An Open Source VIDEO LLM (Apollo Test and Install Tutorial)

แชร์
ฝัง
  • เผยแพร่เมื่อ 16 ม.ค. 2025

ความคิดเห็น • 18

  • @OminousIndustries
    @OminousIndustries  29 วันที่ผ่านมา +1

    Editing mistake: At 10:09, you need to enter "FLASH_ATTENTION_SKIP_CUDA_BUILD=TRUE pip install flash-attn --no-build-isolation" After the pip install -r requirements.txt finishes installing all the reqs.

  • @therion961
    @therion961 3 วันที่ผ่านมา

    Sweet rig bro

  • @jeffwads
    @jeffwads 21 วันที่ผ่านมา

    I set it up to describe the Asteroid City trailer. It was sharded between my CPU and the 3090 which only has 12 GB VRAM. The run took 12 hours and produced a good description of it. You definitely need serious compute to use this thing on anything longer than 5-10 seconds.

    • @OminousIndustries
      @OminousIndustries  21 วันที่ผ่านมา

      I see that's about 2 and a half minutes of video. That's cool to see it gave good info but yes, the performance requirements for any measurable level of time are very steep.

  • @JohnMoney-iy7sy
    @JohnMoney-iy7sy 28 วันที่ผ่านมา +2

    i'm new to the channel, love it tho!
    whats up with the septa shirt in some videos, is that a coded message for help ?

    • @OminousIndustries
      @OminousIndustries  28 วันที่ผ่านมา

      Thanks very much, and no LOL, I just have a few shirts I go between when filming tbh :))

  • @ano0810
    @ano0810 3 วันที่ผ่านมา

    this is so kool. What's the minimum config of the system required to run this model? You mention 3060s, could you please provide a link? Trying to get a minimum setup

    • @OminousIndustries
      @OminousIndustries  วันที่ผ่านมา

      Thanks! I think this needs a 3090 minimum, though for a minimal setup a Zotac 3060 12gb is a very good price/performance option and I have two of those in a machine

  • @marcomerola4271
    @marcomerola4271 28 วันที่ผ่านมา

    High quality

    • @OminousIndustries
      @OminousIndustries  28 วันที่ผ่านมา

      I appreciate it very much! (assuming you meant the video and not the model LOL)

    • @marcomerola4271
      @marcomerola4271 28 วันที่ผ่านมา

      @OminousIndustries lol

  • @pablosanzo5584
    @pablosanzo5584 25 วันที่ผ่านมา

    Great stuff! I don't have a powerful computer, do you have a copy of the original HF Space? 🙏

    • @OminousIndustries
      @OminousIndustries  24 วันที่ผ่านมา +1

      Thanks very much! Someone has mirrored it here: huggingface.co/spaces/GoodiesHere/Apollo-LMMs-Apollo-1-5B-t32

  • @lionmike247
    @lionmike247 28 วันที่ผ่านมา

    wow i love your channel man. thanks for this. Is there any way we can run this on Mac computers?

    • @OminousIndustries
      @OminousIndustries  28 วันที่ผ่านมา +1

      I appreciate it very much! Unfortunately I do not know of any way to run this on Mac, but they are getting more popular for ML so I am hopeful development on mlx continues and we get more options to run repos like this one.

    • @lionmike247
      @lionmike247 28 วันที่ผ่านมา

      @@OminousIndustries appreciate that. nonetheless, really appreciate the value you give out bro. informative and entertaining to watch!

    • @OminousIndustries
      @OminousIndustries  27 วันที่ผ่านมา

      @@lionmike247 Thank you!