Upload 5 files

extractor/weights_1500.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b548d0f1931c68893eff4dfb43b12650eafb5482b519afd5dd866ec645921651
+size 1747896484

mospa/config.json

@@ -0,0 +1 @@
+namespace(arch=namespace(decoder='trans_enc', ff_size=1024, latent_dim=512, num_heads=8, num_layers=4), data='data/sam_train', dataset=namespace(binaural=True, clip_len=240, cond='audio', dataset_name='SAM', move=120, nfeats=6, njoints=25, pose_vec='complex', rot_rep='6d', smpl_dir='smpl_models/smplx'), device=device(type='cuda', index=0), diff=namespace(diffusion_steps=1000, noise_schedule='cosine', sigma_small=True), finetune=False, name='mospa', part=None, save='./save/mospa', ssl='./save/ssl/best.pt', ssl_arch=namespace(audio_dim=512, ch=4, conv_kernel=3, conv_pad=1, latent_dim=256, mode='concat', mp_pad=1, mp_stride=4, num_layers=1, rot_dim=4, rotation=False), test_data='data/sam_test', trainer=namespace(batch_size=128, cond_mask_prob=0.1, ema=True, epoch=10000, eval_freq=500, load_num=-1, loss_terms='11111', lr=0.0001, lr_anneal_steps=0, mask_genre=False, mask_sound_source=False, save_freq=500, use_loss_foot_contact=True, use_loss_geo=True, use_loss_mse=True, use_loss_rotation=True, use_loss_traj=True, weight_decay=0.0, workers=8), val_data='data/sam_val')

mospa/weights_6000.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:156e6c85e6d07edd8f528f82823adb63cb37eedcb6bf0111865bc3121939da65
+size 155523576