Instructions to use tsavage68/Mistral2_700_STEPS_05beta_1e6rate_CDPOSFT with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use tsavage68/Mistral2_700_STEPS_05beta_1e6rate_CDPOSFT with Transformers:

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

pipe = pipeline("text-generation", model="tsavage68/Mistral2_700_STEPS_05beta_1e6rate_CDPOSFT")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("tsavage68/Mistral2_700_STEPS_05beta_1e6rate_CDPOSFT")
model = AutoModelForCausalLM.from_pretrained("tsavage68/Mistral2_700_STEPS_05beta_1e6rate_CDPOSFT", 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/Mistral2_700_STEPS_05beta_1e6rate_CDPOSFT with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "tsavage68/Mistral2_700_STEPS_05beta_1e6rate_CDPOSFT"
# 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/Mistral2_700_STEPS_05beta_1e6rate_CDPOSFT",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/tsavage68/Mistral2_700_STEPS_05beta_1e6rate_CDPOSFT

SGLang

How to use tsavage68/Mistral2_700_STEPS_05beta_1e6rate_CDPOSFT 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/Mistral2_700_STEPS_05beta_1e6rate_CDPOSFT" \
    --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/Mistral2_700_STEPS_05beta_1e6rate_CDPOSFT",
		"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/Mistral2_700_STEPS_05beta_1e6rate_CDPOSFT" \
        --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/Mistral2_700_STEPS_05beta_1e6rate_CDPOSFT",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use tsavage68/Mistral2_700_STEPS_05beta_1e6rate_CDPOSFT with Docker Model Runner:
```
docker model run hf.co/tsavage68/Mistral2_700_STEPS_05beta_1e6rate_CDPOSFT
```

Mistral2_700_STEPS_05beta_1e6rate_CDPOSFT

This model is a fine-tuned version of tsavage68/mistralit2_1000_STEPS_5e7_SFT on an unknown dataset. It achieves the following results on the evaluation set:

Loss: 1.0698
Rewards/chosen: 1.2984
Rewards/rejected: 1.2345
Rewards/accuracies: 0.4593
Rewards/margins: 0.0639
Logps/rejected: -75.0525
Logps/chosen: -72.9052
Logits/rejected: -2.1829
Logits/chosen: -2.1828

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-06
train_batch_size: 4
eval_batch_size: 1
seed: 42
gradient_accumulation_steps: 2
total_train_batch_size: 8
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: cosine
lr_scheduler_warmup_steps: 100
training_steps: 700

Training results

Training Loss	Epoch	Step	Validation Loss	Rewards/chosen	Rewards/rejected	Rewards/accuracies	Rewards/margins	Logps/rejected	Logps/chosen	Logits/rejected	Logits/chosen
0.9416	0.0977	50	0.9097	1.0706	1.0514	0.4022	0.0192	-75.4186	-73.3607	-2.0307	-2.0304
0.8274	0.1953	100	1.2209	-0.4686	-0.4288	0.4418	-0.0397	-78.3791	-76.4391	-2.1855	-2.1854
1.18	0.2930	150	1.2641	0.4071	0.4301	0.4044	-0.0230	-76.6613	-74.6878	-2.5231	-2.5229
1.3266	0.3906	200	1.3043	1.3288	1.3407	0.4088	-0.0119	-74.8401	-72.8443	-1.7175	-1.7174
1.3999	0.4883	250	1.3290	-0.0471	-0.2172	0.4220	0.1701	-77.9559	-75.5961	-2.7522	-2.7522
1.2719	0.5859	300	1.2886	0.8483	0.8336	0.4286	0.0147	-75.8543	-73.8053	-2.3076	-2.3075
0.8563	0.6836	350	1.2006	1.4189	1.4915	0.4220	-0.0726	-74.5385	-72.6641	-1.9980	-1.9979
1.0544	0.7812	400	1.1102	1.3236	1.2452	0.4418	0.0785	-75.0312	-72.8547	-1.9966	-1.9964
1.1287	0.8789	450	1.0957	1.7342	1.6182	0.4462	0.1159	-74.2850	-72.0337	-2.2594	-2.2593
0.9215	0.9766	500	1.0732	1.5069	1.4426	0.4484	0.0643	-74.6362	-72.4881	-2.1957	-2.1956
0.5642	1.0742	550	1.0697	1.3526	1.2850	0.4615	0.0676	-74.9514	-72.7967	-2.1981	-2.1979
0.6954	1.1719	600	1.0710	1.2885	1.2257	0.4593	0.0627	-75.0700	-72.9250	-2.1856	-2.1854
0.4733	1.2695	650	1.0708	1.2971	1.2350	0.4571	0.0621	-75.0515	-72.9078	-2.1832	-2.1830
0.5512	1.3672	700	1.0698	1.2984	1.2345	0.4593	0.0639	-75.0525	-72.9052	-2.1829	-2.1828

Framework versions

Transformers 4.40.1
Pytorch 2.0.0+cu117
Datasets 2.19.0
Tokenizers 0.19.1

Downloads last month: 6

Safetensors

Model size

7B params

Tensor type

F16

Model tree for tsavage68/Mistral2_700_STEPS_05beta_1e6rate_CDPOSFT

Base model

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

Finetuned

tsavage68/mistralit2_1000_STEPS_5e7_SFT

Finetuned

(19)

this model