thek0tyara/llama-cpp-turboquant
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions 6

batch : remove logits_all flag (#14141)

Browse source 
ggml-ci
This commit is contained in:
Georgi Gerganov 2025-06-12 11:49:26 +03:00 • committed by GitHub
parent e2c0b6e46a
commit c3ee46fab4
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
10 changed files with 17 additions and 30 deletions
Download patch file Download diff file Expand all files Collapse all files

6
src/llama-context.cpp
View file

@ -764,7 +764,7 @@ int llama_context::encode(llama_batch & inp_batch) {
const int64_t n_embd = hparams.n_embd;
llama_sbatch sbatch = llama_sbatch(batch, n_embd, /* simple_split */ true, /* logits_all */ true);
llama_sbatch sbatch = llama_sbatch(batch, n_embd, /* simple_split */ true);
const llama_ubatch ubatch = sbatch.split_simple(n_tokens);
@ -976,7 +976,7 @@ int llama_context::decode(llama_batch & inp_batch) {
llama_memory_state_ptr mstate;
while (true) {
mstate = memory->init_batch(batch, cparams.n_ubatch, embd_pooled, /* logits_all */ n_outputs_all == n_tokens_all);
mstate = memory->init_batch(batch, cparams.n_ubatch, embd_pooled);
if (!mstate) {
return -2;
}
@ -2080,7 +2080,7 @@ void llama_context::opt_epoch_iter(
int64_t n_outputs_all = n_tokens_all;
auto mstate = memory->init_batch(batch, cparams.n_ubatch, embd_pooled, /* logits_all */ true);
auto mstate = memory->init_batch(batch, cparams.n_ubatch, embd_pooled);
if (!mstate || mstate->get_status() != LLAMA_MEMORY_STATUS_SUCCESS) {
LLAMA_LOG_ERROR("%s: could not initialize batch\n", __func__);
break;