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

llama : add support for the cohere2 model architecture (#10900)

Browse source
This commit is contained in:
DAN™ 2025-01-04 09:33:31 -05:00 • committed by GitHub
parent 78c6785175
commit 46be942214
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
6 changed files with 221 additions and 0 deletions
Download patch file Download diff file Expand all files Collapse all files

11
src/llama-model.cpp
View file

@ -786,6 +786,16 @@ void llm_load_hparams(llama_model_loader & ml, llama_model & model) {
default: model.type = e_model::MODEL_UNKNOWN;
}
} break;
case LLM_ARCH_COHERE2:
{
ml.get_key(LLM_KV_ATTENTION_SLIDING_WINDOW, hparams.n_swa);
ml.get_key(LLM_KV_LOGIT_SCALE, hparams.f_logit_scale);
ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps);
switch (hparams.n_layer) {
case 32: model.type = e_model::MODEL_8B; break;
default: model.type = e_model::MODEL_UNKNOWN;
}
} break;
case LLM_ARCH_DBRX:
{
ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps);
@ -2031,6 +2041,7 @@ enum llama_rope_type llama_rope_type(const struct llama_model * model) {
case LLM_ARCH_MINICPM:
case LLM_ARCH_XVERSE:
case LLM_ARCH_COMMAND_R:
case LLM_ARCH_COHERE2:
case LLM_ARCH_OLMO:
case LLM_ARCH_ARCTIC:
case LLM_ARCH_DEEPSEEK: