You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Copy file name to clipboardExpand all lines: web/platform/src/content/posts/Finetune_LLM_On_CPU.mdx
+1-1Lines changed: 1 addition & 1 deletion
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -68,7 +68,7 @@ Logging and print statements that track progress (like "Starting to fine tune mo
68
68
69
69
Most cloud infrastructure (including NativeLink) runs on x86_64 processors, while Apple Silicon Macs use ARM64. Dependencies such as PyTorch are built for specified architectures and setting up Bazel to handle platform-specific dependencies is complex.
70
70
71
-
If you want to run this code and you only have a Mac, the simplest way would be to run this code via a cloud-based Linux VM (GCP/AWS). If you don't want to use a cloud server, you could create a minimal docker container for x86_64 (`FROM --platform=linux/amd64 ubuntu:22.04` with just `curl` and `Bazelisk` installed) with **Rosetta enabled** and run from your Mac's terminal using this container. However, we highly recommend against this approach; the correct approach here would be to use toolchain transitions from a local mac platform to the remote Linux runner, but that's outside of the scope of this article.
71
+
If you want to run this code and you only have a Mac, the simplest way would be to run this code via a cloud-based Linux VM (GCP/AWS). If you don't want to use a cloud server, you could create a minimal docker container for x86_64 (`FROM --platform=linux/amd64 ubuntu:24.04` with just `curl` and `Bazelisk` installed) with **Rosetta enabled** and run from your Mac's terminal using this container. However, we highly recommend against this approach; the correct approach here would be to use toolchain transitions from a local mac platform to the remote Linux runner, but that's outside of the scope of this article.
0 commit comments