Instructions to use tsavage68/MedQA_L3_250steps_1e5rate_01beta_CSFTDPO with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use tsavage68/MedQA_L3_250steps_1e5rate_01beta_CSFTDPO with Transformers:

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

pipe = pipeline("text-generation", model="tsavage68/MedQA_L3_250steps_1e5rate_01beta_CSFTDPO")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("tsavage68/MedQA_L3_250steps_1e5rate_01beta_CSFTDPO")
model = AutoModelForCausalLM.from_pretrained("tsavage68/MedQA_L3_250steps_1e5rate_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/MedQA_L3_250steps_1e5rate_01beta_CSFTDPO with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "tsavage68/MedQA_L3_250steps_1e5rate_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/MedQA_L3_250steps_1e5rate_01beta_CSFTDPO",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/tsavage68/MedQA_L3_250steps_1e5rate_01beta_CSFTDPO

SGLang

How to use tsavage68/MedQA_L3_250steps_1e5rate_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/MedQA_L3_250steps_1e5rate_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/MedQA_L3_250steps_1e5rate_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/MedQA_L3_250steps_1e5rate_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/MedQA_L3_250steps_1e5rate_01beta_CSFTDPO",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use tsavage68/MedQA_L3_250steps_1e5rate_01beta_CSFTDPO with Docker Model Runner:
```
docker model run hf.co/tsavage68/MedQA_L3_250steps_1e5rate_01beta_CSFTDPO
```

MedQA_L3_250steps_1e5rate_01beta_CSFTDPO

This model is a fine-tuned version of tsavage68/MedQA_L3_1000steps_1e6rate_SFT on an unknown dataset. It achieves the following results on the evaluation set:

Loss: 0.9785
Rewards/chosen: -1.7641
Rewards/rejected: -1.5433
Rewards/accuracies: 0.4132
Rewards/margins: -0.2209
Logps/rejected: -49.2875
Logps/chosen: -48.9697
Logits/rejected: -1.1405
Logits/chosen: -1.1409

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-05
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: 250

Training results

Training Loss	Epoch	Step	Validation Loss	Rewards/chosen	Rewards/rejected	Rewards/accuracies	Rewards/margins	Logps/rejected	Logps/chosen	Logits/rejected	Logits/chosen
0.671	0.0489	50	1.6433	-6.4141	-6.3515	0.4747	-0.0626	-97.3700	-95.4696	-0.6453	-0.6453
1.0504	0.0977	100	0.8270	-1.6657	-1.8409	0.5385	0.1752	-52.2642	-47.9860	-1.0550	-1.0545
1.3146	0.1466	150	1.0584	-2.1772	-1.8983	0.4110	-0.2789	-52.8378	-53.1002	-1.6449	-1.6452
1.2122	0.1954	200	1.0261	-1.8796	-1.6260	0.4066	-0.2536	-50.1151	-50.1247	-1.1724	-1.1728
0.929	0.2443	250	0.9785	-1.7641	-1.5433	0.4132	-0.2209	-49.2875	-48.9697	-1.1405	-1.1409

Framework versions

Transformers 4.41.1
Pytorch 2.0.0+cu117
Datasets 2.19.1
Tokenizers 0.19.1

Downloads last month: 5

Safetensors

Model size

8B params

Tensor type

F16

Model tree for tsavage68/MedQA_L3_250steps_1e5rate_01beta_CSFTDPO

Base model

meta-llama/Meta-Llama-3-8B-Instruct

Finetuned

tsavage68/MedQA_L3_1000steps_1e6rate_SFT

Finetuned

(20)

this model