Hello,
When training the REPA-E model on the facial dataset using "3. Train the REPA-E model," I encountered an issue with "Speckle Noise" in the generated images (as shown in the attached figure). During training, I used the following parameters:
learning-rate = 5e-5, batch-size = 128, vae = vavae(f16d32)
I would appreciate your guidance on whether a specific parameter setting is causing this issue, or if there are other possible solutions to address it.
Hello,
When training the REPA-E model on the facial dataset using "3. Train the REPA-E model," I encountered an issue with "Speckle Noise" in the generated images (as shown in the attached figure). During training, I used the following parameters:
learning-rate = 5e-5, batch-size = 128, vae = vavae(f16d32)I would appreciate your guidance on whether a specific parameter setting is causing this issue, or if there are other possible solutions to address it.