From 8f429fa5111901f9646cf998643ac5310846d487 Mon Sep 17 00:00:00 2001
From: Georgi Gerganov <ggerganov@gmail.com>
Date: Sun, 3 Sep 2023 13:42:56 +0300
Subject: perplexity : fix ETA by warming up the model with an empty run

---
 common/common.cpp | 8 ++++++++
 1 file changed, 8 insertions(+)

(limited to 'common/common.cpp')

diff --git a/common/common.cpp b/common/common.cpp
index 41fc59ce..a1c3dc78 100644
--- a/common/common.cpp
+++ b/common/common.cpp
@@ -752,6 +752,14 @@ std::tuple<struct llama_model *, struct llama_context *> llama_init_from_gpt_par
         params.logit_bias[llama_token_eos(lctx)] = -INFINITY;
     }
 
+    {
+        LOG("warming up the model with an empty run\n");
+
+        const std::vector<llama_token> tmp = { llama_token_bos(lctx), };
+        llama_eval(lctx, tmp.data(), tmp.size(), 0, params.n_threads);
+        llama_reset_timings(lctx);
+    }
+
     return std::make_tuple(model, lctx);
 }
 
-- 
cgit v1.2.3