Instructions to use tsavage68/Mistral2_1000_STEPS_05beta_1e6rate_CDPOSFT with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use tsavage68/Mistral2_1000_STEPS_05beta_1e6rate_CDPOSFT with Transformers:

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

pipe = pipeline("text-generation", model="tsavage68/Mistral2_1000_STEPS_05beta_1e6rate_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_05beta_1e6rate_CDPOSFT")
model = AutoModelForCausalLM.from_pretrained("tsavage68/Mistral2_1000_STEPS_05beta_1e6rate_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_05beta_1e6rate_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_05beta_1e6rate_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_05beta_1e6rate_CDPOSFT",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/tsavage68/Mistral2_1000_STEPS_05beta_1e6rate_CDPOSFT

SGLang

How to use tsavage68/Mistral2_1000_STEPS_05beta_1e6rate_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_05beta_1e6rate_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_05beta_1e6rate_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_05beta_1e6rate_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_05beta_1e6rate_CDPOSFT",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use tsavage68/Mistral2_1000_STEPS_05beta_1e6rate_CDPOSFT with Docker Model Runner:
```
docker model run hf.co/tsavage68/Mistral2_1000_STEPS_05beta_1e6rate_CDPOSFT
```

Mistral2_1000_STEPS_05beta_1e6rate_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: 1.1230
Rewards/chosen: 0.9764
Rewards/rejected: 0.9267
Rewards/accuracies: 0.4396
Rewards/margins: 0.0496
Logps/rejected: -75.6680
Logps/chosen: -73.5492
Logits/rejected: -2.0157
Logits/chosen: -2.0155

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
0.9416	0.0977	50	0.9097	1.0706	1.0514	0.4022	0.0192	-75.4186	-73.3607	-2.0307	-2.0304
0.8274	0.1953	100	1.2209	-0.4686	-0.4288	0.4418	-0.0397	-78.3791	-76.4391	-2.1855	-2.1854
1.1515	0.2930	150	1.2754	0.1047	0.1835	0.4000	-0.0788	-77.1544	-75.2925	-2.4964	-2.4962
1.3182	0.3906	200	1.2799	1.5118	1.4717	0.4527	0.0401	-74.5782	-72.4784	-1.8645	-1.8645
1.4196	0.4883	250	1.4528	0.0317	-0.0604	0.4286	0.0921	-77.6422	-75.4385	-2.5791	-2.5791
1.1384	0.5859	300	1.4278	-0.7306	-0.7573	0.4044	0.0267	-79.0362	-76.9632	-2.2124	-2.2123
1.0232	0.6836	350	1.2693	1.9620	2.0213	0.4022	-0.0593	-73.4789	-71.5779	-2.2117	-2.2116
1.0455	0.7812	400	1.1135	1.1771	1.0365	0.4374	0.1406	-75.4484	-73.1478	-2.0575	-2.0573
1.2782	0.8789	450	1.1233	1.9456	1.7789	0.4747	0.1667	-73.9637	-71.6107	-2.1983	-2.1983
0.9369	0.9766	500	1.1599	1.7149	1.6877	0.4220	0.0272	-74.1460	-72.0722	-1.9723	-1.9721
0.6002	1.0742	550	1.1711	1.0939	1.0050	0.4462	0.0889	-75.5115	-73.3143	-2.2719	-2.2717
0.9032	1.1719	600	1.1609	1.0215	1.0266	0.4396	-0.0051	-75.4683	-73.4589	-2.2396	-2.2394
0.5823	1.2695	650	1.1563	1.0877	1.0876	0.4330	0.0001	-75.3463	-73.3266	-2.1722	-2.1720
0.5362	1.3672	700	1.1249	1.0802	1.0424	0.4440	0.0377	-75.4367	-73.3417	-1.9807	-1.9804
0.4842	1.4648	750	1.1291	1.0041	0.9592	0.4352	0.0449	-75.6030	-73.4937	-2.0308	-2.0306
0.5886	1.5625	800	1.1233	0.9816	0.9342	0.4330	0.0474	-75.6531	-73.5387	-2.0136	-2.0134
0.5679	1.6602	850	1.1214	0.9789	0.9292	0.4374	0.0497	-75.6630	-73.5441	-2.0189	-2.0187
0.489	1.7578	900	1.1233	0.9752	0.9268	0.4396	0.0484	-75.6680	-73.5516	-2.0159	-2.0157
0.4385	1.8555	950	1.1237	0.9751	0.9265	0.4396	0.0486	-75.6684	-73.5518	-2.0158	-2.0156
0.3933	1.9531	1000	1.1230	0.9764	0.9267	0.4396	0.0496	-75.6680	-73.5492	-2.0157	-2.0155

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_05beta_1e6rate_CDPOSFT

Base model

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

Finetuned

tsavage68/mistralit2_1000_STEPS_5e7_SFT

Finetuned

(19)

this model