Update app.py

Siivottu koodia

app.py

@@ -76,7 +76,7 @@ data_collator = DataCollatorForSeq2Seq(tokenizer=tokenizer, model=model_name)
 import torch
 torch.cuda.empty_cache()
 
-#nvidia-smi
+
 
 #pip install wandb
 import os
@@ -107,7 +107,7 @@ for name, param in model.named_parameters():
 training_args = Seq2SeqTrainingArguments(
     output_dir='./results',
     num_train_epochs=1,
-    per_device_train_batch_size=4,  # Pienennä batch-kokoa
+    per_device_train_batch_size=4,  
     per_device_eval_batch_size=4,
     evaluation_strategy='epoch',
     logging_dir='./logs',
@@ -122,7 +122,7 @@ trainer = Seq2SeqTrainer(
     eval_dataset=tokenized_small_ds.shuffle().select(range(20, 100)),  # Käytetään 200 esimerkkiä arvioimiseen
 )
 
-# Kouluta malli
+# train the model
 trainer.train()
 
 #pip install rouge_score
@@ -170,7 +170,7 @@ new_model = MT5ForConditionalGeneration.from_pretrained(model_name)
 from transformers import pipeline
 import torch
 
-# Syötteesi
+
 # Restructured input
 text = (
     "Summarize the following information regarding psoriasis, its effects on skin health, and its potential health risks:\n\n"
@@ -186,10 +186,10 @@ text = (
 # Määrittele laite (GPU tai CPU)
 device = 0 if torch.cuda.is_available() else -1
 
-# Lataa tiivistämispipeline
+# Load the pipeline
 summarizer = pipeline("summarization", model=new_model, tokenizer=new_tokenizer, device=device)
 
-# Tiivistä teksti
+# Summarize the text
 summary = summarizer(text,
                      max_length=120,
                      min_length=30,
@@ -213,8 +213,7 @@ cleaned_summary = re.sub(pattern, " ", summary).strip()
 print(cleaned_summary)
 
 
-# Niinan koodi
-#pip install gradio PyMuPDF
+
 
 import gradio as gr
 from transformers import T5Tokenizer, MT5ForConditionalGeneration
@@ -286,6 +285,6 @@ interface = gr.Interface(
     description="Upload a PDF file to summarize its content."
 )
 
-# Launch the interface
+
 # Launch the interface with debug mode enabled
 interface.launch(debug=True)