Ts vae (#69)

* start time vae

* vae model

* add fix bugs in timevae

* timevae mlp experiments

* timevae mlp experiments

* adjust timevae

* use different activation for decoder in timevae

* clean up timevae

* add timevae

* reload from checkpoint

* timevae evaluation

dl/notebooks/time_vae.py

@@ -738,9 +738,17 @@ def configure_optimizers(self) -> dict:
 
 # ## Fitted Model
 
-for i in time_vae_dm.predict_dataloader():
-    print(i.size())
-    i_pred = vae_model.model(i)
+IS_RELOAD = True
+
+if IS_RELOAD:
+    checkpoint_path = "lightning_logs/time_vae_naive/version_29/checkpoints/epoch=1999-step=354000.ckpt"
+    vae_model_reloaded = VAEModel.load_from_checkpoint(checkpoint_path, model=vae)
+else:
+    vae_model_reloaded = vae_model
+
+for pred_batch in time_vae_dm.predict_dataloader():
+    print(pred_batch.size())
+    i_pred = vae_model_reloaded.model(pred_batch.float().cuda())
     break
 
 i_pred[0].size()
@@ -752,23 +760,70 @@ def configure_optimizers(self) -> dict:
 
 element = 4
 
-ax.plot(i.detach().numpy()[element, :, 0])
-ax.plot(i_pred[0].detach().numpy()[element, :, 0], "x-")
+ax.plot(pred_batch.detach().numpy()[element, :, 0])
+ax.plot(i_pred[0].cpu().detach().numpy()[element, :, 0], "x-")
 # -
 
 # Data generation using the decoder.
 
-sampling_z = torch.randn(2, vae_model.model.encoder.params.latent_size).type_as(
-    vae_model.model.encoder.z_mean_layer.weight
+sampling_z = torch.randn(
+    pred_batch.size(0), vae_model_reloaded.model.encoder.params.latent_size
+).type_as(vae_model_reloaded.model.encoder.z_mean_layer.weight)
+generated_samples_x = (
+    vae_model_reloaded.model.decoder(sampling_z).cpu().detach().numpy().squeeze()
 )
-sampling_x = vae_model.model.decoder(sampling_z)
 
-sampling_x.size()
+generated_samples_x.size()
 
 # +
 _, ax = plt.subplots()
 
-for i in range(min(len(sampling_x), 4)):
-    ax.plot(sampling_x.detach().numpy()[i, :, 0], "x-")
+for i in range(min(len(generated_samples_x), 4)):
+    ax.plot(generated_samples_x[i, :], "x-")
 
 # -
+from openTSNE import TSNE
+
+n_tsne_samples = 100
+
+original_samples = pred_batch.cpu().detach().numpy().squeeze()[:n_tsne_samples]
+original_samples.shape
+
+tsne = TSNE(
+    perplexity=30,
+    metric="euclidean",
+    n_jobs=8,
+    random_state=42,
+    verbose=True,
+)
+
+original_samples_embedding = tsne.fit(original_samples)
+
+generated_samples_x[:n_tsne_samples]
+
+generated_samples_embedding = original_samples_embedding.transform(
+    generated_samples_x[:n_tsne_samples]
+)
+
+# +
+fig, ax = plt.subplots(figsize=(7, 7))
+
+ax.scatter(
+    original_samples_embedding[:, 0],
+    original_samples_embedding[:, 1],
+    color="black",
+    marker=".",
+    label="original",
+)
+
+ax.scatter(
+    generated_samples_embedding[:, 0],
+    generated_samples_embedding[:, 1],
+    color="red",
+    marker="x",
+    label="generated",
+)
+
+ax.set_title("t-SNE of original and generated samples")
+ax.set_xlabel("t-SNE 1")
+ax.set_ylabel("t-SNE 2")

pyproject.toml

@@ -32,6 +32,7 @@ jupytext = "^1.15.2"
 
 [tool.poetry.group.visualization.dependencies]
 seaborn = "^0.13.2"
+opentsne = "^1.0.2"
 
 
 [tool.poetry.group.data.dependencies]