happyz
/
gemma.cpp
mirror of https://github.com/google/gemma.cpp.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

naming scheme between gemma and gemma2 variants on the command line was not consistent 

PiperOrigin-RevId: 728409104
This commit is contained in:
The gemma.cpp Authors 2025-02-18 16:36:48 -08:00 committed by Copybara-Service
parent 2bdf26d81d
commit b00e8a7bcf
3 changed files with 34 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

14
compression/compress_weights.cc
View File

@ -110,9 +110,17 @@ struct Args : public ArgsBase<Args> {
"Path to model weights (.bin) file.\n" "Path to model weights (.bin) file.\n"
" Required argument."); " Required argument.");
visitor(model_type_str, "model", std::string(), visitor(model_type_str, "model", std::string(),
"Model type\n 2b-it = 2B parameters, instruction-tuned\n " "Model type\n "
"2b-pt = 2B parameters, pretrained\n 7b-it = 7B parameters " "gemma-2b-it = Gemma 2B parameters, instruction-tuned\n "
"instruction-tuned\n 7b-pt = 7B parameters, pretrained\n " "gemma-2b-pt = Gemma 2B parameters, pretrained\n "
"gemma-7b-it = Gemma 7B parameters, instruction-tuned\n "
"gemma-7b-pt = Gemma 7B parameters, pretrained\n "
"gemma2-2b-it = Gemma2 2B parameters, instruction-tuned\n "
"gemma2-2b-pt = Gemma2 2B parameters, pretrained\n "
"gemma2-9b-it = Gemma2 9B parameters, instruction-tuned\n "
"gemma2-9b-pt = Gemma2 9B parameters, pretrained\n "
"gemma2-27b-it = Gemma2 27B parameters, instruction-tuned\n "
"gemma2-27b-pt = Gemma2 27B parameters, pretrained\n "
"gr2b-it = griffin 2B parameters, instruction-tuned\n " "gr2b-it = griffin 2B parameters, instruction-tuned\n "
"gr2b-pt = griffin 2B parameters, pretrained\n " "gr2b-pt = griffin 2B parameters, pretrained\n "
" Required argument."); " Required argument.");

12
gemma/common.cc
View File

@ -44,6 +44,10 @@ constexpr const char* kModelFlags[] = {
"paligemma2-3b-448", // PaliGemma2 3B 448 "paligemma2-3b-448", // PaliGemma2 3B 448
"paligemma2-10b-224", // PaliGemma2 10B 224 "paligemma2-10b-224", // PaliGemma2 10B 224
"paligemma2-10b-448", // PaliGemma2 10B 448 "paligemma2-10b-448", // PaliGemma2 10B 448
"gemma-2b-pt", "gemma-2b-it",
"gemma-7b-pt", "gemma-7b-it",
"gemma2-9b-pt", "gemma2-9b-it",
"gemma2-27b-pt", "gemma2-27b-it",
}; };
constexpr Model kModelTypes[] = { constexpr Model kModelTypes[] = {
Model::GEMMA_2B, Model::GEMMA_2B, // Gemma 2B Model::GEMMA_2B, Model::GEMMA_2B, // Gemma 2B
@ -59,6 +63,10 @@ constexpr Model kModelTypes[] = {
Model::PALIGEMMA2_3B_448, // PaliGemma2 3B 448 Model::PALIGEMMA2_3B_448, // PaliGemma2 3B 448
Model::PALIGEMMA2_10B_224, // PaliGemma2 10B 224 Model::PALIGEMMA2_10B_224, // PaliGemma2 10B 224
Model::PALIGEMMA2_10B_448, // PaliGemma2 10B 448 Model::PALIGEMMA2_10B_448, // PaliGemma2 10B 448
Model::GEMMA_2B, Model::GEMMA_2B, // Gemma 2B
Model::GEMMA_7B, Model::GEMMA_7B, // Gemma 7B
Model::GEMMA2_9B, Model::GEMMA2_9B, // Gemma2 9B
Model::GEMMA2_27B, Model::GEMMA2_27B, // Gemma2 27B
}; };
constexpr PromptWrapping kPromptWrapping[] = { constexpr PromptWrapping kPromptWrapping[] = {
PromptWrapping::GEMMA_PT, PromptWrapping::GEMMA_IT, // Gemma 2B PromptWrapping::GEMMA_PT, PromptWrapping::GEMMA_IT, // Gemma 2B
@ -71,6 +79,10 @@ constexpr PromptWrapping kPromptWrapping[] = {
PromptWrapping::PALIGEMMA, PromptWrapping::PALIGEMMA, // PaliGemma 224/448 PromptWrapping::PALIGEMMA, PromptWrapping::PALIGEMMA, // PaliGemma 224/448
PromptWrapping::PALIGEMMA, PromptWrapping::PALIGEMMA, // PG2 3B 224/448 PromptWrapping::PALIGEMMA, PromptWrapping::PALIGEMMA, // PG2 3B 224/448
PromptWrapping::PALIGEMMA, PromptWrapping::PALIGEMMA, // PG2 10B 224/448 PromptWrapping::PALIGEMMA, PromptWrapping::PALIGEMMA, // PG2 10B 224/448
PromptWrapping::GEMMA_PT, PromptWrapping::GEMMA_IT, // Gemma 2B
PromptWrapping::GEMMA_PT, PromptWrapping::GEMMA_IT, // Gemma 7B
PromptWrapping::GEMMA_PT, PromptWrapping::GEMMA_IT, // Gemma2 9B
PromptWrapping::GEMMA_PT, PromptWrapping::GEMMA_IT, // Gemma2 27B
}; };
constexpr size_t kNumModelFlags = std::size(kModelFlags); constexpr size_t kNumModelFlags = std::size(kModelFlags);

14
util/app.h
View File

@ -206,9 +206,17 @@ struct LoaderArgs : public ArgsBase<LoaderArgs> {
visitor(compressed_weights, "compressed_weights", Path(), visitor(compressed_weights, "compressed_weights", Path(),
"Alias for --weights."); "Alias for --weights.");
visitor(model_type_str, "model", std::string(), visitor(model_type_str, "model", std::string(),
"Model type\n 2b-it = 2B parameters, instruction-tuned\n " "Model type\n "
"2b-pt = 2B parameters, pretrained\n 7b-it = 7B parameters " "gemma-2b-it = Gemma 2B parameters, instruction-tuned\n "
"instruction-tuned\n 7b-pt = 7B parameters, pretrained\n " "gemma-2b-pt = Gemma 2B parameters, pretrained\n "
"gemma-7b-it = Gemma 7B parameters, instruction-tuned\n "
"gemma-7b-pt = Gemma 7B parameters, pretrained\n "
"gemma2-2b-it = Gemma2 2B parameters, instruction-tuned\n "
"gemma2-2b-pt = Gemma2 2B parameters, pretrained\n "
"gemma2-9b-it = Gemma2 9B parameters, instruction-tuned\n "
"gemma2-9b-pt = Gemma2 9B parameters, pretrained\n "
"gemma2-27b-it = Gemma2 27B parameters, instruction-tuned\n "
"gemma2-27b-pt = Gemma2 27B parameters, pretrained\n "
"gr2b-it = griffin 2B parameters, instruction-tuned\n " "gr2b-it = griffin 2B parameters, instruction-tuned\n "
"gr2b-pt = griffin 2B parameters, pretrained."); "gr2b-pt = griffin 2B parameters, pretrained.");
visitor(weight_type_str, "weight_type", std::string("sfp"), visitor(weight_type_str, "weight_type", std::string("sfp"),