From: Minsoo Cheong <redacted>
Date: Tue, 26 Mar 2024 09:11:46 +0000 (+0900)
Subject: embedding : adjust `n_ubatch` value (#6296)
X-Git-Tag: upstream/0.0.4488~1951
X-Git-Url: https://git.djapps.eu/?a=commitdiff_plain;h=deb7240100da99555b9ab9dc635021e591fceaf5;p=pkg%2Fggml%2Fsources%2Fllama.cpp

embedding : adjust `n_ubatch` value (#6296)

* embedding: assign `n_ubatch` value, print error on `n_batch` overflow

* Update examples/embedding/embedding.cpp

Co-authored-by: Xuan Son Nguyen <redacted>

* use %ld instead of %lld

* Revert "use %ld instead of %lld"

This reverts commit ea753ede90a86a0699f65878cc8e2020ff5eabb8.

---------

Co-authored-by: Xuan Son Nguyen <redacted>
---

diff --git a/examples/embedding/embedding.cpp b/examples/embedding/embedding.cpp
index cbf9aa2b..9aede7fa 100644
--- a/examples/embedding/embedding.cpp
+++ b/examples/embedding/embedding.cpp
@@ -61,6 +61,8 @@ int main(int argc, char ** argv) {
     }
 
     params.embedding = true;
+    // For non-causal models, batch size must be equal to ubatch size
+    params.n_ubatch = params.n_batch;
 
     print_build_info();
 
@@ -114,7 +116,9 @@ int main(int argc, char ** argv) {
     for (const auto & prompt : prompts) {
         auto inp = ::llama_tokenize(ctx, prompt, true, false);
         if (inp.size() > n_batch) {
-            inp.resize(n_batch);
+            fprintf(stderr, "%s: error: number of tokens in input line (%lld) exceeds batch size (%lld), increase batch size and re-run\n",
+                    __func__, (long long int) inp.size(), (long long int) n_batch);
+            return 1;
         }
         inputs.push_back(inp);
     }