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

server : improve mtmd ctx checkpoints (#20726) 

* server : improve mtmd ctx checkpoints

* server : fix off-by-one in pos_min_thold
This commit is contained in:
Georgi Gerganov 2026-03-20 11:13:12 +02:00 committed by GitHub
parent 1af9dab32b
commit ab9d4c3678
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 42 additions and 37 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

79
tools/server/server-context.cpp
View File

@ -2307,8 +2307,8 @@ private:
llama_pos pos_next = slot.prompt.tokens.pos_next(n_past);
// note: when n_swa == 0, the model does not use SWA, which is equivalent to a window of 1
const auto n_swa = std::max(1, llama_model_n_swa(model));
// note: when n_swa == 0, the model does not use SWA
const auto n_swa = std::max(0, llama_model_n_swa(model));
// the largest pos_min required for a checkpoint to be useful
const auto pos_min_thold = std::max(0, pos_next - n_swa);
@ -2363,7 +2363,7 @@ private:
SLT_WRN(slot, "%s\n", st1.str().c_str());
}
if (pos_min > pos_min_thold) {
if (pos_min >= pos_min_thold) {
SLT_WRN(slot, "n_past = %d, slot.prompt.tokens.size() = %d, seq_id = %d, pos_min = %d, n_swa = %d\n", n_past, (int) slot.prompt.tokens.size(), slot.id, pos_min, n_swa);
// search for a context checkpoint
@ -2459,8 +2459,39 @@ private:
slot.n_prompt_tokens_cache = 0;
}
// If using an alora, there may be uncached tokens that come
// before the invocation sequence. When this happens, the
// tokens before the invocation sequence need to be
// processed without the adapter in a separate batch, then
// the adapter needs to be enabled for the remaining tokens.
if (lora_all_alora(slot.lora) && slot.alora_invocation_start - 1 > slot.prompt.n_tokens()) {
SLT_DBG(slot, "processing pre-alora tokens without the adapter (n_tokens = %d, alora_invocation_start = %d)\n", slot.prompt.n_tokens(), slot.alora_invocation_start);
const auto & enabled_loras = lora_get_enabled_ids(slot.lora);
GGML_ASSERT(enabled_loras.size() == 1);
alora_scale = slot.lora[enabled_loras[0]].scale;
slot.lora[enabled_loras[0]].scale = 0.0f;
alora_disabled_id = enabled_loras[0];
}
bool do_checkpoint = params_base.n_ctx_checkpoints > 0;
// make checkpoints only for completion tasks
do_checkpoint = do_checkpoint && slot.task->type == SERVER_TASK_TYPE_COMPLETION;
// make a checkpoint of the parts of the memory that cannot be rolled back.
// checkpoints are created only if:
// - the model uses SWA and we are not using `swa_full`
// - the model architecture is marked as recurrent or hybrid
//
// TODO: try to make this conditional on the context or the memory module, instead of the model type
do_checkpoint = do_checkpoint && (
llama_model_is_recurrent(model) ||
llama_model_is_hybrid(model) ||
(llama_model_n_swa(model) > 0 && !params_base.swa_full)
);
bool has_mtmd = false;
// check if we should process the image
if (slot.prompt.n_tokens() < slot.task->n_tokens() && input_tokens[slot.prompt.n_tokens()] == LLAMA_TOKEN_NULL) {
// process the image
@ -2481,38 +2512,9 @@ private:
slot.prompt.tokens.push_back(chunk.get()); // copy
}
do_checkpoint = false; // do not checkpoint right after an image chunk
has_mtmd = true;
}
// If using an alora, there may be uncached tokens that come
// before the invocation sequence. When this happens, the
// tokens before the invocation sequence need to be
// processed without the adapter in a separate batch, then
// the adapter needs to be enabled for the remaining tokens.
if (lora_all_alora(slot.lora) && slot.alora_invocation_start - 1 > slot.prompt.n_tokens()) {
SLT_DBG(slot, "processing pre-alora tokens without the adapter (n_tokens = %d, alora_invocation_start = %d)\n", slot.prompt.n_tokens(), slot.alora_invocation_start);
const auto & enabled_loras = lora_get_enabled_ids(slot.lora);
GGML_ASSERT(enabled_loras.size() == 1);
alora_scale = slot.lora[enabled_loras[0]].scale;
slot.lora[enabled_loras[0]].scale = 0.0f;
alora_disabled_id = enabled_loras[0];
}
// make checkpoints only for completion tasks
do_checkpoint = do_checkpoint && slot.task->type == SERVER_TASK_TYPE_COMPLETION;
// make a checkpoint of the parts of the memory that cannot be rolled back.
// checkpoints are created only if:
// - the model uses SWA and we are not using `swa_full`
// - the model architecture is marked as recurrent or hybrid
//
// TODO: try to make this conditional on the context or the memory module, instead of the model type
do_checkpoint = do_checkpoint && (
llama_model_is_recurrent(model) ||
llama_model_is_hybrid(model) ||
(llama_model_n_swa(model) > 0 && !params_base.swa_full)
);
// add prompt tokens for processing in the current batch
while (slot.prompt.n_tokens() < slot.task->n_tokens() && batch.n_tokens < n_batch) {
// get next token to process
@ -2544,13 +2546,13 @@ private:
// - 4 + n_ubatch
// - 4
// ref: https://github.com/ggml-org/llama.cpp/pull/20288
{
if (do_checkpoint) {
static const int checkpoint_offsets[] = {4 + n_ubatch, 4};
bool should_break = false;
for (int offset : checkpoint_offsets) {
const int n_last = std::min(n_batch, offset);
if (do_checkpoint && slot.task->n_tokens() == slot.prompt.n_tokens() + n_last) {
if (slot.task->n_tokens() == slot.prompt.n_tokens() + n_last) {
should_break = true;
break;
}
@ -2607,10 +2609,13 @@ private:
const auto pos_max = llama_memory_seq_pos_max(llama_get_memory(ctx), slot.id);
// no need for empty or small checkpoints
do_checkpoint = do_checkpoint && (pos_min >= 0 && pos_max >= 64);
do_checkpoint = do_checkpoint && (pos_min >= 0 && slot.prompt.n_tokens() >= 64);
// do not checkpoint after mtmd chunks
do_checkpoint = do_checkpoint && !has_mtmd;
// no need to create checkpoints that are too close together
do_checkpoint = do_checkpoint && (slot.prompt.checkpoints.empty() || pos_max > slot.prompt.checkpoints.back().pos_max + 64);
do_checkpoint = do_checkpoint && (slot.prompt.checkpoints.empty() || slot.prompt.n_tokens() - n_tokens_cur > slot.prompt.checkpoints.back().n_tokens + 64);
// note: we create the checkpoint before calling llama_decode(), so the current batch is not
// yet processed and therefore it is not part of the checkpoint.