Open-Sora: Democratizing Efficient Video Production for All
Python
[docs] update readme
zhengzangw3f848747afbd
diff --git a/README.md b/README.mdindex 7575047..9dc749b 100644--- a/README.md+++ b/README.md@@ -264,8 +264,13 @@ python gradio/app.pyThis will launch a Gradio application on your localhost. If you want to know more about the Gradio applicaiton, you can refer to the [README file](./gradio/README.md).+### Open-Sora 1.2 Command Line Inference+### Open-Sora 1.1 Command Line Inference+<details>+<summary>View more</summary>+Since Open-Sora 1.1 supports inference with dynamic input size, you can pass the input size as an argument.```bash@@ -281,6 +286,8 @@ python scripts/inference.py configs/opensora-v1-1/inference/sample.py --prompt "See [here](docs/commands.md#inference-with-open-sora-11) for more instructions including text-to-image, image-to-video, video-to-video, and infinite time generation.+</details>+### Open-Sora 1.0 Command Line Inference<details>@@ -319,8 +326,13 @@ Also check out the [datasets](docs/datasets.md) we use.## Training+### Open-Sora 1.2 Training+### Open-Sora 1.1 Training+<details>+<summary>View more</summary>+Once you prepare the data in a `csv` file, run the following commands to launch training on a single node.```bash@@ -332,6 +344,8 @@ colossalai run --nproc_per_node 8 --hostfile hostfile scripts/train.py \configs/opensora-v1-1/train/stage1.py --data-path YOUR_CSV_PATH --ckpt-path YOUR_PRETRAINED_CKPT```+</details>+### Open-Sora 1.0 Training<details>@@ -360,6 +374,10 @@ For training other models and advanced usage, see [here](docs/commands.md) for m## Evaluation+### VBench++### VBench-i2v+See [here](eval/README.md) for more instructions.## Contributiondiff --git a/docs/report_03.md b/docs/report_03.mdnew file mode 100644index 0000000..49d022e--- /dev/null+++ b/docs/report_03.md@@ -0,0 +1,16 @@+# Open-Sora 1.2 Report++- [3D VAE](#3d-vae)+- [Rectified flow and model adaptation](#rectified-flow-and-model-adaptation)+- [Training data and stages](#training-data-and-stages)+- [Evaluation](#evaluation)++In Open-Sora 1.2 release, we train a 1.1B models on >20M data, supporting 0s~15s, 144p to 720p, various aspect ratios video generation.++## 3D VAE++## Rectified flow and model adaptation++## Training data and stages++## Evaluation