From 036a0c608e4e43f5da3883c6324a7fca860b908b Mon Sep 17 00:00:00 2001 From: Jules Laplace Date: Sun, 23 Feb 2020 02:58:50 +0100 Subject: more vgg --- cli/app/search/json.py | 2 +- cli/app/search/search_dense.py | 5 +---- 2 files changed, 2 insertions(+), 5 deletions(-) diff --git a/cli/app/search/json.py b/cli/app/search/json.py index 69ce7a5..9b82578 100644 --- a/cli/app/search/json.py +++ b/cli/app/search/json.py @@ -71,7 +71,7 @@ def make_params_dense(tag, folder_id): # "inv_layer": "latent", "decay_lr": True, # "inv_it": 10000, - "inv_it": 12000, + "inv_it": 10000, "generator_path": "https://tfhub.dev/deepmind/biggan-512/2", "attention_map_layer": "Generator_2/attention/Softmax:0", "pre_trained_latent": True, diff --git a/cli/app/search/search_dense.py b/cli/app/search/search_dense.py index 53c548b..7b83952 100644 --- a/cli/app/search/search_dense.py +++ b/cli/app/search/search_dense.py @@ -419,7 +419,6 @@ def find_dense_embedding_for_images(params, opt_tag="inverse_" + timestamp(), op out_fns[:] = sample_fns[:NUM_IMGS_TO_PROCESS] # Gradient descent w.r.t. generator's inputs. - it = 0 out_pos = 0 start_time = time.time() @@ -450,7 +449,7 @@ def find_dense_embedding_for_images(params, opt_tag="inverse_" + timestamp(), op # Main optimization loop. print("Beginning dense iteration...") - for _ in range(params.inv_it): + for it in range(params.inv_it): if it < params.inv_it * 0.7: n = 0.0 @@ -498,8 +497,6 @@ def find_dense_embedding_for_images(params, opt_tag="inverse_" + timestamp(), op inv_batch = vs.grid_transform(inv_batch) vs.save_image('{}/progress_{}_{:04d}.png'.format(SAMPLES_DIR, opt_tag, int(it / 500)), inv_batch) - it += 1 - # Save images that are ready. label_trained, latent_trained = sess.run([label, latent]) if params.inv_layer != 'latent': -- cgit v1.2.3-70-g09d2