hemular/maxduman-flux-lora

maxduman

Prompt

A close-up of maxduman face adorned with intricate black and blue patterns. The left side of the face is predominantly yellow, with symbols and doodles, while the right side is dark, featuring mechanical elements. The eye on the left is a striking shade of yellow, contrasting sharply with the surrounding patterns. The face is partially covered by a hooded garment

Prompt

reate a pop art image featuring maxduman with bold colors and simple shapes, the character should be striking a dynamic pose against a vibrant background, neon-lit scene, graphic novel illustration by mztgr7 henri matisse, flat color, low poly art, high contrast, sharp geometric color blocks are used for shading

Prompt

-

Prompt

-

Prompt

-

Prompt

-

Prompt

-

Prompt

-

Prompt

-

Prompt

Sumo wrestler max duman, sports action shot, shot on sony alpha

Prompt

man A beautiful brunette pilot, wearing headphones and looking at the camera over his shoulder, beautiful, moody lighting, interior of a vintage airplane cockpit, ww2, world war 2, best quality, full body portrait, real picture, intricate details, depth of field, Fujifilm xt3, beautiful lighting, raw photo, 8k uhd, film grain, ray trace, black and white vintage photo

Model Training

A Flux LoRA trained using Fluxgym

Training environment: Runpod A40 (48 GB VRAM, 50GB RAM, 9 vCPU)

Training flags:

  sd-scripts/flux_train_network.py \
  --pretrained_model_name_or_path "/app/fluxgym/models/unet/flux1-dev.sft" \
  --clip_l "/app/fluxgym/models/clip/clip_l.safetensors" \
  --t5xxl "/app/fluxgym/models/clip/t5xxl_fp16.safetensors" \
  --ae "/app/fluxgym/models/vae/ae.sft" \
  --cache_latents_to_disk \
  --save_model_as safetensors \
  --sdpa --persistent_data_loader_workers \
  --max_data_loader_n_workers 2 \
  --seed 42 \
  --gradient_checkpointing \
  --mixed_precision bf16 \
  --save_precision bf16 \
  --network_module networks.lora_flux \
  --network_dim 16 \
  --optimizer_type adamw8bit \
  --learning_rate 1e-3 \
  --cache_text_encoder_outputs \
  --cache_text_encoder_outputs_to_disk \
  --fp8_base \
  --highvram \
  --max_train_epochs 15 \
  --save_every_n_epochs 1 \
  --dataset_config "/app/fluxgym/outputs/maxdumanv4/dataset.toml" \
  --output_dir "/app/fluxgym/outputs/maxdumanv4" \
  --output_name maxdumanv4 \
  --timestep_sampling shift \
  --discrete_flow_shift 3.1582 \
  --model_prediction_type raw \
  --guidance_scale 1 \
  --loss_type l2 \
  --enable_bucket \
  --min_snr_gamma 5 \
  --multires_noise_discount 0.3 \
  --multires_noise_iterations 6 \
  --noise_offset 0.1 \
  --train_batch_size 2

Plots:

Chosen checkpoint: maxdumanv4-000007.safetensor

Download model

Download them in the Files & versions tab.