Instructions to use tsavage68/Summary_L3_1000steps_1e8rate_03beta_CSFTDPO with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use tsavage68/Summary_L3_1000steps_1e8rate_03beta_CSFTDPO with Transformers:

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

pipe = pipeline("text-generation", model="tsavage68/Summary_L3_1000steps_1e8rate_03beta_CSFTDPO")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("tsavage68/Summary_L3_1000steps_1e8rate_03beta_CSFTDPO")
model = AutoModelForCausalLM.from_pretrained("tsavage68/Summary_L3_1000steps_1e8rate_03beta_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/Summary_L3_1000steps_1e8rate_03beta_CSFTDPO with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "tsavage68/Summary_L3_1000steps_1e8rate_03beta_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/Summary_L3_1000steps_1e8rate_03beta_CSFTDPO",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/tsavage68/Summary_L3_1000steps_1e8rate_03beta_CSFTDPO

SGLang

How to use tsavage68/Summary_L3_1000steps_1e8rate_03beta_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/Summary_L3_1000steps_1e8rate_03beta_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/Summary_L3_1000steps_1e8rate_03beta_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/Summary_L3_1000steps_1e8rate_03beta_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/Summary_L3_1000steps_1e8rate_03beta_CSFTDPO",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use tsavage68/Summary_L3_1000steps_1e8rate_03beta_CSFTDPO with Docker Model Runner:
```
docker model run hf.co/tsavage68/Summary_L3_1000steps_1e8rate_03beta_CSFTDPO
```

Summary_L3_1000steps_1e8rate_03beta_CSFTDPO

This model is a fine-tuned version of tsavage68/Summary_L3_1000steps_1e7rate_SFT2 on an unknown dataset. It achieves the following results on the evaluation set:

Loss: 0.6919
Rewards/chosen: -0.0023
Rewards/rejected: -0.0059
Rewards/accuracies: 0.0650
Rewards/margins: 0.0036
Logps/rejected: -15.2835
Logps/chosen: -9.3904
Logits/rejected: -1.0962
Logits/chosen: -1.0977

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
gradient_accumulation_steps: 4
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: 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
0.6866	0.2004	50	0.6914	-0.0024	-0.0068	0.0750	0.0044	-15.2865	-9.3909	-1.0958	-1.0972
0.6966	0.4008	100	0.6896	0.0031	-0.0051	0.0850	0.0082	-15.2806	-9.3724	-1.0965	-1.0979
0.6924	0.6012	150	0.6911	-0.0000	-0.0053	0.0850	0.0053	-15.2813	-9.3828	-1.0957	-1.0972
0.6908	0.8016	200	0.6901	0.0009	-0.0058	0.0900	0.0066	-15.2830	-9.3799	-1.0957	-1.0971
0.6922	1.0020	250	0.6889	0.0008	-0.0086	0.0950	0.0094	-15.2923	-9.3800	-1.0959	-1.0974
0.6944	1.2024	300	0.6906	-0.0011	-0.0069	0.0900	0.0058	-15.2869	-9.3865	-1.0957	-1.0971
0.6919	1.4028	350	0.6878	0.0019	-0.0099	0.0900	0.0117	-15.2966	-9.3766	-1.0961	-1.0975
0.6937	1.6032	400	0.6879	0.0049	-0.0067	0.0900	0.0116	-15.2860	-9.3664	-1.0963	-1.0977
0.6927	1.8036	450	0.6903	0.0001	-0.0065	0.0850	0.0066	-15.2854	-9.3824	-1.0962	-1.0977
0.6917	2.0040	500	0.6922	-0.0002	-0.0030	0.0700	0.0028	-15.2739	-9.3835	-1.0959	-1.0973
0.6983	2.2044	550	0.6911	-0.0014	-0.0068	0.0750	0.0053	-15.2863	-9.3875	-1.0960	-1.0974
0.6901	2.4048	600	0.6902	0.0002	-0.0065	0.0900	0.0067	-15.2854	-9.3820	-1.0967	-1.0982
0.6859	2.6052	650	0.6890	0.0027	-0.0066	0.0950	0.0093	-15.2858	-9.3738	-1.0964	-1.0978
0.694	2.8056	700	0.6910	0.0002	-0.0048	0.0850	0.0050	-15.2799	-9.3823	-1.0963	-1.0978
0.6909	3.0060	750	0.6936	-0.0027	-0.0025	0.0600	-0.0002	-15.2720	-9.3918	-1.0964	-1.0978
0.6909	3.2064	800	0.6912	-0.0017	-0.0065	0.0650	0.0049	-15.2855	-9.3883	-1.0963	-1.0977
0.6929	3.4068	850	0.6914	-0.0008	-0.0054	0.0800	0.0047	-15.2819	-9.3853	-1.0962	-1.0976
0.6938	3.6072	900	0.6919	-0.0023	-0.0059	0.0650	0.0036	-15.2835	-9.3904	-1.0962	-1.0977
0.69	3.8076	950	0.6919	-0.0023	-0.0059	0.0650	0.0036	-15.2835	-9.3904	-1.0962	-1.0977
0.6968	4.0080	1000	0.6919	-0.0023	-0.0059	0.0650	0.0036	-15.2835	-9.3904	-1.0962	-1.0977

Framework versions

Transformers 4.41.2
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/Summary_L3_1000steps_1e8rate_03beta_CSFTDPO

Base model

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

Finetuned

tsavage68/Summary_L3_1000steps_1e7rate_SFT2

Finetuned

(17)

this model