0002-pretokenizer.patch 2.21 KB
Newer Older
1
2
3
4
5
6
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

---
7
 src/llama-vocab.cpp | 14 +++-----------
8
9
 1 file changed, 3 insertions(+), 11 deletions(-)

10
11
12
13
14
15
16
17
diff --git a/src/llama-vocab.cpp b/src/llama-vocab.cpp
index ad9ffe66..a4eee9b8 100644
--- a/src/llama-vocab.cpp
+++ b/src/llama-vocab.cpp
@@ -1468,16 +1468,7 @@ void llama_vocab::impl::load(llama_model_loader & ml, const LLM_KV & kv) {
         if (type == LLAMA_VOCAB_TYPE_BPE) {
             add_space_prefix = false;
             clean_spaces = true;
18
19
20
21
22
23
24
25
-            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__);
26
-                pre_type = LLAMA_VOCAB_PRE_TYPE_DEFAULT;
27
28
-            } else if (tokenizer_pre == "default") {
+            if (tokenizer_pre == "default") {
29
                 pre_type = LLAMA_VOCAB_PRE_TYPE_DEFAULT;
30
31
             } else if (
                     tokenizer_pre == "llama3"   ||
32
@@ -1593,7 +1584,8 @@ void llama_vocab::impl::load(llama_model_loader & ml, const LLM_KV & kv) {
33
                 tokenizer_pre == "megrez") {
34
                 pre_type = LLAMA_VOCAB_PRE_TYPE_QWEN2;
35
36
37
             } 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__);
38
+                pre_type = LLAMA_VOCAB_PRE_TYPE_DEFAULT;
39
             }
40
41
         } else if (type == LLAMA_VOCAB_TYPE_SPM) {
             pre_type = LLAMA_VOCAB_PRE_TYPE_DEFAULT;