Update README.md

README.md

@@ -22,7 +22,7 @@ widget:
 inference:
   parameters:
     no_repeat_ngram_size: 2
-    max_length: 128
+    max_length: 64
     early_stopping: True
 ---
 
@@ -32,6 +32,7 @@ inference:
 - This model was trained a text-to-text task with input text as a summary of a chapter, and the output text as the analysis of that chapter on the [booksum](https://arxiv.org/abs/2105.08209) dataset.
 - it has somewhat learned how to complete literary analysis on an arbitrary input text.
 - **NOTE: this is fairly intensive computationally and recommended to be run on GPU. please see example usage in [this demo notebook](https://colab.research.google.com/gist/pszemraj/8e9cc5bee5cac7916ef9241b66e01b05/demo-t5-large-for-lexical-analysis.ipynb)**
+  - The API is set to return max 64 tokens to avoid timeouts on CPU.
 
 ## Example