Multi GPU #202
Merged
Multi GPU #202
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
…correctly setting the `tensor_splits` collection
|
Note that this PR breaks Newtonsoft.Json serialization of Last time we discussed this I don't think anyone was particularly interested in supporting JSON serialization at all and I only added it because it was very easy to do. |
|
Some discussion about this PR over in #189. |
|
Tested by swiftress on Discord and confirmed working (https://discord.com/channels/1106946823282761851/1106947264938790972/1166937319664795689) |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Added support for multi GPU in the model config.
This is a litte rough at the moment. TheIModelParams.TensorSplitsarray length must be <=NativeApi.llama_max_devices(). If it's too small it's padded up to size (with zeroes). If it's too large theToLlamaModelParamsmethod throws. I'd rather not require the user to query the native API directly, and I'd rather the error were caught at config time.This is exposed through a
TensorSplitsCollectionon theIModelParamsinterface. The collection can be indexed into (like an array) and automatically has the correct size. There is no way to change the size, preventing incorrect usage.