happyz
/
llama.cpp
mirror of https://github.com/ggerganov/llama.cpp.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Merge d7ec22fe46 into 537eadb1b9

This commit is contained in:
Chris Muzyn 2026-02-06 09:44:40 -06:00 committed by GitHub
parent 537eadb1b9 d7ec22fe46
commit 2dcd2b64ff
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 8 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

12
src/llama-context.cpp
View File

@ -1281,6 +1281,8 @@ int llama_context::encode(const llama_batch & batch_inp) {
case LLAMA_POOLING_TYPE_CLS: case LLAMA_POOLING_TYPE_CLS:
case LLAMA_POOLING_TYPE_LAST: case LLAMA_POOLING_TYPE_LAST:
{ {
const int64_t n_embd_tensor = t_embd ? t_embd->ne[0] : hparams.n_embd_inp();
// extract sequence embeddings // extract sequence embeddings
auto & embd_seq_out = embd_seq; auto & embd_seq_out = embd_seq;
@ -1288,8 +1290,8 @@ int llama_context::encode(const llama_batch & batch_inp) {
const llama_seq_id seq_id = ubatch.seq_id_unq[s]; const llama_seq_id seq_id = ubatch.seq_id_unq[s];
const int32_t seq_idx = ubatch.seq_idx[seq_id]; const int32_t seq_idx = ubatch.seq_idx[seq_id];
embd_seq_out[seq_id].resize(n_embd); embd_seq_out[seq_id].resize(n_embd_tensor);
ggml_backend_tensor_get_async(backend_embd, t_embd, embd_seq_out[seq_id].data(), (n_embd*seq_idx)*sizeof(float), n_embd*sizeof(float)); ggml_backend_tensor_get_async(backend_embd, t_embd, embd_seq_out[seq_id].data(), (n_embd_tensor*seq_idx)*sizeof(float), n_embd_tensor*sizeof(float));
} }
} break; } break;
case LLAMA_POOLING_TYPE_RANK: case LLAMA_POOLING_TYPE_RANK:
@ -1708,6 +1710,8 @@ int llama_context::decode(const llama_batch & batch_inp) {
case LLAMA_POOLING_TYPE_CLS: case LLAMA_POOLING_TYPE_CLS:
case LLAMA_POOLING_TYPE_LAST: case LLAMA_POOLING_TYPE_LAST:
{ {
const int64_t n_embd_tensor = t_embd ? t_embd->ne[0] : hparams.n_embd_inp();
// extract sequence embeddings (cleared before processing each batch) // extract sequence embeddings (cleared before processing each batch)
auto & embd_seq_out = embd_seq; auto & embd_seq_out = embd_seq;
@ -1715,8 +1719,8 @@ int llama_context::decode(const llama_batch & batch_inp) {
const llama_seq_id seq_id = ubatch.seq_id_unq[s]; const llama_seq_id seq_id = ubatch.seq_id_unq[s];
const int32_t seq_idx = ubatch.seq_idx[seq_id]; const int32_t seq_idx = ubatch.seq_idx[seq_id];
embd_seq_out[seq_id].resize(n_embd); embd_seq_out[seq_id].resize(n_embd_tensor);
ggml_backend_tensor_get_async(backend_embd, t_embd, embd_seq_out[seq_id].data(), (n_embd*seq_idx)*sizeof(float), n_embd*sizeof(float)); ggml_backend_tensor_get_async(backend_embd, t_embd, embd_seq_out[seq_id].data(), (n_embd_tensor*seq_idx)*sizeof(float), n_embd_tensor*sizeof(float));
} }
} break; } break;
case LLAMA_POOLING_TYPE_RANK: case LLAMA_POOLING_TYPE_RANK: