NeuralChat.yaml

---
# Thank you for contributing!
# In filling out this yaml file, please follow the criteria as described here: 
# https://github.com/opening-up-chatgpt/opening-up-chatgpt.github.io/tree/main/projects#criteria

# You're free to build on this work and reuse the data. It is licensed under CC-BY 4.0, with the
# stipulation that attribution should come in the form of a link to http://opening-up-chatgpt.github.io
# and a citation to the paper in which the initial dataset & criteria were published:

# Liesenfeld, Andreas, Alianda Lopez, and Mark Dingemanse. 2023. “Opening up ChatGPT: Tracking Openness, Transparency, and Accountability in Instruction-Tuned Text Generators.” In CUI '23: Proceedings of the 5th International Conference on Conversational User Interfaces. July 19-21, Eindhoven. doi: 10.1145/3571884.3604316 

system:
    name: NeuralChat
    link: https://huggingface.co/Intel/neural-chat-7b-v3-3
    type: text 
    performanceclass: full
    basemodelname: Mistral-7B-v0.1
    endmodelname: Neural-Chat-7B-v3.3
    endmodellicense: Apache 2.0
    releasedate: 2023-11
    notes: New model. TODO update fields. A mistral-based Orca-finetuned chat model

org:
    name: Intel
    link: https://community.intel.com/t5/Blogs/Tech-Innovation/Artificial-Intelligence-AI/NeuralChat-A-Customizable-Chatbot-Framework/post/1526789
    notes:

# availability:
datasources_basemodel:
    class: closed
    link:
    notes: Mistral has not disclosed anything about its training data

datasources_endmodel:
    class: open
    link: https://huggingface.co/datasets/Open-Orca/SlimOrca
    notes: RL dataset used for post-training is shared and available

weights_basemodel:
    class: open
    link: https://huggingface.co/mistralai/Mistral-7B-v0.1
    notes: Based on Mistral 7B 0.1
    
weights_endmodel:
    class: open
    link: https://huggingface.co/Intel/neural-chat-7b-v3-1/tree/main
    notes: finetuned model made openly available by Intel

trainingcode:
    class: partial
    link: https://github.com/intel/intel-extension-for-transformers/tree/main/intel_extension_for_transformers/neural_chat/examples/finetuning/finetune_neuralchat_v3
    notes: Mistral base model is not open, but repo gives sample code for fine-tuning based on that

# documentation:
code:
    class: partial
    link: https://github.com/intel/intel-extension-for-transformers/tree/main/intel_extension_for_transformers/neural_chat/examples/finetuning/finetune_neuralchat_v3#how-to-train-intelneural-chat-7b-v3-1-on-intel-gaudi2
    notes: Mistral remains closed so only documentation pertains to fine-tuning steps, but that code is quite well documented, so partial.

architecture:
    class: partial
    link: https://medium.com/intel-analytics-software/the-practice-of-supervised-finetuning-and-direct-preference-optimization-on-habana-gaudi2-a1197d8a3cd3
    notes: Described on HuggingFace model card and a Medium post

preprint:
    class: closed
    link: https://medium.com/intel-analytics-software/the-practice-of-supervised-finetuning-and-direct-preference-optimization-on-habana-gaudi2-a1197d8a3cd3
    notes: A medium post is apparently the only scientific documentation of this model

paper:
    class: closed
    link:
    notes: No peer-reviewed paper found

modelcard:
    class: partial
    link: https://huggingface.co/Intel/neural-chat-7b-v3-1
    notes: No model card for Mistral base model portion.

datasheet:
    class: partial
    link: https://huggingface.co/datasets/Open-Orca/OpenOrca
    notes: SlimOrca dataset is described as part of OpenOrca (but partial since Mistral data is incrutable)

# access:
package:
    class: partial
    link: https://huggingface.co/Intel/neural-chat-7b-v3-1#fp32-inference-with-transformers
    notes: Code for running with transformers provided by Intel

api:
    class: closed
    link: 
    notes: Provided through HuggingFace but model too large to run via inference API, local deployment or paid access needed 
    metaprompt: closed

licenses:
    class: open
    link: https://huggingface.co/Intel/neural-chat-7b-v3-1/blob/main/LICENSE
    notes: Apache 2.0