Get up and running with Kimi, GLM, MiniMax, DeepSeek, gpt-oss, Qwen, Gemma and other models.
llama3.1 memory
Michael Yang2003d601594c
diff --git a/llm/ggml.go b/llm/ggml.goindex d7f2eef7c..bde9c3473 100644--- a/llm/ggml.go+++ b/llm/ggml.go@@ -344,11 +344,13 @@ func (llm GGML) GraphSize(context, batch uint64) (partialOffload, fullOffload uiswitch llm.KV().Architecture() {case "llama":- fullOffload = 4 * batch * (1 + 4*embedding + context*(1+heads))+ fullOffload = max(+ 4*batch*(1+4*embedding+context*(1+heads)),+ 4*batch*(embedding+vocab),+ )partialOffload = 4 * batch * embeddingpartialOffload += max(- // 4*batch*(4+6*embedding+context*(2*heads)+llm.KV().GQA()),4*batch*(1+embedding+max(context, embedding))+embedding*embedding*9/16+4*context*(batch*heads+embeddingHeads*headsKV),4*batch*(embedding+vocab)+embedding*vocab*105/128,)