Instructions to use tsavage68/Mistral2_500_STEPS_03beta_1e6_CDPOSFT with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use tsavage68/Mistral2_500_STEPS_03beta_1e6_CDPOSFT with Transformers:

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

pipe = pipeline("text-generation", model="tsavage68/Mistral2_500_STEPS_03beta_1e6_CDPOSFT")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("tsavage68/Mistral2_500_STEPS_03beta_1e6_CDPOSFT")
model = AutoModelForCausalLM.from_pretrained("tsavage68/Mistral2_500_STEPS_03beta_1e6_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_500_STEPS_03beta_1e6_CDPOSFT with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "tsavage68/Mistral2_500_STEPS_03beta_1e6_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_500_STEPS_03beta_1e6_CDPOSFT",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/tsavage68/Mistral2_500_STEPS_03beta_1e6_CDPOSFT

SGLang

How to use tsavage68/Mistral2_500_STEPS_03beta_1e6_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_500_STEPS_03beta_1e6_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_500_STEPS_03beta_1e6_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_500_STEPS_03beta_1e6_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_500_STEPS_03beta_1e6_CDPOSFT",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use tsavage68/Mistral2_500_STEPS_03beta_1e6_CDPOSFT with Docker Model Runner:
```
docker model run hf.co/tsavage68/Mistral2_500_STEPS_03beta_1e6_CDPOSFT
```

Mistral2_500_STEPS_03beta_1e6_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: 0.9078
Rewards/chosen: 0.9010
Rewards/rejected: 0.9103
Rewards/accuracies: 0.4264
Rewards/margins: -0.0093
Logps/rejected: -74.4872
Logps/chosen: -72.4988
Logits/rejected: -2.2208
Logits/chosen: -2.2208

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: 500

Training results

Training Loss	Epoch	Step	Validation Loss	Rewards/chosen	Rewards/rejected	Rewards/accuracies	Rewards/margins	Logps/rejected	Logps/chosen	Logits/rejected	Logits/chosen
0.831	0.0977	50	0.8025	0.6608	0.6344	0.4132	0.0264	-75.4068	-73.2992	-2.0277	-2.0274
0.6868	0.1953	100	0.9417	0.1823	0.1932	0.4198	-0.0109	-76.8774	-74.8943	-2.0626	-2.0624
0.9659	0.2930	150	1.0129	-0.1807	-0.0985	0.4066	-0.0822	-77.8497	-76.1043	-2.2549	-2.2547
1.0344	0.3906	200	0.9907	0.7434	0.7229	0.4154	0.0205	-75.1118	-73.0240	-1.8653	-1.8653
1.0288	0.4883	250	1.0213	0.0313	-0.0408	0.3956	0.0722	-77.6576	-75.3976	-2.9696	-2.9697
0.9039	0.5859	300	1.0169	0.1606	0.1630	0.4176	-0.0024	-76.9781	-74.9667	-2.5569	-2.5569
0.7829	0.6836	350	0.9615	0.7982	0.8506	0.4022	-0.0524	-74.6861	-72.8413	-2.0384	-2.0382
0.8517	0.7812	400	0.9323	0.9042	0.9403	0.4264	-0.0361	-74.3873	-72.4880	-2.2049	-2.2048
0.9601	0.8789	450	0.9082	0.9006	0.9106	0.4286	-0.0100	-74.4862	-72.5	-2.2210	-2.2210
0.7935	0.9766	500	0.9078	0.9010	0.9103	0.4264	-0.0093	-74.4872	-72.4988	-2.2208	-2.2208

Framework versions

Transformers 4.40.1
Pytorch 2.0.0+cu117
Datasets 2.19.0
Tokenizers 0.19.1

Downloads last month: 7

Safetensors

Model size

7B params

Tensor type

F16

Model tree for tsavage68/Mistral2_500_STEPS_03beta_1e6_CDPOSFT

Base model

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

Finetuned

tsavage68/mistralit2_1000_STEPS_5e7_SFT

Finetuned

(19)

this model