facebookresearch / facebookresearch/sam2
Questions about memory usage
- Dominant language
- Jupyter Notebook
- Stars
- 19.9k
- Forks
- 2.5k
- PR merge metrics
- No merged PRs in 30d
Description
Hi there, thanks for your great work!
Recently, I found that in .yaml, the `directly_add_no_mem_embed` and 'use_mask_input_as_output_without_sam' were both set to True which is not consist with the paper, why is this? and is the memory work if both of them set to True?
Also, I found that there is an error if I set the `directly_add_no_mem_embed=False`. The error came from these two:
```
to_cat_memory = [self.no_mem_embed.expand(1, B, self.mem_dim)]
to_cat_memory_pos_embed = [self.no_mem_pos_enc.expand(1, B, self.mem_dim)]
```
Saying the inconsistency dimention of `no_mem_embed` and `1, B, self.mem_dim`. More specifically, 256 vs 64. How could I solve this?
Best,
Yixuan
Contributor guide
Assessment
This issue has not been assessed yet.