cookey39 commited on
Commit
82dbff6
1 Parent(s): 3166ba5

End of training

Browse files
.gitattributes CHANGED
@@ -33,3 +33,7 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
 
 
 
 
 
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
36
+ image_0.png filter=lfs diff=lfs merge=lfs -text
37
+ image_1.png filter=lfs diff=lfs merge=lfs -text
38
+ image_2.png filter=lfs diff=lfs merge=lfs -text
39
+ image_3.png filter=lfs diff=lfs merge=lfs -text
README.md ADDED
@@ -0,0 +1,96 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ tags:
3
+ - stable-diffusion-xl
4
+ - stable-diffusion-xl-diffusers
5
+ - diffusers-training
6
+ - text-to-image
7
+ - diffusers
8
+ - lora
9
+ - template:sd-lora
10
+ widget:
11
+
12
+ - text: 'In the style of Terada,colorful, anime-style, illustration, girl, yellow jacket, blonde hair, two buns, peace sign, hands, rainbow gradient, brightness, cheerfulness, small confetti, festive atmosphere, central position, joy, positivity.'
13
+ output:
14
+ url:
15
+ "image_0.png"
16
+
17
+ - text: 'In the style of Terada,colorful, anime-style, illustration, girl, yellow jacket, blonde hair, two buns, peace sign, hands, rainbow gradient, brightness, cheerfulness, small confetti, festive atmosphere, central position, joy, positivity.'
18
+ output:
19
+ url:
20
+ "image_1.png"
21
+
22
+ - text: 'In the style of Terada,colorful, anime-style, illustration, girl, yellow jacket, blonde hair, two buns, peace sign, hands, rainbow gradient, brightness, cheerfulness, small confetti, festive atmosphere, central position, joy, positivity.'
23
+ output:
24
+ url:
25
+ "image_2.png"
26
+
27
+ - text: 'In the style of Terada,colorful, anime-style, illustration, girl, yellow jacket, blonde hair, two buns, peace sign, hands, rainbow gradient, brightness, cheerfulness, small confetti, festive atmosphere, central position, joy, positivity.'
28
+ output:
29
+ url:
30
+ "image_3.png"
31
+
32
+ base_model: cookey39/aam_xl
33
+ instance_prompt: In the style of Terada,
34
+ license: openrail++
35
+ ---
36
+
37
+ # SDXL LoRA DreamBooth - cookey39/teratera
38
+
39
+ <Gallery />
40
+
41
+ ## Model description
42
+
43
+ ### These are cookey39/teratera LoRA adaption weights for cookey39/aam_xl.
44
+
45
+ ## Download model
46
+
47
+ ### Use it with UIs such as AUTOMATIC1111, Comfy UI, SD.Next, Invoke
48
+
49
+ - **LoRA**: download **[`teratera.safetensors` here 💾](/cookey39/teratera/blob/main/teratera.safetensors)**.
50
+ - Place it on your `models/Lora` folder.
51
+ - On AUTOMATIC1111, load the LoRA by adding `<lora:teratera:1>` to your prompt. On ComfyUI just [load it as a regular LoRA](https://comfyanonymous.github.io/ComfyUI_examples/lora/).
52
+ - *Embeddings*: download **[`teratera_emb.safetensors` here 💾](/cookey39/teratera/blob/main/teratera_emb.safetensors)**.
53
+ - Place it on it on your `embeddings` folder
54
+ - Use it by adding `teratera_emb` to your prompt. For example, `In the style of Terada,`
55
+ (you need both the LoRA and the embeddings as they were trained together for this LoRA)
56
+
57
+
58
+ ## Use it with the [🧨 diffusers library](https://github.com/huggingface/diffusers)
59
+
60
+ ```py
61
+ from diffusers import AutoPipelineForText2Image
62
+ import torch
63
+ from huggingface_hub import hf_hub_download
64
+ from safetensors.torch import load_file
65
+
66
+ pipeline = AutoPipelineForText2Image.from_pretrained('stabilityai/stable-diffusion-xl-base-1.0', torch_dtype=torch.float16).to('cuda')
67
+ pipeline.load_lora_weights('cookey39/teratera', weight_name='pytorch_lora_weights.safetensors')
68
+ embedding_path = hf_hub_download(repo_id='cookey39/teratera', filename='teratera_emb.safetensors', repo_type="model")
69
+ state_dict = load_file(embedding_path)
70
+ pipeline.load_textual_inversion(state_dict["clip_l"], token=[], text_encoder=pipeline.text_encoder, tokenizer=pipeline.tokenizer)
71
+ pipeline.load_textual_inversion(state_dict["clip_g"], token=[], text_encoder=pipeline.text_encoder_2, tokenizer=pipeline.tokenizer_2)
72
+
73
+ image = pipeline('In the style of Terada,colorful, anime-style, illustration, girl, yellow jacket, blonde hair, two buns, peace sign, hands, rainbow gradient, brightness, cheerfulness, small confetti, festive atmosphere, central position, joy, positivity.').images[0]
74
+ ```
75
+
76
+ For more details, including weighting, merging and fusing LoRAs, check the [documentation on loading LoRAs in diffusers](https://huggingface.co/docs/diffusers/main/en/using-diffusers/loading_adapters)
77
+
78
+ ## Trigger words
79
+
80
+ To trigger image generation of trained concept(or concepts) replace each concept identifier in you prompt with the new inserted tokens:
81
+
82
+ to trigger concept `TOK` → use `<s0><s1>` in your prompt
83
+
84
+
85
+
86
+ ## Details
87
+ All [Files & versions](/cookey39/teratera/tree/main).
88
+
89
+ The weights were trained using [🧨 diffusers Advanced Dreambooth Training Script](https://github.com/huggingface/diffusers/blob/main/examples/advanced_diffusion_training/train_dreambooth_lora_sdxl_advanced.py).
90
+
91
+ LoRA for the text encoder was enabled. False.
92
+
93
+ Pivotal tuning was enabled: True.
94
+
95
+ Special VAE used for training: None.
96
+
image_0.png ADDED

Git LFS Details

  • SHA256: ad300c0f95cc8aa486790704bace79045f18472b002aef3cdf9539e66b142e1b
  • Pointer size: 132 Bytes
  • Size of remote file: 1.5 MB
image_1.png ADDED

Git LFS Details

  • SHA256: 6643c9ec8fe3533ade902d4813c2723ff6ead26ffb95545244515fcf0c80a508
  • Pointer size: 132 Bytes
  • Size of remote file: 1.53 MB
image_2.png ADDED

Git LFS Details

  • SHA256: 5e58d4e94e302a47cc8a5a195e4fe1b6db6b1b54706d3fb8a04e18a3a9af8a9f
  • Pointer size: 132 Bytes
  • Size of remote file: 1.45 MB
image_3.png ADDED

Git LFS Details

  • SHA256: 3c006843d57174689b5339737b64d7d197099a43ba076c67667c42eb487009a8
  • Pointer size: 132 Bytes
  • Size of remote file: 1.32 MB
logs/dreambooth-lora-sd-xl/1716938838.3230867/events.out.tfevents.1716938838.cookey39-osagav1ul6md-main.658.1 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:94f54fcf275b5bee29835a1b6cd71d5619b29add2ac75fceb5c6a3f112fbe5af
3
+ size 3792
logs/dreambooth-lora-sd-xl/1716938838.3255281/hparams.yml ADDED
@@ -0,0 +1,80 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ adam_beta1: 0.9
2
+ adam_beta2: 0.99
3
+ adam_epsilon: 1.0e-08
4
+ adam_weight_decay: 0.0001
5
+ adam_weight_decay_text_encoder: null
6
+ allow_tf32: false
7
+ cache_dir: null
8
+ cache_latents: false
9
+ caption_column: prompt
10
+ center_crop: false
11
+ checkpointing_steps: 2000
12
+ checkpoints_total_limit: null
13
+ class_data_dir: null
14
+ class_prompt: null
15
+ dataloader_num_workers: 0
16
+ dataset_config_name: null
17
+ dataset_name: /openbayes/home/img
18
+ do_edm_style_training: false
19
+ enable_xformers_memory_efficient_attention: false
20
+ gradient_accumulation_steps: 1
21
+ gradient_checkpointing: true
22
+ hub_model_id: null
23
+ hub_token: null
24
+ image_column: image
25
+ instance_data_dir: null
26
+ instance_prompt: In the style of Terada,
27
+ learning_rate: 1.0
28
+ local_rank: -1
29
+ logging_dir: logs
30
+ lora_unet_blocks: null
31
+ lr_num_cycles: 1
32
+ lr_power: 1.0
33
+ lr_scheduler: constant
34
+ lr_warmup_steps: 0
35
+ max_grad_norm: 1.0
36
+ max_train_steps: 1000
37
+ mixed_precision: bf16
38
+ noise_offset: 0
39
+ num_class_images: 100
40
+ num_new_tokens_per_abstraction: 2
41
+ num_train_epochs: 143
42
+ num_validation_images: 4
43
+ optimizer: prodigy
44
+ output_dir: teratera
45
+ pretrained_model_name_or_path: cookey39/aam_xl
46
+ pretrained_vae_model_name_or_path: null
47
+ prior_generation_precision: null
48
+ prior_loss_weight: 1.0
49
+ prodigy_beta3: null
50
+ prodigy_decouple: true
51
+ prodigy_safeguard_warmup: true
52
+ prodigy_use_bias_correction: true
53
+ push_to_hub: true
54
+ random_flip: false
55
+ rank: 8
56
+ repeats: 1
57
+ report_to: tensorboard
58
+ resolution: 1024
59
+ resume_from_checkpoint: null
60
+ revision: null
61
+ sample_batch_size: 4
62
+ scale_lr: false
63
+ seed: 0
64
+ snr_gamma: 5.0
65
+ text_encoder_lr: 1.0
66
+ token_abstraction: TOK
67
+ train_batch_size: 3
68
+ train_text_encoder: false
69
+ train_text_encoder_frac: 1.0
70
+ train_text_encoder_ti: true
71
+ train_text_encoder_ti_frac: 0.5
72
+ use_8bit_adam: false
73
+ use_blora: false
74
+ use_dora: false
75
+ validation_epochs: 50
76
+ validation_prompt: In the style of Terada,colorful, anime-style, illustration, girl,
77
+ yellow jacket, blonde hair, two buns, peace sign, hands, rainbow gradient, brightness,
78
+ cheerfulness, small confetti, festive atmosphere, central position, joy, positivity.
79
+ variant: null
80
+ with_prior_preservation: false
logs/dreambooth-lora-sd-xl/events.out.tfevents.1716938838.cookey39-osagav1ul6md-main.658.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b837248e68b36becc2e4a8b194f2e724b53b7ca0160e71f1045b31fda13156b5
3
+ size 24447826
pytorch_lora_weights.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b5d4acd52a84fe291dd28eefc7afb3f8fa9489ff71552747d842b1831ecbe7a9
3
+ size 46615272
teratera.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e411acc582305de46d91d4dfb876d7fdb2c703f41d05749df01ed642a366afd6
3
+ size 46698072
teratera_emb.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6718436631129e2ee1d51a1be1e64280ea36fad5c14d18c6b5b940c8d284af5d
3
+ size 16536