-
### 🐛 Describe the bug
Dtensor shard uses more gpu memory than raw tensor.
With test, Shard gpu mem: 21890MiB > Replicate gpu mem: 17448MiB > Raw tensor gpu mem: 16804MiB.
Confused for a long time…
-
Hi
Posterior sampling with Messenger Pyro guides does not remove observed variables leading to huge memory use.
https://github.com/scverse/scvi-tools/blob/main/scvi/model/base/_pyromixin.py#L184
…
vitkl updated
2 months ago
-
## ❓ Questions and Help
I recived error when try create sqmd mesh on kaggle notebook when flow [Huggingface optimum-tpu](https://github.com/huggingface/optimum-tpu/blob/695ee84d657d9ed2761fcf481685af…
-
From Kaggle submissions (private scores):
Model 3:
- Half of train.csv: 0.4241803
- train_small.csv: 0.4123864
Model 1:
- Half of train.csv: 0.4172254
- train_small.csv: 0.4168748
Run an …
-
The mDNS on the SmartCar cannot be picked up by the android app by using `djsmartcar.local`. One of the two following solutions has to be implemented in order for the app to find the SmartCar on the l…
-
How long will it cost for training AA-Wide-ResNet on CIFAR100 dataset?
And can you share your training device with me?
Thank you!
-
This task is about an algorithm is designed for mobile devices which is [ShuffleNet](https://arxiv.org/pdf/1707.01083.pdf). This task asks you to implement ShuffleNet with Keras. Also, you need to sav…
-
### System Info
If DeepSpeed Config has optimizer/scheduler/fp16 config,will showing warning and **loss** Not Converges in training:
tried to get lr value before scheduler/optimizer started steppi…
-
## 🐛 Bug
For a few models ( Platypus-30B with FSDP zero3, Gemma7b with DDP and vicuna-33b-v1.3 with FSDP zero3) we get segmentation fault error when trying to use fp8 with thunder_cudnn. When usi…
-
During training i have a sporadic message : input tensors must be on the same device. Received cpu and cuda:0
At the end it does not have any influence on the training, it works correctly, i'm just w…