yuchenxie

/

ArlowGPT-8B

like 4

Hi,I'd like to report a conflict license in yuchenxie/ArlowGPT-8B. I noticed that this model was finetuned from meta-llama/Llama-3.1-8B-Instruct, but it's currently published under the mit license. There’s a potential conflict here, as the LLaMA 3.1 Community License includes restrictions on redistribution, sublicensing, and acceptable use, which are not enforced by the MIT License. This can create confusion for downstream users about what they are allowed to do with the model.

⚠️ Key incompatibilities with LLaMA 3.1 Community License:

Clause 1.b.i – Redistribution and Use:
  •  No license file included (should contain the LLaMA 3.1 Community License)
  •  "Built with Llama" is not clearly indicated
  •  Model name does not begin with “Llama 3”, which is required for any derivative
  
Clause 1.b.iii – Required Notice:
  •  Missing the following required text in a "NOTICE" file:
    “Llama 3.1 is licensed under the Llama 3.1 Community License, Copyright © Meta Platforms, Inc. All Rights Reserved.”
    
Clause 1.iv – Acceptable Use Policy:
  •  Meta’s Acceptable Use Policy is not mentioned or passed along to users
  
Clause 2 – Sublicensing and Relicensing:
  •  LLaMA 3.1 license does not allow sublicensing under a more permissive license such as MIT
  •  The MIT License permits nearly unrestricted commercial use, which contradicts Meta’s limits and       conditions (e.g. commercial MAU threshold)

The MIT License allows:

  • Unrestricted commercial use
  • Sublicensing under more permissive terms
  • No obligation to preserve upstream restrictions

But the LLaMA 3.1 Community License explicitly prohibits relicensing, includes enforceable use restrictions (e.g., no surveillance, no discrimination, no legal decision-making), and requires strict downstream compliance — meaning it cannot be replaced or overridden by MIT.

🔹 Suggestion:

To resolve the mismatch, here are a few steps that might help bring things into alignment:

1.  To make sure everything aligns with the LLaMA 3.1 terms, you might want to tweak the licensing setup a bit, like:

  • Maybe include a copy of the LLaMA 3.1 Community License in the repo or model card

  • Include this notice in a “NOTICE” file or the docs:

     > “Llama 3.1 is licensed under the Llama 3.1 Community License, Copyright © Meta Platforms, Inc. All Rights Reserved.”

  • A “Built with LLaMA” note somewhere in the model card could be helpful too

  • Maybe a quick note about usage restrictions, especially for folks using it in commercial settings

  • A statement clarifying that use of the model must comply with Meta’s Acceptable Use Policy
  
2.2. Maybe we can just drop the Apache-2.0 tag and going with the LLaMA 3.1 Community License.  This approach may help reduce potential confusion about redistribution rights and downstream usage conditions.

Hope this helps! 😊 Let me know if you have any questions or need more info.

Thanks for your attention!