Instructions to use tsavage68/Summary4500_L3_1000steps_1e8rate_01beta_CSFTDPO with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use tsavage68/Summary4500_L3_1000steps_1e8rate_01beta_CSFTDPO with Transformers:

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

pipe = pipeline("text-generation", model="tsavage68/Summary4500_L3_1000steps_1e8rate_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_L3_1000steps_1e8rate_01beta_CSFTDPO")
model = AutoModelForCausalLM.from_pretrained("tsavage68/Summary4500_L3_1000steps_1e8rate_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_L3_1000steps_1e8rate_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_L3_1000steps_1e8rate_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_L3_1000steps_1e8rate_01beta_CSFTDPO",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/tsavage68/Summary4500_L3_1000steps_1e8rate_01beta_CSFTDPO

SGLang

How to use tsavage68/Summary4500_L3_1000steps_1e8rate_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_L3_1000steps_1e8rate_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_L3_1000steps_1e8rate_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_L3_1000steps_1e8rate_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_L3_1000steps_1e8rate_01beta_CSFTDPO",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use tsavage68/Summary4500_L3_1000steps_1e8rate_01beta_CSFTDPO with Docker Model Runner:
```
docker model run hf.co/tsavage68/Summary4500_L3_1000steps_1e8rate_01beta_CSFTDPO
```

Hyponatremia_L3_1000steps_1e8rate_01beta_CSFTDPO

This model is a fine-tuned version of tsavage68/Summary4500_L3_100steps_1e6rate_SFT on an unknown dataset. It achieves the following results on the evaluation set:

Loss: 0.6901
Rewards/chosen: 0.0028
Rewards/rejected: -0.0045
Rewards/accuracies: 0.5440
Rewards/margins: 0.0073
Logps/rejected: -133.2426
Logps/chosen: -84.1618
Logits/rejected: -1.0994
Logits/chosen: -1.0693

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-08
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
0.7202	0.0112	50	0.6920	0.0038	0.0005	0.5100	0.0034	-133.1928	-84.1516	-1.0987	-1.0683
0.6983	0.0224	100	0.6938	0.0029	0.0030	0.4940	-0.0001	-133.1671	-84.1607	-1.0980	-1.0678
0.6799	0.0336	150	0.6932	0.0057	0.0046	0.5060	0.0010	-133.1511	-84.1332	-1.0980	-1.0676
0.6921	0.0448	200	0.6896	0.0039	-0.0043	0.5800	0.0081	-133.2399	-84.1511	-1.0984	-1.0683
0.6904	0.0559	250	0.6923	0.0024	-0.0007	0.5280	0.0030	-133.2041	-84.1661	-1.0985	-1.0684
0.6725	0.0671	300	0.6877	0.0016	-0.0105	0.5980	0.0121	-133.3022	-84.1739	-1.0990	-1.0689
0.6848	0.0783	350	0.6888	0.0057	-0.0041	0.5500	0.0099	-133.2388	-84.1326	-1.0992	-1.0690
0.7158	0.0895	400	0.6916	0.0032	-0.0012	0.5400	0.0044	-133.2096	-84.1577	-1.0988	-1.0687
0.6992	0.1007	450	0.6912	0.0007	-0.0043	0.5260	0.0050	-133.2402	-84.1823	-1.0988	-1.0686
0.6827	0.1119	500	0.6885	0.0048	-0.0057	0.5600	0.0105	-133.2546	-84.1417	-1.0988	-1.0687
0.6949	0.1231	550	0.6903	0.0025	-0.0045	0.5440	0.0069	-133.2422	-84.1652	-1.0988	-1.0687
0.7093	0.1343	600	0.6915	0.0015	-0.0031	0.5300	0.0046	-133.2279	-84.1744	-1.0988	-1.0687
0.7026	0.1454	650	0.6894	0.0048	-0.0038	0.5480	0.0086	-133.2351	-84.1415	-1.0992	-1.0691
0.6781	0.1566	700	0.6896	0.0052	-0.0030	0.5400	0.0082	-133.2273	-84.1380	-1.0992	-1.0691
0.7174	0.1678	750	0.6888	0.0036	-0.0063	0.5780	0.0099	-133.2603	-84.1535	-1.0992	-1.0690
0.7065	0.1790	800	0.6895	0.0071	-0.0013	0.5580	0.0084	-133.2102	-84.1191	-1.0992	-1.0691
0.7018	0.1902	850	0.6904	0.0027	-0.0042	0.5280	0.0069	-133.2389	-84.1626	-1.0994	-1.0693
0.6894	0.2014	900	0.6901	0.0028	-0.0045	0.5440	0.0073	-133.2426	-84.1618	-1.0994	-1.0693
0.686	0.2126	950	0.6901	0.0028	-0.0045	0.5440	0.0073	-133.2426	-84.1618	-1.0994	-1.0693
0.6778	0.2238	1000	0.6901	0.0028	-0.0045	0.5440	0.0073	-133.2426	-84.1618	-1.0994	-1.0693

Framework versions

Transformers 4.42.4
Pytorch 2.0.0+cu117
Datasets 2.20.0
Tokenizers 0.19.1

Downloads last month: 6

Safetensors

Model size

8B params

Tensor type

F16

Model tree for tsavage68/Summary4500_L3_1000steps_1e8rate_01beta_CSFTDPO

Base model

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

Finetuned

tsavage68/Summary4500_L3_100steps_1e6rate_SFT

Finetuned

(14)

this model