Upload folder using huggingface_hub
Browse files
1B/stage-2-joint/args.json
ADDED
|
@@ -0,0 +1,85 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"batch_size": 3,
|
| 3 |
+
"accum_iter": 1,
|
| 4 |
+
"epochs": 100,
|
| 5 |
+
"warmup_epochs": 0.0025,
|
| 6 |
+
"lr": 5e-05,
|
| 7 |
+
"min_lr": 5e-05,
|
| 8 |
+
"wd": 0.1,
|
| 9 |
+
"clip_grad": 4.0,
|
| 10 |
+
"init_from": null,
|
| 11 |
+
"data_config": "",
|
| 12 |
+
"cache_ann_on_disk": false,
|
| 13 |
+
"length_clustering": true,
|
| 14 |
+
"num_workers": 32,
|
| 15 |
+
"pin_mem": true,
|
| 16 |
+
"seed": 0,
|
| 17 |
+
"output_dir": "",
|
| 18 |
+
"save_interval": 1,
|
| 19 |
+
"save_iteration_interval": 1000,
|
| 20 |
+
"only_save_trainable": false,
|
| 21 |
+
"ckpt_max_keep": 20,
|
| 22 |
+
"auto_resume": true,
|
| 23 |
+
"resume_path": "",
|
| 24 |
+
"model_parallel_size": 1,
|
| 25 |
+
"data_parallel": "fsdp",
|
| 26 |
+
"precision": "bf16",
|
| 27 |
+
"grad_precision": "fp32",
|
| 28 |
+
"checkpointing": true,
|
| 29 |
+
"max_seq_len": 32768,
|
| 30 |
+
"mask_image_logits": false,
|
| 31 |
+
"dropout": 0.0,
|
| 32 |
+
"z_loss_weight": 0.0,
|
| 33 |
+
"model_size": "1B_MMRoPE",
|
| 34 |
+
"pretrain_task": [
|
| 35 |
+
"predict_video"
|
| 36 |
+
],
|
| 37 |
+
"video_fps": [
|
| 38 |
+
12
|
| 39 |
+
],
|
| 40 |
+
"video_duration": [
|
| 41 |
+
2
|
| 42 |
+
],
|
| 43 |
+
"cfg_mode": [
|
| 44 |
+
"text_only"
|
| 45 |
+
],
|
| 46 |
+
"masking_mode": [
|
| 47 |
+
"text_to_video"
|
| 48 |
+
],
|
| 49 |
+
"text_to_video_prob": 0.5,
|
| 50 |
+
"eval_data_config": "",
|
| 51 |
+
"run_eval": false,
|
| 52 |
+
"eval_in_epoch": null,
|
| 53 |
+
"eval_mode": [],
|
| 54 |
+
"data_video_fps": 12,
|
| 55 |
+
"data_video_frames": 84,
|
| 56 |
+
"MaskedAR": true,
|
| 57 |
+
"mask_schedule": "cosine",
|
| 58 |
+
"min_masking_rate": 0.7,
|
| 59 |
+
"mask_contiguous_region_prob": null,
|
| 60 |
+
"noise_type": "mask",
|
| 61 |
+
"predict_all_tokens": false,
|
| 62 |
+
"frame_closs_recorder": true,
|
| 63 |
+
"mask_type": "tube",
|
| 64 |
+
"train_loss": "CEChunked",
|
| 65 |
+
"decay_start_coef": 1.0,
|
| 66 |
+
"compute_ntp_loss": false,
|
| 67 |
+
"train_with_slim_lm_head": false,
|
| 68 |
+
"train_with_vis_tok": true,
|
| 69 |
+
"vis_tok_start": 65536,
|
| 70 |
+
"visual_tokenizer": "Cosmos-Tokenizer-DV4x8x8",
|
| 71 |
+
"vocab_size": 129536,
|
| 72 |
+
"joint_img_video": true,
|
| 73 |
+
"img_batch_size": 20,
|
| 74 |
+
"img_data_config": "",
|
| 75 |
+
"video_iter_per_img_iter": 1,
|
| 76 |
+
"no_resume_metric_logger": true,
|
| 77 |
+
"no_resume_iterations": false,
|
| 78 |
+
"world_size": 32,
|
| 79 |
+
"rank": 0,
|
| 80 |
+
"gpu": 0,
|
| 81 |
+
"local_rank": 0,
|
| 82 |
+
"dist_url": "env://",
|
| 83 |
+
"distributed": true,
|
| 84 |
+
"dist_backend": "nccl"
|
| 85 |
+
}
|
1B/stage-2-joint/config.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
1B/stage-2-joint/model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b400630366f11af80854888305c9c020e59efa321795b6cf7b9c835d2e10c1dc
|
| 3 |
+
size 3007483424
|