TKCAM / MANIFEST.json
yhz-linearalgebra's picture
TKCAM checkpoints (RealEstate10K-Cap, retrained)
6c8e934 verified
Raw History Blame Contribute Delete
3.03 kB
{
"vq": "rvq_w128_nq4_cb256_25kfixed_60ep",
"tracks": {
"noframe": {
"mask_transformer": "mtrans_25kfixed_noframe_r4",
"residual_transformer": "rtrans_25kfixed_noframe_r4"
},
"firstframe": {
"mask_transformer": "mtrans_25kfixed_1frame_r4",
"residual_transformer": "rtrans_25kfixed_1frame_r4"
},
"xframe": {
"mask_transformer": "mtrans_25kfixed_xframe_r4",
"residual_transformer": "rtrans_25kfixed_xframe_r4"
}
},
"selection": "each transformer: the saved checkpoint (net_best_fid / latest / net_best_acc / net_best_loss) with the best FID on the 2,500-clip validation split under the retrained CLaTr evaluator; saved as net_best_fid.tar, original file in files[*].source",
"pretrained_encoders": {
"text": "t5-base",
"image": "openai/clip-vit-base-patch32"
},
"files": {
"rvq_w128_nq4_cb256_25kfixed_60ep/model/net_best_recon.tar": {
"source": "net_best_recon.tar",
"epoch": 25,
"tensors": 70,
"params_M": 19.19,
"dropped_pretrained_tensors": 0,
"bytes": 76781925,
"sha256": "ab0081a38080959494e9535c1d5b7bc6910ac9b84b057d834a8703964a67ebd7"
},
"mtrans_25kfixed_noframe_r4/net_best_fid.tar": {
"source": "net_best_fid.tar",
"epoch": 555,
"tensors": 86,
"params_M": 10.68,
"dropped_pretrained_tensors": 100,
"bytes": 42761325,
"sha256": "333c824a3a155c4ab937554aed21a6d361640c9b6e0457a3ce19fd8b8b511c8a"
},
"rtrans_25kfixed_noframe_r4/net_best_fid.tar": {
"source": "net_best_fid.tar",
"epoch": 40,
"tensors": 90,
"params_M": 11.44,
"dropped_pretrained_tensors": 100,
"bytes": 45798133,
"sha256": "1600826091cf1f3c43d1407386cf8d02e0c4eb0dafc2784fc03479ea937c46a6"
},
"mtrans_25kfixed_1frame_r4/net_best_fid.tar": {
"source": "net_best_fid.tar",
"epoch": 480,
"tensors": 90,
"params_M": 10.98,
"dropped_pretrained_tensors": 299,
"bytes": 43946741,
"sha256": "f4291f37f47beda194bd318edf0cb61a18fe81e9456d95c7f400a9ddb8d4a338"
},
"rtrans_25kfixed_1frame_r4/net_best_fid.tar": {
"source": "latest.tar",
"epoch": 600,
"tensors": 94,
"params_M": 11.74,
"dropped_pretrained_tensors": 0,
"bytes": 46983613,
"sha256": "8471fc135350e8fc62e794da99b9833fbd85f51b4955f6aabe61041e6fae1d04"
},
"mtrans_25kfixed_xframe_r4/net_best_fid.tar": {
"source": "latest.tar",
"epoch": 600,
"tensors": 91,
"params_M": 11.17,
"dropped_pretrained_tensors": 0,
"bytes": 44715063,
"sha256": "9a2a471ac5a11e2bbee55e3057fe789866ec2470c27c35def43272c1c91c349d"
},
"rtrans_25kfixed_xframe_r4/net_best_fid.tar": {
"source": "net_best_fid.tar",
"epoch": 125,
"tensors": 95,
"params_M": 11.93,
"dropped_pretrained_tensors": 299,
"bytes": 47751871,
"sha256": "bbc578c597c4e47a4d54e95a4b48dc614b816ab5fae2ca315df11b9f17421190"
}
}
}