Push model using huggingface_hub.
Browse files- README.md +13 -0
- config.json +87 -0
- model.safetensors +3 -0
README.md
ADDED
|
@@ -0,0 +1,13 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
tags:
|
| 4 |
+
- arxiv:2406.07550
|
| 5 |
+
- arxiv:2501.07730
|
| 6 |
+
- image-tokenization
|
| 7 |
+
- model_hub_mixin
|
| 8 |
+
- pytorch_model_hub_mixin
|
| 9 |
+
---
|
| 10 |
+
|
| 11 |
+
This model has been pushed to the Hub using the [PytorchModelHubMixin](https://huggingface.co/docs/huggingface_hub/package_reference/mixins#huggingface_hub.PyTorchModelHubMixin) integration:
|
| 12 |
+
- Library: https://github.com/bytedance/1d-tokenizer
|
| 13 |
+
- Docs: [More Information Needed]
|
config.json
ADDED
|
@@ -0,0 +1,87 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"experiment": {
|
| 3 |
+
"project": "tatitok_sl128_vae",
|
| 4 |
+
"name": "tatitok_sl128_vae_run1",
|
| 5 |
+
"output_dir": "tatitok_sl128_vae_run1",
|
| 6 |
+
"max_train_examples": 685800000,
|
| 7 |
+
"save_every": 50000,
|
| 8 |
+
"eval_every": 50000,
|
| 9 |
+
"generate_every": 5000,
|
| 10 |
+
"log_every": 50,
|
| 11 |
+
"log_grad_norm_every": 1000,
|
| 12 |
+
"resume": true,
|
| 13 |
+
"tokenizer_checkpoint": "tatitok_sl128_vae.bin"
|
| 14 |
+
},
|
| 15 |
+
"model": {
|
| 16 |
+
"vq_model": {
|
| 17 |
+
"quantize_mode": "vae",
|
| 18 |
+
"token_size": 16,
|
| 19 |
+
"vit_enc_model_size": "small",
|
| 20 |
+
"vit_dec_model_size": "large",
|
| 21 |
+
"vit_enc_patch_size": 16,
|
| 22 |
+
"vit_dec_patch_size": 16,
|
| 23 |
+
"num_latent_tokens": 128,
|
| 24 |
+
"finetune_decoder": false,
|
| 25 |
+
"is_legacy": false
|
| 26 |
+
}
|
| 27 |
+
},
|
| 28 |
+
"losses": {
|
| 29 |
+
"discriminator_start": 200000,
|
| 30 |
+
"quantizer_weight": 1.0,
|
| 31 |
+
"discriminator_factor": 1.0,
|
| 32 |
+
"discriminator_weight": 0.1,
|
| 33 |
+
"perceptual_loss": "lpips-convnext_s-1.0-0.1",
|
| 34 |
+
"perceptual_weight": 1.1,
|
| 35 |
+
"reconstruction_loss": "l2",
|
| 36 |
+
"reconstruction_weight": 1.0,
|
| 37 |
+
"lecam_regularization_weight": 0.001,
|
| 38 |
+
"kl_weight": 1e-06,
|
| 39 |
+
"logvar_init": 0.0
|
| 40 |
+
},
|
| 41 |
+
"dataset": {
|
| 42 |
+
"params": {
|
| 43 |
+
"train_shards_path_or_url": "datacomp_sharded/train/datacomp-train-{000000..140089}.tar",
|
| 44 |
+
"eval_shards_path_or_url": "imagenet_sharded/val/imagenet-val-{0000..0009}.tar",
|
| 45 |
+
"num_workers_per_gpu": 12,
|
| 46 |
+
"dataset_with_class_label": false,
|
| 47 |
+
"dataset_with_text_label": true
|
| 48 |
+
},
|
| 49 |
+
"preprocessing": {
|
| 50 |
+
"resize_shorter_edge": 256,
|
| 51 |
+
"crop_size": 256,
|
| 52 |
+
"random_crop": true,
|
| 53 |
+
"random_flip": true,
|
| 54 |
+
"res_ratio_filtering": true
|
| 55 |
+
}
|
| 56 |
+
},
|
| 57 |
+
"optimizer": {
|
| 58 |
+
"name": "adamw",
|
| 59 |
+
"params": {
|
| 60 |
+
"learning_rate": 0.0001,
|
| 61 |
+
"discriminator_learning_rate": 0.0001,
|
| 62 |
+
"beta1": 0.9,
|
| 63 |
+
"beta2": 0.999,
|
| 64 |
+
"weight_decay": 0.0001
|
| 65 |
+
}
|
| 66 |
+
},
|
| 67 |
+
"lr_scheduler": {
|
| 68 |
+
"scheduler": "cosine",
|
| 69 |
+
"params": {
|
| 70 |
+
"learning_rate": "${optimizer.params.learning_rate}",
|
| 71 |
+
"warmup_steps": 10000,
|
| 72 |
+
"end_lr": 1e-05
|
| 73 |
+
}
|
| 74 |
+
},
|
| 75 |
+
"training": {
|
| 76 |
+
"gradient_accumulation_steps": 1,
|
| 77 |
+
"per_gpu_batch_size": 32,
|
| 78 |
+
"mixed_precision": "fp16",
|
| 79 |
+
"enable_tf32": true,
|
| 80 |
+
"enable_wandb": true,
|
| 81 |
+
"use_ema": true,
|
| 82 |
+
"seed": 42,
|
| 83 |
+
"max_train_steps": 650000,
|
| 84 |
+
"num_generated_images": 2,
|
| 85 |
+
"max_grad_norm": 1.0
|
| 86 |
+
}
|
| 87 |
+
}
|
model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5bb6275ab4f291cf54a3ed4114db495203e0c771638e16a2338f5980d496ebc5
|
| 3 |
+
size 1321145744
|