huytofu92 commited on
Commit
ca74fff
·
1 Parent(s): 146fc91
Files changed (2) hide show
  1. app.py +0 -4
  2. mini_agents.py +5 -5
app.py CHANGED
@@ -212,10 +212,6 @@ def save_dataset_to_hub(df: pd.DataFrame, dataset_name: str) -> tuple[bool, str]
212
 
213
  # Add metadata with explicit string types for all columns
214
  dataset.info.description = "Agent steps data from evaluation run"
215
- dataset.info.features = {
216
- col: datasets.Value('string') for col in df_to_save.columns
217
- }
218
-
219
  # Save to hub with token
220
  dataset.push_to_hub(
221
  dataset_name,
 
212
 
213
  # Add metadata with explicit string types for all columns
214
  dataset.info.description = "Agent steps data from evaluation run"
 
 
 
 
215
  # Save to hub with token
216
  dataset.push_to_hub(
217
  dataset_name,
mini_agents.py CHANGED
@@ -66,9 +66,9 @@ audio_model = InferenceClientModel(
66
  audio_agent = CodeAgent(
67
  model=audio_model,
68
  tools=[transcribe_audio_tool, get_audio_from_file_path, noise_reduction, audio_segmentation, speaker_diarization],
69
- max_steps=6,
70
  additional_authorized_imports=AUTHORIZED_IMPORTS,
71
- planning_interval=6,
72
  name="audio_agent",
73
  description="This agent is responsible for processing audio, loading mp3 audio and converting it to base64, reducing noise, segmenting audio and transcribing audio (in base64 format). It cannot process videos."
74
  )
@@ -82,9 +82,9 @@ vlm_model = InferenceClientModel(
82
  vlm_agent = CodeAgent(
83
  model=vlm_model,
84
  tools=[image_processing, object_detection_tool, ocr_scan_tool, extract_images_from_video, get_image_from_file_path, get_video_from_file_path],
85
- max_steps=6,
86
  additional_authorized_imports=AUTHORIZED_IMPORTS,
87
- planning_interval=6,
88
  name="vlm_agent",
89
  description="This agent is responsible for downloading images or videos, processing images or videos, detecting objects in them and extracting text from them. It cannot process audios."
90
  )
@@ -167,7 +167,7 @@ class MasterAgentWrapper:
167
  managed_agents=[audio_agent, vlm_agent, arithmetic_agent, pandas_agent],
168
  tools=self.base_tools, # Initialize without browser tools
169
  add_base_tools=False,
170
- max_steps=20,
171
  additional_authorized_imports=AUTHORIZED_IMPORTS,
172
  verbosity_level=logging.INFO,
173
  planning_interval=4,
 
66
  audio_agent = CodeAgent(
67
  model=audio_model,
68
  tools=[transcribe_audio_tool, get_audio_from_file_path, noise_reduction, audio_segmentation, speaker_diarization],
69
+ max_steps=4,
70
  additional_authorized_imports=AUTHORIZED_IMPORTS,
71
+ planning_interval=4,
72
  name="audio_agent",
73
  description="This agent is responsible for processing audio, loading mp3 audio and converting it to base64, reducing noise, segmenting audio and transcribing audio (in base64 format). It cannot process videos."
74
  )
 
82
  vlm_agent = CodeAgent(
83
  model=vlm_model,
84
  tools=[image_processing, object_detection_tool, ocr_scan_tool, extract_images_from_video, get_image_from_file_path, get_video_from_file_path],
85
+ max_steps=4,
86
  additional_authorized_imports=AUTHORIZED_IMPORTS,
87
+ planning_interval=4,
88
  name="vlm_agent",
89
  description="This agent is responsible for downloading images or videos, processing images or videos, detecting objects in them and extracting text from them. It cannot process audios."
90
  )
 
167
  managed_agents=[audio_agent, vlm_agent, arithmetic_agent, pandas_agent],
168
  tools=self.base_tools, # Initialize without browser tools
169
  add_base_tools=False,
170
+ max_steps=8,
171
  additional_authorized_imports=AUTHORIZED_IMPORTS,
172
  verbosity_level=logging.INFO,
173
  planning_interval=4,