[CANN] Adapt to dynamically loadable backends mechanism (ggerganov#9970)

* [CANN] Adapt to dynamically loadable backends mechanism * Fix the Bug: inference running result is garbled in debug running model for LM models who's type is Q4_0 class * Handle the review comments of this pull request build passed
arthw · Nov 15, 2024 · 6fd0666 · 6fd0666
1 parent aec56df
commit 6fd0666
Show file tree

Hide file tree

Showing 5 changed files with 268 additions and 150 deletions.
diff --git a/ggml/include/ggml-cann.h b/ggml/include/ggml-cann.h
@@ -34,6 +34,8 @@ extern "C" {
  */
 #define GGML_CANN_MAX_DEVICES 16
 
+GGML_API ggml_backend_reg_t ggml_backend_cann_reg(void);
+
 /**
  * @brief Initializes the CANN backend for a specified device.
  *

diff --git a/ggml/src/ggml-backend.cpp b/ggml/src/ggml-backend.cpp
@@ -561,6 +561,10 @@ void * ggml_backend_reg_get_proc_address(ggml_backend_reg_t reg, const char * na
 #  include "ggml-amx.h"
 #endif
 
+#ifdef GGML_USE_CANN
+#include "ggml-cann.h"
+#endif
+
 struct ggml_backend_registry {
     std::vector<ggml_backend_reg_t> backends;
     std::vector<ggml_backend_dev_t> devices;
@@ -587,8 +591,11 @@ struct ggml_backend_registry {
 #ifdef GGML_USE_AMX
         register_backend(ggml_backend_amx_reg());
 #endif
+#ifdef GGML_USE_CANN
+        register_backend(ggml_backend_cann_reg());
+#endif
 
-        // TODO: kompute, cann
+        // TODO: kompute
 
         register_backend(ggml_backend_cpu_reg());
     }
-Original file line number
+Diff line change
@@ Expand Up / @@ -34,6 +34,8 @@ extern "C" { @@
      */
     #define GGML_CANN_MAX_DEVICES 16
+    GGML_API ggml_backend_reg_t ggml_backend_cann_reg(void);
     /**
      * @brief Initializes the CANN backend for a specified device.
      *
@@ Expand Down @@