-
-
While testing TPUs provisioning, I noticed that both on-demand and spot TPUs can be deleted right after a successful call to create the TPU. The server correctly fails the job with FAILED_TO_START_DUE…
-
### Community Note
* Please vote on this issue by adding a 👍 [reaction](https://blog.github.com/2016-03-10-add-reactions-to-pull-requests-issues-and-comments/) to the original issue to help the commu…
-
### Description
Hello, I'm running into a core dump when writing TPU kernels. I was testing with interpret on, and the kernel was working. Without it, I get a core dump. Any temporary fix is apprecia…
-
### Describe the bug
ConnectionError: Tried to launch distributed communication on port 29401, but another process is utilizing it. Please specify a different port (such as using the --main_process_p…
-
### Description
Hi,
I recently discovered that jax.numpy.interp has poor performance on TPUs. More specifically, I am trying to perform 1D interpolation on a flattened tensor with an original s…
-
Using the free TPU Colab instances:
```
import jax.tools.colab_tpu
jax.tools.colab_tpu.setup_tpu()
import psutil
def available_ram_mb():
return psutil.virtual_memory().available // (1024…
-
I have fine-tuned the qwen2.5-7b-instruction model using Llama Factory, and now I need to deploy the fine-tuned qwen2.5 model on TPU.
How should I proceed with the support?
I have noticed that…
-
### System Info
```Shell
- `Accelerate` version: 0.34.2
- Platform: Linux-5.15.0-1071-aws-x86_64-with-glibc2.31
- `accelerate` bash location: /home/zeus/miniconda3/envs/cloudspace/bin/accelerate
-…
-
Right now, the program crashes whenever you make a tensor on an XLA device and call a `LazyTensorBarrier()` without setting `wait` to `true`. This can be bypassed by using `LazyTensorBarrier(wait: tru…