Instructions to use tsavage68/Mistral2_1000_STEPS_03beta_1e5rate_CDPOSFT with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use tsavage68/Mistral2_1000_STEPS_03beta_1e5rate_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_1e5rate_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_1e5rate_CDPOSFT")
model = AutoModelForCausalLM.from_pretrained("tsavage68/Mistral2_1000_STEPS_03beta_1e5rate_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_1e5rate_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_1e5rate_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_1e5rate_CDPOSFT",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/tsavage68/Mistral2_1000_STEPS_03beta_1e5rate_CDPOSFT

SGLang

How to use tsavage68/Mistral2_1000_STEPS_03beta_1e5rate_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_1e5rate_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_1e5rate_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_1e5rate_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_1e5rate_CDPOSFT",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use tsavage68/Mistral2_1000_STEPS_03beta_1e5rate_CDPOSFT with Docker Model Runner:
```
docker model run hf.co/tsavage68/Mistral2_1000_STEPS_03beta_1e5rate_CDPOSFT
```

Mistral2_1000_STEPS_03beta_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: 1.1718
Rewards/chosen: 0.7581
Rewards/rejected: 0.6156
Rewards/accuracies: 0.4396
Rewards/margins: 0.1425
Logps/rejected: -75.4695
Logps/chosen: -72.9748
Logits/rejected: -2.0639
Logits/chosen: -2.0639

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-05
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
1.4289	0.0977	50	1.3959	-0.9593	-1.0924	0.3890	0.1331	-81.1629	-78.6996	-2.1377	-2.1377
1.2593	0.1953	100	1.2029	0.0059	-0.1866	0.4198	0.1925	-78.1435	-75.4823	-1.0724	-1.0724
1.4537	0.2930	150	1.1666	0.7872	0.6336	0.4220	0.1536	-75.4096	-72.8781	-1.3308	-1.3308
1.0065	0.3906	200	1.2087	1.4140	1.2976	0.4352	0.1165	-73.1963	-70.7886	-2.3405	-2.3405
1.0856	0.4883	250	1.1566	0.5136	0.2937	0.4374	0.2200	-76.5426	-73.7899	-1.8531	-1.8531
1.3114	0.5859	300	1.1393	0.8895	0.7419	0.4374	0.1476	-75.0486	-72.5371	-1.6986	-1.6986
1.0439	0.6836	350	1.1353	0.8733	0.7520	0.4220	0.1213	-75.0148	-72.5910	-1.9854	-1.9854
1.066	0.7812	400	1.1047	0.6140	0.4163	0.4527	0.1976	-76.1338	-73.4554	-2.0072	-2.0072
1.379	0.8789	450	1.1317	0.9511	0.7423	0.4462	0.2088	-75.0471	-72.3316	-2.0669	-2.0669
1.0806	0.9766	500	1.1449	1.1211	0.9721	0.4418	0.1490	-74.2810	-71.7648	-1.9015	-1.9015
0.9227	1.0742	550	1.2298	0.2455	0.1098	0.4242	0.1358	-77.1555	-74.6835	-2.7262	-2.7262
1.2	1.1719	600	1.2007	0.6867	0.5585	0.4220	0.1282	-75.6600	-73.2131	-1.8276	-1.8276
0.8618	1.2695	650	1.1847	0.8411	0.7278	0.4396	0.1133	-75.0956	-72.6984	-1.9583	-1.9583
0.8477	1.3672	700	1.1792	0.9573	0.8415	0.4418	0.1158	-74.7164	-72.3109	-2.0044	-2.0044
0.645	1.4648	750	1.1888	0.6965	0.5656	0.4286	0.1309	-75.6363	-73.1804	-2.1525	-2.1525
0.7683	1.5625	800	1.1819	0.8138	0.6780	0.4330	0.1358	-75.2613	-72.7893	-2.1381	-2.1381
0.937	1.6602	850	1.1725	0.7651	0.6249	0.4374	0.1402	-75.4385	-72.9517	-2.0730	-2.0730
0.9381	1.7578	900	1.1724	0.7620	0.6205	0.4396	0.1414	-75.4530	-72.9621	-2.0649	-2.0649
0.8179	1.8555	950	1.1720	0.7582	0.6157	0.4396	0.1424	-75.4691	-72.9747	-2.0642	-2.0642
0.6211	1.9531	1000	1.1718	0.7581	0.6156	0.4396	0.1425	-75.4695	-72.9748	-2.0639	-2.0639

Framework versions

Transformers 4.40.1
Pytorch 2.0.0+cu117
Datasets 2.19.0
Tokenizers 0.19.1

Downloads last month: 11

Safetensors

Model size

7B params

Tensor type

F16

Model tree for tsavage68/Mistral2_1000_STEPS_03beta_1e5rate_CDPOSFT

Base model

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

Finetuned

tsavage68/mistralit2_1000_STEPS_5e7_SFT

Finetuned

(19)

this model