davidrd123/

rembrandt_sketches_qwen

Training loss

Training settings

• Training steps: 500
• Learning rate: Automagic (adaptive)
  • Learning rate schedule: Automagic optimizer
  • Warmup steps: 0 (commented out: 100)
• Max grad value: 1.0
• Effective batch size: 4
  • Micro-batch size: 4
  • Gradient accumulation steps: 1
  • Number of GPUs: 1
• Gradient checkpointing: True (unsloth)
• Prediction type: logit_normal
• Optimizer: automagic
• Trainable parameter precision: Pure BF16
• Base model precision: Pure BF16
• Caption dropout probability: 0.0% (not specified)
• LoRA Rank: 64
• LoRA Alpha: 64 (auto-set to rank)
• LoRA Dropout: 0.0
• LoRA initialisation style: default
• LoRA mode: Standard

Datasets

rembrandt-sketch-640

• Repeats: 8
• Total number of images: TBD
• Total number of aspect buckets: 12 (from ar_buckets config)
• Resolution: 640 px
• Cropped: False
• Crop style: aspect ratio bucketing
• Used for regularisation data: No

rembrandt-sketch-1328

• Repeats: 4
• Total number of images: TBD
• Total number of aspect buckets: 12 (from ar_buckets config)
• Resolution: 1328 px
• Cropped: False
• Crop style: aspect ratio bucketing
• Used for regularisation data: No