CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-11-04 09:32:00 +00:00
Code Issues Packages Projects Releases Wiki Activity

llama : add support for EmbeddingGemma 300m (#15798)

Browse Source 
This commit add support for the EmbeddingGemma 300m. This model supports
sliding window attention (SWA) and a new swq_type is introduced to
support symmetric SWA masking.

This commit also extracts the code from the function
llama_is_masked_swa in llama-impl.h, so that the logic can be shared
by both llm_graph_input_attn_no_cache::set_input and
llama_kv_cache::set_input_kq_mask.

With this commit the EmbeddingGemma 300m model can be converted to
to GGUF and used with llama.cpp.

Once the model has been uploaded to HuggingFace it can be used like
this:
```console
./build/bin/llama-cli -hf ggml-org/embeddinggemma-300m-GGUF:Q8_0
```
This commit is contained in:
Daniel Bevenius
2025-09-04 18:10:29 +02:00
committed by GitHub
parent 856ed0947f
commit fb15d649ed
15 changed files with 328 additions and 47 deletions
Download Patch File Download Diff File Expand all files Collapse all files

27
src/llama-kv-cache.cpp
View File

@@ -27,11 +27,10 @@ llama_kv_cache::llama_kv_cache(
uint32_t n_seq_max,
uint32_t n_pad,
uint32_t n_swa,
llama_swa_type swa_type,
const layer_filter_cb & filter,
const layer_reuse_cb & reuse) :
model(model), hparams(model.hparams), v_trans(v_trans),
n_seq_max(n_seq_max), n_stream(unified ? 1 : n_seq_max), n_pad(n_pad), n_swa(n_swa), swa_type(swa_type) {
n_seq_max(n_seq_max), n_stream(unified ? 1 : n_seq_max), n_pad(n_pad), n_swa(n_swa) {
GGML_ASSERT(kv_size % n_pad == 0);
@@ -1393,29 +1392,7 @@ ggml_cgraph * llama_kv_cache::build_graph_shift(llm_graph_result * res, llama_co
}
bool llama_kv_cache::is_masked_swa(llama_pos p0, llama_pos p1) const {
assert(p0 >= 0 && p1 >= 0);
switch (swa_type) {
case LLAMA_SWA_TYPE_NONE:
{
} break;
case LLAMA_SWA_TYPE_STANDARD:
{
if (p1 - p0 >= (int32_t) n_swa) {
return true;
}
} break;
case LLAMA_SWA_TYPE_CHUNKED:
{
const llama_pos pos_chunk_start = (p1 / n_swa) * n_swa;
if (p0 < pos_chunk_start) {
return true;
}
} break;
}
return false;
return hparams.is_masked_swa(p0, p1);
}
void llama_kv_cache::state_write(llama_io_write_i & io, llama_seq_id seq_id, llama_state_seq_flags flags) const {