Merge branch 'develop'

CHANGELOG.md

@@ -6,6 +6,18 @@ Most recent releases are shown at the top. Each release shows:
 - **Changed**: Additional parameters, changes to inputs or outputs, etc
 - **Fixed**: Bug fixes that don't change documented behaviour
 
+## 0.19.1 (2020-07-29)
+
+### New:
+- N/A
+
+### Changed
+- Adjusted `no_grad` scope in `ZeroShotClassifier.predict`
+
+### Fixed:
+- N/A
+
+
 ## 0.19.0 (2020-07-29)
 
 ### New:

ktrain/text/zsl/core.py

@@ -48,27 +48,27 @@ def predict(self, doc, topic_strings=[], include_labels=False, max_length=512, b
           inferred probabilities
         """
         import torch
-        if topic_strings is None or len(topic_strings) == 0:
-            raise ValueError('topic_strings must be a list of strings')
-        if batch_size > len(topic_strings): batch_size = len(topic_strings)
-        topic_chunks = list(U.list2chunks(topic_strings, n=math.ceil(len(topic_strings)/batch_size)))
-        if len(topic_strings) >= 100 and batch_size==8:
-            warnings.warn('TIP: Try increasing batch_size to speedup ZeroShotClassifier predictions')
-        result = []
-        for topics in topic_chunks:
-            pairs = []
-            for topic_string in topics:
-                premise = doc
-                hypothesis = 'This text is about %s.' % (topic_string)
-                pairs.append( (premise, hypothesis) )
-            batch = self.tokenizer.batch_encode_plus(pairs, return_tensors='pt', max_length=max_length, truncation='only_first', padding=True).to(self.torch_device)
-            with torch.no_grad():
+        with torch.no_grad():
+            if topic_strings is None or len(topic_strings) == 0:
+                raise ValueError('topic_strings must be a list of strings')
+            if batch_size > len(topic_strings): batch_size = len(topic_strings)
+            topic_chunks = list(U.list2chunks(topic_strings, n=math.ceil(len(topic_strings)/batch_size)))
+            if len(topic_strings) >= 100 and batch_size==8:
+                warnings.warn('TIP: Try increasing batch_size to speedup ZeroShotClassifier predictions')
+            result = []
+            for topics in topic_chunks:
+                pairs = []
+                for topic_string in topics:
+                    premise = doc
+                    hypothesis = 'This text is about %s.' % (topic_string)
+                    pairs.append( (premise, hypothesis) )
+                batch = self.tokenizer.batch_encode_plus(pairs, return_tensors='pt', max_length=max_length, truncation='only_first', padding=True).to(self.torch_device)
                 logits = self.model(batch['input_ids'], attention_mask=batch['attention_mask'])[0]
-            entail_contradiction_logits = logits[:,[0,2]]
-            probs = entail_contradiction_logits.softmax(dim=1)
-            true_probs = list(probs[:,1].cpu().detach().numpy())
-            if include_labels:
-                true_probs = list(zip(topics, true_probs))
-            result.extend(true_probs)
-        return result
+                entail_contradiction_logits = logits[:,[0,2]]
+                probs = entail_contradiction_logits.softmax(dim=1)
+                true_probs = list(probs[:,1].cpu().detach().numpy())
+                if include_labels:
+                    true_probs = list(zip(topics, true_probs))
+                result.extend(true_probs)
+            return result
 

ktrain/version.py

@@ -1,2 +1,2 @@
 __all__ = ['__version__']
-__version__ = '0.19.0'
+__version__ = '0.19.1'