Instructions to use tsavage68/Transaminitis_L3_400steps_1e7rate_01beta_CSFTDPO with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use tsavage68/Transaminitis_L3_400steps_1e7rate_01beta_CSFTDPO with Transformers:

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

pipe = pipeline("text-generation", model="tsavage68/Transaminitis_L3_400steps_1e7rate_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_400steps_1e7rate_01beta_CSFTDPO")
model = AutoModelForCausalLM.from_pretrained("tsavage68/Transaminitis_L3_400steps_1e7rate_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_400steps_1e7rate_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_400steps_1e7rate_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_400steps_1e7rate_01beta_CSFTDPO",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/tsavage68/Transaminitis_L3_400steps_1e7rate_01beta_CSFTDPO

SGLang

How to use tsavage68/Transaminitis_L3_400steps_1e7rate_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_400steps_1e7rate_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_400steps_1e7rate_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_400steps_1e7rate_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_400steps_1e7rate_01beta_CSFTDPO",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use tsavage68/Transaminitis_L3_400steps_1e7rate_01beta_CSFTDPO with Docker Model Runner:
```
docker model run hf.co/tsavage68/Transaminitis_L3_400steps_1e7rate_01beta_CSFTDPO
```

Transaminitis_L3_400steps_1e7rate_01beta_CSFTDPO

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: 0.6489
Rewards/chosen: 0.0765
Rewards/rejected: -0.0162
Rewards/accuracies: 0.8400
Rewards/margins: 0.0927
Logps/rejected: -18.7162
Logps/chosen: -17.7692
Logits/rejected: -1.0743
Logits/chosen: -1.0727

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-07
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: 400

Training results

Training Loss	Epoch	Step	Validation Loss	Rewards/chosen	Rewards/rejected	Rewards/accuracies	Rewards/margins	Logps/rejected	Logps/chosen	Logits/rejected	Logits/chosen
0.6925	0.2	25	0.6925	-0.0045	-0.0060	0.5500	0.0014	-18.6144	-18.5795	-1.0662	-1.0649
0.6933	0.4	50	0.6925	-0.0166	-0.0181	0.5100	0.0014	-18.7354	-18.7005	-1.0667	-1.0654
0.6915	0.6	75	0.6945	-0.0053	-0.0039	0.4600	-0.0015	-18.5932	-18.5873	-1.0676	-1.0664
0.6761	0.8	100	0.7028	-0.0339	-0.0261	0.4600	-0.0078	-18.8159	-18.8731	-1.0697	-1.0686
0.6884	1.0	125	0.6833	-0.1592	-0.1851	0.5400	0.0259	-20.4055	-20.1260	-1.0741	-1.0726
0.6858	1.2	150	0.6876	0.0184	0.0033	0.4600	0.0152	-18.5221	-18.3500	-1.0698	-1.0685
0.6692	1.4	175	0.6783	-0.0157	-0.0487	0.5200	0.0330	-19.0418	-18.6910	-1.0726	-1.0713
0.6751	1.6	200	0.6672	-0.0238	-0.0778	0.7800	0.0540	-19.3325	-18.7721	-1.0743	-1.0729
0.6668	1.8	225	0.6613	0.0261	-0.0398	0.8400	0.0659	-18.9525	-18.2729	-1.0735	-1.0721
0.6502	2.0	250	0.6564	0.0453	-0.0311	0.8100	0.0764	-18.8662	-18.0815	-1.0744	-1.0729
0.6583	2.2	275	0.6520	0.0709	-0.0148	0.8500	0.0857	-18.7031	-17.8256	-1.0737	-1.0721
0.6453	2.4	300	0.6530	0.0766	-0.0072	0.8300	0.0837	-18.6263	-17.7687	-1.0741	-1.0724
0.6424	2.6	325	0.6489	0.0764	-0.0163	0.8400	0.0927	-18.7181	-17.7702	-1.0743	-1.0726
0.6404	2.8	350	0.6501	0.0702	-0.0199	0.8500	0.0901	-18.7534	-17.8323	-1.0742	-1.0725
0.6484	3.0	375	0.6489	0.0765	-0.0161	0.8400	0.0927	-18.7159	-17.7688	-1.0744	-1.0728
0.6539	3.2	400	0.6489	0.0765	-0.0162	0.8400	0.0927	-18.7162	-17.7692	-1.0743	-1.0727

Framework versions

Transformers 4.40.2
Pytorch 2.0.0+cu117
Datasets 2.19.1
Tokenizers 0.19.1

Downloads last month: 9

Safetensors

Model size

8B params

Tensor type

F16

Model tree for tsavage68/Transaminitis_L3_400steps_1e7rate_01beta_CSFTDPO

Base model

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

Finetuned

tsavage68/Transaminitis_L3_1000rate_1e7_SFT

Finetuned

(17)

this model