extern/whisper.cpp
1
0
Fork 1
mirror of https://github.com/ggerganov/whisper.cpp.git synced 2025-08-09 06:04:37 +02:00
Code Issues Packages Projects Releases Wiki Activity

talk-llama : sync llama.cpp

Browse Source 
ggml-ci
This commit is contained in:
Georgi Gerganov
2025-06-18 10:22:47 +03:00
parent 69061e356f
commit 2f60ebc3c2
26 changed files with 1454 additions and 504 deletions
Download Patch File Download Diff File Expand all files Collapse all files

28
examples/talk-llama/llama-graph.h
View File

@ -200,18 +200,6 @@ public:
const llama_kv_cache_recurrent_state * kv_state;
};
class llm_graph_input_s_mask : public llm_graph_input_i {
public:
llm_graph_input_s_mask(const llama_kv_cache_recurrent_state * kv_state) : kv_state(kv_state) {}
virtual ~llm_graph_input_s_mask() = default;
void set_input(const llama_ubatch * ubatch) override;
ggml_tensor * s_mask; // F32 [1, n_kv]
const llama_kv_cache_recurrent_state * kv_state;
};
class llm_graph_input_cross_embd : public llm_graph_input_i {
public:
llm_graph_input_cross_embd(
@ -390,7 +378,7 @@ struct llm_graph_params {
const llama_memory_state_i * mstate;
const llama_cross * cross;
int32_t n_outputs;
uint32_t n_outputs;
const llm_graph_cb & cb;
};
@ -424,8 +412,8 @@ struct llm_graph_context {
const float norm_eps;
const float norm_rms_eps;
const int32_t n_tokens;
const int32_t n_outputs;
const int64_t n_tokens;
const int64_t n_outputs;
const int32_t n_ctx_orig; // yarn
const enum llama_pooling_type pooling_type;
@ -521,7 +509,6 @@ struct llm_graph_context {
ggml_tensor * build_inp_mean() const;
ggml_tensor * build_inp_cls() const;
ggml_tensor * build_inp_s_copy() const;
ggml_tensor * build_inp_s_mask() const;
ggml_tensor * build_inp_cross_embd() const;
ggml_tensor * build_inp_pos_bucket_enc() const;
@ -606,18 +593,17 @@ struct llm_graph_context {
// recurrent
//
ggml_tensor * build_copy_mask_state(
ggml_tensor * build_recurrent_state(
ggml_cgraph * gf,
ggml_tensor * s,
ggml_tensor * state_copy,
ggml_tensor * state_mask,
int32_t n_state,
int32_t n_seqs) const;
int32_t state_size,
int32_t n_seqs,
bool avoid_copies = false) const;
ggml_tensor * build_rwkv_token_shift_load(
ggml_cgraph * gf,
ggml_tensor * state_copy,
ggml_tensor * state_mask,
const llama_ubatch & ubatch,
int il) const;