Spaces:

numind
/

NuExtract-1.5

Running on L4

Alexandre-Numind commited on Mar 24

Commit

111fbee

verified ·

1 Parent(s): 64d8609

Update app.py

Files changed (1) hide show

app.py CHANGED Viewed

@@ -14,9 +14,9 @@ from examples import examples as input_examples
 from nuextract_logging import log_event
-MAX_INPUT_SIZE = 10_000
-MAX_NEW_TOKENS = 4_000
-MAX_WINDOW_SIZE = 4_000
 markdown_description = """
 <!DOCTYPE html>
@@ -139,13 +139,13 @@ model = AutoModelForCausalLM.from_pretrained(model_name,
 tokenizer = AutoTokenizer.from_pretrained(model_name, use_auth_token=auth_token)
 model.eval()
-def gradio_interface_function(template, text, is_example):
     if len(tokenizer.tokenize(text)) > MAX_INPUT_SIZE:
         yield "", "Input text too long for space. Download model to use unrestricted.", ""
         return  # End the function since there was an error
     # Initialize the sliding window prediction process
-    prediction_generator = sliding_window_prediction(template, text, model, tokenizer, window_size=MAX_WINDOW_SIZE)
     # Iterate over the generator to return values at each step
     for progress, full_pred, html_content in prediction_generator:
@@ -163,6 +163,7 @@ iface = gr.Interface(
     inputs=[
         gr.Textbox(lines=2, placeholder="Enter Template here...", label="Template"),
         gr.Textbox(lines=2, placeholder="Enter input Text here...", label="Input Text"),
         gr.Checkbox(label="Is Example?", visible=False),
     ],
     outputs=[

 from nuextract_logging import log_event
+MAX_INPUT_SIZE = 100_000
+MAX_NEW_TOKENS = 8_000
+MAX_WINDOW_SIZE = 1_000
 markdown_description = """
 <!DOCTYPE html>
 tokenizer = AutoTokenizer.from_pretrained(model_name, use_auth_token=auth_token)
 model.eval()
+def gradio_interface_function(template, text, size, is_example):
     if len(tokenizer.tokenize(text)) > MAX_INPUT_SIZE:
         yield "", "Input text too long for space. Download model to use unrestricted.", ""
         return  # End the function since there was an error
     # Initialize the sliding window prediction process
+    prediction_generator = sliding_window_prediction(template, text, model, tokenizer, window_size=size)
     # Iterate over the generator to return values at each step
     for progress, full_pred, html_content in prediction_generator:
     inputs=[
         gr.Textbox(lines=2, placeholder="Enter Template here...", label="Template"),
         gr.Textbox(lines=2, placeholder="Enter input Text here...", label="Input Text"),
+        gr.Textbox(lines=2, placeholder="Enter windows size here...", label="Size"),
         gr.Checkbox(label="Is Example?", visible=False),
     ],
     outputs=[