feat: Update llama.cpp

abetlen · Mar 15, 2024 · 6eb2523 · 6eb2523
1 parent 20e6815
commit 6eb2523
Show file tree

Hide file tree

Showing 2 changed files with 9 additions and 7 deletions.
diff --git a/llama_cpp/llama_cpp.py b/llama_cpp/llama_cpp.py
@@ -198,13 +198,15 @@ def byref(obj: CtypesCData, offset: Optional[int] = None) -> CtypesRef[CtypesCDa
 
 
 # enum llama_vocab_type {
-#     LLAMA_VOCAB_TYPE_SPM = 0, // SentencePiece
-#     LLAMA_VOCAB_TYPE_BPE = 1, // Byte Pair Encoding
-#     LLAMA_VOCAB_TYPE_WPM = 2, // WordPiece
+#     LLAMA_VOCAB_TYPE_NONE = 0, // For models without vocab
+#     LLAMA_VOCAB_TYPE_SPM  = 1, // SentencePiece
+#     LLAMA_VOCAB_TYPE_BPE  = 2, // Byte Pair Encoding
+#     LLAMA_VOCAB_TYPE_WPM  = 3, // WordPiece
 # };
-LLAMA_VOCAB_TYPE_SPM = 0
-LLAMA_VOCAB_TYPE_BPE = 1
-LLAMA_VOCAB_TYPE_WPM = 2
+LLAMA_VOCAB_TYPE_NONE = 0
+LLAMA_VOCAB_TYPE_SPM = 1
+LLAMA_VOCAB_TYPE_BPE = 2
+LLAMA_VOCAB_TYPE_WPM = 3
 
 
 # // note: these values should be synchronized with ggml_rope

diff --git a/vendor/llama.cpp b/vendor/llama.cpp
+1 −1		Makefile
+73 −53		convert.py
+1 −0		examples/gguf/gguf.cpp
+15 −5		examples/llama-bench/llama-bench.cpp
+22 −18		examples/llava/clip.cpp
+3 −3		examples/llava/clip.h
+1 −1		examples/llava/llava.cpp
+2 −2		examples/llava/llava.h
+4 −1		examples/sycl/build.sh
+13 −3		examples/sycl/run-llama2.sh
+1 −0		examples/train-text-from-scratch/train-text-from-scratch.cpp
+3 −0		ggml-cuda.cu
+244 −79		ggml-sycl.cpp
+5 −0		ggml-sycl.h
+17 −0		ggml.c
+2 −0		ggml.h
+6 −0		gguf-py/gguf/constants.py
+9 −3		gguf-py/gguf/gguf_reader.py
+11 −4		gguf-py/gguf/gguf_writer.py
+1 −1		gguf-py/pyproject.toml
+97 −40		llama.cpp
+4 −3		llama.h
+4 −0		tests/test-chat-template.cpp