Upload model_zoo/mit_customize_img_ids_bs_32_rank_512_usedataset_controlnetuse_original_size_resolution_1024_customize_img_ids_customize_txt_ids/config.yaml with huggingface_hub
Browse files
model_zoo/mit_customize_img_ids_bs_32_rank_512_usedataset_controlnetuse_original_size_resolution_1024_customize_img_ids_customize_txt_ids/config.yaml
ADDED
|
@@ -0,0 +1,88 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
controlnet_dataset_hf_repo_id: lilyecho/lily_controlnet_pose_dataset_ar_prepared_1_2
|
| 2 |
+
anno0507_dataset_hf_repo_id: lilyecho/text2pose_anno0507_dataset_ar_prepared
|
| 3 |
+
report_to: wandb
|
| 4 |
+
resume_wandb: true
|
| 5 |
+
wandb_key: 83965ac150088670531a36955ace77fb9e8468e0
|
| 6 |
+
wandb_entry: LilyDaytoy
|
| 7 |
+
use_wandb: true
|
| 8 |
+
wandb_project: v0_ar_triplelora
|
| 9 |
+
exp_name: mit_customize_img_ids_bs_32_rank_512_usedataset_controlnetuse_original_size_resolution_1024_customize_img_ids_customize_txt_ids
|
| 10 |
+
save_dir: multi_person/two_person
|
| 11 |
+
checkpoint:
|
| 12 |
+
type: full_state_dict
|
| 13 |
+
model_filename: model_double_stream_fsdp-004600.pt
|
| 14 |
+
optimizer_filename: adamw_fsdp-model.pt
|
| 15 |
+
model_save_name: model_double_stream_fsdp
|
| 16 |
+
optimizer_name: adamw_fsdp
|
| 17 |
+
weight_decay: 0.01
|
| 18 |
+
lr: 0.0003
|
| 19 |
+
lr_decay_frac: 0.05
|
| 20 |
+
warmup_iters: 50
|
| 21 |
+
beta1: 0.9
|
| 22 |
+
beta2: 0.95
|
| 23 |
+
model_name: flux-dev
|
| 24 |
+
offload_vae: false
|
| 25 |
+
offload_when_sample: true
|
| 26 |
+
rank: 512
|
| 27 |
+
seed: 42
|
| 28 |
+
single_blocks: null
|
| 29 |
+
double_blocks: null
|
| 30 |
+
use_lora: true
|
| 31 |
+
keep_ori_weights: true
|
| 32 |
+
zero_init: false
|
| 33 |
+
local_batch_size: 16
|
| 34 |
+
gradient_accumulation_steps: 1
|
| 35 |
+
trainable_names:
|
| 36 |
+
- img_in.pose_in
|
| 37 |
+
- final_layer.adaLN_modulation_pose
|
| 38 |
+
- final_layer.linear_pose_img
|
| 39 |
+
max_train_steps: 20000
|
| 40 |
+
is_train: true
|
| 41 |
+
num_workers: 0
|
| 42 |
+
grad_clip: 1.0
|
| 43 |
+
max_norm: 20.0
|
| 44 |
+
pose_id_value: 1.0
|
| 45 |
+
lambda_real: 1.0
|
| 46 |
+
lambda_pose: 1.0
|
| 47 |
+
timestep_to_start_cfg: 5
|
| 48 |
+
true_gs: 3.5
|
| 49 |
+
sample_steps: 50
|
| 50 |
+
sample_width: 1024
|
| 51 |
+
sample_height: 1024
|
| 52 |
+
wandb_log_every: 20
|
| 53 |
+
save_checkpoint_every: 200
|
| 54 |
+
sample_every: 200
|
| 55 |
+
use_attn_mask: false
|
| 56 |
+
time_key: independent
|
| 57 |
+
ratio: 0.5
|
| 58 |
+
weighted_pose_loss: false
|
| 59 |
+
text_lora: false
|
| 60 |
+
pose_lora: true
|
| 61 |
+
real_lora: false
|
| 62 |
+
add_pose_clip: true
|
| 63 |
+
multi_seed_version: false
|
| 64 |
+
sample_prompts_json: /afs/csail.mit.edu/u/t/tianyuan/shared/xflux-iter0/multi_person/two_person/eval/annotated_bboxes
|
| 65 |
+
(4).json
|
| 66 |
+
sample_seeds:
|
| 67 |
+
- 478163327
|
| 68 |
+
- 107420369
|
| 69 |
+
- 1181241943
|
| 70 |
+
- 1051802512
|
| 71 |
+
- 958682846
|
| 72 |
+
resolution: 1024
|
| 73 |
+
customize_txt_ids: true
|
| 74 |
+
customize_img_ids: true
|
| 75 |
+
use_dataset: controlnet
|
| 76 |
+
inference_output_dir: 4600_samples
|
| 77 |
+
resize_to_square: false
|
| 78 |
+
generate_img_ids_type: 1
|
| 79 |
+
background_color: black
|
| 80 |
+
loss_pose_background_lambda: 1.0
|
| 81 |
+
double_real_lora: false
|
| 82 |
+
single_real_lora: false
|
| 83 |
+
real_lr_scale: 1.0
|
| 84 |
+
use_v1_bbox: true
|
| 85 |
+
training_width: 1024
|
| 86 |
+
training_height: 1024
|
| 87 |
+
img_seq_len: 4096
|
| 88 |
+
cond_seq_len: 4096
|