batch: migrate the rest of examples to llama_batch_ext (#29601)

* migrate the rest

* test-thread-safety

* rm common_batch_staged
This commit is contained in:
Xuan-Son Nguyen authored and GitHub committed 2026-09-30 18:08:43 +02:00
1 parent 05af0d2b13
commit 60e9cf7a7b
34 files changed
+513 -631

No files matched your search

+13 -14
View File
@@ -27,27 +27,27 @@ static std::vector<std::string> split_lines(const std::string & s, const std::st
return lines;
}
static void batch_add_seq(llama_batch & batch, const std::vector<int32_t> & tokens, llama_seq_id seq_id) {
static void batch_add_seq(common_batch & batch, const std::vector<int32_t> & tokens, llama_seq_id seq_id) {
size_t n_tokens = tokens.size();
for (size_t i = 0; i < n_tokens; i++) {
common_batch_add(batch, tokens[i], i, { seq_id }, true);
batch.add(tokens[i], i, seq_id, true);
}
}
static void batch_decode(llama_context * ctx, llama_batch & batch, float * output, int n_seq, int n_embd_out, int embd_norm) {
static void batch_decode(llama_context * ctx, common_batch & batch, float * output, int n_seq, int n_embd_out, int embd_norm) {
const enum llama_pooling_type pooling_type = llama_pooling_type(ctx);
// clear previous kv_cache values (irrelevant for embeddings)
llama_memory_clear(llama_get_memory(ctx), true);
// run model
LOG_INF("%s: n_tokens = %d, n_seq = %d\n", __func__, batch.n_tokens, n_seq);
if (llama_decode(ctx, batch) < 0) {
LOG_INF("%s: n_tokens = %d, n_seq = %d\n", __func__, batch.size(), n_seq);
if (llama_process(ctx, LLAMA_PROCESS_TYPE_DECODE, batch.get()) < 0) {
LOG_ERR("%s : failed to process\n", __func__);
}
for (int i = 0; i < batch.n_tokens; i++) {
if (!batch.logits[i]) {
for (int i = 0; i < batch.size(); i++) {
if (!batch.tokens[i].output) {
continue;
}
@@ -61,8 +61,8 @@ static void batch_decode(llama_context * ctx, llama_batch & batch, float * outpu
GGML_ASSERT(embd != NULL && "failed to get token embeddings");
} else {
// try to get sequence embeddings - supported only when pooling_type is not NONE
embd = llama_get_embeddings_seq(ctx, batch.seq_id[i][0]);
embd_pos = batch.seq_id[i][0];
embd = llama_get_embeddings_seq(ctx, batch.tokens[i].seq_id);
embd_pos = batch.tokens[i].seq_id;
GGML_ASSERT(embd != NULL && "failed to get sequence embeddings");
}
@@ -242,7 +242,7 @@ int main(int argc, char ** argv) {
// initialize batch
const int n_prompts = prompts.size();
struct llama_batch batch = llama_batch_init(n_batch, 0, 1);
common_batch batch(ctx);
// count number of embeddings
int n_embd_count = 0;
@@ -269,12 +269,12 @@ int main(int argc, char ** argv) {
const uint64_t n_toks = inp.size();
// encode if at capacity
if (batch.n_tokens + n_toks > n_batch || s >= n_seq_max) {
if (batch.size() + n_toks > n_batch || s >= n_seq_max) {
float * out = emb + e * n_embd_out;
batch_decode(ctx, batch, out, s, n_embd_out, params.embd_normalize);
e += pooling_type == LLAMA_POOLING_TYPE_NONE ? batch.n_tokens : s;
e += pooling_type == LLAMA_POOLING_TYPE_NONE ? batch.size() : s;
s = 0;
common_batch_clear(batch);
batch.clear();
}
// add to batch
@@ -407,7 +407,6 @@ int main(int argc, char ** argv) {
llama_perf_context_print(ctx);
// clean up
llama_batch_free(batch);
llama_backend_free();
return 0;