Hugging Face's logo

Spaces:
rerun
/
Vista
Build error

App Files Community
1
Vista
File size: 1,328 Bytes 
d323598
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
 
## Trouble Shooting

1. #### Out of memory during sampling.

   - Possible reason:
     - Too many high-resolution frames for parallel decoding. The default setting will request ca. 66 GB peak VARM.

   - Try this:
     - Reduce the number of jointly decoded frames *en_and_decode_n_samples_a_time* in `inference/vista.yaml`.

2. #### Get stuck at loading FrozenCLIPEmbedder or FrozenOpenCLIPImageEmbedder.

   - Possible reason:
     - A network failure.
   
   - Try this:
     1. Download [openai/clip-vit-large-patch14](https://huggingface.co/openai/clip-vit-large-patch14/tree/main) and [laion/CLIP-ViT-H-14-laion2B-s32B-b79K](https://huggingface.co/laion/CLIP-ViT-H-14-laion2B-s32B-b79K/tree/main) in advance.
     2. Set *version* of FrozenCLIPEmbedder and FrozenOpenCLIPImageEmbedder in `vwm/modules/encoders/modules.py` to the new paths of `pytorch_model.bin`.

3. #### Datasets not yet available during training.

   - Possible reason:

     - The installed [sdata](https://github.com/Stability-AI/datapipelines) is not detected.

   - Try this:

     - Reinstall in the current project directory.

       ````shell
       pip3 install -e git+https://github.com/Stability-AI/datapipelines.git@main#egg=sdata
       ````

---

<= Previous: [[Sampling](https://github.com/OpenDriveLab/Vista/blob/main/docs/SAMPLING.md)]