Trained for 288 epochs and 8000 steps.

Trained with datasets ['text-embeds', 'graffito']
Learning rate 0.0001, batch size 1, and 4 gradient accumulation steps.
Trained with flow_matching prediction type and rescaled_betas_zero_snr=False
Using 'trailing' timestep spacing.
Base model: Qwen/Qwen-Image
VAE: Qwen/Qwen-Image

pytorch_lora_weights.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:99345db831a086ed421729e299a395a998a1d6991de7521ed4edf717420c5410
+oid sha256:44850aeceb0fd5e765e843ce2b8ebfa8bd912bb0148b0a18cd46c52ca4ae5e5d
 size 319046320