public_git/whisper.cpp
1
0
Fork 0
mirror of https://github.com/ggml-org/whisper.cpp.git synced 2026-10-18 04:35:33 +02:00
Code Issues Packages Projects Releases Wiki Activity

talk-llama : sync llama.cpp

Browse Source
This commit is contained in:
Georgi Gerganov
2026-06-26 16:03:57 +03:00
parent 0279b5386a
commit 0ae02cdb2c
17 changed files with 101 additions and 59 deletions
Download Patch File Download Diff File Expand all files Collapse all files
examples/talk-llama/models/lfm2.cpp
+15 -4
View File
@@ -13,6 +13,7 @@ void llama_model_lfm2::load_arch_hparams(llama_model_loader & ml) {
hparams.n_layer_dense_lead = hparams.n_layer();
switch (hparams.n_ff()) {
case 2560: type = LLM_TYPE_230M; break;
case 4608: type = LLM_TYPE_350M; break;
case 6912: type = LLM_TYPE_700M; break;
case 8192: type = LLM_TYPE_1_2B; break;
@@ -190,7 +191,15 @@ llama_model_lfm2::graph<iswa>::graph(const llama_model & model, const llm_graph_
auto * conv_rs = build_rs(inp_recr, conv_state, hparams.n_embd_r(), n_seqs);
auto * conv = ggml_reshape_3d(ctx0, conv_rs, d_conv, hparams.n_embd, n_seqs);
bx = ggml_concat(ctx0, conv, bx, 0);
// causal prepends the state, non-causal pads symmetrically for a centered window
if (hparams.causal_attn) {
bx = ggml_concat(ctx0, conv, bx, 0);
} else {
const int64_t pad = (hparams.n_shortconv_l_cache - 1) / 2;
auto * left = ggml_cont(ctx0,
ggml_view_3d(ctx0, conv, pad, hparams.n_embd, n_seqs, conv->nb[1], conv->nb[2], (d_conv - pad) * conv->nb[0]));
bx = ggml_pad_ext(ctx0, ggml_concat(ctx0, left, bx, 0), 0, pad, 0, 0, 0, 0, 0, 0);
}
GGML_ASSERT(bx->ne[0] > conv->ne[0]);
// last d_conv columns is a new conv state
@@ -266,10 +275,12 @@ llama_model_lfm2::graph<iswa>::graph(const llama_model & model, const llm_graph_
cb(cur, "result_norm", -1);
res->t_embd = cur;
cur = build_lora_mm(model.output, cur, model.output_s);
cb(cur, "result_output", -1);
if (!cparams.embeddings) {
cur = build_lora_mm(model.output, cur, model.output_s);
cb(cur, "result_output", -1);
res->t_logits = cur;
res->t_logits = cur;
}
ggml_build_forward_expand(gf, cur);
}