-
- [ ] [S-LoRA: Serving Thousands of Models From One GPU for Fun and Profit - OpenPipe](https://openpipe.ai/blog/s-lora)
# S-LoRA: Serving Thousands of Models From One GPU for Fun and Profit - OpenPi…
-
# Effort
1 day
# Skills
curiosity, attention to detail, spreadsheets
# Description
There are several classes of errors in DBpedia. Data may be incorrect or missing. Errors may be caused by differ…
-
### 🐛 Describe the bug
Hello there 👋
From my understanding, the code from the [docs](https://pytorch.org/docs/stable/generated/torch.nn.functional.scaled_dot_product_attention.html) should produc…
-
The license scan has been failing for a while. Fossa shows an issue with wicg-inert's W3C license being incompatible with Apache 2.0 .
Since my GitHub account is no longer associated with this orga…
ca-d updated
8 months ago
-
The changes include the QK normlizaiton, Parallel layers and etc. It would be cool to see how CLIP performs by applying those changes to VIT-L VIT-B VIT-H
-
Durable input (#652) and deduplicated output (#655) apply to Kafka partitions. To use this for Feldera, we need to map Kafka partitions to workers for input and for output.
The pipeline manager curre…
blp updated
11 months ago
-
Because the model is too big for my machine, I get
```sh
RuntimeError: CUDA out of memory. Tried to allocate 100.00 MiB (GPU 0; 7.80 GiB total capacity; 7.19 GiB already allocated; 76.00 MiB free;…
-
Thank you for sharing the code for fine-tuning Roberta.
I've been trying to train Roberta-base on squad v2 and I am getting the following error with this command:
```
python ./fairseq_train.py …
-
```
From Bradley Eltoch:
2007.09.12 21:35
And I think it's great
I still am having problems though with it crashing when loading a large
number of market reports
```
Original issue reported on cod…
-
Grade: Meets Expectations
Great job on this project! You've included all the necessary elements and you're getting a great handle on HTML/CSS. Your code is well-formatted and easy to read as a resu…