Instructions to use tsavage68/Mistral2_1000_STEPS_03beta_1e6_CDPOSFT with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use tsavage68/Mistral2_1000_STEPS_03beta_1e6_CDPOSFT with Transformers:

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

pipe = pipeline("text-generation", model="tsavage68/Mistral2_1000_STEPS_03beta_1e6_CDPOSFT")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("tsavage68/Mistral2_1000_STEPS_03beta_1e6_CDPOSFT")
model = AutoModelForCausalLM.from_pretrained("tsavage68/Mistral2_1000_STEPS_03beta_1e6_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/Mistral2_1000_STEPS_03beta_1e6_CDPOSFT with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "tsavage68/Mistral2_1000_STEPS_03beta_1e6_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/Mistral2_1000_STEPS_03beta_1e6_CDPOSFT",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/tsavage68/Mistral2_1000_STEPS_03beta_1e6_CDPOSFT

SGLang

How to use tsavage68/Mistral2_1000_STEPS_03beta_1e6_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/Mistral2_1000_STEPS_03beta_1e6_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/Mistral2_1000_STEPS_03beta_1e6_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/Mistral2_1000_STEPS_03beta_1e6_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/Mistral2_1000_STEPS_03beta_1e6_CDPOSFT",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use tsavage68/Mistral2_1000_STEPS_03beta_1e6_CDPOSFT with Docker Model Runner:
```
docker model run hf.co/tsavage68/Mistral2_1000_STEPS_03beta_1e6_CDPOSFT
```

Mistral2_1000_STEPS_03beta_1e6_CDPOSFT

This model is a fine-tuned version of tsavage68/mistralit2_1000_STEPS_5e7_SFT on an unknown dataset. It achieves the following results on the evaluation set:

Loss: 0.9147
Rewards/chosen: 0.2853
Rewards/rejected: 0.2117
Rewards/accuracies: 0.4637
Rewards/margins: 0.0736
Logps/rejected: -76.8158
Logps/chosen: -74.5509
Logits/rejected: -1.8957
Logits/chosen: -1.8954

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: 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
0.831	0.0977	50	0.8025	0.6608	0.6344	0.4132	0.0264	-75.4068	-73.2992	-2.0277	-2.0274
0.6868	0.1953	100	0.9417	0.1823	0.1932	0.4198	-0.0109	-76.8774	-74.8943	-2.0626	-2.0624
1.1447	0.2930	150	1.0449	0.0804	0.1916	0.4000	-0.1112	-76.8828	-75.2339	-2.0660	-2.0660
1.0588	0.3906	200	1.0433	0.5444	0.5437	0.4176	0.0007	-75.7091	-73.6874	-1.8690	-1.8690
1.1749	0.4883	250	1.0509	0.0937	0.0800	0.3780	0.0138	-77.2550	-75.1895	-2.7221	-2.7221
0.9602	0.5859	300	1.0556	0.6428	0.6497	0.3978	-0.0069	-75.3558	-73.3592	-2.1885	-2.1884
0.8567	0.6836	350	1.0032	0.8514	0.9163	0.4022	-0.0649	-74.4671	-72.6639	-1.9197	-1.9197
0.8145	0.7812	400	0.9071	0.7911	0.7180	0.4549	0.0730	-75.1281	-72.8651	-2.1063	-2.1063
0.9805	0.8789	450	0.9092	1.0927	0.9910	0.4549	0.1017	-74.2182	-71.8597	-2.3062	-2.3062
0.8022	0.9766	500	0.8968	1.2157	1.1916	0.4396	0.0241	-73.5496	-71.4498	-1.9867	-1.9865
0.4835	1.0742	550	0.9087	0.3603	0.2979	0.4396	0.0624	-76.5285	-74.3010	-2.4092	-2.4089
0.7127	1.1719	600	0.9140	0.2002	0.1567	0.4374	0.0435	-76.9992	-74.8348	-2.1858	-2.1855
0.4928	1.2695	650	0.9377	0.3603	0.3349	0.4396	0.0253	-76.4051	-74.3011	-2.0564	-2.0560
0.5228	1.3672	700	0.9233	0.3468	0.2928	0.4462	0.0541	-76.5456	-74.3459	-1.8095	-1.8091
0.4985	1.4648	750	0.9155	0.3134	0.2441	0.4484	0.0693	-76.7079	-74.4573	-1.9045	-1.9041
0.5495	1.5625	800	0.9141	0.2956	0.2238	0.4593	0.0717	-76.7754	-74.5168	-1.8841	-1.8837
0.518	1.6602	850	0.9136	0.2853	0.2115	0.4637	0.0737	-76.8164	-74.5511	-1.8972	-1.8968
0.5009	1.7578	900	0.9149	0.2859	0.2124	0.4637	0.0735	-76.8134	-74.5489	-1.8954	-1.8950
0.4334	1.8555	950	0.9148	0.2846	0.2116	0.4659	0.0730	-76.8163	-74.5534	-1.8957	-1.8953
0.3651	1.9531	1000	0.9147	0.2853	0.2117	0.4637	0.0736	-76.8158	-74.5509	-1.8957	-1.8954

Framework versions

Transformers 4.40.1
Pytorch 2.0.0+cu117
Datasets 2.19.0
Tokenizers 0.19.1

Downloads last month: 6

Safetensors

Model size

7B params

Tensor type

F16

Model tree for tsavage68/Mistral2_1000_STEPS_03beta_1e6_CDPOSFT

Base model

mistralai/Mistral-7B-Instruct-v0.2

Finetuned

tsavage68/mistralit2_1000_STEPS_5e7_SFT

Finetuned

(19)

this model