-
### Problem Description
Even on Real World Llama 2 70B Training Shapes, TE Linear FP8 is 1.5 to 2x slower than AMP BF16 Linear. Do you have any suggestions or magic env flags on how to improve perf…
-
When opening the URL (http://0.0.0.0:7860) I get the "can't reach this page" message. I don't get any errors while loading, apart from the "No module named 'triton'" one, which I assume is normal on …
-
### 🐛 Describe the bug
Was doing accuracy aware quantization for the model Marqo/marqo-fashionSigLIP . It was working for more than 2 hours+ and suddenly generated error. It was also not converging v…
-
So with fp32... Sampler fails with... (using gguf i2v model) seems somewhere your doing a operation where element types don't match, (not in an autocast?), sadly its crashing out at the shadergraph so…
-
## error log | 日志或报错信息 | ログ
## model | 模型 | モデル
1. rtdetr-l model 下载路径: https://github.com/ultralytics/assets/releases/download/v8.2.0/rtdetr-l.pt
## how to reproduce | 复现步骤 | 再現方法
1.rtdetr-l.…
-
Hello,
I am very interested in your research and am currently trying to run some experiments based on it. However, I encountered an issue while running the program from the HuggingFace_EncDec directo…
-
All of the following are linear transformers:
- `center()`
- `standardize()`
- `slide()`
- `reverse()`
- `rescale()`
- `normalize()`
So hypothetically we should be able to simplify their co…
-
PEFT finetuning (LoRA, adapter) raises the following warning for each FSDP-wrapped layer (transformer block in our case):
```python
The following parameters have requires_grad=True:
['transformer…
-
I am using LoRA for my custom model. Within the model i have trasnformer encoder block and a series of linear layers. I want to implement LoRA in the custom model specifically targeting the q,k and v …
-
The canny is not even working at all. Please find the workflow attached, Am I doing something wrong?
…
LiJT updated
3 months ago