add dry sampler (#513)

* add dry sampler

* use vocab instead of model in dry_init function

* fix compile error for build test

---------

Co-authored-by: firecoperana <firecoperana>

1 files changed, 2 insertions, 2 deletions

diff --git a/examples/speculative/speculative.cpp b/examples/speculative/speculative.cpp
index b051a18f..3063f0b6 100644
--- a/examples/speculative/speculative.cpp
+++ b/examples/speculative/speculative.cpp
@@ -179,7 +179,7 @@ int main(int argc, char ** argv) {
     bool has_eos = false;
 
     // target model sampling context
-    struct llama_sampling_context * ctx_sampling = llama_sampling_init(params.sparams);
+    struct llama_sampling_context * ctx_sampling = llama_sampling_init(llama_get_model_vocab(model_tgt), params.sparams);
 
     // draft sequence data
     std::vector<seq_draft> drafts(n_seq_dft);
@@ -190,7 +190,7 @@ int main(int argc, char ** argv) {
     }
 
     for (int s = 0; s < n_seq_dft; ++s) {
-        drafts[s].ctx_sampling = llama_sampling_init(params.sparams);
+        drafts[s].ctx_sampling = llama_sampling_init(llama_get_model_vocab(model_dft), params.sparams);
     }
 
     llama_batch batch_dft = llama_batch_init(params.n_ctx, 0, 1);