mirror of
https://github.com/ggerganov/llama.cpp.git
synced 2024-12-27 06:39:25 +01:00
llama : updated llm_build_ffn() calls to new API in build_t5()
This commit is contained in:
parent
1c8d37a267
commit
bad0cafee9
12
llama.cpp
12
llama.cpp
@ -12395,9 +12395,9 @@ struct llm_build_context {
|
|||||||
|
|
||||||
// T5 uses relu, flan-T5 uses gelu-gated
|
// T5 uses relu, flan-T5 uses gelu-gated
|
||||||
cur = llm_build_ffn(ctx0, cur,
|
cur = llm_build_ffn(ctx0, cur,
|
||||||
model.layers[il].enc_ffn_up, NULL,
|
model.layers[il].enc_ffn_up, NULL, NULL,
|
||||||
model.layers[il].enc_ffn_gate, NULL,
|
model.layers[il].enc_ffn_gate, NULL, NULL,
|
||||||
model.layers[il].enc_ffn_down, NULL,
|
model.layers[il].enc_ffn_down, NULL, NULL,
|
||||||
NULL,
|
NULL,
|
||||||
model.layers[il].enc_ffn_gate ? LLM_FFN_GELU : LLM_FFN_RELU,
|
model.layers[il].enc_ffn_gate ? LLM_FFN_GELU : LLM_FFN_RELU,
|
||||||
model.layers[il].enc_ffn_gate ? LLM_FFN_PAR : LLM_FFN_SEQ,
|
model.layers[il].enc_ffn_gate ? LLM_FFN_PAR : LLM_FFN_SEQ,
|
||||||
@ -12573,9 +12573,9 @@ struct llm_build_context {
|
|||||||
|
|
||||||
// T5 uses relu, flan-T5 uses gelu-gated
|
// T5 uses relu, flan-T5 uses gelu-gated
|
||||||
cur = llm_build_ffn(ctx0, cur,
|
cur = llm_build_ffn(ctx0, cur,
|
||||||
model.layers[il].ffn_up, NULL,
|
model.layers[il].ffn_up, NULL, NULL,
|
||||||
model.layers[il].ffn_gate, NULL,
|
model.layers[il].ffn_gate, NULL, NULL,
|
||||||
model.layers[il].ffn_down, NULL,
|
model.layers[il].ffn_down, NULL, NULL,
|
||||||
NULL,
|
NULL,
|
||||||
model.layers[il].enc_ffn_gate ? LLM_FFN_GELU : LLM_FFN_RELU,
|
model.layers[il].enc_ffn_gate ? LLM_FFN_GELU : LLM_FFN_RELU,
|
||||||
model.layers[il].enc_ffn_gate ? LLM_FFN_PAR : LLM_FFN_SEQ,
|
model.layers[il].enc_ffn_gate ? LLM_FFN_PAR : LLM_FFN_SEQ,
|
||||||
|
Loading…
Reference in New Issue
Block a user