Instructions to use tsavage68/chat_1000_STEPS_03beta_5e7rate_CDPOSFT with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use tsavage68/chat_1000_STEPS_03beta_5e7rate_CDPOSFT with Transformers:

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

pipe = pipeline("text-generation", model="tsavage68/chat_1000_STEPS_03beta_5e7rate_CDPOSFT")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("tsavage68/chat_1000_STEPS_03beta_5e7rate_CDPOSFT")
model = AutoModelForCausalLM.from_pretrained("tsavage68/chat_1000_STEPS_03beta_5e7rate_CDPOSFT", 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/chat_1000_STEPS_03beta_5e7rate_CDPOSFT with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "tsavage68/chat_1000_STEPS_03beta_5e7rate_CDPOSFT"
# 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/chat_1000_STEPS_03beta_5e7rate_CDPOSFT",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/tsavage68/chat_1000_STEPS_03beta_5e7rate_CDPOSFT

SGLang

How to use tsavage68/chat_1000_STEPS_03beta_5e7rate_CDPOSFT 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/chat_1000_STEPS_03beta_5e7rate_CDPOSFT" \
    --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/chat_1000_STEPS_03beta_5e7rate_CDPOSFT",
		"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/chat_1000_STEPS_03beta_5e7rate_CDPOSFT" \
        --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/chat_1000_STEPS_03beta_5e7rate_CDPOSFT",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use tsavage68/chat_1000_STEPS_03beta_5e7rate_CDPOSFT with Docker Model Runner:
```
docker model run hf.co/tsavage68/chat_1000_STEPS_03beta_5e7rate_CDPOSFT
```

chat_1000_STEPS_03beta_5e7rate_CDPOSFT

This model is a fine-tuned version of tsavage68/chat_600STEPS_1e8rate_SFT on an unknown dataset. It achieves the following results on the evaluation set:

Loss: 0.6933
Rewards/chosen: 0.0002
Rewards/rejected: -0.0002
Rewards/accuracies: 0.4242
Rewards/margins: 0.0004
Logps/rejected: -18.8025
Logps/chosen: -16.7538
Logits/rejected: -0.5979
Logits/chosen: -0.5978

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: 4
eval_batch_size: 1
seed: 42
gradient_accumulation_steps: 2
total_train_batch_size: 8
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
0.6928	0.0977	50	0.6947	0.0015	0.0040	0.4264	-0.0024	-18.7888	-16.7496	-0.5974	-0.5972
0.6895	0.1953	100	0.6950	-0.0009	0.0022	0.3890	-0.0030	-18.7948	-16.7576	-0.5980	-0.5979
0.6932	0.2930	150	0.6934	0.0018	0.0016	0.4308	0.0002	-18.7966	-16.7487	-0.5984	-0.5983
0.695	0.3906	200	0.6947	0.0002	0.0027	0.3868	-0.0025	-18.7930	-16.7541	-0.5982	-0.5981
0.6971	0.4883	250	0.6938	0.0028	0.0034	0.4220	-0.0006	-18.7907	-16.7453	-0.5980	-0.5979
0.6891	0.5859	300	0.6934	-0.0003	-0.0005	0.4396	0.0002	-18.8036	-16.7557	-0.5988	-0.5987
0.6872	0.6836	350	0.6939	0.0002	0.0010	0.4088	-0.0009	-18.7985	-16.7541	-0.5983	-0.5982
0.6953	0.7812	400	0.6964	-0.0050	0.0010	0.3846	-0.0060	-18.7985	-16.7713	-0.5980	-0.5979
0.689	0.8789	450	0.6949	0.0002	0.0031	0.4088	-0.0029	-18.7915	-16.7539	-0.5983	-0.5982
0.6935	0.9766	500	0.6920	0.0025	-0.0003	0.4154	0.0028	-18.8029	-16.7461	-0.5985	-0.5983
0.6931	1.0742	550	0.6931	0.0014	0.0007	0.4220	0.0007	-18.7996	-16.7500	-0.5979	-0.5978
0.6919	1.1719	600	0.6951	-0.0001	0.0032	0.3780	-0.0032	-18.7914	-16.7548	-0.5982	-0.5980
0.6916	1.2695	650	0.6930	0.0027	0.0019	0.4110	0.0009	-18.7958	-16.7455	-0.5982	-0.5980
0.6914	1.3672	700	0.6936	0.0017	0.0020	0.4198	-0.0003	-18.7954	-16.7489	-0.5975	-0.5974
0.6937	1.4648	750	0.6937	0.0001	0.0006	0.4352	-0.0005	-18.7999	-16.7542	-0.5977	-0.5976
0.6953	1.5625	800	0.6948	-0.0003	0.0025	0.3934	-0.0028	-18.7936	-16.7556	-0.5977	-0.5976
0.6914	1.6602	850	0.6932	0.0004	-0.0002	0.4264	0.0006	-18.8026	-16.7534	-0.5979	-0.5978
0.6936	1.7578	900	0.6933	0.0002	-0.0002	0.4242	0.0004	-18.8025	-16.7538	-0.5979	-0.5978
0.6925	1.8555	950	0.6933	0.0002	-0.0002	0.4242	0.0004	-18.8025	-16.7538	-0.5979	-0.5978
0.6919	1.9531	1000	0.6933	0.0002	-0.0002	0.4242	0.0004	-18.8025	-16.7538	-0.5979	-0.5978

Framework versions

Transformers 4.40.1
Pytorch 2.0.0+cu117
Datasets 2.19.1
Tokenizers 0.19.1

Downloads last month: 6

Safetensors

Model size

7B params

Tensor type

F16

Model tree for tsavage68/chat_1000_STEPS_03beta_5e7rate_CDPOSFT

Base model

meta-llama/Llama-2-7b-chat-hf

Finetuned

tsavage68/chat_600STEPS_1e8rate_SFT

Finetuned

(23)

this model