prepping final release of code

README.md

@@ -14,9 +14,13 @@ Here are instructions on how to run the repository:
     ```
 2. Run the evaluation script:
     ```shell
-    conda activate keypoint_correspondences_ldm
+    conda activate LDM_correspondences
     python3 -m eval.eval
     ```
+3. More options can be found with 
+    ```shell
+    python3 -m eval.eval --help
+    ```
 
 ## Visualizing Attention Maps
 
@@ -25,7 +29,7 @@ The project includes an interactive local website for visualizing attention maps
 1. Activate the conda environment and run the evaluation script with visualization:
 
     ```shell
-    conda activate keypoint_correspondences_ldm
+    conda activate LDM_correspondences
     python3 -m eval.eval --visualize
     ```
 

environment.yml

@@ -1,6 +1,5 @@
-name: keypoint_correspondences_ldm
+name: LDM_correspondences
 channels:
-  - xformers/label/dev
   - pytorch
   - nvidia
   - conda-forge
@@ -154,7 +153,6 @@ dependencies:
   - urllib3=1.26.14=pyhd8ed1ab_0
   - wcwidth=0.2.5=pyhd3eb1b0_0
   - wheel=0.37.1=pyhd3eb1b0_0
-  - xformers=0.0.17.dev435+git.5df1f0b=py310_cu11.6_pyt1.12.1
   - xz=5.2.8=h5eee18b_0
   - zeromq=4.3.4=h2531618_0
   - zlib=1.2.13=h5eee18b_0
@@ -240,4 +238,3 @@ dependencies:
       - werkzeug==2.3.0
       - widgetsnbextension==4.0.5
       - zipp==3.11.0
-prefix: /scratch/iamerich/miniconda3/envs/prompt-to-prompt

eval/eval.py

@@ -32,8 +32,6 @@
                         choices=['spair', 'pfwillow', 'cubs', 'custom'], default='custom')
     parser.add_argument('--thres', type=str, default='auto',
                         choices=['auto', 'img', 'bbox'])
-    parser.add_argument('--alpha', type=float, default=0.1,
-                        help='alpha for the pck threshold')
     parser.add_argument('--sub_class', type=str, default="all", choices=['aeroplane', 'bicycle', 'bird', 'boat', 'bottle', 'bus',
                                                                          'car', 'cat', 'chair', 'cow', 'dog', 'horse', 'motorbike', 'person', 'pottedplant', 'sheep', 'train', 'tvmonitor', 'all'])
     parser.add_argument('--split', type=str, default="test",
@@ -65,7 +63,6 @@
                         default='CompVis/stable-diffusion-v1-4', help='ldm model type')
     parser.add_argument('--upsample_res', type=int, default=512,
                         help='Resolution to upsample the attention maps to')
-    parser.add_argument('--num_words', type=int, default=2)
 
     # Run details
     parser.add_argument('--wandb_log', action='store_true',
@@ -83,9 +80,9 @@
                         help='what epoch of the model to load')
     parser.add_argument('--save_loc', type=str, default='outputs',
                         help='save location for the trained model')
-    parser.add_argument('--results_loc', type=str, default='/scratch/iamerich/prompt-to-prompt/outputs/ldm_visualization_020',
+    parser.add_argument('--results_loc', type=str, default='outputs/',
                         help='save location for the trained model')
-    parser.add_argument('--seed', type=int, default=2023,
+    parser.add_argument('--seed', type=int, default=-1,
                         help='Pseudo-RNG seed')
     parser.add_argument('--ablate', action='store_true',
                         help='evaluate over a smaller number of points')
@@ -94,10 +91,6 @@
 
     args = parser.parse_args()
 
-    # from glob import glob
-    # if args.item_index != -1 and len(glob(f"/home/iamerich/burst/pfwillow_no_crop_ablation/{args.item_index}/*.txt")) > 0:
-    #     exit()
-
 
     if args.seed != -1:
         random.seed(args.seed)

utils/optimize.py

@@ -149,7 +149,7 @@ def retest(ldm,
             item_index = -1,
             ablate_results = False,
             num_iterations = 20,
-            results_loc = "/scratch/iamerich/prompt-to-prompt/outputs/ldm_visualization_020",
+            results_loc = "outputs/",
             save_folder = "outputs"):
     """
     Takes the saved text embeddings and re-evaluates them