Skip to content

Open-Sora: Democratizing Efficient Video Production for All

Python

debug

Shen-Chenhuidba5de3d90f7
diff --git a/scripts/inference-debug.py b/scripts/inference-debug.py
index e565e14..bd69b4d 100644
--- a/scripts/inference-debug.py
+++ b/scripts/inference-debug.py
@@ -59,8 +59,6 @@ def main():
-
-
# 2.3 DEBUG: USE BOOSTER
# 2.3. initialize ColossalAI booster
if cfg.plugin == "zero2":
@@ -158,8 +156,11 @@ def main():
### TODO: DEBUG, USE booster
torch.set_default_dtype(dtype)
- vae, optimizer, _, dataloader, lr_scheduler = booster.boost(
- model=vae, optimizer=optimizer, lr_scheduler=lr_scheduler, dataloader=dataloader
+ # vae, optimizer, _, dataloader, lr_scheduler = booster.boost(
+ # model=vae, optimizer=optimizer, lr_scheduler=lr_scheduler, dataloader=dataloader
+ # )
+ vae, _, _, dataloader, _ = booster.boost(
+ model=vae, dataloader=dataloader
)
torch.set_default_dtype(torch.float)
@@ -167,9 +168,9 @@ def main():
# load model using booster
print("loading:", cfg.model["from_pretrained"])
booster.load_model(vae, os.path.join(cfg.model["from_pretrained"], "model"))
- booster.load_optimizer(optimizer, os.path.join(cfg.model["from_pretrained"], "optimizer"))
- if lr_scheduler is not None:
- booster.load_lr_scheduler(lr_scheduler, os.path.join(cfg.model["from_pretrained"], "lr_scheduler"))
+ # booster.load_optimizer(optimizer, os.path.join(cfg.model["from_pretrained"], "optimizer"))
+ # if lr_scheduler is not None:
+ # booster.load_lr_scheduler(lr_scheduler, os.path.join(cfg.model["from_pretrained"], "lr_scheduler"))
# 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"]