-
### System Info
It was [suggested](https://github.com/OpenGVLab/InternImage/issues/111#issuecomment-1514210563) that there is an incompatibility between accelerate and deepspeed on newer versions. …
-
## 🚀 Feature Request
Use torch.fx to automatically detect residual blocks and residual connections in a model, then manipulate these components to perform stochastic depth.
## Motivation
Rig…
-
## Description
There are a lot of pieces to putting a pull request together and the information that code management needs isn't always readily available and can take developers a long time to gather…
-
As discussed before, our initial step towards supporting archetypes will be to provide `.json` templates that can be imported with all necessary parameter relationships defined. This issue serves as a…
-
In GitLab by @DillonJ on Jun 19, 2020, 10:09
The purpose of this issue is to discuss our approach to Decomposition in SpineOpt
### Introduction :
Decomposition approaches take advantage of certain p…
-
Hi, its me again 😄
This is rather a bizarre problem that I haven't been able to triangulate, let alone reproduce.
Effectively, I'm serialising both my `model` and optimizer_state in a single Py…
-
**Describe the bug**
I cannot initialize an inference engine from Hugging Face's GPT2 implementation.
**To Reproduce**
```
from transformers import AutoTokenizer, AutoModelForCausalLM
model…
-
I used the linked data you provided to train Train_C2FViT_pairwise.py without modifying the code. At the beginning, loss was negative, but sometimes it would become positive and finally become 0. Do y…
-
At the behest of @manuelma, I've been testing the v0.8 branches of Toolbox, SpineInterface, and SpineOpt using a [workflow I used for a conference paper earlier this year](https://zenodo.org/records/7…
Tasqu updated
9 months ago
-
matplotlib 3.8.3 and pytest fails on scanning units with
```console
+ PYTHONPATH=/home/tkloczko/rpmbuild/BUILDROOT/python-networkx-3.2.1-5.fc36.x86_64/usr/lib64/python3.9/site-packages:/home/tkloczk…