fix: transpose vocoder buffers (filter, basis) and upscaler kernel to MLX channels-last layout

spatial_upscaler_x1_5_v1_0.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ce9c1080ac3bdaddc25438825274400c3bbf999bb6d36b7d41df8b9908d68c83
-size 1090127524
+oid sha256:e186732be14a71b953efed15181cbb94595fa836555a07b3b7560d6cf5deeb15
+size 1090127476

spatial_upscaler_x2_v1_1.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6845344671406444e2b0e58d5d05403f0d07218f790e7e68bc272ea6957149eb
-size 995745121
+oid sha256:755f69dcb8a3f9589045925eb4a2237e9969f9c026a016393bc395e65ebd5788
+size 995745061

split_model.json

@@ -8,8 +8,8 @@
     "vae_encoder",
     "audio_vae",
     "vocoder",
-    "spatial_upscaler_x1_5_v1_0",
     "spatial_upscaler_x2_v1_1",
+    "spatial_upscaler_x1_5_v1_0",
     "temporal_upscaler_x2_v1_0"
   ],
   "source": "Lightricks/LTX-2.3",

temporal_upscaler_x2_v1_0.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:644552138e183b7d2c235518ddc62a53313bfd7f37eb155d443a799d051a7cba
-size 261945645
+oid sha256:982bc28b455ed22de1e39a9fcf5213299de4e15f948ae3bd2cee65550e019628
+size 261945581

vocoder.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:bd64374df11a3099ecc781e89526fb18abd879240294917c8ffd13abec0fcd71
+oid sha256:a19289832ac2a4e644d896d1b416fe8db1fa64d1d333bd6b1c4038a3789c4f72
 size 258313851