Update app.py

app.py

@@ -4,17 +4,15 @@ import gradio as gr
 import pandas as pd
 import requests
 
-from pyabsa import ATEPCCheckpointManager
-from pyabsa.functional.dataset.dataset_manager import download_datasets_from_github, ABSADatasetList, detect_infer_dataset
+from pyabsa import download_all_available_datasets, AspectTermExtraction as ATEPC, TaskCodeOption
+from pyabsa.utils.data_utils.dataset_manager import detect_infer_dataset
 
-download_datasets_from_github(os.getcwd())
+download_all_available_datasets()
 
-dataset_items = {dataset.name: dataset for dataset in ABSADatasetList()}
-
-URL = 'https://api.visitorbadge.io/api/combined?path=https%3A%2F%2Fhuggingface.co%2Fspaces%2Fyangheng%2Fpyabsa_inference&label=Inference%20Count&labelColor=%2337d67a&countColor=%23f47373&style=flat&labelStyle=none'
+dataset_items = {dataset.name: dataset for dataset in ATEPC.ATEPCDatasetList()}
 
 def get_example(dataset):
-    task = 'apc'
+    task = TaskCodeOption.Aspect_Polarity_Classification
     dataset_file = detect_infer_dataset(dataset_items[dataset], task)
 
     for fname in dataset_file:
@@ -28,27 +26,27 @@ def get_example(dataset):
             lines.extend(fin.readlines())
             fin.close()
         for i in range(len(lines)):
-            lines[i] = lines[i][:lines[i].find('!sent!')].replace('[ASP]', '')
+            lines[i] = lines[i][:lines[i].find('$LABEL$')].replace('[B-ASP]', '').replace('[E-ASP]', '').strip()
         return sorted(set(lines), key=lines.index)
 
 
-dataset_dict = {dataset.name: get_example(dataset.name) for dataset in ABSADatasetList()}
-aspect_extractor = ATEPCCheckpointManager.get_aspect_extractor(checkpoint='multilingual-256-2')
+dataset_dict = {dataset.name: get_example(dataset.name) for dataset in ATEPC.ATEPCDatasetList()}
+aspect_extractor = ATEPC.AspectExtractor(checkpoint='multilingual')
 
 
 def perform_inference(text, dataset):
     if not text:
         text = dataset_dict[dataset][random.randint(0, len(dataset_dict[dataset]) - 1)]
 
-    result = aspect_extractor.extract_aspect(inference_source=[text],
-                                             pred_sentiment=True)
+    result = aspect_extractor.predict(text,
+                                      pred_sentiment=True)
 
     result = pd.DataFrame({
-        'aspect': result[0]['aspect'],
-        'sentiment': result[0]['sentiment'],
+        'aspect': result['aspect'],
+        'sentiment': result['sentiment'],
         # 'probability': result[0]['probs'],
-        'confidence': [round(x, 4) for x in result[0]['confidence']],
-        'position': result[0]['position']
+        'confidence': [round(x, 4) for x in result['confidence']],
+        'position': result['position']
     })
     return result, '{}'.format(text)
 
@@ -57,22 +55,20 @@ demo = gr.Blocks()
 
 with demo:
     gr.Markdown("# <p align='center'>Multilingual Aspect-based Sentiment Analysis !</p>")
-    gr.Markdown("""### Repo: [PyABSA](https://github.com/yangheng95/PyABSA)
+    gr.Markdown("""### Repo: [PyABSA V2](https://github.com/yangheng95/PyABSA)
                 ### Author: [Heng Yang](https://github.com/yangheng95) (杨恒)
                 [![Downloads](https://pepy.tech/badge/pyabsa)](https://pepy.tech/project/pyabsa) 
                 [![Downloads](https://pepy.tech/badge/pyabsa/month)](https://pepy.tech/project/pyabsa)
                 """
                 )
-    gr.Markdown("Your input text should be no more than 80 words, that's the longest text we used in training. However, you can train your own model using 512 max length")
-    gr.Markdown("**You don't need to split each Chinese (Korean, etc.) token as the provided examples, just input the natural language text.**")
-    gr.Markdown("请确保输入的文本长度不超过200词，这是训练时的最大文本长度，过长将不会获得结果")
-    gr.Markdown("**提供的中文等其他非拉丁语系数据集采用了空格分字，这是早期数据集的遗留问题，预测时不用对中文等语言进行空格分字**")
+    gr.Markdown("Your input text should be no more than 80 words, that's the longest text we used in trainer. However, you can try longer text in self-trainer ")
+    gr.Markdown("**You don't need to split each Chinese (Korean, etc.) token as the provided, just input the natural language text.**")
     output_dfs = []
     with gr.Row():
         with gr.Column():
             input_sentence = gr.Textbox(placeholder='Leave this box blank and choose a dataset will give you a random example...', label="Example:")
             gr.Markdown("You can find the datasets at [github.com/yangheng95/ABSADatasets](https://github.com/yangheng95/ABSADatasets/tree/v1.2/datasets/text_classification)")
-            dataset_ids = gr.Radio(choices=[dataset.name for dataset in ABSADatasetList()[:-1]], value='Laptop14', label="Datasets")
+            dataset_ids = gr.Radio(choices=[dataset.name for dataset in ATEPC.ATEPCDatasetList()[:-1]], value='Laptop14', label="Datasets")
             inference_button = gr.Button("Let's go!")
             gr.Markdown("There is a [demo](https://huggingface.co/spaces/yangheng/PyABSA-ATEPC-Chinese) specialized for the Chinese langauge")
             gr.Markdown("This demo support many other language as well, you can try and explore the results of other languages by yourself.")
@@ -84,10 +80,6 @@ with demo:
 
         inference_button.click(fn=perform_inference,
                                inputs=[input_sentence, dataset_ids],
-                               outputs=[output_df, output_text],
-                               api_name='inference')
-
-    gr.Markdown("![visitor badge](https://visitor-badge.glitch.me/badge?page_id=https://huggingface.co/spaces/yangheng/Multilingual-Aspect-Based-Sentiment-Analysis)")
-    gr.Markdown("![Visitors]({})".format(URL))
+                               outputs=[output_df, output_text])
 
 demo.launch()