Re: Building llama.cpp for AMD GPU using only Debian packages?

To: debian-ai@lists.debian.org
Subject: Re: Building llama.cpp for AMD GPU using only Debian packages?
From: Petter Reinholdtsen <pere@hungry.com>
Date: Mon, 01 Jul 2024 14:52:57 +0200
Message-id: <[🔎] sa6wmm5s28m.fsf@hjemme.reinholdtsen.name>
In-reply-to: <[🔎] a2e3c24b-6459-41cb-995f-93bb4bdfee97@slerp.xyz>
References: <[🔎] sa6sewtv5km.fsf@hjemme.reinholdtsen.name> <[🔎] a2e3c24b-6459-41cb-995f-93bb4bdfee97@slerp.xyz>

[Cordell Bloor]
> Here is a new set of instructions for building and running a recent
> llama-cpp version on Debian Unstable:

Thank you.  I got it building.  But I guess I miss some preparation
step, as I get this when running llama-cli:

  rocBLAS error: Could not initialize Tensile host: No devices found
  Avbrutt (SIGABRT)

"sudo rocminfo" report Agent 2 name as gfx1100 with marketing Name "AMD
Radeon RX 7900 XTX", so I guess something is working as it should.  Any
idea what is missing?

-- 
Happy hacking
Petter Reinholdtsen

Reply to:

Follow-Ups:
- Re: Building llama.cpp for AMD GPU using only Debian packages?
  - From: Petter Reinholdtsen <pere@hungry.com>
- Re: Building llama.cpp for AMD GPU using only Debian packages?
  - From: Cordell Bloor <cgmb@slerp.xyz>

References:
- Building llama.cpp for AMD GPU using only Debian packages?
  - From: Petter Reinholdtsen <pere@hungry.com>
- Re: Building llama.cpp for AMD GPU using only Debian packages?
  - From: Cordell Bloor <cgmb@slerp.xyz>

Prev by Date: Processing of hipfft_6.1.2-1_amd64.changes
Next by Date: Re: Building llama.cpp for AMD GPU using only Debian packages?
Previous by thread: Re: Building llama.cpp for AMD GPU using only Debian packages?
Next by thread: Re: Building llama.cpp for AMD GPU using only Debian packages?
Index(es):
- Date
- Thread