Instructions to use tsavage68/Transaminitis_L3_1000steps_1e8rate_01beta_CSFTDPO with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use tsavage68/Transaminitis_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/Transaminitis_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/Transaminitis_L3_1000steps_1e8rate_01beta_CSFTDPO")
model = AutoModelForCausalLM.from_pretrained("tsavage68/Transaminitis_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/Transaminitis_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/Transaminitis_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/Transaminitis_L3_1000steps_1e8rate_01beta_CSFTDPO",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/tsavage68/Transaminitis_L3_1000steps_1e8rate_01beta_CSFTDPO

SGLang

How to use tsavage68/Transaminitis_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/Transaminitis_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/Transaminitis_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/Transaminitis_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/Transaminitis_L3_1000steps_1e8rate_01beta_CSFTDPO",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use tsavage68/Transaminitis_L3_1000steps_1e8rate_01beta_CSFTDPO with Docker Model Runner:
```
docker model run hf.co/tsavage68/Transaminitis_L3_1000steps_1e8rate_01beta_CSFTDPO
```

Transaminitis_L3_1000steps_1e8rate_01beta_DPO

This model is a fine-tuned version of tsavage68/Transaminitis_L3_1000rate_1e7_SFT on an unknown dataset. It achieves the following results on the evaluation set:

Loss: 0.6939
Rewards/chosen: 0.0011
Rewards/rejected: 0.0026
Rewards/accuracies: 0.4100
Rewards/margins: -0.0014
Logps/rejected: -18.5291
Logps/chosen: -18.5229
Logits/rejected: -1.0656
Logits/chosen: -1.0644

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: 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: 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
0.6937	0.2	25	0.6931	0.0001	0.0001	0.0100	0.0000	-18.5542	-18.5333	-1.0657	-1.0646
0.6937	0.4	50	0.6931	0.0014	0.0012	0.5400	0.0002	-18.5426	-18.5205	-1.0657	-1.0645
0.6937	0.6	75	0.6938	0.0004	0.0017	0.4600	-0.0013	-18.5374	-18.5302	-1.0653	-1.0643
0.6941	0.8	100	0.6929	0.0003	-0.0003	0.5	0.0006	-18.5573	-18.5312	-1.0667	-1.0656
0.6922	1.0	125	0.6934	0.0022	0.0026	0.4800	-0.0004	-18.5288	-18.5123	-1.0666	-1.0654
0.6945	1.2	150	0.6937	0.0009	0.0020	0.4500	-0.0011	-18.5347	-18.5251	-1.0648	-1.0637
0.6934	1.4	175	0.6927	0.0058	0.0049	0.5600	0.0010	-18.5061	-18.4759	-1.0650	-1.0639
0.6934	1.6	200	0.6937	0.0009	0.0021	0.4200	-0.0011	-18.5342	-18.5251	-1.0652	-1.0640
0.6953	1.8	225	0.6935	-0.0007	-0.0002	0.4700	-0.0006	-18.5563	-18.5415	-1.0650	-1.0638
0.6906	2.0	250	0.6935	0.0008	0.0014	0.4900	-0.0006	-18.5411	-18.5264	-1.0657	-1.0645
0.693	2.2	275	0.6935	0.0028	0.0035	0.5100	-0.0007	-18.5196	-18.5059	-1.0662	-1.0650
0.6945	2.4	300	0.6934	0.0013	0.0018	0.5300	-0.0005	-18.5368	-18.5211	-1.0658	-1.0646
0.6934	2.6	325	0.6933	0.0002	0.0005	0.5	-0.0002	-18.5500	-18.5320	-1.0657	-1.0646
0.6914	2.8	350	0.6933	-0.0038	-0.0036	0.4900	-0.0003	-18.5903	-18.5727	-1.0655	-1.0643
0.6914	3.0	375	0.6935	0.0004	0.0011	0.4900	-0.0007	-18.5435	-18.5301	-1.0665	-1.0654
0.6914	3.2	400	0.6927	0.0048	0.0038	0.4900	0.0009	-18.5165	-18.4865	-1.0655	-1.0643
0.6949	3.4	425	0.6933	0.0020	0.0023	0.4900	-0.0003	-18.5321	-18.5146	-1.0660	-1.0649
0.6922	3.6	450	0.6937	-0.0020	-0.0009	0.5	-0.0011	-18.5634	-18.5540	-1.0653	-1.0642
0.6926	3.8	475	0.6927	0.0040	0.0030	0.4800	0.0010	-18.5242	-18.4937	-1.0656	-1.0645
0.693	4.0	500	0.6942	0.0022	0.0042	0.4400	-0.0020	-18.5124	-18.5118	-1.0658	-1.0646
0.693	4.2	525	0.6932	0.0030	0.0031	0.4500	-0.0000	-18.5239	-18.5038	-1.0662	-1.0649
0.6922	4.4	550	0.6936	0.0028	0.0036	0.5100	-0.0009	-18.5182	-18.5066	-1.0651	-1.0640
0.6934	4.6	575	0.6938	0.0014	0.0027	0.4800	-0.0013	-18.5278	-18.5202	-1.0656	-1.0645
0.6937	4.8	600	0.6941	0.0023	0.0041	0.4500	-0.0019	-18.5132	-18.5113	-1.0653	-1.0642
0.691	5.0	625	0.6936	0.0024	0.0033	0.5100	-0.0009	-18.5219	-18.5103	-1.0654	-1.0642
0.6926	5.2	650	0.6942	0.0006	0.0027	0.4100	-0.0021	-18.5279	-18.5280	-1.0655	-1.0643
0.6953	5.4	675	0.6938	0.0027	0.0040	0.4400	-0.0013	-18.5149	-18.5071	-1.0656	-1.0645
0.6937	5.6	700	0.6930	0.0042	0.0038	0.5	0.0004	-18.5169	-18.4921	-1.0657	-1.0645
0.693	5.8	725	0.6935	0.0022	0.0027	0.4600	-0.0006	-18.5272	-18.5127	-1.0656	-1.0644
0.6937	6.0	750	0.6935	0.0014	0.0022	0.4400	-0.0008	-18.5327	-18.5198	-1.0656	-1.0645
0.6918	6.2	775	0.6936	0.0017	0.0024	0.4300	-0.0008	-18.5303	-18.5175	-1.0655	-1.0644
0.6934	6.4	800	0.6938	0.0008	0.0021	0.4200	-0.0013	-18.5333	-18.5261	-1.0655	-1.0644
0.6902	6.6	825	0.6939	0.0011	0.0026	0.4100	-0.0014	-18.5291	-18.5229	-1.0656	-1.0644
0.6937	6.8	850	0.6939	0.0011	0.0026	0.4100	-0.0014	-18.5291	-18.5229	-1.0656	-1.0644
0.6949	7.0	875	0.6939	0.0011	0.0026	0.4100	-0.0014	-18.5291	-18.5229	-1.0656	-1.0644
0.693	7.2	900	0.6939	0.0011	0.0026	0.4100	-0.0014	-18.5291	-18.5229	-1.0656	-1.0644
0.6941	7.4	925	0.6939	0.0011	0.0026	0.4100	-0.0014	-18.5291	-18.5229	-1.0656	-1.0644
0.6937	7.6	950	0.6939	0.0011	0.0026	0.4100	-0.0014	-18.5291	-18.5229	-1.0656	-1.0644
0.6926	7.8	975	0.6939	0.0011	0.0026	0.4100	-0.0014	-18.5291	-18.5229	-1.0656	-1.0644
0.6918	8.0	1000	0.6939	0.0011	0.0026	0.4100	-0.0014	-18.5291	-18.5229	-1.0656	-1.0644

Framework versions

Transformers 4.40.2
Pytorch 2.0.0+cu117
Datasets 2.19.1
Tokenizers 0.19.1

Downloads last month: 6

Safetensors

Model size

8B params

Tensor type

F16

Model tree for tsavage68/Transaminitis_L3_1000steps_1e8rate_01beta_CSFTDPO

Base model

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

Finetuned

tsavage68/Transaminitis_L3_1000rate_1e7_SFT

Finetuned

(17)

this model