fixes and retrieval model
Browse files
tmr/config.json
ADDED
@@ -0,0 +1,124 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"data": {
|
3 |
+
"motion_loader": {
|
4 |
+
"_target_": "src.data.amass_motion.AMASSMotionLoader_custom",
|
5 |
+
"base_dir": "datasets/motions/amass_feats",
|
6 |
+
"normalizer": {
|
7 |
+
"_target_": "src.data.motion.Normalizer",
|
8 |
+
"base_dir": "stats/humanml3d/amass_feats",
|
9 |
+
"eps": 1e-12
|
10 |
+
},
|
11 |
+
"nfeats": 135
|
12 |
+
},
|
13 |
+
"_target_": "src.data.text_motion.TextMotionDataset",
|
14 |
+
"path": "datasets/annotations/humanml3d",
|
15 |
+
"text_to_token_emb": {
|
16 |
+
"_target_": "src.data.text.TokenEmbeddings",
|
17 |
+
"path": "datasets/annotations/humanml3d",
|
18 |
+
"modelname": "distilbert-base-uncased",
|
19 |
+
"preload": true
|
20 |
+
},
|
21 |
+
"text_to_sent_emb": {
|
22 |
+
"_target_": "src.data.text.SentenceEmbeddings",
|
23 |
+
"path": "datasets/annotations/humanml3d",
|
24 |
+
"modelname": "sentence-transformers/all-mpnet-base-v2",
|
25 |
+
"preload": true
|
26 |
+
},
|
27 |
+
"preload": true
|
28 |
+
},
|
29 |
+
"model": {
|
30 |
+
"_target_": "src.model.TMR",
|
31 |
+
"motion_encoder": {
|
32 |
+
"_target_": "src.model.ACTORStyleEncoder",
|
33 |
+
"nfeats": 135,
|
34 |
+
"vae": true,
|
35 |
+
"latent_dim": 256,
|
36 |
+
"ff_size": 1024,
|
37 |
+
"num_layers": 6,
|
38 |
+
"num_heads": 4,
|
39 |
+
"dropout": 0.1,
|
40 |
+
"activation": "gelu"
|
41 |
+
},
|
42 |
+
"text_encoder": {
|
43 |
+
"_target_": "src.model.ACTORStyleEncoder",
|
44 |
+
"nfeats": 768,
|
45 |
+
"vae": true,
|
46 |
+
"latent_dim": 256,
|
47 |
+
"ff_size": 1024,
|
48 |
+
"num_layers": 6,
|
49 |
+
"num_heads": 4,
|
50 |
+
"dropout": 0.1,
|
51 |
+
"activation": "gelu"
|
52 |
+
},
|
53 |
+
"motion_decoder": {
|
54 |
+
"_target_": "src.model.ACTORStyleDecoder",
|
55 |
+
"nfeats": 135,
|
56 |
+
"latent_dim": 256,
|
57 |
+
"ff_size": 1024,
|
58 |
+
"num_layers": 6,
|
59 |
+
"num_heads": 4,
|
60 |
+
"dropout": 0.1,
|
61 |
+
"activation": "gelu"
|
62 |
+
},
|
63 |
+
"vae": true,
|
64 |
+
"lmd": {
|
65 |
+
"recons": 1.0,
|
66 |
+
"latent": 1e-05,
|
67 |
+
"kl": 1e-05,
|
68 |
+
"contrastive": 0.1
|
69 |
+
},
|
70 |
+
"lr": 0.0001,
|
71 |
+
"temperature": 0.1,
|
72 |
+
"threshold_selfsim": 0.8,
|
73 |
+
"threshold_selfsim_metrics": 0.95
|
74 |
+
},
|
75 |
+
"trainer": {
|
76 |
+
"_target_": "pytorch_lightning.Trainer",
|
77 |
+
"max_epochs": 1000,
|
78 |
+
"log_every_n_steps": 50,
|
79 |
+
"num_sanity_val_steps": 0,
|
80 |
+
"check_val_every_n_epoch": 1,
|
81 |
+
"accelerator": "gpu",
|
82 |
+
"devices": 1,
|
83 |
+
"callbacks": [
|
84 |
+
{
|
85 |
+
"_target_": "pytorch_lightning.callbacks.ModelCheckpoint",
|
86 |
+
"filename": "latest-{epoch}",
|
87 |
+
"every_n_epochs": 1,
|
88 |
+
"save_top_k": 1,
|
89 |
+
"save_last": true
|
90 |
+
},
|
91 |
+
{
|
92 |
+
"_target_": "pytorch_lightning.callbacks.ModelCheckpoint",
|
93 |
+
"filename": "latest-{epoch}",
|
94 |
+
"monitor": "step",
|
95 |
+
"mode": "max",
|
96 |
+
"every_n_epochs": 100,
|
97 |
+
"save_top_k": -1,
|
98 |
+
"save_last": false
|
99 |
+
},
|
100 |
+
{
|
101 |
+
"_target_": "src.callback.progress.ProgressLogger",
|
102 |
+
"precision": 3
|
103 |
+
},
|
104 |
+
{
|
105 |
+
"_target_": "src.callback.tqdmbar.TQDMProgressBar"
|
106 |
+
}
|
107 |
+
],
|
108 |
+
"logger": {
|
109 |
+
"_target_": "src.logger.csv.CSVLogger",
|
110 |
+
"save_dir": "outputs/tmr_humanml3d_amass_feats",
|
111 |
+
"name": "logs"
|
112 |
+
}
|
113 |
+
},
|
114 |
+
"run_dir": "outputs/tmr_humanml3d_amass_feats",
|
115 |
+
"seed": 1234,
|
116 |
+
"logger_level": "INFO",
|
117 |
+
"ckpt": "last",
|
118 |
+
"resume_dir": null,
|
119 |
+
"dataloader": {
|
120 |
+
"_target_": "torch.utils.data.DataLoader",
|
121 |
+
"batch_size": 32,
|
122 |
+
"num_workers": 8
|
123 |
+
}
|
124 |
+
}
|
last.ckpt → tmr/last_weights/motion_decoder.pt
RENAMED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
-
size
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:9cdab3049d62fc5039704826d1c69934ea31c9cd0f1edd669560a5599dbe87ce
|
3 |
+
size 25456757
|
tmr/last_weights/motion_encoder.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:ed070ae24b355a829543dd0cb3a3afdd5ca0f942cf2bbe3269fbadaaf73cacfe
|
3 |
+
size 19119391
|
tmr/last_weights/text_encoder.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:bdcd6151acd3b44ed76330e5dae747e6cb144c18293c847573a5d182a4437e87
|
3 |
+
size 19767429
|