0006-embeddings.patch

From c9a6ca9fc039233dee746a4da9705762cd9e515d Mon Sep 17 00:00:00 2001
From: Michael Yang <mxyng@pm.me>
Date: Mon, 16 Sep 2024 15:53:14 -0700
Subject: [PATCH] 06-embeddings.diff

---
 src/llama.cpp | 17 ++++++++++-------
 1 file changed, 10 insertions(+), 7 deletions(-)

diff --git a/src/llama.cpp b/src/llama.cpp
index d1e959fc..f79bd782 100644
--- a/src/llama.cpp
+++ b/src/llama.cpp
@@ -15898,7 +15898,7 @@ static size_t llama_output_reserve(llama_context & lctx, size_t n_outputs) {
     const auto n_embd  = hparams.n_embd;
 
     // TODO: use a per-batch flag for logits presence instead
-    const bool has_logits = !cparams.embeddings;
+    const bool has_logits =  cparams.causal_attn;
     const bool has_embd   =  cparams.embeddings && (cparams.pooling_type == LLAMA_POOLING_TYPE_NONE);
 
     const size_t logits_size = has_logits ? n_vocab*n_outputs_max : 0;
@@ -16167,20 +16167,23 @@ static int llama_decode_internal(
             // no output
             res  = nullptr;
             embd = nullptr;
-        } else if (cparams.embeddings) {
-            res  = nullptr; // do not extract logits for embedding case
-            embd = nullptr;
+        }
+
+        if (cparams.embeddings) {
             for (int i = gf->n_nodes - 1; i >= 0; --i) {
-                if (strcmp(gf->nodes[i]->name, "result_embd_pooled") == 0) {
-                    embd = gf->nodes[i];
+                embd = gf->nodes[i];
+                if (strcmp(embd->name, "result_embd_pooled") == 0) {
                     break;
                 }
             }
-            GGML_ASSERT(embd != nullptr && "missing embeddings tensor");
         } else {
             embd = nullptr; // do not extract embeddings when not needed
             GGML_ASSERT(strcmp(res->name, "result_output") == 0 && "missing result_output tensor");
         }
+
+        if (!cparams.causal_attn) {
+            res = nullptr; // do not extract logits when not needed
+        }
         // LLAMA_LOG_INFO("graph build time: %.3f ms (%d nodes, %d leafs)\n", (ggml_time_us() - t_start_us)/1000.0, gf->n_nodes, gf->n_leafs);
 
         ggml_backend_sched_alloc_graph(lctx.sched, gf);
-- 
2.46.0