tbelway/immich-quadlet
1
0
Fork 0
forked from Cutlery/immich
Code Pull Requests Activity

Improve TensorFlow docs (#1138)

Browse Source 
* Improve grep

* Machine learning is optional

* Improved wording

* Clarify how to remove
This commit is contained in:
bt90 2022-12-20 17:27:41 +01:00 committed by GitHub
parent 38bf310eac
commit 354593a70d
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 5 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
docs/docs/installation/requirements.md
View File

@ -23,15 +23,15 @@ You can also use Podman to run the application. However, additional configuratio
:::info Machine Learning on older CPU
TensorFlow doesn't run with older CPU architecture, it requires a CPU with AVX and AVX2 instruction set. If you encounter the error `illegal instruction core dump` check for your CPU flags with the command and make sure you see `AVX` and `AVX2`:
The TensorFlow version used by Immich doesn't run on older CPU architectures. It requires a CPU with AVX and AVX2 instruction sets. If you encounter the error `illegal instruction core dump` check your CPU flags with the command below and make sure you see `avx` and `avx2`:
```bash
more /proc/cpuinfo | grep flags
grep -E 'avx2?' /proc/cpuinfo
```
#### Promox
If you are running virtualization in Proxmox, the VM doesn't have the flag enabled.
If you are running virtualization in Proxmox, the CPU type of the VM is probably configured incorrectly.
You need to change the CPU type from `kvm64` to `host` under VMs hardware tab.
@ -42,6 +42,8 @@ You need to change the CPU type from `kvm64` to `host` under VMs hardware tab.
You can use the machine learning image that is built for Non-AVX CPU. The image is community maintained and can be found in the repository below
https://github.com/bertmelis/immich-machine-learning-no-avx
Otherwise, you can safely remove the `immich-machine-learning` service if you do not intend to use Immich's object detection features. Simply remove or comment out the declaration of the service in your compose file.
:::
## Installation methods