You can deploy this model and start earning money today!

PJMixers-Images/
Qwen-Image-SynthSat-v1.0-warmup-stage2.5-nsfw

Name: Qwen-Image-SynthSat-v1.0-warmup-stage2.5-nsfw
Availability: InStock

not-for-all-audiences

Views

No views yet

Qwen-Image-SynthSat-v1.0-warmup-stage2.5-nsfw

The idea of the "warmup stages" is to sort of pretrain the LoRA weights so that they can be further trained with more refined data, but maybe have a boost in overall knowledge and stuff.

This stage is 256p+640p, re-using the weights from stage2-step1400, using only NSFW sources (which is a mix of NSFW+SFW).

Training Stages

SynthSat-v1.0-warmup-stage1 is 256p only, using freshly initialized weights.
SynthSat-v1.0-warmup-stage2 is 640p only, re-using the weights from SynthSat-v1.0-warmup-stage1.
SynthSat-v1.0-warmup-stage2.5-nsfw is 256p+640p, re-using the weights from SynthSat-v1.0-warmup-stage2-step1400, using only NSFW sources (which is a mix of NSFW+SFW).
Next stage is either warmup-stage3 doing 640p+1328p or v1.0 doing 640p+1328p+1536p on a more refined dataset.

Current Stage Details

INFO:musubi_tuner.hv_train_network:Using timestep bucketing. Number of buckets: 10
INFO:musubi_tuner.hv_train_network:Load dataset config from /media/xzuyn/Toshiba1/musubi-stuff/dataset_configs/SynthSat-v1.0-warmup-stage2.5-nsfw.toml
INFO:musubi_tuner.dataset.image_video_dataset:glob images in /media/xzuyn/Toshiba1/000_SynthSatNSFWOnly
INFO:musubi_tuner.dataset.image_video_dataset:found 5326 images
INFO:musubi_tuner.dataset.image_video_dataset:glob images in /media/xzuyn/Toshiba1/000_SynthSatNSFWOnly
INFO:musubi_tuner.dataset.image_video_dataset:found 5326 images
INFO:musubi_tuner.dataset.config_utils:[Dataset 0]
  is_image_dataset: True
  resolution: (256, 256)
  batch_size: 2
  num_repeats: 1
  caption_extension: ".txt"
  enable_bucket: True
  bucket_no_upscale: True
  cache_directory: "/media/xzuyn/NVMe/LClones/musubi-tuner/dataset_cache/SynthSat-v1.0-warmup-stage2.5-nsfw-256"
  debug_dataset: False
    image_directory: "/media/xzuyn/Toshiba1/000_SynthSatNSFWOnly"
    image_jsonl_file: "None"
    fp_latent_window_size: 9
    fp_1f_clean_indices: None
    fp_1f_target_index: None
    fp_1f_no_post: False
    flux_kontext_no_resize_control: False
    qwen_image_edit_no_resize_control: False
    qwen_image_edit_control_resolution: None
[Dataset 1]
  is_image_dataset: True
  resolution: (640, 640)
  batch_size: 2
  num_repeats: 1
  caption_extension: ".txt"
  enable_bucket: True
  bucket_no_upscale: True
  cache_directory: "/media/xzuyn/NVMe/LClones/musubi-tuner/dataset_cache/SynthSat-v1.0-warmup-stage2.5-nsfw-640"
  debug_dataset: False
    image_directory: "/media/xzuyn/Toshiba1/000_SynthSatNSFWOnly"
    image_jsonl_file: "None"
    fp_latent_window_size: 9
    fp_1f_clean_indices: None
    fp_1f_target_index: None
    fp_1f_no_post: False
    flux_kontext_no_resize_control: False
    qwen_image_edit_no_resize_control: False
    qwen_image_edit_control_resolution: None
INFO:musubi_tuner.dataset.image_video_dataset:bucket: (160, 400), count: 1
INFO:musubi_tuner.dataset.image_video_dataset:bucket: (176, 368), count: 20
INFO:musubi_tuner.dataset.image_video_dataset:bucket: (192, 336), count: 325
INFO:musubi_tuner.dataset.image_video_dataset:bucket: (208, 304), count: 2198
INFO:musubi_tuner.dataset.image_video_dataset:bucket: (224, 288), count: 2114
INFO:musubi_tuner.dataset.image_video_dataset:bucket: (240, 272), count: 73
INFO:musubi_tuner.dataset.image_video_dataset:bucket: (256, 256), count: 65
INFO:musubi_tuner.dataset.image_video_dataset:bucket: (272, 240), count: 10
INFO:musubi_tuner.dataset.image_video_dataset:bucket: (288, 224), count: 250
INFO:musubi_tuner.dataset.image_video_dataset:bucket: (304, 208), count: 254
INFO:musubi_tuner.dataset.image_video_dataset:bucket: (336, 192), count: 15
INFO:musubi_tuner.dataset.image_video_dataset:bucket: (368, 176), count: 1
INFO:musubi_tuner.dataset.image_video_dataset:total batches: 2666
INFO:musubi_tuner.dataset.image_video_dataset:bucket: (400, 1024), count: 1
INFO:musubi_tuner.dataset.image_video_dataset:bucket: (416, 976), count: 1
INFO:musubi_tuner.dataset.image_video_dataset:bucket: (432, 944), count: 6
INFO:musubi_tuner.dataset.image_video_dataset:bucket: (448, 912), count: 8
INFO:musubi_tuner.dataset.image_video_dataset:bucket: (464, 880), count: 27
INFO:musubi_tuner.dataset.image_video_dataset:bucket: (480, 848), count: 278
INFO:musubi_tuner.dataset.image_video_dataset:bucket: (496, 816), count: 24
INFO:musubi_tuner.dataset.image_video_dataset:bucket: (512, 800), count: 40
INFO:musubi_tuner.dataset.image_video_dataset:bucket: (528, 768), count: 2110
INFO:musubi_tuner.dataset.image_video_dataset:bucket: (544, 752), count: 1603
INFO:musubi_tuner.dataset.image_video_dataset:bucket: (560, 720), count: 209
INFO:musubi_tuner.dataset.image_video_dataset:bucket: (576, 704), count: 361
INFO:musubi_tuner.dataset.image_video_dataset:bucket: (592, 688), count: 37
INFO:musubi_tuner.dataset.image_video_dataset:bucket: (608, 672), count: 24
INFO:musubi_tuner.dataset.image_video_dataset:bucket: (624, 656), count: 13
INFO:musubi_tuner.dataset.image_video_dataset:bucket: (640, 640), count: 49
INFO:musubi_tuner.dataset.image_video_dataset:bucket: (656, 624), count: 5
INFO:musubi_tuner.dataset.image_video_dataset:bucket: (672, 608), count: 7
INFO:musubi_tuner.dataset.image_video_dataset:bucket: (688, 592), count: 1
INFO:musubi_tuner.dataset.image_video_dataset:bucket: (704, 576), count: 14
INFO:musubi_tuner.dataset.image_video_dataset:bucket: (720, 560), count: 225
INFO:musubi_tuner.dataset.image_video_dataset:bucket: (752, 544), count: 19
INFO:musubi_tuner.dataset.image_video_dataset:bucket: (768, 528), count: 244
INFO:musubi_tuner.dataset.image_video_dataset:bucket: (800, 512), count: 4
INFO:musubi_tuner.dataset.image_video_dataset:bucket: (848, 480), count: 15
INFO:musubi_tuner.dataset.image_video_dataset:bucket: (880, 464), count: 1
INFO:musubi_tuner.dataset.image_video_dataset:total batches: 2671
INFO:musubi_tuner.hv_train_network:preparing accelerator
accelerator device: cuda
INFO:musubi_tuner.hv_train_network:DiT precision: torch.bfloat16, weight precision: None
INFO:musubi_tuner.hv_train_network:Loading DiT model from /media/xzuyn/NVMe/LClones/musubi-tuner/source_models/qwen_image_bf16.safetensors
INFO:musubi_tuner.qwen_image.qwen_image_model:Creating QwenImageTransformer2DModel
INFO:musubi_tuner.qwen_image.qwen_image_model:Loading DiT model from /media/xzuyn/NVMe/LClones/musubi-tuner/source_models/qwen_image_bf16.safetensors, device=cpu
INFO:musubi_tuner.utils.lora_utils:Loading model files: ['/media/xzuyn/NVMe/LClones/musubi-tuner/source_models/qwen_image_bf16.safetensors']
INFO:musubi_tuner.utils.lora_utils:Loading state dict with FP8 optimization. Dtype of weight: None, hook enabled: False
Loading qwen_image_bf16.safetensors: 100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1933/1933 [02:55<00:00, 11.01key/s]
INFO:musubi_tuner.modules.fp8_optimization_utils:Number of optimized Linear layers: 840
INFO:musubi_tuner.modules.fp8_optimization_utils:Number of monkey-patched Linear layers: 840
INFO:musubi_tuner.qwen_image.qwen_image_model:Loaded DiT model from /media/xzuyn/NVMe/LClones/musubi-tuner/source_models/qwen_image_bf16.safetensors, info=<All keys matched successfully>
INFO:musubi_tuner.hv_train_network:enable swap 5 blocks to CPU from device: cuda
QwenModel: Block swap enabled. Swapping 5 blocks out of 60 blocks. Supports backward: True
import network module: networks.lora_qwen_image
INFO:musubi_tuner.networks.lora:create LoRA network. base dim (rank): 16, alpha: 4.0
INFO:musubi_tuner.networks.lora:neuron dropout: p=0.125, rank dropout: p=0.0, module dropout: p=0.0
INFO:musubi_tuner.networks.lora:create LoRA for U-Net/DiT: 840 modules.
INFO:musubi_tuner.networks.lora:enable LoRA for U-Net: 840 modules
load network weights from /media/xzuyn/NVMe/LClones/musubi-tuner/source_models/Qwen-Image-SynthSat-v1.0-warmup-stage2-step00001400.safetensors: <All keys matched successfully>
QwenModel: Gradient checkpointing enabled. Activation CPU offloading: True
prepare optimizer, data loader etc.
INFO:musubi_tuner.hv_train_network:use came_pytorch.CAME | {'weight_decay': 0.01, 'enable_8bit': True, 'enable_cautious': True, 'enable_stochastic_rounding': True}
==== CAME Modifications ====
- Stochastic Rounding enabled.
- Cautious Masking enabled.
- 8-bit enabled: block_size=2048, min_8bit_size=16384.
==== CAME Modifications ====
override steps. steps for 1 epochs is / 指定エポックまでのステップ数: 5337
INFO:musubi_tuner.hv_train_network:preparing fused backward pass stuff
running training / 学習開始
  num train items / 学習画像、動画数: 10652
  num batches per epoch / 1epochのバッチ数: 5337
  num epochs / epoch数: 1
  batch size per device / バッチサイズ: 2, 2
  gradient accumulation steps / 勾配を合計するステップ数 = 1
  total optimization steps / 学習ステップ数: 5337
INFO:musubi_tuner.hv_train_network:set DiT model name for metadata: /media/xzuyn/NVMe/LClones/musubi-tuner/source_models/qwen_image_bf16.safetensors
INFO:musubi_tuner.hv_train_network:set VAE model name for metadata: /media/xzuyn/NVMe/LClones/musubi-tuner/source_models/vae_diffusion_pytorch_model.safetensors

PJMixers-Images/Qwen-Image-SynthSat-v1.0-warmup-stage2.5-nsfw

Qwen-Image-SynthSat-v1.0-warmup-stage2.5-nsfw

Training Stages

Current Stage Details

PJMixers-Images/
Qwen-Image-SynthSat-v1.0-warmup-stage2.5-nsfw