add base model for faster rt

app.py

@@ -43,6 +43,7 @@ def truncate_word_count(text, max_words=512):
 
 def proc_submission(
     input_text: str,
+    model_size: str,
     num_beams,
     token_batch_length,
     length_penalty,
@@ -74,6 +75,7 @@ def proc_submission(
 
     history = {}
     clean_text = clean(input_text, lower=False)
+    max_input_length = 1024 if model_size == "base" else max_input_length
     processed = truncate_word_count(clean_text, max_input_length)
     if processed["was_truncated"]:
         tr_in = processed["truncated_text"]
@@ -86,8 +88,8 @@ def proc_submission(
 
     _summaries = summarize_via_tokenbatches(
         tr_in,
-        model,
-        tokenizer,
+        model_sm if model_size == "base" else model,
+        tokenizer_sm if model_size == "base" else tokenizer,
         batch_length=token_batch_length,
         **settings,
     )
@@ -128,6 +130,7 @@ def load_examples(examples_dir="examples"):
 if __name__ == "__main__":
 
     model, tokenizer = load_model_and_tokenizer("pszemraj/led-large-book-summary")
+    model_sm, tokenizer_sm = load_model_and_tokenizer("pszemraj/led-base-book-summary")
     title = "Long-Form Summarization: LED & BookSum"
     description = "A simple demo of how to use a fine-tuned LED model to summarize long-form text. [This model](https://huggingface.co/pszemraj/led-large-book-summary) is a fine-tuned version of [allenai/led-large-16384](https://huggingface.co/allenai/led-large-16384) on the [BookSum dataset](https://arxiv.org/abs/2105.08209). The goal was to create a model that can generalize well and is useful in summarizing lots of text in academic and daily usage."
 
@@ -137,8 +140,9 @@ if __name__ == "__main__":
             gr.inputs.Textbox(
                 lines=10,
                 label="input text",
-                placeholder="Enter text to summarize, the text will be cleaned and truncated to 512 words on Spaces. Narrative, academic (both papers and lecture transcription), and article text work well.",
+                placeholder="Enter text to summarize, the text will be cleaned and truncated on Spaces. Narrative, academic (both papers and lecture transcription), and article text work well. May take a bit to generate depending on the input text :)",
             ),
+            gr.inputs.radio(choices=['base', 'large'], label='model size', default='base'),
             gr.inputs.Slider(
                 minimum=1, maximum=6, label="num_beams", default=4, step=1
             ),