Instructions to use tsavage68/Summary_L3_1000steps_1e6rate_03beta_CSFTDPO with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use tsavage68/Summary_L3_1000steps_1e6rate_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_1e6rate_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_1e6rate_03beta_CSFTDPO")
model = AutoModelForCausalLM.from_pretrained("tsavage68/Summary_L3_1000steps_1e6rate_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_1e6rate_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_1e6rate_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_1e6rate_03beta_CSFTDPO",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/tsavage68/Summary_L3_1000steps_1e6rate_03beta_CSFTDPO

SGLang

How to use tsavage68/Summary_L3_1000steps_1e6rate_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_1e6rate_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_1e6rate_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_1e6rate_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_1e6rate_03beta_CSFTDPO",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use tsavage68/Summary_L3_1000steps_1e6rate_03beta_CSFTDPO with Docker Model Runner:
```
docker model run hf.co/tsavage68/Summary_L3_1000steps_1e6rate_03beta_CSFTDPO
```

Summary_L3_1000steps_1e6rate_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: 0.5961
Rewards/chosen: 0.0294
Rewards/rejected: -2.5656
Rewards/accuracies: 0.1400
Rewards/margins: 2.5950
Logps/rejected: -23.8158
Logps/chosen: -9.2849
Logits/rejected: -1.1435
Logits/chosen: -1.1436

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: 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
0.5553	0.2004	50	0.5962	0.0778	-1.2696	0.1400	1.3473	-19.4956	-9.1236	-1.1038	-1.1053
0.6585	0.4008	100	0.5962	0.0854	-1.4439	0.1400	1.5292	-20.0766	-9.0982	-1.1078	-1.1092
0.6238	0.6012	150	0.5961	0.0687	-2.1556	0.1400	2.2243	-22.4490	-9.1538	-1.1298	-1.1306
0.6065	0.8016	200	0.5961	0.0322	-2.5726	0.1400	2.6048	-23.8390	-9.2754	-1.1437	-1.1438
0.6238	1.0020	250	0.5961	0.0294	-2.5678	0.1400	2.5971	-23.8230	-9.2849	-1.1438	-1.1440
0.6238	1.2024	300	0.5961	0.0279	-2.5674	0.1400	2.5953	-23.8219	-9.2899	-1.1439	-1.1440
0.6238	1.4028	350	0.5961	0.0304	-2.5648	0.1400	2.5952	-23.8131	-9.2814	-1.1438	-1.1439
0.5718	1.6032	400	0.5961	0.0304	-2.5648	0.1400	2.5952	-23.8131	-9.2814	-1.1438	-1.1439
0.5892	1.8036	450	0.5961	0.0338	-2.5715	0.1400	2.6052	-23.8353	-9.2702	-1.1435	-1.1436
0.5718	2.0040	500	0.5961	0.0279	-2.5720	0.1400	2.5999	-23.8372	-9.2897	-1.1434	-1.1435
0.5718	2.2044	550	0.5961	0.0266	-2.5750	0.1400	2.6016	-23.8472	-9.2942	-1.1438	-1.1440
0.5545	2.4048	600	0.5961	0.0271	-2.5761	0.1400	2.6032	-23.8507	-9.2925	-1.1438	-1.1440
0.5199	2.6052	650	0.5961	0.0271	-2.5761	0.1400	2.6032	-23.8507	-9.2925	-1.1438	-1.1440
0.6238	2.8056	700	0.5961	0.0270	-2.5764	0.1400	2.6035	-23.8519	-9.2928	-1.1438	-1.1440
0.6065	3.0060	750	0.5961	0.0315	-2.5674	0.1400	2.5989	-23.8216	-9.2777	-1.1434	-1.1436
0.6412	3.2064	800	0.5961	0.0276	-2.5662	0.1400	2.5937	-23.8176	-9.2909	-1.1434	-1.1436
0.6585	3.4068	850	0.5961	0.0277	-2.5666	0.1400	2.5943	-23.8191	-9.2903	-1.1434	-1.1436
0.6238	3.6072	900	0.5961	0.0281	-2.5670	0.1400	2.5952	-23.8205	-9.2891	-1.1434	-1.1436
0.5372	3.8076	950	0.5961	0.0310	-2.5656	0.1400	2.5966	-23.8159	-9.2795	-1.1435	-1.1436
0.6238	4.0080	1000	0.5961	0.0294	-2.5656	0.1400	2.5950	-23.8158	-9.2849	-1.1435	-1.1436

Framework versions

Transformers 4.41.2
Pytorch 2.0.0+cu117
Datasets 2.20.0
Tokenizers 0.19.1

Downloads last month: 5

Safetensors

Model size

8B params

Tensor type

F16

Model tree for tsavage68/Summary_L3_1000steps_1e6rate_03beta_CSFTDPO

Base model

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

Finetuned

tsavage68/Summary_L3_1000steps_1e7rate_SFT2

Finetuned

(17)

this model