Any plans on changing LLAMA3.1 to something with Apache 2.0 license?
Since LLAMA 3.1 has a restrictive license, are there any plans on finetuning your model with an open license LLM such as Qwen, IBM-Granite, etc which has Apache 2.0 license. Then we could have a truly open source model where every component is open source.
@natalie5 Thanks. Removing Llama in the current version is challenging. Simply dropping it leads to poor results, and replacing it with other LLMs hasn't been investigated and we don't have short-term plans to finetune this model version. If you've experimented with this, please share your findings.
@natalie5 Thanks. Removing Llama in the current version is challenging. Simply dropping it leads to poor results, and replacing it with other LLMs hasn't been investigated and we don't have short-term plans to finetune this model version. If you've experimented with this, please share your findings.
The model with MIT license is great, but LLAMA having a restrictive license prevents anyone from using it commercially. I was hoping for you guys to use a LLM that doesn't have a restrictive license as a text encoder. Using LLAMA was a plain mistake to be honest. Many researchers are not happy with the model using a LLM that has a restrictive license. The only major problem with your model currently is LLAMA. I was hoping you guys would be able to adopt a better licensed LLM for it since it requires a lot of compute.
Thanks for your work cai-qi! Lots of complaining in the comments, lol.
Just wanted to say the model is fantastic and very much appreciated!!
Thanks for your work cai-qi! Lots of complaining in the comments, lol.
Just wanted to say the model is fantastic and very much appreciated!!
I literally said it's a great model in the first line. Pointing out and questioning about issues isn't complaining, it's just a discussion with the dev.
The model with MIT license is great,
The models use also appears to be governed by:You own all content you create with this model. You can use your generated content freely, but you must comply with this license agreement. You are responsible for how you use the models. Do not create illegal content, harmful material, personal information that could harm others, false information, or content targeting vulnerable groups.
https://huggingface.co/HiDream-ai/HiDream-I1-Full/discussions/2#67f4d1ca93b8bb3454d04c2c
LLAma 3.1 License：
“2. Additional Commercial Terms. If, on the Llama 3.1 version release date, the monthly active users
of the products or services made available by or for Licensee, or Licensee’s affiliates, is greater than 700
million monthly active users in the preceding calendar month, you must request a license from Meta,
which Meta may grant to you in its sole discretion, and you are not authorized to exercise any of the
rights under this Agreement unless or until Meta otherwise expressly grants you such rights.”
I tried to use Mistral-7B-v0.3(Apache-2 license) as a text encoder. It can generate image, but the image is a noise image... Maybe we need to fully retrain the DiT model...
The LLama 3.1 license isn't commercial restrictive, it simply has a scale clause that if you have more than 700 million active users utilizing the model or outputs of the model, you need to request a license. So unless folks here are running businesses with almost a billion people using them, you should not be concerned about the Llama 3.1 commercial restrictions...
I can't guess what other community members are thinking, but for me, Llama's license only 'opensources the model weights' – it's not even as fully-opensourced as Gemma2 used by Lumina. What I truly want is a thoroughly open-source model (where all components are open-sourced under permissive licenses).