Update README.md

README.md

@@ -120,7 +120,7 @@ data curation, model training, and model evaluation.
 
 ## Generation
 
-You can generate your own examples as follows:
+You can generate your own examples as follows.  For a "complete" patent, you'll want to extend the `max_new_tokens` value to the biggest number you can fit in your available VRAM.
 
 ```python
 import json
@@ -130,23 +130,22 @@ from transformers import pipeline
 p = pipeline('text-generation', 'alea-institute/kl3m-002-170m-patent', device='cpu')
 
 # Example usage on CPU
-text = "# Title\n"
+text = "# Patent\n\n## Title"
 print(
     json.dumps(
         [
             r.get("generated_text")
-            for r in p(text, do_sample=True, temperature=0.5, num_return_sequences=3, max_new_tokens=64)
+            for r in p(text, do_sample=True, temperature=0.5, num_return_sequences=3, max_new_tokens=32)
         ], 
         indent=2
     )
 )
 ```
 
-```json
 [
-  "# Title\nElectrode array and method for producing electrode array and electrode array using same\n\n## Abstract\nAn electrode array includes a plurality of electro",
-  "# Title\nMethods and systems for using a neural network to perform a plurality of operations on\na multi-layer path in a distributed manner in a ",
-  "# Title\nAnti-soluble compositions and methods for their use in the treatment of an eye\ndisease or condition mediated by soluble or non-soluble "
+  "# Patent\n\n## Title\nMethod for manufacturing a temperature-controllable polyurethane composition and method",
+  "# Patent\n\n## Title\nElectronic device\n\n## Abstract\nAn electronic device includes a display panel and a",
+  "# Patent\n\n## Title\nMethods and devices for tissue repair using a neural network\n\n## Abstract"
 ]
 ```