jeans/ollama
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

llama: fix crash on snowflake embedding model (#10690)

Browse Source
This commit is contained in:
Jeffrey Morgan 2025-05-13 13:11:11 -07:00 committed by GitHub
parent c7f4ae7b9c
commit 4b903f088a
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
2 changed files with 8 additions and 8 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
llama/llama.cpp/src/llama-vocab.cpp vendored
View File

@ -1469,8 +1469,6 @@ void llama_vocab::impl::load(llama_model_loader & ml, const LLM_KV & kv) {
const int precompiled_charsmap_keyidx = gguf_find_key(ctx, kv(LLM_KV_TOKENIZER_PRECOMPILED_CHARSMAP).c_str()); const int precompiled_charsmap_keyidx = gguf_find_key(ctx, kv(LLM_KV_TOKENIZER_PRECOMPILED_CHARSMAP).c_str());
if (precompiled_charsmap_keyidx != -1) { if (precompiled_charsmap_keyidx != -1) {
const gguf_type pc_type = gguf_get_arr_type(ctx, precompiled_charsmap_keyidx); const gguf_type pc_type = gguf_get_arr_type(ctx, precompiled_charsmap_keyidx);
GGML_ASSERT(pc_type == GGUF_TYPE_INT8 || pc_type == GGUF_TYPE_UINT8);
const size_t n_precompiled_charsmap = gguf_get_arr_data_n(ctx, precompiled_charsmap_keyidx); const size_t n_precompiled_charsmap = gguf_get_arr_data_n(ctx, precompiled_charsmap_keyidx);
const char * pc = (const char *) gguf_get_arr_data(ctx, precompiled_charsmap_keyidx); const char * pc = (const char *) gguf_get_arr_data(ctx, precompiled_charsmap_keyidx);
precompiled_charsmap.assign(pc, pc + n_precompiled_charsmap); precompiled_charsmap.assign(pc, pc + n_precompiled_charsmap);

14
llama/patches/0014-fix-string-arr-kv-loading.patch
View File

@ -9,8 +9,8 @@ such as vocab fields
--- ---
ggml/include/gguf.h | 1 + ggml/include/gguf.h | 1 +
ggml/src/gguf.cpp | 7 +++++-- ggml/src/gguf.cpp | 7 +++++--
src/llama-vocab.cpp | 2 +- src/llama-vocab.cpp | 4 +---
3 files changed, 7 insertions(+), 3 deletions(-) 3 files changed, 7 insertions(+), 5 deletions(-)
diff --git a/ggml/include/gguf.h b/ggml/include/gguf.h diff --git a/ggml/include/gguf.h b/ggml/include/gguf.h
index 79ee2020..3efb22f0 100644 index 79ee2020..3efb22f0 100644
@ -53,13 +53,15 @@ index 381a9c7d..e45b453d 100644
} }
diff --git a/src/llama-vocab.cpp b/src/llama-vocab.cpp diff --git a/src/llama-vocab.cpp b/src/llama-vocab.cpp
index 10f34d33..b098bb25 100644 index 10f34d33..9f5fd57b 100644
--- a/src/llama-vocab.cpp --- a/src/llama-vocab.cpp
+++ b/src/llama-vocab.cpp +++ b/src/llama-vocab.cpp
@@ -1471,7 +1471,7 @@ void llama_vocab::impl::load(llama_model_loader & ml, const LLM_KV & kv) { @@ -1469,9 +1469,7 @@ void llama_vocab::impl::load(llama_model_loader & ml, const LLM_KV & kv) {
const int precompiled_charsmap_keyidx = gguf_find_key(ctx, kv(LLM_KV_TOKENIZER_PRECOMPILED_CHARSMAP).c_str());
if (precompiled_charsmap_keyidx != -1) {
const gguf_type pc_type = gguf_get_arr_type(ctx, precompiled_charsmap_keyidx); const gguf_type pc_type = gguf_get_arr_type(ctx, precompiled_charsmap_keyidx);
GGML_ASSERT(pc_type == GGUF_TYPE_INT8 || pc_type == GGUF_TYPE_UINT8); - GGML_ASSERT(pc_type == GGUF_TYPE_INT8 || pc_type == GGUF_TYPE_UINT8);
-
- const size_t n_precompiled_charsmap = gguf_get_arr_n(ctx, precompiled_charsmap_keyidx); - const size_t n_precompiled_charsmap = gguf_get_arr_n(ctx, precompiled_charsmap_keyidx);
+ const size_t n_precompiled_charsmap = gguf_get_arr_data_n(ctx, precompiled_charsmap_keyidx); + const size_t n_precompiled_charsmap = gguf_get_arr_data_n(ctx, precompiled_charsmap_keyidx);
const char * pc = (const char *) gguf_get_arr_data(ctx, precompiled_charsmap_keyidx); const char * pc = (const char *) gguf_get_arr_data(ctx, precompiled_charsmap_keyidx);