Skip to content

Open-Sora: Democratizing Efficient Video Production for All

Python

add lecam support

Shen-Chenhui1d6cee302f07
diff --git a/configs/vae_magvit_v2/train/16x128x128_pixabay_8_GPU_pipeline.py b/configs/vae_magvit_v2/train/16x128x128_pixabay_8_GPU_pipeline.py
index b356c46..8405bfa 100644
--- a/configs/vae_magvit_v2/train/16x128x128_pixabay_8_GPU_pipeline.py
+++ b/configs/vae_magvit_v2/train/16x128x128_pixabay_8_GPU_pipeline.py
@@ -84,9 +84,9 @@ magvit uses about # samples (K) * epochs ~ 2-5 K, num_frames = 4, reso = 128
epochs = 200
log_every = 1
-ckpt_every = 50
+ckpt_every = 1 # 50
load = None
-batch_size = 32
+batch_size = 4 # 32
lr = 1e-4
grad_clip = 1.0
diff --git a/scripts/train-vae-v2.py b/scripts/train-vae-v2.py
index fb18ca5..b6cdd25 100644
--- a/scripts/train-vae-v2.py
+++ b/scripts/train-vae-v2.py
@@ -225,11 +225,13 @@ def main():
if disc_lr_scheduler is not None:
booster.load_lr_scheduler(disc_lr_scheduler, os.path.join(cfg.load, "disc_lr_scheduler"))
# LeCam EMA for discriminator
- lecam_path = os.path.join(cfg.load, "lecam_state.json")
+ lecam_path = os.path.join(cfg.load, "lecam_states.json")
if cfg.lecam_loss_weight is not None and os.path.exists(lecam_path):
lecam_state = load_json(lecam_path)
lecam_ema_real, lecam_ema_fake = lecam_state["lecam_ema_real"], lecam_state["lecam_ema_fake"]
lecam_ema = LeCamEMA(decay=cfg.ema_decay, ema_real=lecam_ema_real, ema_fake=lecam_ema_fake, dtype=dtype, device=device)
+ else:
+ print(f"lecan not loaded, path: {lecam_path}, lecame loss weight {cfg.lecam_loss_weight}")
running_states = load_json(os.path.join(cfg.load, "running_states.json"))
dist.barrier()
start_epoch, start_step, sampler_start_idx = running_states["epoch"], running_states["step"], running_states["sample_start_index"]
@@ -403,7 +405,7 @@ def main():
real_video = real_video if cfg.gradient_penalty_loss_weight is not None else None,
)
disc_loss = weighted_d_adversarial_loss + lecam_loss + gradient_penalty_loss
- if cfg.ema_decay is not None:
+ if cfg.lecam_loss_weight is not None:
# SCH: TODO: is this written properly like this for moving average? e.g. distributed training etc.
# lecam_ema_real = lecam_ema_real * cfg.ema_decay + (1 - cfg.ema_decay) * torch.mean(real_logits.clone().detach())
# lecam_ema_fake = lecam_ema_fake * cfg.ema_decay + (1 - cfg.ema_decay) * torch.mean(fake_logits.clone().detach())
@@ -478,6 +480,7 @@ def main():
"global_step": global_step+1,
"sample_start_index": (step+1) * cfg.batch_size,
}
+
lecam_ema_real, lecam_ema_fake = lecam_ema.get()
lecam_state = {
"lecam_ema_real": lecam_ema_real.item(),
@@ -485,7 +488,8 @@ def main():
}
if coordinator.is_master():
save_json(running_states, os.path.join(save_dir, "running_states.json"))
- save_json(lecam_state, os.path.join(save_dir, "lecam_states.json"))
+ if cfg.lecam_loss_weight is not None:
+ save_json(lecam_state, os.path.join(save_dir, "lecam_states.json"))
dist.barrier()
logger.info(
f"Saved checkpoint at epoch {epoch} step {step + 1} global_step {global_step + 1} to {exp_dir}"