ghost
Sign In
homemodelsimagesvideospostsarticlesbountieschallengeseventsupdatesshop

We Are ⚡Electric [XYZ]

1

idle's Avatar

idle

Oct 24, 2023

(Updated: 2 years ago)

workflows
lora training
We Are ⚡Electric [XYZ]

As requested by @denrakeiw
Here is an XYZ for

🚗 We Are Electric ⚡


Loras:

  1. Lora FA <lora:FF-We-Are-Electric-XL-Cr2-v0564-FA:1>

    • Module: networks.lora_fa

    • Dim: 64

    • Alpha: 64.0

  2. Lycoris - preset FULL <lora:WeAreElectric-FFai-LycoLora:1>

    • Module: lycoris.kohya

    • Resolution: 1024x1024

    • Preset: full

    • Algo: lora

    • Dropout: 0.0

  3. Lycroris LoKR preset FULL with SDXL v-prediction <lora:WeAreElectric-FFai-LoKr-vpred:1>

    • Preset: full

    • Algo: lokr

    • Factor: 8

    • Dim: 100,000,000

    • Alpha: 0.69

  4. LoRA - High Alpha Experiment (1.7GB oversized lora)

    • Module: networks.lora

    • Dim: 256

    • Alpha: 512.0

  5. CivitAI Lora <lora:We_are_Electric:1>

    • Resolution: 1024x1024

    • Architecture: stable-diffusion-xl-v1-base/lora

    • Learning Rate: 0.0005

    • UNet LR: 0.0005

    • TE LR: 5e-05

    • Network Dim/Rank: 32.0 Alpha: 1.0

    • Optimizer: bitsandbytes.optim.adamw.AdamW8bit(weight_decay=0.1)

    • Epoch: 15

    • Batches per epoch: 455

    • Gradient accumulation steps: 1

      Train images: 1820

    • Regularization images: 0

      Multires noise iterations: 6.0

    • Multires noise discount: 0.3

    • Min SNR gamma: 5.0

      UNet weight average magnitude: 4.496516309802129

      UNet weight average strength: 0.016011949140159916

      Text Encoder (1) weight average magnitude: 1.8189308640704058

      Text Encoder (1) weight average strength: 0.009030077111409775

      Text Encoder (2) weight average magnitude: 1.835224805119421

      Text Encoder (2) weight average strength: 0.007039581994775502

  6. FFusion LoRA <lora:FFusion-We-Are-Electric.64LORA:1>:

    • Description: Merged and scaled using dynamic weights from Loras 1-5.

1