vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Respect tokenizer.ggml.add_bos_token value when tokenizing (#4040)

Browse source 
* gguf-py: gguf-dump: Respect --no-tensor flag in JSON mode.

* Respect add_bos_token GGUF metadata value

* gguf-py: Try to fix SpecialVocab giving up too easily for the Nth time
This commit is contained in:
Kerfuffle 2023-11-16 19:14:37 -07:00 • committed by GitHub
parent 8da46278e1
commit 91f6499393
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
12 changed files with 85 additions and 29 deletions
Download patch file Download diff file Expand all files Collapse all files

6
llama.h
View file

@ -517,6 +517,12 @@ extern "C" {
LLAMA_API llama_token llama_token_eos(const struct llama_model * model); // end-of-sentence
LLAMA_API llama_token llama_token_nl (const struct llama_model * model); // next-line
// Returns -1 if unknown, 1 for true or 0 for false.
LLAMA_API int llama_add_bos_token(const struct llama_model * model);
// Returns -1 if unknown, 1 for true or 0 for false.
LLAMA_API int llama_add_eos_token(const struct llama_model * model);
// codellama infill tokens
LLAMA_API llama_token llama_token_prefix(const struct llama_model * model); // Beginning of infill prefix
LLAMA_API llama_token llama_token_middle(const struct llama_model * model); // Beginning of infill middle