feat(llama.cp logging): Add some helpful logs when tensors are missing

NOTE: This also adds the LLAMA_LOG_DEBUG macro for use in llama.cpp. This may have been omitted for a good reason! If so, this change is certainly optional. Branch: GraniteMoE Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>
ggerganov · Sep 11, 2024 · 5f37be3 · 5f37be3
1 parent 3219f58
commit 5f37be3
Show file tree

Hide file tree

Showing 2 changed files with 6 additions and 0 deletions.
diff --git a/src/llama-impl.h b/src/llama-impl.h
@@ -24,6 +24,7 @@ LLAMA_ATTRIBUTE_FORMAT(2, 3)
 void llama_log_internal        (ggml_log_level level, const char * format, ...);
 void llama_log_callback_default(ggml_log_level level, const char * text, void * user_data);
 
+#define LLAMA_LOG_DEBUG(...) llama_log_internal(GGML_LOG_LEVEL_DEBUG , __VA_ARGS__)
 #define LLAMA_LOG_INFO(...)  llama_log_internal(GGML_LOG_LEVEL_INFO , __VA_ARGS__)
 #define LLAMA_LOG_WARN(...)  llama_log_internal(GGML_LOG_LEVEL_WARN , __VA_ARGS__)
 #define LLAMA_LOG_ERROR(...) llama_log_internal(GGML_LOG_LEVEL_ERROR, __VA_ARGS__)

diff --git a/src/llama.cpp b/src/llama.cpp
@@ -1484,34 +1484,39 @@ struct LLM_TN {
 
     std::string operator()(llm_tensor tensor) const {
         if (LLM_TENSOR_NAMES.at(arch).find(tensor) == LLM_TENSOR_NAMES.at(arch).end()) {
+            LLAMA_LOG_DEBUG("%s: Missing tensor %d for arch %s\n", __func__, tensor, LLM_ARCH_NAMES.at(arch));
             return "__missing__";
         }
         return LLM_TENSOR_NAMES.at(arch).at(tensor);
     }
 
     std::string operator()(llm_tensor tensor, const std::string & suffix) const {
         if (LLM_TENSOR_NAMES.at(arch).find(tensor) == LLM_TENSOR_NAMES.at(arch).end()) {
+            LLAMA_LOG_DEBUG("%s(%s): Missing tensor %d for arch %s\n", __func__, suffix.c_str(), tensor, LLM_ARCH_NAMES.at(arch));
             return "__missing__";
         }
         return LLM_TENSOR_NAMES.at(arch).at(tensor) + "." + suffix;
     }
 
     std::string operator()(llm_tensor tensor, int bid) const {
         if (LLM_TENSOR_NAMES.at(arch).find(tensor) == LLM_TENSOR_NAMES.at(arch).end()) {
+            LLAMA_LOG_DEBUG("%s(%d): Missing tensor %d for arch %s\n", __func__, bid, tensor, LLM_ARCH_NAMES.at(arch));
             return "__missing__";
         }
         return ::format(LLM_TENSOR_NAMES.at(arch).at(tensor).c_str(), bid);
     }
 
     std::string operator()(llm_tensor tensor, const std::string & suffix, int bid) const {
         if (LLM_TENSOR_NAMES.at(arch).find(tensor) == LLM_TENSOR_NAMES.at(arch).end()) {
+            LLAMA_LOG_DEBUG("%s(%s, %d): Missing tensor %d for arch %s\n", __func__, suffix.c_str(), bid, tensor, LLM_ARCH_NAMES.at(arch));
             return "__missing__";
         }
         return ::format(LLM_TENSOR_NAMES.at(arch).at(tensor).c_str(), bid) + "." + suffix;
     }
 
     std::string operator()(llm_tensor tensor, const std::string & suffix, int bid, int xid) const {
         if (LLM_TENSOR_NAMES.at(arch).find(tensor) == LLM_TENSOR_NAMES.at(arch).end()) {
+            LLAMA_LOG_DEBUG("%s(%s, %d, %d): Missing tensor %d for arch %s\n", __func__, suffix.c_str(), bid, xid, tensor, LLM_ARCH_NAMES.at(arch));
             return "__missing__";
         }
         return ::format(LLM_TENSOR_NAMES.at(arch).at(tensor).c_str(), bid, xid) + "." + suffix;