vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

plamo : fix tensor names for correct GPU offload

Browse source
This commit is contained in:
Georgi Gerganov 2023-12-24 15:33:31 +02:00
parent eedd434575
commit 1949c95598
No known key found for this signature in database
GPG key ID: 449E073F9DC10735
1 changed files with 4 additions and 3 deletions
Download patch file Download diff file Expand all files Collapse all files

7
llama.cpp
View file

@ -5704,13 +5704,14 @@ struct llm_build_context {
model.layers[il].ffn_gate, NULL, model.layers[il].ffn_gate, NULL,
model.layers[il].ffn_down, NULL, model.layers[il].ffn_down, NULL,
LLM_FFN_SILU, LLM_FFN_PAR, cb, il); LLM_FFN_SILU, LLM_FFN_PAR, cb, il);
cb(cur, "mlp_out", il); cb(cur, "ffn_out", il);
} }
cur = ggml_add(ctx0, cur, sa_out); cur = ggml_add(ctx0, cur, sa_out);
cb(cur, "mlp_out + sa_out", il); cb(cur, "l_out", il);
cur = ggml_add(ctx0, cur, inpL); cur = ggml_add(ctx0, cur, inpL);
cb(cur, "mlp_out + sa_out + inpL", il); cb(cur, "l_out", il);
// input for next layer // input for next layer
inpL = cur; inpL = cur;