CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-11-16 11:27:03 +00:00
Code Issues Packages Projects Releases Wiki Activity

push more fixes

Browse Source
This commit is contained in:
younesbelkada
2025-07-03 15:05:01 +04:00
parent 991de6cbe4
commit f897efdaf6
9 changed files with 504 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
src/llama-graph.cpp
View File

@@ -545,6 +545,10 @@ ggml_tensor * llm_graph_context::build_ffn(
case LLM_FFN_PAR:
{
cur = build_lora_mm(gate, cur);
if (arch == LLM_ARCH_FALCON_H1) {
cur = ggml_scale(ctx0, cur, hparams.mlp_gate_multiplier);
}
cb(cur, "ffn_gate", il);
} break;
}
@@ -631,6 +635,9 @@ ggml_tensor * llm_graph_context::build_ffn(
// GLM4 seems to have numerical issues with half-precision accumulators
ggml_mul_mat_set_prec(cur, GGML_PREC_F32);
}
if (arch == LLM_ARCH_FALCON_H1) {
cur = ggml_scale(ctx0, cur, hparams.mlp_down_multiplier);
}
}
if (down_b) {