From 786798a7f4deba0a81cdfc94a1fb98773e9cfb01 Mon Sep 17 00:00:00 2001 From: Owen Law Date: Sun, 12 Nov 2023 09:47:04 -0500 Subject: [PATCH] Update intro.md and faq.md - OA complete --- docs/docs/faq.md | 186 +++++++-------------------------------------- docs/docs/intro.md | 6 ++ 2 files changed, 35 insertions(+), 157 deletions(-) diff --git a/docs/docs/faq.md b/docs/docs/faq.md index 91c277cb66..8566b9dfe2 100644 --- a/docs/docs/faq.md +++ b/docs/docs/faq.md @@ -15,12 +15,12 @@ In this page, there are some of the most frequently asked questions. -We have released candidate supervised finetuning (SFT) models using both Pythia -and LLaMa, as well as candidate reward models for reinforcement learning from -human feedback training using Pythia, which you can try, and are beginning the -process of applying (RLHF). We have also released the first version of the -OpenAssistant Conversations dataset -[here](https://huggingface.co/datasets/OpenAssistant/oasst1). +This project has concluded. We have released supervised finetuning (SFT) models +using Llama 2, LLaMa, Falcon, Pythia, and StabeLM as well as reinforcement +learning from human feedback trained models and reward models, all of which are +available at [here](https://huggingface.co/OpenAssistant). In addition to our +models, we have released three datasets from OpenAssistant conversations, and a +[research paper](https://arxiv.org/abs/2304.07327). @@ -31,9 +31,8 @@ OpenAssistant Conversations dataset -You can play with our best candidate model -[here](https://open-assistant.io/chat) and provide thumbs up/down responses to -help us improve the model in future! +Our online demonstration is no longer available, but the models remain available +to download [here](https://huggingface.co/OpenAssistant). @@ -44,37 +43,18 @@ help us improve the model in future! -The candidate Pythia SFT models are +All of our models are [available on HuggingFace](https://huggingface.co/OpenAssistant) and can be -loaded via the HuggingFace Transformers library. As such you may be able to use -them with sufficient hardware. There are also spaces on HF which can be used to -chat with the OA candidate without your own hardware. However, these models are -not final and can produce poor or undesirable outputs. +loaded via the HuggingFace Transformers library or other runners if converted. +As such you may be able to use them with sufficient hardware. There are also +spaces on HF which can be used to chat with the OA candidate without your own +hardware. However, some of these models are not final and can produce poor or +undesirable outputs. -LLaMa SFT models cannot be released directly due to Meta's license but XOR +LLaMa (v1) SFT models cannot be released directly due to Meta's license but XOR weights are released on the HuggingFace org. Follow the process in the README -there to obtain a full model from these XOR weights. - - - -
- - -### Is there an API available? - - - -There is no API currently available for Open Assistant. Any mention of an API in -documentation is referencing the website's internal API. We understand that an -API is a highly requested feature, but unfortunately, we can't provide one at -this time due to a couple of reasons. Firstly, the inference system is already -under high load and running off of compute from our sponsors. Secondly, the -project's primary goal is currently data collection and model training, not -providing a product. - -However, if you're looking to run inference, you can host the model yourself -either on your own hardware or with a cloud provider. We appreciate your -understanding and patience as we continue to develop this project. +there to obtain a full model from these XOR weights. Llama 2 models are not +required to be XORed.
@@ -102,15 +82,13 @@ inference setup and UI locally unless you wish to assist in development. All Open Assistant code is licensed under Apache 2.0. This means it is available for a wide range of uses including commercial use. -The Open Assistant Pythia based models are released as full weights and will be -licensed under the Apache 2.0 license. - -The Open Assistant LLaMa based models will be released only as delta weights -meaning you will need the original LLaMa weights to use them, and the license -restrictions will therefore be those placed on the LLaMa weights. +Open Assistant models are released under the license of their respective base +models, be that Llama 2, Falcon, Pythia, or StableLM. LLaMa (not 2) models are +only released as XOR weights, meaning you will need the original LLaMa weights +to use them. -The Open Assistant data is released under a Creative Commons license allowing a -wide range of uses including commercial use. +The Open Assistant data is released under Apache-2.0 allowing a wide range of +uses including commercial use. @@ -138,9 +116,8 @@ you to everyone who has taken part! -The model code, weights, and data are free. We are additionally hosting a free -public instance of our best current model for as long as we can thanks to -compute donation from Stability AI via LAION! +The model code, weights, and data are free. Our free public instance of our best +models is not longer available due to the project's conclusion. @@ -151,10 +128,9 @@ compute donation from Stability AI via LAION! -The current smallest (Pythia) model is 12B parameters and is challenging to run -on consumer hardware, but can run on a single professional GPU. In future there -may be smaller models and we hope to make progress on methods like integer -quantisation which can help run the model on smaller hardware. +The current smallest models are 7B parameters and are challenging to run on +consumer hardware, but can run on a single professional GPU or be quantized to +run on more widely available hardware. @@ -165,13 +141,7 @@ quantisation which can help run the model on smaller hardware. -If you want to help in the data collection for training the model, go to the -website [https://open-assistant.io/](https://open-assistant.io/). - -If you want to contribute code, take a look at the -[tasks in GitHub](https://github.com/orgs/LAION-AI/projects/3) and comment on an -issue stating your wish to be assigned. You can also take a look at this -[contributing guide](https://github.com/LAION-AI/Open-Assistant/blob/main/CONTRIBUTING.md). +This project has now concluded. @@ -190,104 +160,6 @@ well as accelerate, DeepSpeed, bitsandbytes, NLTK, and other libraries. -## Questions about the data collection website - -
- - -### Can I use ChatGPT to help in training Open Assistant, for instance, by generating answers? - - - -No, it is against their terms of service to use it to help train other models. -See -[this issue](https://github.com/LAION-AI/Open-Assistant/issues/471#issuecomment-1374392299). -ChatGPT-like answers will be removed. - -
- -
- - -### What should I do if I don't know how to complete the task as an assistant? - - -Skip it. -
- -
- - -### Should I fact check the answers by the assistant? - - - -Yes, you should try. If you are not sure, skip the task. - -
- -
- - -### How can I see my score? - - - -In your [account settings](https://open-assistant.io/account). - -
- -
- - -### Can we see how many data points have been collected? - - - -You can see a regularly updated interface at -[https://open-assistant.io/stats](https://open-assistant.io/stats). - -
- -
- - -### How do I write and label prompts? - - - -Check the -[guidelines](https://projects.laion.ai/Open-Assistant/docs/guides/guidelines). - -
- -
- - -### Where can I report a bug or create a new feature request? - - - -In the [GitHub issues](https://github.com/LAION-AI/Open-Assistant/issues). - -
- -
- - -### Why am I not allowed to write about this topic, even though it isn't illegal? - - - -We want to ensure that the Open Assistant dataset is as accessible as possible. -As such, it's necessary to avoid any harmful or offensive content that could be -grounds for removal on sites such as Hugging Face. Likewise, we want the model -to be trained to reject as few questions as possible, so it's important to not -include prompts that leave the assistant with no other choice but to refuse in -order to avoid the generation of harmful content. - -
- ## Questions about the development process
diff --git a/docs/docs/intro.md b/docs/docs/intro.md index 326502bfe3..98f50762ca 100644 --- a/docs/docs/intro.md +++ b/docs/docs/intro.md @@ -1,3 +1,9 @@ +# Notice + +**Open Assistant has now concluded.** Please see +[this video](https://www.youtube.com/watch?v=gqtmUHhaplo) for more information. +Thanks you to all those who made this project possible. + # Introduction > The FAQ page is available at