How to inference on CPU? RuntimeError: "addmm_impl_cpu_" not implemented for 'Half'

#41
by demetera - opened

Trying to play with the default script using stable_audio_tools on CPU.

I'm getting : RuntimeError: "addmm_impl_cpu_" not implemented for 'Half'
When executing the generation block :

output = generate_diffusion_cond(
    model,
    steps=100,
    cfg_scale=7,
    conditioning=conditioning,
    sample_size=sample_size,
    sigma_min=0.3,
    sigma_max=500,
    sampler_type="dpmpp-3m-sde",
    device='cpu'
)

I found a solution to use float32 torch, but no idea, how to implement it using Stable Audio Tools.

demetera changed discussion title from How to inference on CPU? to How to inference on CPU? RuntimeError: "addmm_impl_cpu_" not implemented for 'Half'

model = model.float() # Convert model to float
conditioning = conditioning.float() # Convert conditioning to float

output = generate_diffusion_cond(
model,
steps=100,
cfg_scale=7,
conditioning=conditioning,
sample_size=sample_size,
sigma_min=0.3,
sigma_max=500,
sampler_type="dpmpp-3m-sde",
device='cpu'
)

maybe?

total guess

Sign up or log in to comment