Bump llama.cpp to b1662 and set n_parallel=1

llm/ext_server.go

@@ -160,7 +160,7 @@ func newExtServer(server extServer, model string, adapters, projectors []string,
 	sparams.n_batch = C.uint(opts.NumBatch)
 	sparams.n_gpu_layers = C.int(numGPU)
 	sparams.main_gpu = C.int(opts.MainGPU)
-	sparams.n_parallel = 2 // TODO - wire up concurrency
+	sparams.n_parallel = 1 // TODO - wire up concurrency
 
 	// Always use the value encoded in the model
 	sparams.rope_freq_base = 0.0

llm/llama.cpp/gguf

@@ -1 +1 @@
-Subproject commit a7aee47b98e45539d491071b25778b833b77e387
+Subproject commit 328b83de23b33240e28f4e74900d1d06726f5eb1

llm/llama.cpp/patches/0001-Expose-callable-API-for-server.patch

@@ -1,4 +1,4 @@
-From 087cf3300e973d7790db8f7cad01d2a790de38be Mon Sep 17 00:00:00 2001
+From b5e195803e2a989e57eef0010adce778df1e2d01 Mon Sep 17 00:00:00 2001
 From: Daniel Hiltgen <daniel@ollama.com>
 Date: Mon, 13 Nov 2023 12:25:58 -0800
 Subject: [PATCH] Expose callable API for server
@@ -46,7 +46,7 @@ index 859cd12..4ea47a7 100644
 +endif()
 \ No newline at end of file
 diff --git a/examples/server/server.cpp b/examples/server/server.cpp
-index d0cd8e1..5f5d4c5 100644
+index 0403853..2084fd8 100644
 --- a/examples/server/server.cpp
 +++ b/examples/server/server.cpp
 @@ -5,6 +5,9 @@
@@ -59,15 +59,15 @@ index d0cd8e1..5f5d4c5 100644
  
  #ifndef NDEBUG
  // crash the server in debug mode, otherwise send an http 500 error
-@@ -2632,6 +2635,7 @@ static void append_to_generated_text_from_generated_token_probs(llama_server_con
+@@ -2643,6 +2646,7 @@ static void append_to_generated_text_from_generated_token_probs(llama_server_con
      }
  }
  
 +#ifndef LLAMA_SERVER_LIBRARY
  int main(int argc, char **argv)
  {
-     // own arguments required by this example
-@@ -3066,3 +3070,273 @@ int main(int argc, char **argv)
+ #if SERVER_VERBOSE != 1
+@@ -3123,3 +3127,273 @@ int main(int argc, char **argv)
      llama_backend_free();
      return 0;
  }
@@ -439,10 +439,10 @@ index 0000000..d22f1b6
 +#endif // LLAMA_SERVER_LIBRARY
 \ No newline at end of file
 diff --git a/ggml-cuda.cu b/ggml-cuda.cu
-index 9e1acd3..ea64b55 100644
+index f20846f..9640cf3 100644
 --- a/ggml-cuda.cu
 +++ b/ggml-cuda.cu
-@@ -6505,6 +6505,7 @@ static cudaError_t ggml_cuda_cpy_tensor_2d(
+@@ -6757,6 +6757,7 @@ static cudaError_t ggml_cuda_cpy_tensor_2d(
          CUDA_CHECK(cudaGetDevice(&id));
          src_ptr = (char *) extra->data_device[id];
      } else {