Skip to content

Open-Sora: Democratizing Efficient Video Production for All

Python

[feat] add score for inference

zhengzangw6094f5fe92d5
diff --git a/configs/opensora-v1-2/inference/sample.py b/configs/opensora-v1-2/inference/sample.py
index abc43b5..68956f0 100644
--- a/configs/opensora-v1-2/inference/sample.py
+++ b/configs/opensora-v1-2/inference/sample.py
@@ -48,3 +48,6 @@ scheduler = dict(
num_sampling_steps=30,
cfg_scale=7.0,
)
+
+aes = None
+flow = None
diff --git a/opensora/utils/ckpt_utils.py b/opensora/utils/ckpt_utils.py
index 4d4fde3..8ecd350 100644
--- a/opensora/utils/ckpt_utils.py
+++ b/opensora/utils/ckpt_utils.py
@@ -41,9 +41,10 @@ pretrained_models = {
def reparameter(ckpt, name=None, model=None):
+ model_name = name
name = os.path.basename(name)
if not dist.is_initialized() or dist.get_rank() == 0:
- get_logger().info("loading pretrained model: %s", name)
+ get_logger().info("loading pretrained model: %s", model_name)
if name in ["DiT-XL-2-512x512.pt", "DiT-XL-2-256x256.pt"]:
ckpt["x_embedder.proj.weight"] = ckpt["x_embedder.proj.weight"].unsqueeze(2)
del ckpt["pos_embed"]
diff --git a/opensora/utils/config_utils.py b/opensora/utils/config_utils.py
index d0fa845..503e06b 100644
--- a/opensora/utils/config_utils.py
+++ b/opensora/utils/config_utils.py
@@ -62,6 +62,8 @@ def parse_args(training=False):
parser.add_argument("--condition-frame-length", default=None, type=int, help="condition frame length")
parser.add_argument("--reference-path", default=None, type=str, nargs="+", help="reference path")
parser.add_argument("--mask-strategy", default=None, type=str, nargs="+", help="mask strategy")
+ parser.add_argument("--aes", default=None, type=float, help="aesthetic score")
+ parser.add_argument("--flow", default=None, type=float, help="flow score")
# ======================================================
# Training
# ======================================================
diff --git a/scripts/inference.py b/scripts/inference.py
index 68f3862..982994c 100644
--- a/scripts/inference.py
+++ b/scripts/inference.py
@@ -110,6 +110,14 @@ def main():
if prompts is None:
assert cfg.get("prompt_path", None) is not None, "Prompt or prompt_path must be provided"
prompts = load_prompts(cfg.prompt_path, start_idx, cfg.get("end_index", None))
+ score_prompts = []
+ if cfg.get("aes", None) is not None:
+ score_prompts.append(f"{prompt} aesthetic score: {cfg.aes:.1f}" for prompt in prompts)
+ if cfg.get("flow", None) is not None:
+ score_prompts.append(f"{prompt} motion score: {cfg.flow:.1f}" for prompt in prompts)
+ if len(score_prompts) > 0:
+ score_text = ", ".join(score_prompts)
+ prompts = [f"{prompt} [{score_text}]" for prompt in prompts]
# == prepare reference ==
reference_path = cfg.get("reference_path", [""] * len(prompts))