Merge pull request XingangPan#79 from ochafik/mps-support

Support & instructions for MPS (MacOS w/ GPU support on M1/M2) and CPU

README.md

@@ -35,7 +35,19 @@
 
 ## Requirements
 
-Please follow the requirements of [NVlabs/stylegan3](https://github.com/NVlabs/stylegan3#requirements).
+If you have CUDA graphic card, please follow the requirements of [NVlabs/stylegan3](https://github.com/NVlabs/stylegan3#requirements).  
+
+Otherwise (for GPU acceleration on MacOS with Silicon Mac M1/M2, or just CPU) try the following:
+
+```sh
+cat environment.yml | \
+  grep -v -E 'nvidia|cuda' > environment-no-nvidia.yml && \
+    conda env create -f environment-no-nvidia.yml
+conda activate stylegan3
+
+# On MacOS
+export PYTORCH_ENABLE_MPS_FALLBACK=1
+```
 
 ## Download pre-trained StyleGAN2 weights
 

environment.yml

@@ -5,20 +5,23 @@ channels:
 dependencies:
   - python >= 3.8
   - pip
-  - numpy>=1.20
+  - numpy>=1.25
   - click>=8.0
-  - pillow=8.3.1
-  - scipy=1.7.1
-  - pytorch=1.9.1
+  - pillow=9.4.0
+  - scipy=1.11.0
+  - pytorch>=2.0.1
+  - torchvision>=0.15.2
   - cudatoolkit=11.1
   - requests=2.26.0
   - tqdm=4.62.2
   - ninja=1.10.2
   - matplotlib=3.4.2
   - imageio=2.9.0
   - pip:
-    - imgui==1.3.0
-    - glfw==2.2.0
+    - imgui==2.0.0
+    - glfw==2.6.1
+    - gradio==3.35.2
     - pyopengl==3.1.5
     - imageio-ffmpeg==0.4.3
-    - pyspng
+    # pyspng is currently broken on MacOS (see https://github.com/nurpax/pyspng/pull/6 for instance)
+    - pyspng-seunglab

gen_images.py

@@ -103,9 +103,10 @@ def generate_images(
     """
 
     print('Loading networks from "%s"...' % network_pkl)
-    device = torch.device('cuda')
+    device = torch.device('cuda' if torch.cuda.is_available() else 'mps' if torch.backends.mps.is_available() else 'cpu')
+    dtype = torch.float32 if device.type == 'mps' else torch.float64
     with dnnlib.util.open_url(network_pkl) as f:
-        G = legacy.load_network_pkl(f)['G_ema'].to(device) # type: ignore
+        G = legacy.load_network_pkl(f)['G_ema'].to(device, dtype=dtype) # type: ignore
         # import pickle
         # G = legacy.load_network_pkl(f)
         # output = open('checkpoints/stylegan2-car-config-f-pt.pkl', 'wb')
@@ -126,7 +127,7 @@ def generate_images(
     # Generate images.
     for seed_idx, seed in enumerate(seeds):
         print('Generating image for seed %d (%d/%d) ...' % (seed, seed_idx, len(seeds)))
-        z = torch.from_numpy(np.random.RandomState(seed).randn(1, G.z_dim)).to(device)
+        z = torch.from_numpy(np.random.RandomState(seed).randn(1, G.z_dim)).to(device, dtype=dtype)
 
         # Construct an inverse rotation/translation matrix and pass to the generator.  The
         # generator expects this matrix as an inverse to avoid potentially failing numerical

stylegan_human/generate.py

@@ -63,9 +63,10 @@ def generate_images(
 
     else: 
         import torch
-        device = torch.device('cuda')
+        device = torch.device('cuda' if torch.cuda.is_available() else 'mps' if torch.backends.mps.is_available() else 'cpu')
+        dtype = torch.float32 if device.type == 'mps' else torch.float64
         with dnnlib.util.open_url(network_pkl) as f:
-            G = legacy.load_network_pkl(f)['G_ema'].to(device) # type: ignore
+            G = legacy.load_network_pkl(f)['G_ema'].to(device, dtype=dtype) # type: ignore
     os.makedirs(outdir, exist_ok=True)
 
 
@@ -92,7 +93,7 @@ def generate_images(
 
         else: ## stylegan v2/v3
             label = torch.zeros([1, G.c_dim], device=device)
-            z = torch.from_numpy(np.random.RandomState(seed).randn(1, G.z_dim)).to(device)
+            z = torch.from_numpy(np.random.RandomState(seed).randn(1, G.z_dim)).to(device, dtype=dtype)
             if target_z.size==0:
                 target_z= z.cpu()
             else:

stylegan_human/interpolation.py

@@ -116,9 +116,10 @@ def main(
 ):
 
 
-    device = torch.device('cuda')
+    device = torch.device('cuda' if torch.cuda.is_available() else 'mps' if torch.backends.mps.is_available() else 'cpu')
+    dtype = torch.float32 if device.type == 'mps' else torch.float64
     with dnnlib.util.open_url(network_pkl) as f:
-        G = legacy.load_network_pkl(f)['G_ema'].to(device) # type: ignore
+        G = legacy.load_network_pkl(f)['G_ema'].to(device, dtype=dtype) # type: ignore
 
     outdir = os.path.join(outdir)
     if not os.path.exists(outdir):
@@ -132,8 +133,8 @@ def main(
         print('Require two seeds, randomly generate two now.')
         seeds = [seeds[0],random.randint(0,10000)]
 
-    z1 = torch.from_numpy(np.random.RandomState(seeds[0]).randn(1, G.z_dim)).to(device)
-    z2 = torch.from_numpy(np.random.RandomState(seeds[1]).randn(1, G.z_dim)).to(device)
+    z1 = torch.from_numpy(np.random.RandomState(seeds[0]).randn(1, G.z_dim)).to(device, dtype=dtype)
+    z2 = torch.from_numpy(np.random.RandomState(seeds[1]).randn(1, G.z_dim)).to(device, dtype=dtype)
     img1 = generate_image_from_z(G, z1, noise_mode, truncation_psi, device)
     img2 = generate_image_from_z(G, z2, noise_mode, truncation_psi, device)
     img1.save(f'{outdir}/seed{seeds[0]:04d}.png')

stylegan_human/style_mixing.py

@@ -49,16 +49,17 @@ def generate_style_mix(
 ):
 
     print('Loading networks from "%s"...' % network_pkl)
-    device = torch.device('cuda')
+    device = torch.device('cuda' if torch.cuda.is_available() else 'mps' if torch.backends.mps.is_available() else 'cpu')
+    dtype = torch.float32 if device.type == 'mps' else torch.float64
     with dnnlib.util.open_url(network_pkl) as f:
-        G = legacy.load_network_pkl(f)['G_ema'].to(device)
+        G = legacy.load_network_pkl(f)['G_ema'].to(device, dtype=dtype)
 
     os.makedirs(outdir, exist_ok=True)
 
     print('Generating W vectors...')
     all_seeds = list(set(row_seeds + col_seeds))
     all_z = np.stack([np.random.RandomState(seed).randn(G.z_dim) for seed in all_seeds])
-    all_w = G.mapping(torch.from_numpy(all_z).to(device), None)
+    all_w = G.mapping(torch.from_numpy(all_z).to(device, dtype=dtype), None)
     w_avg = G.mapping.w_avg
     all_w = w_avg + (all_w - w_avg) * truncation_psi
     w_dict = {seed: w for seed, w in zip(all_seeds, list(all_w))}

stylegan_human/stylemixing_video.py

@@ -65,9 +65,10 @@ def style_mixing_video(network_pkl: str,
     print('col_seeds: ', dst_seeds)
     num_frames = int(np.rint(duration_sec * mp4_fps))
     print('Loading networks from "%s"...' % network_pkl)
-    device = torch.device('cuda')
+    device = torch.device('cuda' if torch.cuda.is_available() else 'mps' if torch.backends.mps.is_available() else 'cpu')
+    dtype = torch.float32 if device.type == 'mps' else torch.float64
     with dnnlib.util.open_url(network_pkl) as f:
-        Gs = legacy.load_network_pkl(f)['G_ema'].to(device) 
+        Gs = legacy.load_network_pkl(f)['G_ema'].to(device, dtype=dtype) 
 
     print(Gs.num_ws, Gs.w_dim, Gs.img_resolution) 
     max_style = int(2 * np.log2(Gs.img_resolution)) - 3
@@ -80,14 +81,14 @@ def style_mixing_video(network_pkl: str,
     src_z = scipy.ndimage.gaussian_filter(src_z, [smoothing_sec * mp4_fps] + [0] * (2- 1), mode="wrap") 
     src_z /= np.sqrt(np.mean(np.square(src_z)))
     # Map into the detangled latent space W and do truncation trick
-    src_w = Gs.mapping(torch.from_numpy(src_z).to(device), None)
+    src_w = Gs.mapping(torch.from_numpy(src_z).to(device, dtype=dtype), None)
     w_avg = Gs.mapping.w_avg
     src_w = w_avg + (src_w - w_avg) * truncation_psi
 
     # Top row latents (fixed reference)
     print('Generating Destination W vectors...')  
     dst_z = np.stack([np.random.RandomState(seed).randn(Gs.z_dim) for seed in dst_seeds]) 
-    dst_w = Gs.mapping(torch.from_numpy(dst_z).to(device), None)
+    dst_w = Gs.mapping(torch.from_numpy(dst_z).to(device, dtype=dtype), None)
     dst_w = w_avg + (dst_w - w_avg) * truncation_psi
     # Get the width and height of each image:
     H = Gs.img_resolution       # 1024
@@ -120,7 +121,7 @@ def make_frame(t):
         for col, dst_image in enumerate(list(dst_images)):
             # Select the pertinent latent w column:
             w_col = np.stack([dst_w[col].cpu()]) # [18, 512] -> [1, 18, 512]
-            w_col = torch.from_numpy(w_col).to(device)
+            w_col = torch.from_numpy(w_col).to(device, dtype=dtype)
             # Replace the values defined by col_styles:
             w_col[:, col_styles] = src_w[frame_idx, col_styles]#.cpu()
             # Generate these synthesized images:

viz/renderer.py

@@ -69,7 +69,8 @@ def add_watermark_np(input_image_array, watermark_text="AI Generated"):
 
 class Renderer:
     def __init__(self, disable_timing=False):
-        self._device        = torch.device('cuda')
+        self._device        = torch.device('cuda' if torch.cuda.is_available() else 'mps' if torch.backends.mps.is_available() else 'cpu')
+        self._dtype         = torch.float32 if self._device.type == 'mps' else torch.float64
         self._pkl_data      = dict()    # {pkl: dict | CapturedException, ...}
         self._networks      = dict()    # {cache_key: torch.nn.Module, ...}
         self._pinned_bufs   = dict()    # {(shape, dtype): torch.Tensor, ...}
@@ -241,7 +242,7 @@ def init_network(self, res,
 
         if self.w_load is None:
             # Generate random latents.
-            z = torch.from_numpy(np.random.RandomState(w0_seed).randn(1, 512)).to(self._device).float()
+            z = torch.from_numpy(np.random.RandomState(w0_seed).randn(1, 512)).to(self._device, dtype=self._dtype)
 
             # Run mapping network.
             label = torch.zeros([1, G.c_dim], device=self._device)