kmpartner commited on
Commit
ef27bda
·
1 Parent(s): 95dfbd3

Upload folder using huggingface_hub

Browse files
checkpoint-500/unet/config.json ADDED
@@ -0,0 +1,73 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_class_name": "UNet2DConditionModel",
3
+ "_diffusers_version": "0.42.0.dev0",
4
+ "_name_or_path": "segmind/Segmind-Vega",
5
+ "act_fn": "silu",
6
+ "addition_embed_type": "text_time",
7
+ "addition_embed_type_num_heads": 64,
8
+ "addition_time_embed_dim": 256,
9
+ "attention_head_dim": [
10
+ 5,
11
+ 10,
12
+ 20
13
+ ],
14
+ "attention_type": "default",
15
+ "block_out_channels": [
16
+ 320,
17
+ 640,
18
+ 1280
19
+ ],
20
+ "center_input_sample": false,
21
+ "class_embed_type": null,
22
+ "class_embeddings_concat": false,
23
+ "conv_in_kernel": 3,
24
+ "conv_out_kernel": 3,
25
+ "cross_attention_dim": 2048,
26
+ "cross_attention_norm": null,
27
+ "down_block_types": [
28
+ "DownBlock2D",
29
+ "CrossAttnDownBlock2D",
30
+ "CrossAttnDownBlock2D"
31
+ ],
32
+ "downsample_padding": 1,
33
+ "dropout": 0.0,
34
+ "dual_cross_attention": false,
35
+ "encoder_hid_dim": null,
36
+ "encoder_hid_dim_type": null,
37
+ "flip_sin_to_cos": true,
38
+ "freq_shift": 0,
39
+ "in_channels": 4,
40
+ "layers_per_block": 2,
41
+ "mid_block_only_cross_attention": null,
42
+ "mid_block_scale_factor": 1,
43
+ "mid_block_type": "UNetMidBlock2D",
44
+ "norm_eps": 1e-05,
45
+ "norm_num_groups": 32,
46
+ "num_attention_heads": null,
47
+ "num_class_embeds": null,
48
+ "only_cross_attention": false,
49
+ "out_channels": 4,
50
+ "projection_class_embeddings_input_dim": 2816,
51
+ "resnet_out_scale_factor": 1.0,
52
+ "resnet_skip_time_act": false,
53
+ "resnet_time_scale_shift": "default",
54
+ "reverse_transformer_layers_per_block": null,
55
+ "sample_size": 128,
56
+ "time_cond_proj_dim": null,
57
+ "time_embedding_act_fn": null,
58
+ "time_embedding_dim": null,
59
+ "time_embedding_type": "positional",
60
+ "timestep_post_act": null,
61
+ "transformer_layers_per_block": [
62
+ 1,
63
+ 1,
64
+ 2
65
+ ],
66
+ "up_block_types": [
67
+ "CrossAttnUpBlock2D",
68
+ "CrossAttnUpBlock2D",
69
+ "UpBlock2D"
70
+ ],
71
+ "upcast_attention": null,
72
+ "use_linear_projection": true
73
+ }
logs/text2image-fine-tune-sdxl/1793014130.8738775/hparams.yml ADDED
@@ -0,0 +1,59 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ adam_beta1: 0.9
2
+ adam_beta2: 0.999
3
+ adam_epsilon: 1.0e-08
4
+ adam_weight_decay: 0.01
5
+ allow_tf32: false
6
+ cache_dir: null
7
+ caption_column: text
8
+ center_crop: true
9
+ checkpointing_steps: 500
10
+ checkpoints_total_limit: 2
11
+ dataloader_num_workers: 0
12
+ dataset_config_name: null
13
+ dataset_name: ./img-dataset
14
+ enable_npu_flash_attention: false
15
+ enable_xformers_memory_efficient_attention: false
16
+ gradient_accumulation_steps: 8
17
+ gradient_checkpointing: true
18
+ hub_model_id: null
19
+ hub_token: null
20
+ image_column: image
21
+ image_interpolation_mode: lanczos
22
+ learning_rate: 1.0e-06
23
+ local_rank: -1
24
+ logging_dir: logs
25
+ lr_scheduler: constant
26
+ lr_warmup_steps: 0
27
+ max_grad_norm: 1.0
28
+ max_train_samples: null
29
+ max_train_steps: 3000
30
+ mixed_precision: null
31
+ noise_offset: 0
32
+ num_train_epochs: 100
33
+ num_validation_images: 4
34
+ output_dir: k7m-vgft-test
35
+ prediction_type: null
36
+ pretrained_model_name_or_path: segmind/Segmind-Vega
37
+ pretrained_vae_model_name_or_path: null
38
+ proportion_empty_prompts: 0
39
+ push_to_hub: false
40
+ random_flip: false
41
+ report_to: tensorboard
42
+ resolution: 512
43
+ resume_from_checkpoint: null
44
+ revision: null
45
+ scale_lr: false
46
+ seed: 42
47
+ snr_gamma: null
48
+ timestep_bias_begin: 0
49
+ timestep_bias_end: 1000
50
+ timestep_bias_multiplier: 1.0
51
+ timestep_bias_portion: 0.25
52
+ timestep_bias_strategy: none
53
+ train_batch_size: 1
54
+ train_data_dir: null
55
+ use_8bit_adam: true
56
+ use_ema: false
57
+ validation_epochs: 1
58
+ validation_prompt: null
59
+ variant: null
logs/text2image-fine-tune-sdxl/1793014268.9098377/hparams.yml ADDED
@@ -0,0 +1,59 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ adam_beta1: 0.9
2
+ adam_beta2: 0.999
3
+ adam_epsilon: 1.0e-08
4
+ adam_weight_decay: 0.01
5
+ allow_tf32: false
6
+ cache_dir: null
7
+ caption_column: text
8
+ center_crop: true
9
+ checkpointing_steps: 250
10
+ checkpoints_total_limit: 2
11
+ dataloader_num_workers: 0
12
+ dataset_config_name: null
13
+ dataset_name: ./img-dataset
14
+ enable_npu_flash_attention: false
15
+ enable_xformers_memory_efficient_attention: false
16
+ gradient_accumulation_steps: 8
17
+ gradient_checkpointing: true
18
+ hub_model_id: null
19
+ hub_token: null
20
+ image_column: image
21
+ image_interpolation_mode: lanczos
22
+ learning_rate: 1.0e-06
23
+ local_rank: -1
24
+ logging_dir: logs
25
+ lr_scheduler: constant
26
+ lr_warmup_steps: 0
27
+ max_grad_norm: 1.0
28
+ max_train_samples: null
29
+ max_train_steps: 2000
30
+ mixed_precision: null
31
+ noise_offset: 0
32
+ num_train_epochs: 100
33
+ num_validation_images: 4
34
+ output_dir: k7m-vgft-test
35
+ prediction_type: null
36
+ pretrained_model_name_or_path: segmind/Segmind-Vega
37
+ pretrained_vae_model_name_or_path: null
38
+ proportion_empty_prompts: 0
39
+ push_to_hub: false
40
+ random_flip: false
41
+ report_to: tensorboard
42
+ resolution: 512
43
+ resume_from_checkpoint: null
44
+ revision: null
45
+ scale_lr: false
46
+ seed: 42
47
+ snr_gamma: null
48
+ timestep_bias_begin: 0
49
+ timestep_bias_end: 1000
50
+ timestep_bias_multiplier: 1.0
51
+ timestep_bias_portion: 0.25
52
+ timestep_bias_strategy: none
53
+ train_batch_size: 1
54
+ train_data_dir: null
55
+ use_8bit_adam: true
56
+ use_ema: false
57
+ validation_epochs: 1
58
+ validation_prompt: null
59
+ variant: null