Mirrors/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/llama.cpp.git synced 2024-12-25 13:58:46 +01:00
Code Issues Packages Projects Releases Wiki Activity

Add more tokenizer tests (#3742)

Browse Source 
* Add more tokenizer tests

* Add starcoder

* Update test vocab files

* Restrict bpe tokenizer tests to unicode planes

* Update comment

* Comment cosmetics

* Remove bloom vocab/test
This commit is contained in:
Galunid 2023-10-24 09:17:17 +02:00 committed by GitHub
parent 469c9addef
commit daab3d7f45
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
6 changed files with 16 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

BIN
models/ggml-vocab-baichuan.gguf Normal file
View File

Binary file not shown.

BIN
models/ggml-vocab-gpt-neox.gguf Normal file
View File

Binary file not shown.

BIN
models/ggml-vocab-refact.gguf Normal file
View File

Binary file not shown.

BIN
models/ggml-vocab-starcoder.gguf Normal file
View File

Binary file not shown.

4
tests/CMakeLists.txt
View File

@ -28,10 +28,14 @@ llama_build_executable(test-tokenizer-0-falcon.cpp)
llama_test_executable (test-tokenizer-0-falcon test-tokenizer-0-falcon.cpp ${CMAKE_CURRENT_SOURCE_DIR}/../models/ggml-vocab-falcon.gguf) llama_test_executable (test-tokenizer-0-falcon test-tokenizer-0-falcon.cpp ${CMAKE_CURRENT_SOURCE_DIR}/../models/ggml-vocab-falcon.gguf)
llama_build_executable(test-tokenizer-1-llama.cpp) llama_build_executable(test-tokenizer-1-llama.cpp)
llama_test_executable (test-tokenizer-1-llama test-tokenizer-1-llama.cpp ${CMAKE_CURRENT_SOURCE_DIR}/../models/ggml-vocab-llama.gguf) llama_test_executable (test-tokenizer-1-llama test-tokenizer-1-llama.cpp ${CMAKE_CURRENT_SOURCE_DIR}/../models/ggml-vocab-llama.gguf)
llama_test_executable(test-tokenizer-1-baichuan test-tokenizer-1-llama.cpp ${CMAKE_CURRENT_SOURCE_DIR}/../models/ggml-vocab-baichuan.gguf)
llama_build_executable(test-tokenizer-1-bpe.cpp) llama_build_executable(test-tokenizer-1-bpe.cpp)
llama_test_executable (test-tokenizer-1-falcon test-tokenizer-1-bpe.cpp ${CMAKE_CURRENT_SOURCE_DIR}/../models/ggml-vocab-falcon.gguf) llama_test_executable (test-tokenizer-1-falcon test-tokenizer-1-bpe.cpp ${CMAKE_CURRENT_SOURCE_DIR}/../models/ggml-vocab-falcon.gguf)
llama_test_executable(test-tokenizer-1-aquila test-tokenizer-1-bpe.cpp ${CMAKE_CURRENT_SOURCE_DIR}/../models/ggml-vocab-aquila.gguf) llama_test_executable(test-tokenizer-1-aquila test-tokenizer-1-bpe.cpp ${CMAKE_CURRENT_SOURCE_DIR}/../models/ggml-vocab-aquila.gguf)
llama_test_executable(test-tokenizer-1-mpt test-tokenizer-1-bpe.cpp ${CMAKE_CURRENT_SOURCE_DIR}/../models/ggml-vocab-mpt.gguf) llama_test_executable(test-tokenizer-1-mpt test-tokenizer-1-bpe.cpp ${CMAKE_CURRENT_SOURCE_DIR}/../models/ggml-vocab-mpt.gguf)
llama_test_executable(test-tokenizer-1-gpt-neox test-tokenizer-1-bpe.cpp ${CMAKE_CURRENT_SOURCE_DIR}/../models/ggml-vocab-gpt-neox.gguf)
llama_test_executable(test-tokenizer-1-refact test-tokenizer-1-bpe.cpp ${CMAKE_CURRENT_SOURCE_DIR}/../models/ggml-vocab-refact.gguf)
llama_test_executable(test-tokenizer-1-starcoder test-tokenizer-1-bpe.cpp ${CMAKE_CURRENT_SOURCE_DIR}/../models/ggml-vocab-starcoder.gguf)
llama_build_and_test_executable(test-grammar-parser.cpp) llama_build_and_test_executable(test-grammar-parser.cpp)
llama_build_and_test_executable(test-llama-grammar.cpp) llama_build_and_test_executable(test-llama-grammar.cpp)
llama_build_and_test_executable(test-grad0.cpp) # SLOW llama_build_and_test_executable(test-grad0.cpp) # SLOW

15
tests/test-tokenizer-1-bpe.cpp
View File

@ -91,9 +91,19 @@ int main(int argc, char **argv) {
} }
} }
} }
// TODO: why doesn't this work for the full range of Unicodes? // Restrict to assigned unicode planes
// for (uint32_t cp = 0x10000; cp < 0x0010ffff; ++cp) { // for (uint32_t cp = 0x10000; cp < 0x0010ffff; ++cp) {
for (uint32_t cp = 0x10000; cp < 0x00080000; ++cp) { for (uint32_t cp = 0x10000; cp < 0x00040000; ++cp) {
std::string str = codepoint_to_utf8(cp);
std::vector<llama_token> tokens = llama_tokenize(ctx, str, false);
std::string check = llama_detokenize_bpe(ctx, tokens);
if (str != check) {
fprintf(stderr, "%s : error: codepoint %x detokenizes to '%s'(%zu) instead of '%s'(%zu)\n",
__func__, cp, check.c_str(), check.length(), str.c_str(), str.length());
return 4;
}
}
for (uint32_t cp = 0x000e0000; cp < 0x0010ffff; ++cp) {
std::string str = codepoint_to_utf8(cp); std::string str = codepoint_to_utf8(cp);
std::vector<llama_token> tokens = llama_tokenize(ctx, str, false); std::vector<llama_token> tokens = llama_tokenize(ctx, str, false);
std::string check = llama_detokenize_bpe(ctx, tokens); std::string check = llama_detokenize_bpe(ctx, tokens);
@ -103,7 +113,6 @@ int main(int argc, char **argv) {
return 4; return 4;
} }
} }
llama_free_model(model); llama_free_model(model);
llama_free(ctx); llama_free(ctx);