revert

app.py

@@ -6,12 +6,6 @@ from typing import Literal, Optional
 
 import gradio as gr
 import torch
-import numpy as np
-import matplotlib.pyplot as plt
-import librosa
-import soundfile as sf
-from matplotlib.figure import Figure
-from matplotlib.backends.backend_agg import FigureCanvasAgg as FigureCanvas
 
 from NatureLM.config import Config
 from NatureLM.models.NatureLM import NatureLM
@@ -19,71 +13,6 @@ from NatureLM.utils import generate_sample_batches, prepare_sample_waveforms
 import spaces
 
 
-def create_spectrogram(audio_path: str, sr: int = None, n_fft: int = 2048, hop_length: int = 512) -> str:
-    """
-    Create a spectrogram visualization from an audio file.
-    
-    Args:
-        audio_path: Path to the audio file
-        sr: Sample rate (if None, uses original sample rate)
-        n_fft: FFT window size
-        hop_length: Hop length for STFT
-        
-    Returns:
-        Path to the saved spectrogram image
-    """
-    try:
-        # Load audio
-        audio, orig_sr = sf.read(audio_path)
-        if len(audio.shape) == 2:  # stereo to mono
-            audio = audio.mean(axis=1)
-        
-        # Use original sample rate if not specified
-        if sr is None:
-            sr = orig_sr
-        
-        # Create spectrogram using original sample rate
-        D = librosa.stft(audio, n_fft=n_fft, hop_length=hop_length, sr=orig_sr)
-        S_db = librosa.amplitude_to_db(np.abs(D), ref=np.max)
-        
-        # Create figure
-        fig, ax = plt.subplots(figsize=(10, 6))
-        img = librosa.display.specshow(S_db, sr=orig_sr, hop_length=hop_length, x_axis='time', y_axis='hz', ax=ax)
-        ax.set_title('Spectrogram')
-        fig.colorbar(img, ax=ax, format='%+2.0f dB')
-        
-        # Save to temporary file
-        temp_path = tempfile.mktemp(suffix='.png')
-        fig.savefig(temp_path, dpi=150, bbox_inches='tight')
-        plt.close(fig)
-        
-        return temp_path
-    except Exception as e:
-        print(f"Error creating spectrogram: {e}")
-        return None
-
-
-def audio_visualization_component(audio_path: str, view_type: str = "waveform") -> str:
-    """
-    Create an audio visualization component that can show either waveform or spectrogram.
-    
-    Args:
-        audio_path: Path to the audio file
-        view_type: Either "waveform" or "spectrogram"
-        
-    Returns:
-        Path to the visualization image
-    """
-    if not audio_path:
-        return None
-    
-    if view_type == "spectrogram":
-        return create_spectrogram(audio_path)
-    else:
-        # For waveform, we'll use the default Gradio audio component
-        return audio_path
-
-
 class ModelManager:
     """Manages model loading and state"""
     
@@ -256,18 +185,6 @@ def _chat_tab(examples):
         visible=True
     )
     
-    # Add visualization options for chat
-    with gr.Row():
-        view_selector = gr.Radio(
-            choices=["Waveform", "Spectrogram"],
-            value="Waveform",
-            label="Audio Visualization Type",
-            interactive=True
-        )
-    
-    # Add visualization display
-    visualization_display = gr.Image(label="Audio Visualization", visible=True)
-    
     chatbot = gr.Chatbot(
         label="Model inputs",
         elem_id="chatbot",
@@ -281,33 +198,6 @@ def _chat_tab(examples):
     send_all = gr.Button("Send all", elem_id="send-all")
     clear_button = gr.ClearButton(components=[chatbot, chat_input], visible=False)
 
-    # Function to update audio visualization in chat
-    def update_chat_visualization(message, view_type):
-        if not message or not message.get("files"):
-            return None
-        
-        # Get the first audio file from the message
-        audio_files = message.get("files", [])
-        if not audio_files:
-            return None
-            
-        audio_path = audio_files[0]
-        return audio_visualization_component(audio_path, view_type.lower())
-    
-    # Connect visualization update for chat
-    view_selector.change(
-        update_chat_visualization,
-        [chat_input, view_selector],
-        [visualization_display]
-    )
-    
-    # Also update when audio is uploaded
-    chat_input.change(
-        update_chat_visualization,
-        [chat_input, view_selector],
-        [visualization_display]
-    )
-
     chat_input.submit(add_message, [chatbot, chat_input], [chatbot, chat_input])
     bot_msg = send_all.click(
         bot_response,

requirements.txt

@@ -9,7 +9,6 @@ gradio>=5.10.0
 google-cloud-aiplatform>=1.76.0
 Levenshtein>=0.25.1
 librosa>=0.9.2
-matplotlib>=3.9.0
 memoization>=0.4.0
 mir-eval>=0.7
 numpy>=1.26.4