fixes for next_token_lm (#85)

* fixes for next_token_lm * update CHANGELOG
allenai · Jun 30, 2020 · 0ce14da · 0ce14da
1 parent 37136f8
commit 0ce14da
Show file tree

Hide file tree

Showing 3 changed files with 5 additions and 2 deletions.
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -10,6 +10,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 ### Fixed
 
 - `CopyNetSeq2Seq` model now works with pretrained transformers.
+- A bug with `NextTokenLM` that caused simple gradient interpreters to fail.
 
 ### Changed
 

diff --git a/allennlp_models/lm/models/next_token_lm.py b/allennlp_models/lm/models/next_token_lm.py
@@ -131,7 +131,6 @@ def make_output_human_readable(
  )
  output_dict["words"] = top_words
  tokens = []
- print(output_dict["token_ids"])
  for instance_tokens in output_dict["token_ids"]:
  tokens.append(
  [

diff --git a/allennlp_models/lm/predictors/next_token_lm.py b/allennlp_models/lm/predictors/next_token_lm.py
@@ -20,7 +20,10 @@ def predictions_to_labeled_instances(
  ):
  new_instance = instance.duplicate()
  token_field: TextField = instance["tokens"] # type: ignore
- mask_targets = [Token(target_top_k[0]) for target_top_k in outputs["words"]]
+ mask_targets = [
+ Token(target_top_k_text[0], text_id=target_top_id_id)
+ for (target_top_k_text, target_top_id_id) in zip(outputs["words"], outputs["token_ids"])
+ ]
 
  new_instance.add_field(
  "target_ids",