update model

{diffusion_posiguide_wav2veclarge_tf2 → diffusion_posiguide_hubertlarge_mamba2}/args.txt

@@ -7,7 +7,7 @@
   "batch_size": 20,
   "gpu0_bs": 2,
   "valid_batch_size": 32,
-  "experiment_name": "diffusion_posiguide_wav2veclarge_tf2",
+  "experiment_name": "diffusion_posiguide_hubertlarge_mamba2",
   "data_root": "/mnt/ssd/PianoPose-new",
   "preload": false,
   "tiny": false,
@@ -21,7 +21,7 @@
   ],
   "continue_train": false,
   "wav2vec_path": "facebook/hubert-large-ls960-ft",
-  "piano2posi_path": "logs/piano2posi_wav2veclarge_tf",
+  "piano2posi_path": "logs/piano2posi_mamba",
   "timesteps": 1000,
   "fine_map": 0,
   "unet_dim": 256,
@@ -30,8 +30,8 @@
   "RAG": false,
   "hidden_type": "audio_f",
   "latest_layer": "tanh",
-  "encoder_type": "transformer",
-  "num_layer": 8,
+  "encoder_type": "mamba",
+  "num_layer": 16,
   "loss_mode": "naive_l1",
   "weight_rec": 1.0,
   "weight_vel": 1.0,

diffusion_posiguide_wav2veclarge_tf2/piano2pose-iter=99000-val_loss=0.0359656028449535.ckpt → diffusion_posiguide_hubertlarge_mamba2/piano2pose-iter=99999-val_loss=0.0372750014066696.ckpt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ce042fe9fbd4b0020040033f5748c34675354955bf0280f1e13b52d133923850
-size 2970320221
+oid sha256:c19475e0c7c2bd7a4531fbc2fc357a6f4ee1e4f5fea8afe9f927184d9582062c
+size 2213668834