Added local version of CLAP checkpoint

audiocaptioner.py

@@ -65,4 +65,4 @@ class AudioCaptioner(torch.nn.Module):
             amodel = 'HTSAT-base'
         )
         with contextlib.redirect_stdout(io.StringIO()):
-            self.clap_model.load_ckpt(ckpt = '/graft1/datasets/kechen/clap_ckpt/music_audioset_epoch_15_esc_90.14.pt')
+            self.clap_model.load_ckpt(ckpt = '/checkpoints/music_audioset_epoch_15_esc_90.14.pt')

checkpoints/music_audioset_epoch_15_esc_90.14.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fae3e9c087f2909c28a09dc31c8dfcdacbc42ba44c70e972b58c1bd1caf6dedd
+size 2352471003

infer.py

@@ -18,7 +18,6 @@ def infer(input_filename):
     print(f'Loaded from {checkpoint}')
     model.eval()
     # read in the wav file and precompute neighbors
-    #dataset_path = '/graft1/datasets/kechen/audiostock-full'
     dataset_path = ''
     train_dataset = AudiostockDataset(
             dataset_path=dataset_path,