Upload folder using huggingface_hub
Browse files- config.json +89 -0
- model.safetensors +3 -0
config.json
ADDED
|
@@ -0,0 +1,89 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"bottleneck": {
|
| 3 |
+
"args": {
|
| 4 |
+
"bottleneck_dim": 16,
|
| 5 |
+
"norm": "none",
|
| 6 |
+
"regularizer": {
|
| 7 |
+
"args": {
|
| 8 |
+
"codebook_loss_weight": 1.0,
|
| 9 |
+
"codebook_size": 8192,
|
| 10 |
+
"commitment_loss_weight": 0.25,
|
| 11 |
+
"entropy_loss_temperature": 0.01,
|
| 12 |
+
"entropy_loss_weight": 0.0,
|
| 13 |
+
"eval_deterministic": true,
|
| 14 |
+
"l2_normalized": true,
|
| 15 |
+
"stochastic": true,
|
| 16 |
+
"stochastic_temperature": 0.03,
|
| 17 |
+
"train_deterministic": true
|
| 18 |
+
},
|
| 19 |
+
"name": "vq"
|
| 20 |
+
}
|
| 21 |
+
},
|
| 22 |
+
"name": "bottleneck"
|
| 23 |
+
},
|
| 24 |
+
"bottleneck_token_num": 2048,
|
| 25 |
+
"decode_mode": "with_drop",
|
| 26 |
+
"decoder_attn_type": "full_causal_type2",
|
| 27 |
+
"decoder_depth": 12,
|
| 28 |
+
"decoder_hidden_size": 768,
|
| 29 |
+
"decoder_name": "none",
|
| 30 |
+
"decoder_num_heads": 12,
|
| 31 |
+
"decoder_patch_size": 8,
|
| 32 |
+
"decoder_temporal_patch_size": 4,
|
| 33 |
+
"encoder_attn_type": "full_causal_type1",
|
| 34 |
+
"encoder_depth": 12,
|
| 35 |
+
"encoder_hidden_size": 768,
|
| 36 |
+
"encoder_name": "none",
|
| 37 |
+
"encoder_num_heads": 12,
|
| 38 |
+
"encoder_query_gaussian_init": true,
|
| 39 |
+
"frame_num": 16,
|
| 40 |
+
"in_channels": 3,
|
| 41 |
+
"input_size": 128,
|
| 42 |
+
"latent_pe_scale_factor": 10000,
|
| 43 |
+
"learned_decoder_latent_pe": false,
|
| 44 |
+
"learned_decoder_patch_query_embed": false,
|
| 45 |
+
"learned_encoder_latent_query_embed": true,
|
| 46 |
+
"learned_encoder_patch_pe": false,
|
| 47 |
+
"mask_generator": {
|
| 48 |
+
"mask_type": "left_masking_by_group_normal",
|
| 49 |
+
"max_toks": 512,
|
| 50 |
+
"mean_toks": 256,
|
| 51 |
+
"min_first_toks": 1,
|
| 52 |
+
"min_toks": 32,
|
| 53 |
+
"std_toks": 128,
|
| 54 |
+
"tot_groups": 4,
|
| 55 |
+
"total_toks": 512
|
| 56 |
+
},
|
| 57 |
+
"mode": "train_online_scorer",
|
| 58 |
+
"patch_size": 8,
|
| 59 |
+
"prior_model": {
|
| 60 |
+
"args": {
|
| 61 |
+
"l2_normalized": true
|
| 62 |
+
},
|
| 63 |
+
"avg_loss_over_rounds": true,
|
| 64 |
+
"latent_ce_temperature": 1.0,
|
| 65 |
+
"mix_ss_max_ratio": 0.5,
|
| 66 |
+
"mix_ss_peak_steps_ratio": 0.3,
|
| 67 |
+
"n_rounds": 2,
|
| 68 |
+
"name": "gptc-S",
|
| 69 |
+
"no_dropout": false,
|
| 70 |
+
"no_grad_before_last_round": false,
|
| 71 |
+
"use_mix_ss": true
|
| 72 |
+
},
|
| 73 |
+
"query_init_std": 0.02,
|
| 74 |
+
"scorer_attn_type": "full_causal_type1",
|
| 75 |
+
"scorer_depth": 12,
|
| 76 |
+
"scorer_hidden_size": 768,
|
| 77 |
+
"scorer_name": "transformer_scorer",
|
| 78 |
+
"scorer_num_heads": 12,
|
| 79 |
+
"scorer_step": 64,
|
| 80 |
+
"temporal_patch_size": 4,
|
| 81 |
+
"token_select_mode": "none",
|
| 82 |
+
"token_select_num": 10000,
|
| 83 |
+
"transformer_name": "transformer_encoder_parallel",
|
| 84 |
+
"transformers_version": "4.50.0",
|
| 85 |
+
"use_decoder_latent_token_type_embed": false,
|
| 86 |
+
"use_decoder_patch_query_token_type_embed": true,
|
| 87 |
+
"use_encoder_latent_query_token_type_embed": false,
|
| 88 |
+
"use_encoder_patch_token_type_embed": false
|
| 89 |
+
}
|
model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f4c48c40171aa7ad58e6f77d363545ce7dc819b7c0d2df698fb8a472f97af4db
|
| 3 |
+
size 1148743316
|