_1000steps_1e5rate_SFT

Finetuned

(12)

this model