ikawrakow/ik_llama.cpp
1
0
Fork 0
mirror of https://github.com/ikawrakow/ik_llama.cpp.git synced 2026-02-24 23:24:13 +00:00
Code Issues Packages Projects Releases Wiki Activity

More fix

Browse Source
This commit is contained in:
Saood Karim
2025-03-25 12:16:43 -05:00
parent 8ab6b155d5
commit 9b6b55f441
1 changed files with 2 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
src/llama.cpp
View File

@@ -3273,7 +3273,7 @@ static bool llama_kv_cache_init(
const uint32_t n_embd_head_qk_rope = hparams.n_rot;
const uint32_t n_embd_head_qk_nope = hparams.n_embd_head_k - hparams.n_rot;
const uint32_t kv_lora_rank = hparams.n_lora_kv;
LLAMA_LOG_INFO("%s: layer %d: n_embd_head_qk_rope = %d, kv_lora_rank = %d\n", __func__, i, n_embd_head_qk_rope, kv_lora_rank);
//LLAMA_LOG_INFO("%s: layer %d: n_embd_head_qk_rope = %d, kv_lora_rank = %d\n", __func__, i, n_embd_head_qk_rope, kv_lora_rank);
if (cparams.flash_attn) {
ggml_tensor * kv = ggml_new_tensor_2d(ctx, cache.type_k, kv_lora_rank + n_embd_head_qk_rope, kv_size);
ggml_format_name(kv, "cache_kv_l%d", i);
@@ -3299,7 +3299,7 @@ static bool llama_kv_cache_init(
cache.k_l.push_back(k);
cache.v_l.push_back(v);
//Commented out old method
struct ggml_context * ctx = offload ? ctx_map.at(model.buft_layer[i].buft) : cache.ctxs.front();
//struct ggml_context * ctx = offload ? ctx_map.at(model.buft_layer[i].buft) : cache.ctxs.front();
//ggml_backend_buffer_type_t buft;
//ggml_context * ctx;