From 888d4f591b246ea74016a54f2a289aab24731744 Mon Sep 17 00:00:00 2001 From: Branden Butler Date: Mon, 30 Oct 2023 10:50:20 -0500 Subject: [PATCH] Update MPI code to new KV seq rm and bos/eos model APIs --- examples/mpi/mpi.cpp | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) diff --git a/examples/mpi/mpi.cpp b/examples/mpi/mpi.cpp index 3030918bf..b4944099e 100644 --- a/examples/mpi/mpi.cpp +++ b/examples/mpi/mpi.cpp @@ -284,7 +284,7 @@ int main(int argc, char ** argv) { // Should not run without any tokens if (embd_inp.empty()) { - embd_inp.push_back(llama_token_bos(ctx)); + embd_inp.push_back(llama_token_bos(model)); LOG("embd_inp was considered empty and bos was added: %s\n", LOG_TOKENS_TOSTR_PRETTY(ctx, embd_inp).c_str()); } @@ -334,7 +334,7 @@ int main(int argc, char ** argv) { } // remove any "future" tokens that we might have inherited from the previous session - llama_kv_cache_tokens_rm(ctx, n_matching_session_tokens, -1); + llama_kv_cache_seq_rm(ctx, -1, n_matching_session_tokens, -1); } LOGLN( @@ -729,7 +729,7 @@ int main(int argc, char ** argv) { } // deal with end of text token in interactive mode - if (llama_sampling_last(ctx_sampling) == llama_token_eos(ctx)) { + if (llama_sampling_last(ctx_sampling) == llama_token_eos(model)) { LOG("found EOS token\n"); if (params.interactive) { @@ -756,7 +756,7 @@ int main(int argc, char ** argv) { if (params.input_prefix_bos) { LOG("adding input prefix BOS token\n"); - embd_inp.push_back(llama_token_bos(ctx)); + embd_inp.push_back(llama_token_bos(model)); } std::string buffer; @@ -840,7 +840,7 @@ int main(int argc, char ** argv) { } // end of text token - if (!embd.empty() && embd.back() == llama_token_eos(ctx) && !(params.instruct || params.interactive)) { + if (!embd.empty() && embd.back() == llama_token_eos(model) && !(params.instruct || params.interactive)) { LOG_TEE(" [end of text]\n"); break; }