0002-pretokenizer.patch 2.16 KB
Newer Older
1
2
3
4
5
6
7
8
9
From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
From: Michael Yang <mxyng@pm.me>
Date: Mon, 16 Sep 2024 15:53:13 -0700
Subject: [PATCH] pretokenizer

---
 src/llama.cpp | 14 +++-----------
 1 file changed, 3 insertions(+), 11 deletions(-)

10
diff --git a/src/llama.cpp b/src/llama.cpp
11
index 4c0a1bb6..800dfb95 100644
12
13
--- a/src/llama.cpp
+++ b/src/llama.cpp
14
@@ -6287,16 +6287,7 @@ static void llm_load_vocab(
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
         if (vocab.type == LLAMA_VOCAB_TYPE_BPE) {
             vocab.tokenizer_add_space_prefix = false;
             vocab.tokenizer_clean_spaces = true;
-            if (tokenizer_pre.empty()) {
-                LLAMA_LOG_WARN("%s: missing pre-tokenizer type, using: 'default'\n", __func__);
-                LLAMA_LOG_WARN("%s:                                             \n", __func__);
-                LLAMA_LOG_WARN("%s: ************************************        \n", __func__);
-                LLAMA_LOG_WARN("%s: GENERATION QUALITY WILL BE DEGRADED!        \n", __func__);
-                LLAMA_LOG_WARN("%s: CONSIDER REGENERATING THE MODEL             \n", __func__);
-                LLAMA_LOG_WARN("%s: ************************************        \n", __func__);
-                LLAMA_LOG_WARN("%s:                                             \n", __func__);
-                vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_DEFAULT;
-            } else if (tokenizer_pre == "default") {
+            if (tokenizer_pre == "default") {
                 vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_DEFAULT;
             } else if (
                     tokenizer_pre == "llama3"   ||
32
33
34
@@ -6398,7 +6389,8 @@ static void llm_load_vocab(
                 vocab.tokenizer_add_bos = true;
                 vocab.tokenizer_clean_spaces = false;
35
36
37
38
39
40
41
             } else {
-                throw std::runtime_error(format("unknown pre-tokenizer type: '%s'", tokenizer_pre.c_str()));
+                LLAMA_LOG_WARN("%s: missing or unrecognized pre-tokenizer type, using: 'default'\n", __func__);
+                vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_DEFAULT;
             }
         } else if (vocab.type == LLAMA_VOCAB_TYPE_SPM) {
             vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_DEFAULT;