Upload folder using huggingface_hub

README.md

@@ -8,16 +8,11 @@ tags:
 - fluxgym
 widget:
 -   output:
-        url: sample/q-avatar_001200_00_20241017140310.png
-    text: a Q avatar of
--   output:
-        url: sample/q-avatar_001200_01_20241017140329.png
-    text: in style with Q avatar
-base_model: black-forest-labs/FLUX.1-dev
-instance_prompt: a Q avatar of
-license: other
-license_name: flux-1-dev-non-commercial-license
-license_link: https://huggingface.co/black-forest-labs/FLUX.1-dev/blob/main/LICENSE.md
+        url: sample/q-avatar_001200_00_20241018075331.png
+    text: a Q avatar
+base_model: stabilityai/stable-diffusion-xl-base-1.0
+instance_prompt: a Q avatar
+license: openrail++
 ---
 
 # q_avatar
@@ -28,7 +23,7 @@ A Flux LoRA trained on a local computer with [Fluxgym](https://github.com/cockta
 
 ## Trigger words
 
-You should use `a Q avatar of` to trigger the image generation.
+You should use `a Q avatar` to trigger the image generation.
 
 ## Download model and use it with ComfyUI, AUTOMATIC1111, SD.Next, Invoke AI, Forge, etc.
 

dataset.toml

@@ -10,5 +10,5 @@ keep_tokens = 1
 
   [[datasets.subsets]]
   image_dir = '/app/fluxgym/datasets/q-avatar'
-  class_tokens = 'a Q avatar of'
+  class_tokens = 'a Q avatar'
   num_repeats = 10

q-avatar-000004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:93a2a6295f1811cab771a57f8ffce195274028d116318fa9eeb9ac152d03db3a
-size 39761288
+oid sha256:6adb4c46e22678dc09b72e7151e88fb7ab8b8e963a42b1da87582abd45ebe941
+size 25635684

q-avatar-000008.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1deb62709affe885424843f659018bd11aa9e4191ab9b243a3de2e6743ccb236
-size 39761288
+oid sha256:b2bec8785973c44e1cd4a5d486c6e43dfb1833f236ce4541cfe4cfaa8843e675
+size 25635684

q-avatar-000012.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1a13a5f410c156a82cee7e13bdb3f71be9e7393c71df74c26bffd436423b1dba
-size 39761288
+oid sha256:7ded7d821cf86f975c343e4fee3afb74dc539aceebf2b88a32e3305878d29745
+size 25635684

q-avatar.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:046033446bbe6d18ef205282f20e16f6cabc011953e85cfa207faf187278849c
-size 39761288
+oid sha256:c19ced7aa4689bab9b7226c5cb9e366d437eced074fa1442188bfd876fda1a95
+size 25635684

sample_prompts.txt

@@ -1,2 +1 @@
-a Q avatar of
-in style with Q avatar
+a Q avatar

train.sh

@@ -1,34 +1,18 @@
 accelerate launch \
-  --mixed_precision bf16 \
   --num_cpu_threads_per_process 1 \
-  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 \
+  sd-scripts/sdxl_train_network.py \
+  --pretrained_model_name_or_path "/app/fluxgym/models/unet/stabilityai/stable-diffusion-xl-base-1.0/sd_xl_base_1.0.safetensors" \
   --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 \
+  --mixed_precision fp16 \
+  --save_precision fp16 \
+  --network_module networks.lora \
   --network_dim 4 \
   --optimizer_type adamw8bit \--sample_prompts="/app/fluxgym/outputs/q-avatar/sample_prompts.txt" --sample_every_n_steps="100" \
   --learning_rate 8e-4 \
-  --cache_text_encoder_outputs \
-  --cache_text_encoder_outputs_to_disk \
-  --fp8_base \
-  --highvram \
   --max_train_epochs 16 \
   --save_every_n_epochs 4 \
   --dataset_config "/app/fluxgym/outputs/q-avatar/dataset.toml" \
   --output_dir "/app/fluxgym/outputs/q-avatar" \
   --output_name q-avatar \
-  --timestep_sampling shift \
-  --discrete_flow_shift 3.1582 \
-  --model_prediction_type raw \
-  --guidance_scale 1 \
-  --loss_type l2 \
+  --xformers --no_half_vae \
+