diff --git a/llama.cpp b/llama.cpp index d490d4e95..e3c3568c8 100644 --- a/llama.cpp +++ b/llama.cpp @@ -4072,7 +4072,8 @@ static int llama_decode_internal( #ifdef GGML_USE_MPI // TODO: needs fix after #3228 - ggml_mpi_eval_init(lctx.ctx_mpi, &n_tokens, &n_past, &n_threads); + GGML_ASSERT(false && "not implemented"); + //ggml_mpi_eval_init(lctx.ctx_mpi, &n_tokens, &n_past, &n_threads); #endif GGML_ASSERT(n_threads > 0); @@ -6846,8 +6847,10 @@ struct llama_context * llama_new_context_with_model( if (ggml_mpi_rank(ctx->ctx_mpi) > 0) { // Enter a blocking eval loop with dummy input, letting rank=0 drive the process - const std::vector tmp(ctx->model.hparams.n_ctx, llama_token_bos(ctx)); - while (!llama_eval(ctx, tmp.data(), tmp.size(), 0, 0)) {}; + // TODO: needs fix after #3228 + GGML_ASSERT(false && "not implemented"); + //const std::vector tmp(ctx->model.hparams.n_ctx, llama_token_bos(ctx)); + //while (!llama_eval(ctx, tmp.data(), tmp.size(), 0, 0)) {}; llama_backend_free(); exit(1); }