---
language:
- en
license: other
tags:
- causal-lm
datasets:
- HuggingFaceH4/ultrachat_200k
- allenai/ultrafeedback_binarized_cleaned
- meta-math/MetaMathQA
- WizardLM/WizardLM_evol_instruct_V2_196k
- openchat/openchat_sharegpt4_dataset
- LDJnr/Capybara
- Intel/orca_dpo_pairs
- hkust-nlp/deita-10k-v0
- Anthropic/hh-rlhf
extra_gated_fields:
  Name: text
  Email: text
  Country: text
  Organization or Affiliation: text
  I ALLOW Stability AI to email me about new model releases: checkbox
---
# `StableLM 2 12B Chat`

## Model Description

`Stable LM 2 12B Chat` is a 12 billion parameter instruction tuned language model trained on a mix of publicly available datasets and synthetic datasets, utilizing [Direct Preference Optimization (DPO)](https://arxiv.org/abs/2305.18290).

## Usage

`StableLM 2 12B Chat` uses the following instruction ChatML format
This format is also available through the tokenizer's `apply_chat_template` method:

```python
from transformers import AutoModelForCausalLM, AutoTokenizer

tokenizer = AutoTokenizer.from_pretrained('stabilityai/stablelm-2-chat', trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained(
    'stabilityai/stablelm-2-chat',
    device_map="auto",
    trust_remote_code=True,
)

prompt = [{'role': 'user', 'content': 'How to achieve multiple rows of data into one row of data in Excel?'}]
inputs = tokenizer.apply_chat_template(
    prompt,
    add_generation_prompt=True,
    return_tensors='pt'
)

tokens = model.generate(
    inputs.to(model.device),
    max_new_tokens=100,
    temperature=0.7,
    do_sample=True
)
output = tokenizer.decode(tokens[:, inputs.input_ids.shape[-1]:][0], skip_special_tokens=False)

print(output)
```

StableLM 2 12B Chat also supports function call usage this is an example how you can use it:
```python
system_prompt_func = """\
You are a helpful assistant with access to the following functions. You must use them if required -\n
[
  {
    "type": "function",
    "function": {
      "name": "TextToImage",
      "description": "This function able to creating, drawing, or illustrating an image from a text prompt.",
      "parameters": {
        "type": "object",
        "properties": {
          "prompt": {
            "type": "string",
            "description": "The description of image that user wanto to create."
          }
        },
        "required": [
          "prompt"
        ]
      }
    }
  }
]
"""
messages = [
    {'role': 'system', 'content': system_prompt},
    {'role': "user", 'content': "Help me to generate a picture of Eiffel Tower in the night!"}
]

inputs = tokenizer.apply_chat_template(
    messages,
    add_generation_prompt=True,
    return_tensors='pt'
)

tokens = model.generate(
    inputs.to(model.device),
    max_new_tokens=1024,
    temperature=0.5,
    do_sample=True
)
output = tokenizer.decode(tokens[:, inputs.input_ids.shape[-1]:][0], skip_special_tokens=False)

print(output)
"""
[
  {
    "name": "TextToImage",
    "arguments": {
      "prompt": "Eiffel Tower in the night"
    }
  }
]
"""

```


## Model Details

* **Developed by**: [Stability AI](https://stability.ai/)
* **Model type**: `StableLM 2 12B Chat` model is an auto-regressive language model based on the transformer decoder architecture.
* **Language(s)**: English
TODO: Check if we want to keep paper link since it's not mentioned in that paper.
* **Paper**: [Stable LM 2 Chat Technical Report](https://drive.google.com/file/d/1JYJHszhS8EFChTbNAf8xmqhKjogWRrQF/view?usp=sharing)
* **Library**: [Alignment Handbook](https://github.com/huggingface/alignment-handbook.git)
* **Finetuned from model**: 
* **License**: [StabilityAI Non-Commercial Research Community License](https://huggingface.co/stabilityai/stablelm-2-zephyr-1_6b/blob/main/LICENSE). If you want to use this model for your commercial products or purposes, please contact us [here](https://stability.ai/contact) to learn more.
* **Contact**: For questions and comments about the model, please email `lm@stability.ai`

### Training Dataset

The dataset is comprised of a mixture of open datasets large-scale datasets available on the [HuggingFace Hub](https://huggingface.co/datasets) as well as an internal safety dataset:
1. SFT Datasets
- HuggingFaceH4/ultrachat_200k
- meta-math/MetaMathQA
- WizardLM/WizardLM_evol_instruct_V2_196k
- Open-Orca/SlimOrca
- openchat/openchat_sharegpt4_dataset
- LDJnr/Capybara
- hkust-nlp/deita-10k-v0

2. Safety Datasets:
- Anthropic/hh-rlhf
- Internal Safety Dataset

3. Preference Datasets:


## Performance

### MT-Bench



### OpenLLM Leaderboard


### Training Infrastructure

TODO: Fix this
* **Hardware**: `StableLM 2 12B Chat` was trained on the Stability AI cluster across 8 nodes with 8 A100 80GBs GPUs for each nodes.
* **Code Base**: We use our internal script for SFT training and [HuggingFace Alignment Handbook](https://github.com/huggingface/alignment-handbook) for DPO training.

## Use and Limitations

### Intended Use

The model is intended to be used in chat-like applications. Developers must evaluate the model for safety performance in their specific use case. Read more about [safety and limitations](#limitations-and-bias) below.

### Limitations and Bias

TODO: Do we need or have a standard template to throw in here now?

We strongly recommend pairing this model with an input and output classifier to prevent harmful responses.
Using this model will require guardrails around your inputs and outputs to ensure that any outputs returned are not hallucinations.
Additionally, as each use case is unique, we recommend running your own suite of tests to ensure proper performance of this model.
Finally, do not use the models if they are unsuitable for your application, or for any applications that may cause deliberate or unintentional harm to others.


## How to Cite