-
### 🐛 Describe the bug
我使用examples/language/llama2中的代码预训练llama2-70b。使用gemini.sh直接跑benchmark.py是成功的,但是我想基于训好的模型进行增量预训练,训练参数和gemini.sh中给出的参数一致,只是修改了如下代码读取已有的模型:
with init_ctx:
# model = L…
-
Hello,
I'm encountering issues related to the dependencies between JAX and Orbax in my current setup. My environment uses CUDA 11.8, and after running `pip install -e .`, I configured the GPU versi…
-
Hi there,
I was curious if there are any ways to bench mark landscape evolution models in landlab. I've been running some large, high-resolution models that i've had to start all the way over a co…
-
**Describe the bug**
- I used a customized datasets to train and test yolox+bytetrack model.
- I used the training script to train bytetrack and got `epoch_80.pth`.
- Then I used it as checkpoint a…
-
Some errors appeared as follows:
Setting up [LPIPS] perceptual loss: trunk [alex], v[0.1], spatial [off]
/home/yunfei/anaconda3/envs/BBDM/lib/python3.9/site-packages/torchvision/models/_utils.py:2…
-
### Question
I have successfully done the pretrain stage, while for fintuning, i encounter following issues.
```
(llava2) wangyh@A16:/data/wangyh/mllms/LLaVA$ bash finetune2.sh
[2023-08-12 15:3…
-
Hello !
Regarding MCTformer+, in the previous code training, "la_crf_dir" and "ha_crf_dir" were not generated. Where do they obtain them from? Can you provide relevant code?
-
Hello,
I have run inference on the clean and final passes of MPI-Sintel providing a custom save path (with the `--save` flag) and noticed that all outputs are written to the same directory instead …
-
I0922 11:46:38.663871 140308634334976 run_squad_v2.py:505] ***** Final Eval results *****
INFO:tensorflow: exact = 50.09685841825992
I0922 11:46:38.663987 140308634334976 run_squad_v2.py:507] exa…
-
This might be more of a general question, but is it possible to use [FlashAttention](https://github.com/Dao-AILab/flash-attention/tree/v1.0.9) with QLoRA in order to further decrease memory requiremen…