public_git/whisper.cpp
1
0
Fork 0
mirror of https://github.com/ggml-org/whisper.cpp.git synced 2026-10-17 20:03:45 +02:00
Code Issues Packages Projects Releases Wiki Activity

talk-llama : sync llama.cpp

Browse Source
This commit is contained in:
Georgi Gerganov
2026-03-16 13:10:15 +02:00
parent ab1252c19e
commit 2bc630f197
131 changed files with 2374 additions and 1519 deletions
Download Patch File Download Diff File Expand all files Collapse all files
examples/talk-llama/models/plamo2.cpp
+6 -4
View File
@@ -27,7 +27,7 @@ llm_build_plamo2::llm_build_plamo2(const llama_model & model, const llm_graph_pa
cur = build_norm(inpL, model.layers[il].attn_norm, NULL, LLM_NORM_RMS, il);
// check if this layer is Mamba or Attention
bool is_mamba_layer = hparams.is_recurrent(il);
const bool is_mamba_layer = hparams.is_recurrent(il);
if (is_mamba_layer) {
// PLaMo-2 Mamba layer
@@ -106,9 +106,9 @@ ggml_tensor * llm_build_plamo2::build_plamo2_attn_layer(llm_graph_input_attn_kv
cb(qkv, "wqkv", il);
// split QKV tensor into Q, K, V
const int64_t n_embd_head_q = hparams.n_embd_head_k;
const int64_t n_embd_head_k = hparams.n_embd_head_k;
const int64_t n_embd_head_v = hparams.n_embd_head_v;
const int64_t n_embd_head_q = hparams.n_embd_head_k();
const int64_t n_embd_head_k = hparams.n_embd_head_k();
const int64_t n_embd_head_v = hparams.n_embd_head_v();
int32_t n_head = hparams.n_head(il);
int32_t n_head_kv = hparams.n_head_kv(il);
@@ -171,6 +171,8 @@ ggml_tensor * llm_build_plamo2::build_plamo2_mamba_layer(llm_graph_input_rs * in
GGML_ASSERT(n_seqs != 0);
GGML_ASSERT(ubatch.equal_seqs());
GGML_ASSERT(ubatch.n_tokens == n_seq_tokens * n_seqs);
GGML_ASSERT(d_inner % n_head == 0);
GGML_ASSERT(n_group == 0);
ggml_tensor * conv_states_all = mctx_cur->get_r_l(il);
ggml_tensor * ssm_states_all = mctx_cur->get_s_l(il);