vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

examples : do not assume BOS when shifting context

Browse source
This commit is contained in:
Jared Van Bortel 2024-02-20 22:48:15 -05:00
parent 6560bed3f0
commit f921fc3ecd
2 changed files with 15 additions and 11 deletions
Download patch file Download diff file Expand all files Collapse all files

12
examples/main/main.cpp
View file

@ -334,6 +334,8 @@ int main(int argc, char ** argv) {
// number of tokens to keep when resetting context
if (params.n_keep < 0 || params.n_keep > (int) embd_inp.size() || params.instruct || params.chatml) {
params.n_keep = (int)embd_inp.size();
} else {
params.n_keep += add_bos; // always keep the BOS token
}
// prefix & suffix for instruct mode
@ -383,8 +385,8 @@ int main(int argc, char ** argv) {
}
}
if (params.n_keep > 0) {
LOG_TEE("%s: static prompt based on n_keep: '", __func__);
if (params.n_keep > add_bos) {
LOG_TEE("%s: static prompt based on n_keep: '", __func__);
for (int i = 0; i < params.n_keep; i++) {
LOG_TEE("%s", llama_token_to_piece(ctx, embd_inp[i]).c_str());
}
@ -540,14 +542,14 @@ int main(int argc, char ** argv) {
break;
}
const int n_left = n_past - params.n_keep - 1;
const int n_left = n_past - params.n_keep;
const int n_discard = n_left/2;
LOG("context full, swapping: n_past = %d, n_left = %d, n_ctx = %d, n_keep = %d, n_discard = %d\n",
n_past, n_left, n_ctx, params.n_keep, n_discard);
llama_kv_cache_seq_rm (ctx, 0, params.n_keep + 1 , params.n_keep + n_discard + 1);
llama_kv_cache_seq_shift(ctx, 0, params.n_keep + 1 + n_discard, n_past, -n_discard);
llama_kv_cache_seq_rm (ctx, 0, params.n_keep , params.n_keep + n_discard);
llama_kv_cache_seq_shift(ctx, 0, params.n_keep + n_discard, n_past, -n_discard);
n_past -= n_discard;

14
examples/server/server.cpp
View file

@ -1440,6 +1440,8 @@ struct llama_server_context
task.target_id = -1;
queue_tasks.post(task);
int32_t n_keep = slot.params.n_keep + add_bos_token;
for (llama_client_slot &slot : slots)
{
if (slot.ga_n == 1)
@ -1447,14 +1449,14 @@ struct llama_server_context
if (slot.is_processing() && system_tokens.size() + slot.cache_tokens.size() >= (size_t) slot.n_ctx)
{
// Shift context
const int n_left = system_tokens.size() + slot.n_past - slot.params.n_keep - 1;
const int n_left = system_tokens.size() + slot.n_past - n_keep;
const int n_discard = n_left / 2;
LOG_TEE("slot %d: context shift - n_keep = %d, n_left = %d, n_discard = %d\n", slot.id, slot.params.n_keep, n_left, n_discard);
llama_kv_cache_seq_rm (ctx, slot.id, slot.params.n_keep + 1 , slot.params.n_keep + n_discard + 1);
llama_kv_cache_seq_shift(ctx, slot.id, slot.params.n_keep + 1 + n_discard, system_tokens.size() + slot.n_past, -n_discard);
LOG_TEE("slot %d: context shift - n_keep = %d, n_left = %d, n_discard = %d\n", slot.id, n_keep, n_left, n_discard);
llama_kv_cache_seq_rm (ctx, slot.id, n_keep , n_keep + n_discard);
llama_kv_cache_seq_shift(ctx, slot.id, n_keep + n_discard, system_tokens.size() + slot.n_past, -n_discard);
for (size_t i = slot.params.n_keep + 1 + n_discard; i < slot.cache_tokens.size(); i++)
for (size_t i = n_keep + n_discard; i < slot.cache_tokens.size(); i++)
{
slot.cache_tokens[i - n_discard] = slot.cache_tokens[i];
}
@ -1467,7 +1469,7 @@ struct llama_server_context
LOG_VERBOSE("context shift", {
{ "n_ctx", n_ctx },
{ "n_keep", params.n_keep },
{ "n_keep", n_keep },
{ "n_left", n_left },
});
}