BasicSR executing: 'python3 inference/inference

Downloaded BasicSR-1.4.2 release
Downloaded pretrained models and placed them in 'experiments/pretrained_models' subfolder: experiments/pretrained_models/BasicVSR/EDVR_REDS_pretrained_for_IconVSR-f62a2f1e.pth experiments/pretrained_models/BasicVSR/BasicVSR_Vimeo90K_BIx4-2a29695a.pth experiments/pretrained_models/BasicVSR/IconVSR_Vimeo90K_BIx4-35fec07c.pth experiments/pretrained_models/BasicVSR/BasicVSR_Vimeo90K_BDx4-e9bf46eb.pth experiments/pretrained_models/BasicVSR/EDVR_Vimeo90K_pretrained_for_IconVSR-ee48ee92.pth experiments/pretrained_models/BasicVSR/BasicVSR_REDS4-543c8261.pth experiments/pretrained_models/BasicVSR/IconVSR_Vimeo90K_BDx4-cfcb7e00.pth experiments/pretrained_models/BasicVSR/IconVSR_REDS-aaa5367f.pth experiments/pretrained_models/EDVR/EDVR_L_deblur_REDS_official-ca46bd8c.pth experiments/pretrained_models/EDVR/EDVR_L_deblurcomp_REDS_official-0e988e5c.pth experiments/pretrained_models/EDVR/EDVR_M_x4_SR_REDS_official-32075921.pth experiments/pretrained_models/EDVR/EDVR_L_x4_SR_Vimeo90K_official-162b54e4.pth experiments/pretrained_models/EDVR/EDVR_L_x4_SR_REDS_official-9f5f5039.pth experiments/pretrained_models/EDVR/EDVR_M_woTSA_x4_SR_REDS_official-1edf645c.pth experiments/pretrained_models/EDVR/EDVR_L_x4_SRblur_REDS_official-983d7b8e.pth experiments/pretrained_models/TOF/tof_x4_vimeo90k_official-32c9e01f.pth experiments/pretrained_models/DFDNet/DFDNet_official-d1fa5650.pth experiments/pretrained_models/DFDNet/DFDNet_dict_512-f79685f0-001.pth experiments/pretrained_models/DUF/DUF_x3_16L_official-34ce53ec.pth experiments/pretrained_models/DUF/DUF_x4_28L_official-cbada450.pth experiments/pretrained_models/DUF/DUF_x4_52L_official-483d2c78.pth experiments/pretrained_models/DUF/DUF_x2_16L_official-39537cb9.pth experiments/pretrained_models/DUF/DUF_x4_16L_official-bf8f0cfa.pth experiments/pretrained_models/dlib/shape_predictor_5_face_landmarks-c4b1e980.dat experiments/pretrained_models/dlib/mmod_human_face_detector-4cb19393.dat experiments/pretrained_models/dlib/shape_predictor_68_face_landmarks-fbdc2cb8.dat experiments/pretrained_models/flownet/spynet_sintel_final-3d2a1287.pth experiments/pretrained_models/EDSR/EDSR_Lx2_f256b32_DIV2K_official-be38e77d.pth experiments/pretrained_models/EDSR/EDSR_Lx3_f256b32_DIV2K_official-3660f70d.pth experiments/pretrained_models/EDSR/EDSR_Mx4_f64b16_DIV2K_official-0c287733.pth experiments/pretrained_models/EDSR/EDSR_Lx4_f256b32_DIV2K_official-76ee1c8f.pth experiments/pretrained_models/EDSR/EDSR_Mx3_f64b16_DIV2K_official-6908f88a.pth experiments/pretrained_models/EDSR/EDSR_Mx2_f64b16_DIV2K_official-3ba7b086.pth experiments/pretrained_models/README.md experiments/pretrained_models/StyleGAN/stylegan2_horse_config_f_256_discriminator_official-be6c4c33.pth experiments/pretrained_models/StyleGAN/stylegan2_cat_config_f_256_discriminator_official-2c97fd08.pth experiments/pretrained_models/StyleGAN/stylegan2_car_config_f_512_discriminator_official-5008e3d1.pth experiments/pretrained_models/StyleGAN/stylegan2_horse_config_f_256_official-26d57fee.pth experiments/pretrained_models/StyleGAN/stylegan2_cat_config_f_256_official-0a9173ad.pth experiments/pretrained_models/StyleGAN/stylegan2_ffhq_config_f_1024_official-3ab41b38.pth experiments/pretrained_models/StyleGAN/stylegan2_church_config_f_256_official-44ba63bf.pth experiments/pretrained_models/StyleGAN/stylegan2_church_config_f_256_discriminator_official-20cd675b.pth experiments/pretrained_models/StyleGAN/stylegan2_ffhq_config_f_1024_discriminator_official-a386354a.pth experiments/pretrained_models/StyleGAN/stylegan2_car_config_f_512_official-e8fcab4f.pth experiments/pretrained_models/ESRGAN/ESRGAN_PSNR_SRx4_DF2K_official-150ff491.pth experiments/pretrained_models/ESRGAN/ESRGAN_SRx4_DF2KOST_official-ff704c30.pth
executed: 'python3 inference/inference_stylegan2.py'
got the following error: /opt/venv/lib/python3.11/site-packages/torchvision/transforms/functional_tensor.py:5: UserWarning: The torchvision.transforms.functional_tensor module is deprecated in 0.15 and will be **removed in 0.17**. Please don't rely on it. You probably just need to use APIs in torchvision.transforms.functional or in torchvision.transforms.v2.functional. warnings.warn( Traceback (most recent call last): File "/opt/BasicSR-1.4.2/inference/inference_stylegan2.py", line 70, in <module> generate(args, g_ema, device, mean_latent, args.randomize_noise) File "/opt/BasicSR-1.4.2/inference/inference_stylegan2.py", line 18, in generate sample, _ = g_ema([sample_z], ^^^^^^^^^^^^^^^^^ File "/opt/venv/lib/python3.11/site-packages/torch/nn/modules/module.py", line 1501, in _call_impl return forward_call(*args, **kwargs) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/opt/venv/lib/python3.11/site-packages/basicsr/archs/stylegan2_arch.py", line 539, in forward styles = [self.style_mlp(s) for s in styles] ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/opt/venv/lib/python3.11/site-packages/basicsr/archs/stylegan2_arch.py", line 539, in <listcomp> styles = [self.style_mlp(s) for s in styles] ^^^^^^^^^^^^^^^^^ File "/opt/venv/lib/python3.11/site-packages/torch/nn/modules/module.py", line 1501, in _call_impl return forward_call(*args, **kwargs) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/opt/venv/lib/python3.11/site-packages/torch/nn/modules/container.py", line 217, in forward input = module(input) ^^^^^^^^^^^^^ File "/opt/venv/lib/python3.11/site-packages/torch/nn/modules/module.py", line 1501, in _call_impl return forward_call(*args, **kwargs) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/opt/venv/lib/python3.11/site-packages/basicsr/archs/stylegan2_arch.py", line 172, in forward out = fused_leaky_relu(out, bias) ^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/opt/venv/lib/python3.11/site-packages/basicsr/ops/fused_act/fused_act.py", line 95, in fused_leaky_relu return FusedLeakyReLUFunction.apply(input, bias, negative_slope, scale) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/opt/venv/lib/python3.11/site-packages/torch/autograd/function.py", line 506, in apply return super().apply(*args, **kwargs) # type: ignore[misc] ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/opt/venv/lib/python3.11/site-packages/basicsr/ops/fused_act/fused_act.py", line 65, in forward out = fused_act_ext.fused_bias_act(input, bias, empty, 3, 0, negative_slope, scale) ^^^^^^^^^^^^^ NameError: name 'fused_act_ext' is not defined

May 19 '23 22:05 igoryonya

add BASICSR_JIT=True and use python

Oct 17 '24 08:10 qianx77

if you still have this problem, the fix is to either install the extension, which didnt work for me on windows, or use this code workaround with the same functionality: fused_act.py

# modify from https://github.com/rosinality/stylegan2-pytorch/blob/master/op/fused_act.py # noqa:E501

import os
import torch
from torch import nn
from torch.autograd import Function

BASICSR_JIT = os.getenv('BASICSR_JIT')
if BASICSR_JIT == 'True':
    from torch.utils.cpp_extension import load
    module_path = os.path.dirname(__file__)
    fused_act_ext = load(
        'fused',
        sources=[
            os.path.join(module_path, 'src', 'fused_bias_act.cpp'),
            os.path.join(module_path, 'src', 'fused_bias_act_kernel.cu'),
        ],
    )
else:
    try:
        from . import fused_act_ext
    except ImportError:
        pass
        # avoid annoying print output
        # print(f'Cannot import deform_conv_ext. Error: {error}. You may need to: \n '
        #       '1. compile with BASICSR_EXT=True. or\n '
        #       '2. set BASICSR_JIT=True during running')


class FusedLeakyReLUFunctionBackward(Function):

    @staticmethod
    def forward(ctx, grad_output, out, negative_slope, scale):
        ctx.save_for_backward(out)
        ctx.negative_slope = negative_slope
        ctx.scale = scale

        empty = grad_output.new_empty(0)

        grad_input = fused_act_ext.fused_bias_act(grad_output, empty, out, 3, 1, negative_slope, scale)

        dim = [0]

        if grad_input.ndim > 2:
            dim += list(range(2, grad_input.ndim))

        grad_bias = grad_input.sum(dim).detach()

        return grad_input, grad_bias

    @staticmethod
    def backward(ctx, gradgrad_input, gradgrad_bias):
        out, = ctx.saved_tensors
        gradgrad_out = fused_act_ext.fused_bias_act(gradgrad_input, gradgrad_bias, out, 3, 1, ctx.negative_slope,
                                                    ctx.scale)

        return gradgrad_out, None, None, None


def FusedLeakyReLUFunction(input, bias, negative_slope=0.2, scale=2**0.5):
    """
    Fused bias addition and activation (LeakyReLU) in one function.
    Mimics the fused operation in CUDA but using PyTorch.
    
    Args:
        input (Tensor): The input tensor.
        bias (Tensor): The bias term, typically a learned parameter.
        negative_slope (float): The negative slope for LeakyReLU.
        scale (float): A scaling factor applied to the output after activation.
        
    Returns:
        Tensor: The result after bias addition and LeakyReLU activation.
    """
    # Bias addition
    out = input + bias

    # Apply Leaky ReLU activation
    out = torch.nn.functional.leaky_relu(out, negative_slope=negative_slope)

    # Apply scaling factor (if needed)
    out = out * scale
    
    return out

class FusedLeakyReLU(nn.Module):
    def __init__(self, channel, negative_slope=0.2, scale=2**0.5):
        super().__init__()
        self.bias = nn.Parameter(torch.zeros(channel))  # Shape (channel,)
        self.negative_slope = negative_slope
        self.scale = scale

        # Using standard PyTorch LeakyReLU
        self.leaky_relu = nn.LeakyReLU(negative_slope=negative_slope)

    def forward(self, input):
        # Reshape the bias to match input size
        bias = self.bias.view(1, -1, 1, 1)  # Shape: (1, channel, 1, 1)
        
        # Apply the bias and LeakyReLU activation
        return self.leaky_relu(input + bias)
    

def fused_leaky_relu(input, bias, negative_slope=0.2, scale=2**0.5):
    return FusedLeakyReLUFunction(input, bias, negative_slope, scale)

Nov 19 '24 16:11 pyphan1

executing: 'python3 inference/inference_stylegan2.py' results in error