update huggingface app to use the new models

Author: GreenWizard2015

configs/experiments/diffusion/ddim.json

@@ -8,7 +8,8 @@
     "restorator": {
       "sampler": {
         "name": "DDIM",
-        "stochasticity": 0.0,
+        "stochasticity": 1.0,
+        "project noise": true,
         "steps skip type": {
           "name": "uniform",
           "K": 5

huggingface/app.py

@@ -35,36 +35,33 @@ def run2inference(run, runName=None):
   runConfig['huggingface'] = { "name": runName, "wandb": run.fullId }
   return list(inference_from_config(runConfig))
 
-def SPModels(project):
-  bestPerGroup = project.groups(onlyBest=True)
-  bestPerGroup = {k: v for k, v in bestPerGroup.items() if k.startswith('Single-pass | ')}
-  for run in bestPerGroup.values():
-    for model in run2inference(run):
-      yield model
-    continue
-  return
 
 def main(args):
   WBProject = CWBProject('green_wizard/FranNet')
   folder = os.path.dirname(os.path.abspath(__file__))
   # load list of models from the folder "configs"
-  models = modelsFrom(os.path.join(folder, 'configs'))
-  # add some models from W&B
-  models.extend( list(SPModels(WBProject)) )
+  models = []
   bestGroups = WBProject.groups(onlyBest=True)
   customRuns = [
-    'DDPM v2 | Basic',
-    # some of "DDPM-V, encoder v2, masking-8, residual"
-    'DDPM-V, encoder v2, masking-8, residual, RGB',
-    'DDPM-V, encoder v2, masking-8, residual, LAB',
-    'DDPM-V, encoder v2, masking-8, residual, LAB, structured',
-    # autorregressive direction
+    'Single-pass | Basic+masking-8',
+    'Single-pass | Basic+masking-uniform',
+    'Single-pass | Basic+masking-dynamic',
+    'Single-pass | Basic',
+    'Single-pass | Basic+masking-4',
+    'Single-pass | Basic+masking-32',
+    'Single-pass | Basic+masking-16',
     'AR | direction',
     'AR direction, encoder v2, masking-8, residual, LAB, structured',
     'AR direction, masking-8, RGB, extra(grayscale), structured',
-
+    'DDPM-V, encoder v2, residual, RGB, extra(grayscale), structured, T-start',
+    'DDPM-V, encoder v2, residual, RGB, extra(grayscale), structured, huge',
+    'Single-pass | Basic+masking-16, encoder v2, huber, RGB, residual, extra(grayscale), structured',
+    'Huge++',
+    'DDPM-V | ONLY SR | huber, RGB, residual, extra(grayscale), structured, T start',
+    'DDPM-V | ONLY SR | encoder v2, huber, RGB, residual, extra(grayscale), structured, T start',
+    'AR direction | ONLY SR | masking-16, encoder v2, huber, RGB, residual, extra(grayscale), structured',
+    'Single-pass | ONLY SR | Basic+masking-16, encoder v2, huber, RGB, residual, extra(grayscale), structured',
     'huge++, only SR',
-    'Huge++'
   ]
   for runName in customRuns:
     models.extend( run2inference(bestGroups[runName]) )