README

README.ipynb

@@ -53,7 +53,7 @@
     "from PIL import Image\n",
     "from IPython.display import display\n",
     "from torchvision.transforms import ToPILImage, PILToTensor\n",
-    "from walloc.walloc import Walloc\n",
+    "from walloc import walloc\n",
     "class Args: pass"
    ]
   },
@@ -77,7 +77,7 @@
     "device = \"cpu\"\n",
     "checkpoint = torch.load(\"v0.6.3_ext.pth\",map_location=\"cpu\")\n",
     "args = checkpoint['args']\n",
-    "codec = Walloc(\n",
+    "codec = walloc.Walloc(\n",
     "    channels = args.channels,\n",
     "    J = args.J,\n",
     "    N = args.N,\n",
@@ -258,65 +258,6 @@
   {
    "cell_type": "code",
    "execution_count": 8,
-   "id": "563fb76e-404f-4a99-b99a-b2f2496fe614",
-   "metadata": {},
-   "outputs": [],
-   "source": [
-    "def concatenate_channels(x):\n",
-    "    batch_size, N, h, w = x.shape\n",
-    "    n = int(N**0.5)\n",
-    "    if n*n != N:\n",
-    "        raise ValueError(\"Number of channels must be a perfect square.\")\n",
-    "    \n",
-    "    x = x.view(batch_size, n, n, h, w)\n",
-    "    x = x.permute(0, 1, 3, 2, 4).contiguous()\n",
-    "    x = x.view(batch_size, 1, n*h, n*w)\n",
-    "    return x\n",
-    "\n",
-    "def split_channels(x, N):\n",
-    "    batch_size, _, H, W = x.shape\n",
-    "    n = int(N**0.5)\n",
-    "    h = H // n\n",
-    "    w = W // n\n",
-    "    \n",
-    "    x = x.view(batch_size, n, h, n, w)\n",
-    "    x = x.permute(0, 1, 3, 2, 4).contiguous()\n",
-    "    x = x.view(batch_size, N, h, w)\n",
-    "    return x\n",
-    "\n",
-    "def to_bytes(x, n_bits):\n",
-    "    max_value = 2**(n_bits - 1) - 1\n",
-    "    min_value = -max_value - 1\n",
-    "    if x.min() < min_value or x.max() > max_value:\n",
-    "        raise ValueError(f\"Tensor values should be in the range [{min_value}, {max_value}].\")\n",
-    "    return (x + (max_value + 1)).to(torch.uint8)\n",
-    "\n",
-    "def from_bytes(x, n_bits):\n",
-    "    max_value = 2**(n_bits - 1) - 1\n",
-    "    return (x.to(torch.float32) - (max_value + 1))\n",
-    "\n",
-    "def latent_to_pil(latent, n_bits):\n",
-    "    latent_bytes = to_bytes(latent, n_bits)\n",
-    "    concatenated_latent = concatenate_channels(latent_bytes)\n",
-    "    \n",
-    "    pil_images = []\n",
-    "    for i in range(concatenated_latent.shape[0]):\n",
-    "        pil_image = Image.fromarray(concatenated_latent[i][0].numpy(), mode='L')\n",
-    "        pil_images.append(pil_image)\n",
-    "    \n",
-    "    return pil_images\n",
-    "\n",
-    "def pil_to_latent(pil_images, N, n_bits):\n",
-    "    tensor_images = [PILToTensor()(img).unsqueeze(0) for img in pil_images]\n",
-    "    tensor_images = torch.cat(tensor_images, dim=0)\n",
-    "    split_latent = split_channels(tensor_images, N)\n",
-    "    latent = from_bytes(split_latent, n_bits)\n",
-    "    return latent"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": 9,
    "id": "60fc51ff-cb67-45ec-929d-d35af77dc3fb",
    "metadata": {},
    "outputs": [
@@ -328,19 +269,19 @@
        "<PIL.Image.Image image mode=L size=384x256>"
       ]
      },
-     "execution_count": 9,
+     "execution_count": 8,
      "metadata": {},
      "output_type": "execute_result"
     }
    ],
    "source": [
-    "Y_pil = latent_to_pil(Y,5)\n",
+    "Y_pil = walloc.latent_to_pil(Y,5)\n",
     "Y_pil[0]"
    ]
   },
   {
    "cell_type": "code",
-   "execution_count": 10,
+   "execution_count": 9,
    "id": "b027bdc2-f843-40a4-b20f-3c30fa13a3d8",
    "metadata": {},
    "outputs": [
@@ -359,7 +300,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 11,
+   "execution_count": 10,
    "id": "ed1fadec-ba42-4cc0-9a4c-2161b23de5a9",
    "metadata": {},
    "outputs": [
@@ -369,36 +310,35 @@
        "tensor(1.)"
       ]
      },
-     "execution_count": 11,
+     "execution_count": 10,
      "metadata": {},
      "output_type": "execute_result"
     }
    ],
    "source": [
-    "Y2 = pil_to_latent(Y_pil, 16, 5)\n",
+    "Y2 = walloc.pil_to_latent(Y_pil, 16, 5)\n",
     "(Y == Y2).sum()/Y.numel()"
    ]
   },
   {
    "cell_type": "code",
-   "execution_count": 13,
+   "execution_count": null,
    "id": "0ae9e5bb-d616-4d22-a9c9-6743e4216802",
    "metadata": {},
-   "outputs": [
-    {
-     "name": "stdout",
-     "output_type": "stream",
-     "text": [
-      "[NbConvertApp] Converting notebook README.ipynb to markdown\n",
-      "[NbConvertApp] Support files will be in README_files/\n",
-      "[NbConvertApp] Making directory README_files\n",
-      "[NbConvertApp] Writing 5751 bytes to README.md\n"
-     ]
-    }
-   ],
+   "outputs": [],
    "source": [
     "!jupyter nbconvert --to markdown README.ipynb"
    ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "3d4794d5-a26a-4789-804e-7aa543db89d6",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "!sed -i 's|!\\[png](README_files/\\(README_[0-9]*_[0-9]*\\.png\\))|![png](https://huggingface.co/danjacobellis/walloc/resolve/main/README_files/\\1)|g' README.md"
+   ]
   }
  ],
  "metadata": {

README.md

@@ -33,7 +33,7 @@ import numpy as np
 from PIL import Image
 from IPython.display import display
 from torchvision.transforms import ToPILImage, PILToTensor
-from walloc.walloc import Walloc
+from walloc import walloc
 class Args: pass
 ```
 
@@ -46,7 +46,7 @@ class Args: pass
 device = "cpu"
 checkpoint = torch.load("v0.6.3_ext.pth",map_location="cpu")
 args = checkpoint['args']
-codec = Walloc(
+codec = walloc.Walloc(
     channels = args.channels,
     J = args.J,
     N = args.N,
@@ -71,7 +71,7 @@ img
 
 
     
-![png](README_files/README_6_0.png)
+![png](https://huggingface.co/danjacobellis/walloc/resolve/main/README_files/README_6_0.png)
     
 
 
@@ -96,7 +96,7 @@ ToPILImage()(x_hat[0]+0.5)
 
 
     
-![png](README_files/README_8_0.png)
+![png](https://huggingface.co/danjacobellis/walloc/resolve/main/README_files/README_8_0.png)
     
 
 
@@ -147,7 +147,7 @@ plt.xticks(range(-15,16,5));
 
 
     
-![png](README_files/README_12_0.png)
+![png](https://huggingface.co/danjacobellis/walloc/resolve/main/README_files/README_12_0.png)
     
 
 
@@ -155,61 +155,7 @@ plt.xticks(range(-15,16,5));
 
 
 ```python
-def concatenate_channels(x):
-    batch_size, N, h, w = x.shape
-    n = int(N**0.5)
-    if n*n != N:
-        raise ValueError("Number of channels must be a perfect square.")
-    
-    x = x.view(batch_size, n, n, h, w)
-    x = x.permute(0, 1, 3, 2, 4).contiguous()
-    x = x.view(batch_size, 1, n*h, n*w)
-    return x
-
-def split_channels(x, N):
-    batch_size, _, H, W = x.shape
-    n = int(N**0.5)
-    h = H // n
-    w = W // n
-    
-    x = x.view(batch_size, n, h, n, w)
-    x = x.permute(0, 1, 3, 2, 4).contiguous()
-    x = x.view(batch_size, N, h, w)
-    return x
-
-def to_bytes(x, n_bits):
-    max_value = 2**(n_bits - 1) - 1
-    min_value = -max_value - 1
-    if x.min() < min_value or x.max() > max_value:
-        raise ValueError(f"Tensor values should be in the range [{min_value}, {max_value}].")
-    return (x + (max_value + 1)).to(torch.uint8)
-
-def from_bytes(x, n_bits):
-    max_value = 2**(n_bits - 1) - 1
-    return (x.to(torch.float32) - (max_value + 1))
-
-def latent_to_pil(latent, n_bits):
-    latent_bytes = to_bytes(latent, n_bits)
-    concatenated_latent = concatenate_channels(latent_bytes)
-    
-    pil_images = []
-    for i in range(concatenated_latent.shape[0]):
-        pil_image = Image.fromarray(concatenated_latent[i][0].numpy(), mode='L')
-        pil_images.append(pil_image)
-    
-    return pil_images
-
-def pil_to_latent(pil_images, N, n_bits):
-    tensor_images = [PILToTensor()(img).unsqueeze(0) for img in pil_images]
-    tensor_images = torch.cat(tensor_images, dim=0)
-    split_latent = split_channels(tensor_images, N)
-    latent = from_bytes(split_latent, n_bits)
-    return latent
-```
-
-
-```python
-Y_pil = latent_to_pil(Y,5)
+Y_pil = walloc.latent_to_pil(Y,5)
 Y_pil[0]
 ```
 
@@ -217,7 +163,7 @@ Y_pil[0]
 
 
     
-![png](README_files/README_15_0.png)
+![png](https://huggingface.co/danjacobellis/walloc/resolve/main/README_files/README_14_0.png)
     
 
 
@@ -233,7 +179,7 @@ print("compression_ratio: ", x.numel()/os.path.getsize("latent.png"))
 
 
 ```python
-Y2 = pil_to_latent(Y_pil, 16, 5)
+Y2 = walloc.pil_to_latent(Y_pil, 16, 5)
 (Y == Y2).sum()/Y.numel()
 ```
 
@@ -249,7 +195,7 @@ Y2 = pil_to_latent(Y_pil, 16, 5)
 !jupyter nbconvert --to markdown README.ipynb
 ```
 
-    [NbConvertApp] Converting notebook README.ipynb to markdown
-    [NbConvertApp] Support files will be in README_files/
-    [NbConvertApp] Writing 5751 bytes to README.md
 
+```python
+!sed -i 's|!\[png](README_files/\(README_[0-9]*_[0-9]*\.png\))|![png](https://huggingface.co/danjacobellis/walloc/resolve/main/README_files/\1)|g' README.md
+```