Install Failed

[Gle2xn]

Installing causal-conv1d and mamba-ssm failed.

[albertfgu]

I don't have enough information. What OS are you using?

[getorca]

post the full stack trace. But kinda looks like you're missing nvcc

[alexanderwerning]

I had the same issue, bare_metal_version is only defined in the setup.py if the CUDA_HOME environment variable is set, but the 11.8 version check is outside of that if block:

mamba/setup.py

Lines 100 to 113 in 009bec5

	if CUDA_HOME is not None:
	_, bare_metal_version = get_cuda_bare_metal_version(CUDA_HOME)
	if bare_metal_version < Version("11.6"):
	raise RuntimeError(
	f"{PACKAGE_NAME} is only supported on CUDA 11.6 and above. "
	"Note: make sure nvcc has a supported version by running nvcc -V."
	)
	cc_flag.append("-gencode")
	cc_flag.append("arch=compute_70,code=sm_70")
	cc_flag.append("-gencode")
	cc_flag.append("arch=compute_80,code=sm_80")
	if bare_metal_version >= Version("11.8"):
	cc_flag.append("-gencode")

CUDA 11.8 was installed on my machine, so once I set the CUDA_HOME variable I could install the package.
Maybe an error message about CUDA_HOME would be helpful?

[Rhys-McAlister]

I had the same issue, bare_metal_version is only defined in the setup.py if the CUDA_HOME environment variable is set, but the 11.8 version check is outside of that if block:

mamba/setup.py

Lines 100 to 113 in 009bec5

	if CUDA_HOME is not None:
	_, bare_metal_version = get_cuda_bare_metal_version(CUDA_HOME)
	if bare_metal_version < Version("11.6"):
	raise RuntimeError(
	f"{PACKAGE_NAME} is only supported on CUDA 11.6 and above. "
	"Note: make sure nvcc has a supported version by running nvcc -V."
	)
	cc_flag.append("-gencode")
	cc_flag.append("arch=compute_70,code=sm_70")
	cc_flag.append("-gencode")
	cc_flag.append("arch=compute_80,code=sm_80")
	if bare_metal_version >= Version("11.8"):
	cc_flag.append("-gencode")

CUDA 11.8 was installed on my machine, so once I set the CUDA_HOME variable I could install the package. Maybe an error message about CUDA_HOME would be helpful?

how do you set the home variable? I'm having this same issue. Thanks in advance

[alexanderwerning]

It depends on the location and version of your cuda installation, for me it was this:
export CUDA_HOME=/usr/local/cuda-11.8

[lxaw]

Has this been solved? Thanks!