Link https://abetlen.github.io/llama-cpp-python/whl/cu125 returns 404 Not Found.

[wimtdw]

Link https://abetlen.github.io/llama-cpp-python/whl/cu125 returns 404 Not Found.
I've been following instructions in readme on how to install a pre-built wheel with CUDA support. Links for versions 12.1-12.4 work fine, but not for 12.5.

Readme section:

"Pre-built Wheel (New)

It is also possible to install a pre-built wheel with CUDA support. As long as your system meets some requirements:

• CUDA Version is 12.1, 12.2, 12.3, 12.4 or 12.5
• Python Version is 3.10, 3.11 or 3.12

pip install llama-cpp-python \
  --extra-index-url https://abetlen.github.io/llama-cpp-python/whl/<cuda-version>

Where <cuda-version> is one of the following:

• cu121: CUDA 12.1
• cu122: CUDA 12.2
• cu123: CUDA 12.3
• cu124: CUDA 12.4
• cu125: CUDA 12.5

For example, to install the CUDA 12.1 wheel:

pip install llama-cpp-python \
  --extra-index-url https://abetlen.github.io/llama-cpp-python/whl/cu121"

[anr2me]

Only wheels that are listed here will be available:

llama-cpp-python/.github/workflows/generate-index-from-release.yaml

Lines 42 to 49 in c37132b

	./scripts/releases-to-pep-503.sh index/whl/cpu '^[v]?[0-9]+\.[0-9]+\.[0-9]+$'
	./scripts/releases-to-pep-503.sh index/whl/cu121 '^[v]?[0-9]+\.[0-9]+\.[0-9]+-cu121$'
	./scripts/releases-to-pep-503.sh index/whl/cu122 '^[v]?[0-9]+\.[0-9]+\.[0-9]+-cu122$'
	./scripts/releases-to-pep-503.sh index/whl/cu123 '^[v]?[0-9]+\.[0-9]+\.[0-9]+-cu123$'
	./scripts/releases-to-pep-503.sh index/whl/cu124 '^[v]?[0-9]+\.[0-9]+\.[0-9]+-cu124$'
	# ./scripts/releases-to-pep-503.sh index/whl/cu125 '^[v]?[0-9]+\.[0-9]+\.[0-9]+-cu124$'
	# ./scripts/releases-to-pep-503.sh index/whl/cu126 '^[v]?[0-9]+\.[0-9]+\.[0-9]+-cu124$'
	./scripts/releases-to-pep-503.sh index/whl/metal '^[v]?[0-9]+\.[0-9]+\.[0-9]+-metal$'

And currently cu125 & cu126 are being commented, thus not included.

Anyway, you can still use the wheel for cu124 even if you have cuda 12.9

[Kaszanas]

Only wheels that are listed here will be available:

llama-cpp-python/.github/workflows/generate-index-from-release.yaml

Lines 42 to 49 in c37132b

       ./scripts/releases-to-pep-503.sh index/whl/cpu '^[v]?[0-9]+\.[0-9]+\.[0-9]+$' 
       ./scripts/releases-to-pep-503.sh index/whl/cu121 '^[v]?[0-9]+\.[0-9]+\.[0-9]+-cu121$' 
       ./scripts/releases-to-pep-503.sh index/whl/cu122 '^[v]?[0-9]+\.[0-9]+\.[0-9]+-cu122$' 
       ./scripts/releases-to-pep-503.sh index/whl/cu123 '^[v]?[0-9]+\.[0-9]+\.[0-9]+-cu123$' 
       ./scripts/releases-to-pep-503.sh index/whl/cu124 '^[v]?[0-9]+\.[0-9]+\.[0-9]+-cu124$' 
       # ./scripts/releases-to-pep-503.sh index/whl/cu125 '^[v]?[0-9]+\.[0-9]+\.[0-9]+-cu124$' 
       # ./scripts/releases-to-pep-503.sh index/whl/cu126 '^[v]?[0-9]+\.[0-9]+\.[0-9]+-cu124$' 
       ./scripts/releases-to-pep-503.sh index/whl/metal '^[v]?[0-9]+\.[0-9]+\.[0-9]+-metal$' 

And currently cu125 & cu126 are being commented, thus not included.
Anyway, you can still use the wheel for cu124 even if you have cuda 12.9

Great to know, the README states something entirely different:

To install with CUDA support, set the GGML_CUDA=on environment variable before installing:

CMAKE_ARGS="-DGGML_CUDA=on" pip install llama-cpp-python
Pre-built Wheel (New)

It is also possible to install a pre-built wheel with CUDA support. As long as your system meets some requirements:

CUDA Version is 12.1, 12.2, 12.3, 12.4 or 12.5
Python Version is 3.10, 3.11 or 3.12
pip install llama-cpp-python \
  --extra-index-url https://abetlen.github.io/llama-cpp-python/whl/<cuda-version>
Where <cuda-version> is one of the following:

cu121: CUDA 12.1
cu122: CUDA 12.2
cu123: CUDA 12.3
cu124: CUDA 12.4
cu125: CUDA 12.5

[anr2me]

Yeah, the readme probably need to be updated. May be they had an issue with cu125 & cu126 and decided to disabled the wheels.

Btw, i've tried modifying the wheel building workflow on my fork for:

• CUDA: 12.5, 12.6, 12.8, 13.0 (had issue with 12.9 and 13.1 so i disabled them)
• Python: 3.9, 3.10, 3.11, 3.12, 3.13, 3.14
• OS: Windows (x64), Linux (x86_64 and aarch64)

For example, you can try cu125 using this link https://anr2me.github.io/llama-cpp-python/whl/cu125

[Kaszanas]

Yeah, the readme probably need to be updated. May be they had an issue with cu125 & cu126 and decided to disabled the wheels.

Btw, i've tried modifying the wheel building workflow on my fork for:

• CUDA: 12.5, 12.6, 12.8, 13.0 (had issue with 12.9 and 13.1 so i disabled them)
• Python: 3.9, 3.10, 3.11, 3.12, 3.13, 3.14
• OS: Windows (x64), Linux (x86_64)

For example, you can try cu125 using this link https://anr2me.github.io/llama-cpp-python/whl/cu125

I found out that for Windows wheels I need to use a fork with them being pre-built. In the end it errors out anyway on .dll load. This one is quite unusable.

[anr2me]

I see, i don't have a Windows PC to tested it.