32353912

diffusion.py

@@ -234,7 +234,7 @@ class TrainConfig:
     ###########################
     ## hardcoding these here ##
     ###########################
-    push_to_hub = True 
+    push_to_hub = False #True 
     hub_model_id = "Xsmos/ml21cm"
     hub_private_repo = False
     dataset_name = "/storage/home/hcoda1/3/bxia34/scratch/LEN128-DIM64-CUB8.h5"
@@ -438,13 +438,13 @@ class DDPM21CM:
 
         del dataset
 
-    def transform(self, img, idx):
+    def transform(self, img, idx=0):
         #flip along x or y or both
-        flip_xy = [i+1 for i in range(2) if getrandbits(1)]
+        flip_xy = [i+2 for i in range(2) if getrandbits(1)]
         img[idx] = torch.flip(img[idx], dims=flip_xy) 
         # flip diagonally 
         if getrandbits(1):
-            img[idx] = img[idx].clone().transpose(1,2)
+            img = img.transpose(2,3)
             #print(f"transform: img.shape={img.shape}, idx={idx}, flip_xy={flip_xy}, w/ transpose")
         #else:
             #print(f"transform: img.shape={img.shape}, idx={idx}, flip_xy={flip_xy}, w/o tranpose")
@@ -499,8 +499,9 @@ class DDPM21CM:
             epoch_start = time()
             for i, (x, c) in enumerate(self.dataloader):
                 if self.config.dim == 3:
-                    for idx in range(len(x)):
-                        x = self.transform(x, idx)
+                    x = self.transform(x)
+                    #for idx in range(len(x)):
+                    #    x = self.transform(x, idx)
 
                 x = x.to(self.config.device)#.to(self.config.dtype)
                 # autocast forward propogation

perlmutter_diffusion.sbatch

@@ -3,9 +3,9 @@
 #SBATCH -J diffusion
 #SBATCH -C gpu&hbm80g
 #SBATCH -q regular #shared 
-#SBATCH -N4
+#SBATCH -N1
 #SBATCH --gpus-per-node=4
-#SBATCH -t 16:00:00
+#SBATCH -t 16:00
 #SBATCH --ntasks-per-node=1
 #SBATCH -oReport-%j 
 #SBATCH --mail-type=BEGIN,END,FAIL

tensorboard.ipynb

@@ -2,21 +2,12 @@
  "cells": [
   {
    "cell_type": "code",
-   "execution_count": 4,
+   "execution_count": 3,
    "id": "ae45e44e-a11c-43ef-b830-c7a58a72f51e",
    "metadata": {
     "tags": []
    },
-   "outputs": [
-    {
-     "name": "stdout",
-     "output_type": "stream",
-     "text": [
-      "The tensorboard extension is already loaded. To reload it, use:\n",
-      "  %reload_ext tensorboard\n"
-     ]
-    }
-   ],
+   "outputs": [],
    "source": [
     "import nersc_tensorboard_helper\n",
     "%load_ext tensorboard"
@@ -24,30 +15,21 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 5,
+   "execution_count": 4,
    "id": "a5c088b8-5051-402f-b4ec-2b684ad5a952",
    "metadata": {},
    "outputs": [
-    {
-     "data": {
-      "text/plain": [
-       "Reusing TensorBoard on port 33249 (pid 774528), started 4:01:53 ago. (Use '!kill 774528' to kill it.)"
-      ]
-     },
-     "metadata": {},
-     "output_type": "display_data"
-    },
     {
      "data": {
       "text/html": [
        "\n",
-       "      <iframe id=\"tensorboard-frame-a8fadb66b88e501a\" width=\"100%\" height=\"800\" frameborder=\"0\">\n",
+       "      <iframe id=\"tensorboard-frame-7f9f2d9b643f0d7b\" width=\"100%\" height=\"800\" frameborder=\"0\">\n",
        "      </iframe>\n",
        "      <script>\n",
        "        (function() {\n",
-       "          const frame = document.getElementById(\"tensorboard-frame-a8fadb66b88e501a\");\n",
+       "          const frame = document.getElementById(\"tensorboard-frame-7f9f2d9b643f0d7b\");\n",
        "          const url = new URL(\"/\", window.location);\n",
-       "          const port = 33249;\n",
+       "          const port = 40725;\n",
        "          if (port) {\n",
        "            url.port = port;\n",
        "          }\n",
@@ -70,14 +52,14 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 6,
+   "execution_count": 5,
    "id": "2f76c0a9-2218-4073-86aa-f4f655d7642f",
    "metadata": {},
    "outputs": [
     {
      "data": {
       "text/html": [
-       "<a href=\"https://jupyter.nersc.gov/user/binxia/perlmutter-login-node-base/proxy/33249/\">https://jupyter.nersc.gov/user/binxia/perlmutter-login-node-base/proxy/33249/</a>"
+       "<a href=\"https://jupyter.nersc.gov/user/binxia/perlmutter-login-node-base/proxy/40725/\">https://jupyter.nersc.gov/user/binxia/perlmutter-login-node-base/proxy/40725/</a>"
       ],
       "text/plain": [
        "<IPython.core.display.HTML object>"