diff options
| author | Kawrakow <iwankawrakow@gmail.com> | 2025-03-21 13:23:01 +0100 |
|---|---|---|
| committer | GitHub <noreply@github.com> | 2025-03-21 13:23:01 +0100 |
| commit | 3d6e25c82db5510df483185b8a20f0ce01136dd7 (patch) | |
| tree | 76d4512a875d82c7bd397355edbdbb3d81239fbf | |
| parent | 022660f7aba973c149e011eac5c4b3dfea02618d (diff) | |
Fix bug: missing parentheses in logical expression (#275)
This results in GGGGGGGGGGGGG when generating with
mla = 3, fa = 0.
Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
| -rw-r--r-- | src/llama.cpp | 2 |
1 files changed, 1 insertions, 1 deletions
diff --git a/src/llama.cpp b/src/llama.cpp index dfe445b8..186cb5a5 100644 --- a/src/llama.cpp +++ b/src/llama.cpp @@ -13869,7 +13869,7 @@ struct llm_build_context { ggml_tensor * q = ggml_concat(ctx0, q_nope2, ggml_permute(ctx0, q_rope, 0, 2, 1, 3), 0); cb(q, "q", il); - if (lctx.cparams.flash_attn && lctx.cparams.mla_attn == 1 || lctx.cparams.mla_attn == 3) { + if (lctx.cparams.flash_attn && (lctx.cparams.mla_attn == 1 || lctx.cparams.mla_attn == 3)) { ggml_tensor * kv_cache_lora = ggml_view_2d(ctx0, kv_self.kv_l[il], kv_lora_rank, n_kv, ggml_row_size(kv_self.kv_l[il]->type, kv_lora_rank + n_embd_head_qk_rope), 0); |