chore: bump llama.cpp

Signed-off-by: thxCode <thxcode0824@gmail.com>
gpustack · Jan 14, 2025 · ccf473b · ccf473b
1 parent 7a2ddd6
commit ccf473b
Show file tree

Hide file tree

Showing 4 changed files with 7 additions and 7 deletions.
diff --git a/llama-box/patches/llama.cpp/model.patch b/llama-box/patches/llama.cpp/model.patch
@@ -133,10 +133,10 @@ index f90f5e74..818caf0b 100644
          case LLM_ARCH_NOMIC_BERT:
          case LLM_ARCH_STABLELM:
 diff --git a/src/llama-vocab.cpp b/src/llama-vocab.cpp
-index 96b74e93..6cc05279 100644
+index 4969d262..3c178c90 100644
 --- a/src/llama-vocab.cpp
 +++ b/src/llama-vocab.cpp
-@@ -1429,7 +1429,7 @@ void llama_vocab::impl::load(llama_model_loader & ml, const LLM_KV & kv) {
+@@ -1430,7 +1430,7 @@ void llama_vocab::impl::load(llama_model_loader & ml, const LLM_KV & kv) {
              special_mask_id = LLAMA_TOKEN_NULL;
 
              const int precompiled_charsmap_keyidx = gguf_find_key(ctx, kv(LLM_KV_TOKENIZER_PRECOMPILED_CHARSMAP).c_str());

diff --git a/llama-box/patches/llama.cpp/model_py.patch b/llama-box/patches/llama.cpp/model_py.patch
@@ -1,5 +1,5 @@
 diff --git a/convert_hf_to_gguf.py b/convert_hf_to_gguf.py
-index 4dc9837a..ba252f57 100755
+index 4dc9837a..d488122e 100755
 --- a/convert_hf_to_gguf.py
 +++ b/convert_hf_to_gguf.py
 @@ -4,6 +4,7 @@

diff --git a/llama-box/patches/llama.cpp/vocab.patch b/llama-box/patches/llama.cpp/vocab.patch
@@ -11,10 +11,10 @@ index a184884c..ec814882 100644
      LLAMA_API bool llama_vocab_get_add_bos(const struct llama_vocab * vocab);
      LLAMA_API bool llama_vocab_get_add_eos(const struct llama_vocab * vocab);
 diff --git a/src/llama-vocab.cpp b/src/llama-vocab.cpp
-index 96b74e93..c57a8648 100644
+index 4969d262..44b68cbf 100644
 --- a/src/llama-vocab.cpp
 +++ b/src/llama-vocab.cpp
-@@ -1586,7 +1586,8 @@ void llama_vocab::impl::load(llama_model_loader & ml, const LLM_KV & kv) {
+@@ -1587,7 +1587,8 @@ void llama_vocab::impl::load(llama_model_loader & ml, const LLM_KV & kv) {
                  tokenizer_pre == "megrez") {
                  pre_type = LLAMA_VOCAB_PRE_TYPE_QWEN2;
              } else {
@@ -24,7 +24,7 @@ index 96b74e93..c57a8648 100644
              }
          } else if (type == LLAMA_VOCAB_TYPE_SPM) {
              pre_type = LLAMA_VOCAB_PRE_TYPE_DEFAULT;
-@@ -3074,6 +3075,10 @@ llama_token llama_vocab_pad(const struct llama_vocab * vocab) {
+@@ -3075,6 +3076,10 @@ llama_token llama_vocab_pad(const struct llama_vocab * vocab) {
      return vocab->token_pad();
  }
 

diff --git a/llama.cpp b/llama.cpp
+33 −33		ci/run.sh
+ −		examples/server/public/index.html.gz
+2 −2		examples/server/webui/src/main.js
+3 −2		src/llama-vocab.cpp
+162 −132		tests/test-chat-template.cpp