Updating the sample

jloveric · Nov 19, 2023 · a165987 · a165987
1 parent cea24ab
commit a165987
Showing 1 changed file with 27 additions and 15 deletions.
diff --git a/language_interpolation/utils.py b/language_interpolation/utils.py
@@ -107,24 +107,24 @@ def generate_transformer_text(
     model.eval()
 
     for index, text in enumerate(text_list):
-        just = ((len(text) // characters_per_feature)+1) * characters_per_feature
+        just = ((len(text) // characters_per_feature) + 1) * characters_per_feature
         text_list[index] = text.rjust(just)
 
-    print('text_list', text_list)
-    print('text lengths', [len(text) for text in text_list])
+    print("text_list", text_list)
+    print("text lengths", [len(text) for text in text_list])
     results = []
     for text_in in text_list:
         for i in range(output_size):
             encoding, text_used = encode_input_from_text(
                 text_in=text_in, features=max_characters
             )
-            print('encoding length', len(encoding), encoding.shape)
+            print("encoding length", len(encoding), encoding.shape)
             encoding = (
                 ascii_to_float(encoding)
                 .to(model._device)
                 .reshape(1, -1, characters_per_feature)
             )
-            print('encoding', encoding)
+            print("encoding", encoding)
             model.eval()
             output = model(encoding)
             values, indices, ascii = decode_output_to_text(
@@ -135,10 +135,11 @@ def generate_transformer_text(
             # prevents the same response for every query.
             actual = random.choices(ascii, values.tolist())
             text_in = text_in + actual[0]
-            just = ((len(text_in) // characters_per_feature)+1) * characters_per_feature
+            just = (
+                (len(text_in) // characters_per_feature) + 1
+            ) * characters_per_feature
             text_in = text_in.rjust(just)
 
-
         results.append(text_in.replace("\n", " "))
 
     return results
@@ -152,14 +153,25 @@ def __init__(self, cfg):
     def on_train_epoch_end(self, trainer, pl_module, outputs=None):
         with torch.no_grad():
             for topk in range(1, self._cfg.topk + 1):
-                predictions = generate_text(
-                    pl_module,
-                    features=self._cfg.net.features,
-                    text_list=self._cfg.prompts,
-                    output_size=self._cfg.num_predict,
-                    topk=topk,
-                    add_channel_dimension=self._cfg.data.add_channel_dimension,
-                )
+                if self._cfg.model_type == "high_order_transformer":
+                    predictions = generate_transformer_text(
+                        pl_module,
+                        characters_per_feature=self._cfg.data.characters_per_feature,
+                        max_characters=self._cfg.data.characters_per_feature
+                        * self._cfg.data.max_features,
+                        text_list=self._cfg.prompts,
+                        output_size=self._cfg.num_predict,
+                        topk=topk,
+                    )
+                else:
+                    predictions = generate_text(
+                        pl_module,
+                        features=self._cfg.net.features,
+                        text_list=self._cfg.prompts,
+                        output_size=self._cfg.num_predict,
+                        topk=topk,
+                        add_channel_dimension=self._cfg.data.add_channel_dimension,
+                    )
 
                 for index, text in enumerate(predictions):
                     trainer.logger.experiment.add_text(