fix auto device detection in voxtral

internal/transcription/adapters/py/voxtral/voxtral_transcribe.py

@@ -35,8 +35,9 @@ def transcribe_audio(
         Dictionary containing transcription results
     """
     # Determine device
-    if device == "auto":
-        device = "cuda" if torch.cuda.is_available() else "cpu"
+    # if device == "auto":
+    #     device = "cuda" if torch.cuda.is_available() else "cpu"
+    device = "cuda" if torch.cuda.is_available() else "cpu"
 
     print(f"Loading Voxtral model on {device}...", file=sys.stderr)
 
@@ -57,9 +58,7 @@ def transcribe_audio(
 
     # Prepare transcription request using the proper method
     inputs = processor.apply_transcription_request(
-        language=language,
-        audio=audio_path,
-        model_id=model_id
+        language=language, audio=audio_path, model_id=model_id
     )
 
     # Move inputs to device with correct dtype
@@ -76,8 +75,7 @@ def transcribe_audio(
 
     # Decode only the newly generated tokens (skip the input prompt)
     decoded_outputs = processor.batch_decode(
-        outputs[:, inputs.input_ids.shape[1]:],
-        skip_special_tokens=True
+        outputs[:, inputs.input_ids.shape[1] :], skip_special_tokens=True
     )
 
     transcription_text = decoded_outputs[0]
@@ -94,7 +92,7 @@ def transcribe_audio(
                 "start": 0.0,
                 "end": 0.0,  # Duration unknown without audio analysis
                 "text": transcription_text,
-                "words": []  # Voxtral doesn't provide word-level timestamps
+                "words": [],  # Voxtral doesn't provide word-level timestamps
             }
         ],
         "language": language,
@@ -104,7 +102,7 @@ def transcribe_audio(
 
     # Write output
     output_file = Path(output_path)
-    with output_file.open('w', encoding='utf-8') as f:
+    with output_file.open("w", encoding="utf-8") as f:
         json.dump(result, f, ensure_ascii=False, indent=2)
 
     print(f"Results written to {output_path}", file=sys.stderr)
@@ -116,40 +114,29 @@ def main():
     parser = argparse.ArgumentParser(
         description="Transcribe audio using Voxtral-mini model"
     )
+    parser.add_argument("audio_path", type=str, help="Path to input audio file")
+    parser.add_argument("output_path", type=str, help="Path to output JSON file")
     parser.add_argument(
-        "audio_path",
-        type=str,
-        help="Path to input audio file"
-    )
-    parser.add_argument(
-        "output_path",
-        type=str,
-        help="Path to output JSON file"
-    )
-    parser.add_argument(
-        "--language",
-        type=str,
-        default="en",
-        help="Language code (default: en)"
+        "--language", type=str, default="en", help="Language code (default: en)"
     )
     parser.add_argument(
         "--model-id",
         type=str,
         default="mistralai/Voxtral-mini",
-        help="HuggingFace model ID (default: mistralai/Voxtral-mini)"
+        help="HuggingFace model ID (default: mistralai/Voxtral-mini)",
     )
     parser.add_argument(
         "--device",
         type=str,
         default="auto",
         choices=["cpu", "cuda", "auto"],
-        help="Device to use (default: auto)"
+        help="Device to use (default: auto)",
     )
     parser.add_argument(
         "--max-new-tokens",
         type=int,
         default=8192,
-        help="Maximum number of tokens to generate (default: 8192)"
+        help="Maximum number of tokens to generate (default: 8192)",
     )
 
     args = parser.parse_args()
@@ -166,6 +153,7 @@ def main():
     except Exception as e:
         print(f"Error: {e}", file=sys.stderr)
         import traceback
+
         traceback.print_exc(file=sys.stderr)
         sys.exit(1)
 

internal/transcription/adapters/py/voxtral/voxtral_transcribe_buffered.py

@@ -29,11 +29,13 @@ def split_audio_file(audio_path, chunk_duration_secs=1500):
         end_sample = min(start_sample + chunk_samples, len(audio))
         chunk_audio = audio[start_sample:end_sample]
         start_time = start_sample / sr
-        chunks.append({
-            'audio': chunk_audio,
-            'start_time': start_time,
-            'duration': len(chunk_audio) / sr
-        })
+        chunks.append(
+            {
+                "audio": chunk_audio,
+                "start_time": start_time,
+                "duration": len(chunk_audio) / sr,
+            }
+        )
 
     return chunks, sr
 
@@ -51,8 +53,9 @@ def transcribe_buffered(
     Transcribe long audio by splitting into chunks and merging results.
     """
     # Determine device
-    if device == "auto":
-        device = "cuda" if torch.cuda.is_available() else "cpu"
+    # if device == "auto":
+    #     device = "cuda" if torch.cuda.is_available() else "cpu"
+    device = "cuda" if torch.cuda.is_available() else "cpu"
 
     print(f"Loading Voxtral model on {device}...", file=sys.stderr)
 
@@ -77,18 +80,19 @@ def transcribe_buffered(
     full_text = []
 
     for i, chunk_info in enumerate(chunks):
-        print(f"Transcribing chunk {i+1}/{len(chunks)} (duration: {chunk_info['duration']:.1f}s)...", file=sys.stderr)
+        print(
+            f"Transcribing chunk {i + 1}/{len(chunks)} (duration: {chunk_info['duration']:.1f}s)...",
+            file=sys.stderr,
+        )
 
         # Save chunk to temporary file
         chunk_path = f"/tmp/voxtral_chunk_{i}.wav"
-        sf.write(chunk_path, chunk_info['audio'], sr)
+        sf.write(chunk_path, chunk_info["audio"], sr)
 
         try:
             # Prepare transcription request for this chunk
             inputs = processor.apply_transcription_request(
-                language=language,
-                audio=chunk_path,
-                model_id=model_id
+                language=language, audio=chunk_path, model_id=model_id
             )
 
             # Move inputs to device with correct dtype
@@ -103,14 +107,15 @@ def transcribe_buffered(
 
             # Decode only the newly generated tokens (skip the input prompt)
             decoded_outputs = processor.batch_decode(
-                outputs[:, inputs.input_ids.shape[1]:],
-                skip_special_tokens=True
+                outputs[:, inputs.input_ids.shape[1] :], skip_special_tokens=True
             )
 
             chunk_text = decoded_outputs[0]
             full_text.append(chunk_text)
 
-            print(f"Chunk {i+1} complete: {len(chunk_text)} characters", file=sys.stderr)
+            print(
+                f"Chunk {i + 1} complete: {len(chunk_text)} characters", file=sys.stderr
+            )
 
         finally:
             # Clean up temp file
@@ -119,7 +124,9 @@ def transcribe_buffered(
 
     # Concatenate all chunks
     final_text = " ".join(full_text)
-    print(f"Transcription complete: {len(final_text)} characters total", file=sys.stderr)
+    print(
+        f"Transcription complete: {len(final_text)} characters total", file=sys.stderr
+    )
 
     # Prepare output in Scriberr format
     # Note: Voxtral doesn't provide word-level timestamps
@@ -131,7 +138,7 @@ def transcribe_buffered(
                 "start": 0.0,
                 "end": 0.0,  # Duration unknown without audio analysis
                 "text": final_text,
-                "words": []  # Voxtral doesn't provide word-level timestamps
+                "words": [],  # Voxtral doesn't provide word-level timestamps
             }
         ],
         "language": language,
@@ -144,7 +151,7 @@ def transcribe_buffered(
 
     # Write output
     output_file_path = Path(output_file)
-    with output_file_path.open('w', encoding='utf-8') as f:
+    with output_file_path.open("w", encoding="utf-8") as f:
         json.dump(result, f, ensure_ascii=False, indent=2)
 
     print(f"Results written to {output_file}", file=sys.stderr)
@@ -156,46 +163,35 @@ def main():
     parser = argparse.ArgumentParser(
         description="Transcribe long audio using Voxtral with chunking"
     )
+    parser.add_argument("audio_path", type=str, help="Path to input audio file")
+    parser.add_argument("output_path", type=str, help="Path to output JSON file")
     parser.add_argument(
-        "audio_path",
-        type=str,
-        help="Path to input audio file"
-    )
-    parser.add_argument(
-        "output_path",
-        type=str,
-        help="Path to output JSON file"
-    )
-    parser.add_argument(
-        "--language",
-        type=str,
-        default="en",
-        help="Language code (default: en)"
+        "--language", type=str, default="en", help="Language code (default: en)"
     )
     parser.add_argument(
         "--model-id",
         type=str,
         default="mistralai/Voxtral-mini",
-        help="HuggingFace model ID (default: mistralai/Voxtral-mini)"
+        help="HuggingFace model ID (default: mistralai/Voxtral-mini)",
     )
     parser.add_argument(
         "--device",
         type=str,
         default="auto",
         choices=["cpu", "cuda", "auto"],
-        help="Device to use (default: auto)"
+        help="Device to use (default: auto)",
     )
     parser.add_argument(
         "--max-new-tokens",
         type=int,
         default=8192,
-        help="Maximum number of tokens to generate per chunk (default: 8192)"
+        help="Maximum number of tokens to generate per chunk (default: 8192)",
     )
     parser.add_argument(
         "--chunk-len",
         type=float,
         default=1500,
-        help="Chunk duration in seconds (default: 1500 = 25 minutes)"
+        help="Chunk duration in seconds (default: 1500 = 25 minutes)",
     )
 
     args = parser.parse_args()
@@ -217,6 +213,7 @@ def main():
     except Exception as e:
         print(f"Error: {e}", file=sys.stderr)
         import traceback
+
         traceback.print_exc(file=sys.stderr)
         sys.exit(1)