enhancement:improve startup time and minor improvement.

metagpt/environment/android/text_icon_localization.py

@@ -8,7 +8,21 @@
 import cv2
 import groundingdino.datasets.transforms as T
 import numpy as np
-import torch
+class LazyTorch:
+    def __init__(self):
+        self._torch = None
+
+    def _import_torch(self):
+        if self._torch is None:
+            import torch
+            self._torch = torch
+
+    def __getattr__(self, item):
+        self._import_torch()
+        return getattr(self._torch, item)
+
+# Create the LazyTorch instance
+torch = LazyTorch()
 from groundingdino.models import build_model
 from groundingdino.util.slconfig import SLConfig
 from groundingdino.util.utils import clean_state_dict, get_phrases_from_posmap

metagpt/repo_parser.py

@@ -18,7 +18,6 @@
 from pathlib import Path
 from typing import Dict, List, Optional
 
-import pandas as pd
 from pydantic import BaseModel, Field, field_validator
 
 from metagpt.const import AGGREGATION, COMPOSITION, GENERALIZATION
@@ -508,6 +507,8 @@ def generate_dataframe_structure(self, output_path: Path):
         Args:
             output_path (Path): The path to the CSV file to be generated.
         """
+        import pandas as pd
+
         files_classes = [i.model_dump() for i in self.generate_symbols()]
         df = pd.DataFrame(files_classes)
         df.to_csv(output_path, index=False)

metagpt/utils/text.py

@@ -59,18 +59,19 @@ def generate_prompt_chunk(
     max_token = TOKEN_MAX.get(model_name, 2048) - reserved - 100
 
     while paragraphs:
-        paragraph = paragraphs.pop(0)
-        token = count_output_tokens(paragraph, model_name)
-        if current_token + token <= max_token:
-            current_lines.append(paragraph)
-            current_token += token
-        elif token > max_token:
-            paragraphs = split_paragraph(paragraph) + paragraphs
-            continue
-        else:
-            yield prompt_template.format("".join(current_lines))
-            current_lines = [paragraph]
-            current_token = token
+        paragraph = paragraphs.pop(0).strip()
+        if len(paragraph) != 0:
+            token = count_output_tokens(paragraph, model_name)
+            if current_token + token <= max_token:
+                current_lines.append(paragraph)
+                current_token += token
+            elif token > max_token:
+                paragraphs = split_paragraph(paragraph) + paragraphs
+                continue
+            else:
+                yield prompt_template.format("".join(current_lines))
+                current_lines = [paragraph]
+                current_token = token
 
     if current_lines:
         yield prompt_template.format("".join(current_lines))