Hugging Face's logo

terminusresearch
/
pixart-900m-1024-ft-v0.7-stage2

Text-to-Image
Diffusers
Safetensors
PixArtSigmaPipeline
stable-diffusion
stable-diffusion-diffusers
simpletuner
full
Model card Files Community
PseudoTerminal X commited on
Commit
047a367
•
1 Parent(s): 9083b4f

Trained for 9 epochs and 20000 steps.

Browse files

Trained with datasets ['text-embeds', 'shutterstock', 'nijijourney', 'bg20k-1024', 'photo-aesthetics', 'text-1mp', 'cinemamix-1mp']
Learning rate 1e-06, batch size 16, and 1 gradient accumulation steps.
Used DDPM noise scheduler for training with epsilon prediction type and rescaled_betas_zero_snr=False
Using 'linspace' timestep spacing.
Base model: terminusresearch/pixart-900m-1024-ft-v0.6
VAE: madebyollin/sdxl-vae-fp16-fix

Files changed (13) hide show
  1. README.md +2 -2
  2. optimizer.bin +1 -1
  3. random_states_0.pkl +1 -1
  4. scheduler.bin +1 -1
  5. training_state-bg20k-1024.json +2 -2
  6. training_state-cinemamix-1mp.json +0 -0
  7. training_state-nijijourney.json +0 -0
  8. training_state-photo-aesthetics.json +0 -0
  9. training_state-shutterstock.json +0 -0
  10. training_state-text-1mp.json +0 -0
  11. training_state.json +1 -1
  12. transformer/config.json +1 -1
  13. transformer/diffusion_pytorch_model.safetensors +1 -1
README.md CHANGED
@@ -552,7 +552,7 @@ You may reuse the base model text encoder for inference.
552
  ## Training settings
553
 
554
  - Training epochs: 9
555
- - Training steps: 19500
556
  - Learning rate: 1e-06
557
  - Effective batch size: 16
558
  - Micro-batch size: 16
@@ -593,7 +593,7 @@ You may reuse the base model text encoder for inference.
593
  - Crop aspect: square
594
  ### photo-aesthetics
595
  - Repeats: 0
596
- - Total number of images: 33104
597
  - Total number of aspect buckets: 3
598
  - Resolution: 1.0 megapixels
599
  - Cropped: True
 
552
  ## Training settings
553
 
554
  - Training epochs: 9
555
+ - Training steps: 20000
556
  - Learning rate: 1e-06
557
  - Effective batch size: 16
558
  - Micro-batch size: 16
 
593
  - Crop aspect: square
594
  ### photo-aesthetics
595
  - Repeats: 0
596
+ - Total number of images: 33120
597
  - Total number of aspect buckets: 3
598
  - Resolution: 1.0 megapixels
599
  - Cropped: True
optimizer.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:620d6048a732cb2a8be9efd3cbf3d3173e314092738b7cb55d699a9c1b61330a
3
  size 5451415117
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fd17b631b90b2d1a778eb85fa3ecffbf01f365be762ceaec87f33079daedfa5b
3
  size 5451415117
random_states_0.pkl CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:49072f5975971477bf33556cfd7b511e714bbce16d41f950e0d8f9cbfb8e14d5
3
  size 14344
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e0384caf9c3024414cd1fb7a988e30c4ac871501012c72308b860aeb756e2899
3
  size 14344
scheduler.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:6803d33ca8df001910c8ae963ffc4af4df0382fa313920e2d980a2a4edd08989
3
  size 1000
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c4e75ee4d38c1b2615ee55dec73b7f55789603c9bfe6c5405a928299b4f089eb
3
  size 1000
training_state-bg20k-1024.json CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:548ec2aa970ae128454546bb3f7b4b8a153ca98cd922a1fbcf48ca766e4ed082
3
- size 15646268
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:978e67a40ba7303e937fa17129f4b3af28dc7ec38cedcb90db6ccbe1b8e14d27
3
+ size 15893492
training_state-cinemamix-1mp.json CHANGED
The diff for this file is too large to render. See raw diff
 
training_state-nijijourney.json CHANGED
The diff for this file is too large to render. See raw diff
 
training_state-photo-aesthetics.json CHANGED
The diff for this file is too large to render. See raw diff
 
training_state-shutterstock.json CHANGED
The diff for this file is too large to render. See raw diff
 
training_state-text-1mp.json CHANGED
The diff for this file is too large to render. See raw diff
 
training_state.json CHANGED
@@ -1 +1 @@
1
- {"global_step": 19500, "epoch_step": 19500, "epoch": 10, "exhausted_backends": ["celebrities", "gay", "moviecollection", "ethnic", "cinemamix-1mp"], "repeats": {"movieposters": 20, "celebrities": 0, "normalnudes": 3, "moviecollection": 0, "gay": 0, "ethnic": 0, "nsfw-1024": 0, "experimental": 0, "cinemamix-1mp": 0}}
 
1
+ {"global_step": 20000, "epoch_step": 20000, "epoch": 10, "exhausted_backends": ["celebrities", "gay", "moviecollection", "ethnic", "cinemamix-1mp"], "repeats": {"movieposters": 20, "celebrities": 0, "normalnudes": 3, "moviecollection": 0, "gay": 0, "ethnic": 0, "nsfw-1024": 0, "experimental": 0, "cinemamix-1mp": 0}}
transformer/config.json CHANGED
@@ -1,7 +1,7 @@
1
  {
2
  "_class_name": "PixArtTransformer2DModel",
3
  "_diffusers_version": "0.30.0.dev0",
4
- "_name_or_path": "/notebooks/datasets/lite-models/checkpoint-18500",
5
  "activation_fn": "gelu-approximate",
6
  "attention_bias": true,
7
  "attention_head_dim": 72,
 
1
  {
2
  "_class_name": "PixArtTransformer2DModel",
3
  "_diffusers_version": "0.30.0.dev0",
4
+ "_name_or_path": "/notebooks/datasets/lite-models/checkpoint-19500",
5
  "activation_fn": "gelu-approximate",
6
  "attention_bias": true,
7
  "attention_head_dim": 72,
transformer/diffusion_pytorch_model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:04b94317e7f37f2b6b9ed88002a655068c46b95b7da01c33a0a33fad54fe025d
3
  size 1816969728
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0cfc3b55c41026d5b780c3d8120f2a6e5e1b4f84c86950698461a2f6437cc0b1
3
  size 1816969728