Upload README.md with huggingface_hub
Browse files
README.md
CHANGED
|
@@ -32,8 +32,6 @@ blockmask_bf16/
|
|
| 32 |
└── pytorch_model.bin.index.json
|
| 33 |
```
|
| 34 |
|
| 35 |
-
The **same checkpoint serves both 8- and 16-frame inference** — the block-sparse temporal attention and positional encodings are frame-count-agnostic (up to 32 frames). Only the config changes.
|
| 36 |
-
|
| 37 |
## Usage
|
| 38 |
|
| 39 |
Download the checkpoint:
|
|
@@ -52,8 +50,6 @@ python inference_4d.py \
|
|
| 52 |
--output_dir ./inference_output/door
|
| 53 |
```
|
| 54 |
|
| 55 |
-
For 16-frame generation, use `--config configs/4d_config_16.yaml` with the same `--ckpt_path`.
|
| 56 |
-
|
| 57 |
## Citation
|
| 58 |
|
| 59 |
```bibtex
|
|
|
|
| 32 |
└── pytorch_model.bin.index.json
|
| 33 |
```
|
| 34 |
|
|
|
|
|
|
|
| 35 |
## Usage
|
| 36 |
|
| 37 |
Download the checkpoint:
|
|
|
|
| 50 |
--output_dir ./inference_output/door
|
| 51 |
```
|
| 52 |
|
|
|
|
|
|
|
| 53 |
## Citation
|
| 54 |
|
| 55 |
```bibtex
|