work with tuple shape (#393)

whisper/requirements.txt

@@ -1,4 +1,4 @@
-mlx
+mlx>=0.1
 numba
 numpy
 torch

whisper/test.py

@@ -303,7 +303,7 @@ class TestWhisper(unittest.TestCase):
         def check_segment(seg, expected):
             for k, v in expected.items():
                 if isinstance(v, float):
-                    self.assertAlmostEqual(seg[k], v, places=3)
+                    self.assertAlmostEqual(seg[k], v, places=2)
                 else:
                     self.assertEqual(seg[k], v)
 

whisper/whisper/decoding.py

@@ -50,7 +50,7 @@ def detect_language(
         mel = mel[None]
 
     # skip encoder forward pass if already-encoded audio features were given
-    if mel.shape[-2:] != [model.dims.n_audio_ctx, model.dims.n_audio_state]:
+    if mel.shape[-2:] != (model.dims.n_audio_ctx, model.dims.n_audio_state):
         mel = model.encoder(mel)
 
     # forward pass using a single token, startoftranscript