在TensorFlow中使用RBM处理MIDI文件时遭遇TypeError错误

Hey there! Let's work through that TypeError you're facing while moving your notebook to Python 3.6 with the python3-midi library. Python 2 and 3 have key differences around type handling (especially strings/bytes) and library APIs, which are almost certainly the root cause here. Here are targeted fixes and debugging steps based on common porting issues:

Common Causes & Solutions #

1. Bytes vs. String Type Mismatch #

Python 2 blurred lines between str and bytes, but Python 3 enforces strict separation. The python3-midi library likely returns byte data where your old Py2 code expected strings, or vice versa.

• If your code tries to perform string operations on byte data (like concatenation or string checks), use .decode('utf-8') (or appropriate encoding) to convert bytes to strings:

  # Example: Convert MIDI event data from bytes to string
  event_str = event.data.decode('utf-8')
  

• If you need to pass string data to the library, encode it to bytes first with .encode('utf-8').

2. Numpy Array Type Incompatibility #

When converting MIDI data to numpy arrays, mixing incompatible types (like bytes and integers) will throw a TypeError.

• Ensure all raw MIDI data is converted to integers before creating the numpy array:

  # Convert raw byte data to integer list, then to numpy array
  raw_midi_bytes = your_parsed_midi_data
  midi_int_list = [int(byte) for byte in raw_midi_bytes]
  midi_array = np.array(midi_int_list, dtype=np.int32)  # Use appropriate dtype
  

• If you're combining numpy arrays with MIDI event data, convert the event data to a numpy array first:

  # Instead of directly adding bytes to an array
  updated_array = midi_array + np.array(list(event.data), dtype=midi_array.dtype)
  

3. API Differences in python3-midi #

The python3-midi library might have changed function signatures or return structures compared to the old Py2 MIDI library you used.

• Double-check how you're parsing MIDI files: if your old code used midi.read_midifile(), confirm the returned Pattern and Track objects have the same attributes (like events, data) as in Py2.
• Print the type and structure of your parsed MIDI objects to spot discrepancies:

  pattern = midi.read_midifile("your_file.mid")
  print(type(pattern))
  print(type(pattern[0].events[0]))  # Inspect first event in first track
  print(pattern[0].events[0].data)  # Check the data type here
  

4. Debug the Specific Error Line #

Since the error traces back to <ipython-input-62-f35c20bfe55b>, focus on that code block:

• Add print statements to check the type of every variable involved in the line that throws the error. For example, if the line is result = process_data(midi_data, numpy_arr), print type(midi_data) and type(numpy_arr) to spot mismatches.
• Look for operations like addition, concatenation, or function calls that mix incompatible types (e.g., numpy arrays with bytes, integers with strings).

Final Tips #

• Test small chunks of code separately: parse a single MIDI file, inspect its events, then convert to a numpy array step-by-step to isolate where the error occurs.
• Ensure all dependencies (like numpy) are updated to versions compatible with Python 3.6—old numpy versions might have quirks with Py3 type handling.

内容的提问来源于stack exchange，提问作者MrL