ml/backend/ggml/ggml.go · ad95d5b30be7b1e8d845697722a05f9e5d52b8c7 · OpenDAS / ollama · GitLab

"megatron/inference/batching.py" did not exist on "554d1cc04242bc2273630c672af5ac5f6e2883fe"

Find file Blame History Permalink

use split activations when possible (#12293) · ad95d5b3
Michael Yang authored Sep 16, 2025
```
* use ggml_*_split activations when possible

* forward qkv
```
ad95d5b3

ggml.go 40.9 KB