Instructions to use tsavage68/Summary4500_M2_1000steps_1e7rate_01beta_CSFTDPO with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use tsavage68/Summary4500_M2_1000steps_1e7rate_01beta_CSFTDPO with Transformers:

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

pipe = pipeline("text-generation", model="tsavage68/Summary4500_M2_1000steps_1e7rate_01beta_CSFTDPO")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("tsavage68/Summary4500_M2_1000steps_1e7rate_01beta_CSFTDPO")
model = AutoModelForCausalLM.from_pretrained("tsavage68/Summary4500_M2_1000steps_1e7rate_01beta_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/Summary4500_M2_1000steps_1e7rate_01beta_CSFTDPO with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "tsavage68/Summary4500_M2_1000steps_1e7rate_01beta_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/Summary4500_M2_1000steps_1e7rate_01beta_CSFTDPO",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/tsavage68/Summary4500_M2_1000steps_1e7rate_01beta_CSFTDPO

SGLang

How to use tsavage68/Summary4500_M2_1000steps_1e7rate_01beta_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/Summary4500_M2_1000steps_1e7rate_01beta_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/Summary4500_M2_1000steps_1e7rate_01beta_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/Summary4500_M2_1000steps_1e7rate_01beta_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/Summary4500_M2_1000steps_1e7rate_01beta_CSFTDPO",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use tsavage68/Summary4500_M2_1000steps_1e7rate_01beta_CSFTDPO with Docker Model Runner:
```
docker model run hf.co/tsavage68/Summary4500_M2_1000steps_1e7rate_01beta_CSFTDPO
```

Hyponatremia_M2_1000steps_1e7rate_01beta_CSFTDPO

This model is a fine-tuned version of tsavage68/Summary4500_M2_200steps_1e7rate_SFT on an unknown dataset. It achieves the following results on the evaluation set:

Loss: 0.0020
Rewards/chosen: -4.7124
Rewards/rejected: -19.2249
Rewards/accuracies: 0.9980
Rewards/margins: 14.5125
Logps/rejected: -344.9792
Logps/chosen: -140.8642
Logits/rejected: -2.0739
Logits/chosen: -2.0387

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: 1
eval_batch_size: 1
seed: 42
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
0.5233	0.0112	50	0.5232	-0.0091	-0.3925	0.9980	0.3833	-156.6543	-93.8310	-2.3405	-2.2932
0.0383	0.0224	100	0.0430	-0.7708	-4.7747	0.9980	4.0039	-200.4769	-101.4475	-2.2287	-2.1837
0.0	0.0336	150	0.0070	-2.4385	-10.5541	0.9980	8.1156	-258.2711	-118.1250	-2.1599	-2.1186
0.0016	0.0448	200	0.0031	-3.4937	-14.8511	0.9980	11.3574	-301.2408	-128.6765	-2.1144	-2.0760
0.0002	0.0559	250	0.0029	-3.4966	-15.1134	0.9980	11.6168	-303.8634	-128.7055	-2.1174	-2.0790
0.0	0.0671	300	0.0026	-3.6799	-15.9182	0.9980	12.2384	-311.9122	-130.5386	-2.1042	-2.0666
0.0012	0.0783	350	0.0024	-3.9841	-16.8321	0.9980	12.8480	-321.0512	-133.5813	-2.0951	-2.0582
0.0001	0.0895	400	0.0022	-4.3249	-17.8761	0.9980	13.5512	-331.4908	-136.9887	-2.0841	-2.0480
0.0	0.1007	450	0.0022	-4.4809	-18.3463	0.9980	13.8653	-336.1925	-138.5490	-2.0802	-2.0445
0.0	0.1119	500	0.0022	-4.5041	-18.4203	0.9980	13.9162	-336.9331	-138.7807	-2.0778	-2.0423
0.0	0.1231	550	0.0021	-4.5894	-18.8015	0.9980	14.2121	-340.7446	-139.6336	-2.0767	-2.0412
0.0	0.1343	600	0.0021	-4.6515	-19.0151	0.9980	14.3636	-342.8809	-140.2545	-2.0750	-2.0398
0.0	0.1454	650	0.0020	-4.6765	-19.1006	0.9980	14.4240	-343.7354	-140.5052	-2.0753	-2.0401
0.0	0.1566	700	0.0020	-4.6869	-19.1397	0.9980	14.4528	-344.1270	-140.6091	-2.0750	-2.0398
0.0	0.1678	750	0.0020	-4.6998	-19.1976	0.9980	14.4978	-344.7062	-140.7377	-2.0747	-2.0396
0.0	0.1790	800	0.0020	-4.7132	-19.2365	0.9980	14.5233	-345.0950	-140.8720	-2.0739	-2.0389
0.0096	0.1902	850	0.0020	-4.7099	-19.2301	0.9980	14.5202	-345.0307	-140.8386	-2.0740	-2.0389
0.0	0.2014	900	0.0020	-4.7077	-19.2206	0.9980	14.5129	-344.9359	-140.8168	-2.0737	-2.0386
0.0	0.2126	950	0.0020	-4.7125	-19.2249	0.9980	14.5125	-344.9792	-140.8644	-2.0739	-2.0387
0.0	0.2238	1000	0.0020	-4.7124	-19.2249	0.9980	14.5125	-344.9792	-140.8642	-2.0739	-2.0387

Framework versions

Transformers 4.42.4
Pytorch 2.0.0+cu117
Datasets 2.20.0
Tokenizers 0.19.1

Downloads last month: 5

Safetensors

Model size

7B params

Tensor type

F16

Model tree for tsavage68/Summary4500_M2_1000steps_1e7rate_01beta_CSFTDPO

Base model

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

Finetuned

tsavage68/Summary4500_M2_200steps_1e7rate_SFT

Finetuned

(14)

this model