Merge pull request #60 from jalammar/hotfix-jan-2022-1

Hotfix jan 2022 1

setup.py

@@ -25,7 +25,7 @@ def read(*names, **kwargs):
 
 setup(
     name='ecco',
-    version='0.1.0',
+    version='0.1.1',
     license='BSD-3-Clause',
     description='Visualization tools for NLP machine learning models.',
     long_description='%s\n%s' % (

src/ecco/__init__.py

@@ -12,7 +12,7 @@
 """
 
 
-__version__ = '0.0.15'
+__version__ = '0.1.1'
 from ecco.lm import LM
 from transformers import AutoTokenizer, AutoModelForCausalLM, AutoModel, AutoModelForSeq2SeqLM
 from typing import Any, Dict, Optional, List

src/ecco/lm.py

@@ -199,7 +199,10 @@ def generate(self, input_str: str,
         # Get decoder input ids
         if self.model_type == 'enc-dec': # FIXME: only done because causal LMs like GPT-2 have the _prepare_decoder_input_ids_for_generation method but do not use it
             assert len(input_ids.size()) == 2 # will break otherwise
-            decoder_input_ids = self.model._prepare_decoder_input_ids_for_generation(input_ids, None, None)
+            if transformers.__version__ >= '4.13': # ALSO FIXME: awful hack. But seems to work?
+                decoder_input_ids = self.model._prepare_decoder_input_ids_for_generation(input_ids.shape[0], None, None)
+            else:
+                decoder_input_ids = self.model._prepare_decoder_input_ids_for_generation(input_ids, None, None)
         else:
             decoder_input_ids = None