Adding the openBLAS

gguf_engine.py

@@ -92,6 +92,7 @@ def _load_text_model(adapter_name: str) -> Llama:
             n_ctx      = N_CTX,
             n_batch    = N_BATCH,
             n_threads  = N_THREADS,
+            n_threads_batch=2,    # Crucial for OpenBLAS prompt evaluation
             use_mmap   = True,
             verbose    = False,
         )
@@ -125,6 +126,7 @@ def _load_vision_model(vision_module: str) -> Llama:
             n_ctx        = 1024,
             n_batch      = N_BATCH,
             n_threads    = N_THREADS,
+            n_threads_batch=2,    # Crucial for OpenBLAS prompt evaluation
             use_mmap     = True,
             verbose      = False,
         )

packages.txt

@@ -0,0 +1 @@
+libopenblas0-pthread