fix warnings

Author: Judd

Makefile

@@ -211,9 +211,7 @@ MK_CFLAGS += \
 	-Werror=implicit-function-declaration
 
 MK_CXXFLAGS += \
-	$(WARN_FLAGS) \
-	-Wmissing-declarations \
-	-Wmissing-noreturn
+	$(WARN_FLAGS)
 
 ifeq ($(CHATLLM_FATAL_WARNINGS),1)
 	MK_CFLAGS   += -Werror

src/layers.cpp

@@ -999,7 +999,7 @@ namespace chatllm
 
     ggml::tensor *BaichuanSelfAttention::apply_pos_embedding_kq(ComputeContext *ctx, ggml::tensor *kq, int hidden_size, int qlen, ggml::tensor *past) const
     {
-        return ggml::map_custom1(ctx, kq, ggml_compute_forward_custom_alibi, GGML_N_TASKS_MAX, (void *)&alibi);
+        return ggml::map_custom1(ctx, kq, ggml_compute_forward_custom_alibi, GGML_N_TASKS_MAX, const_cast<void *>((const void *)&alibi));
     }
 
     QWenSelfAttention::QWenSelfAttention(InitContext *ctx, int hidden_size, int num_attention_heads, int max_length)

src/layers.h

@@ -133,6 +133,16 @@ namespace chatllm
             CHATLLM_THROW << "forward(ComputeContext *ctx, ggml::tensor *input, int n_past): not implemented";
             return NULL;
         }
+        virtual ggml::tensor *forward(ComputeContext *ctx, ggml::tensor *hidden_states, ggml::tensor *_)
+        {
+            CHATLLM_THROW << "forward(ComputeContext *ctx, ggml::tensor *hidden_states, ggml::tensor *_): not implemented";
+            return NULL;
+        }
+        virtual ggml::tensor *forward(ComputeContext *ctx, ggml::tensor *patches, int patches_per_row, ggml::tensor *text_input)
+        {
+            CHATLLM_THROW << "forward(ComputeContext *ctx, ggml::tensor *patches, int patches_per_row, ggml::tensor *text_input): not implemented";
+            return NULL;
+        }
         virtual void set_ctx(int n_ctx) { }
         virtual void shift_cache(int shift, int total) { }
 
@@ -220,8 +230,6 @@ namespace chatllm
     public:
         VisualEmbedding(InitContext *ctx, int num_embeddings, int embedding_dim)
             : Embedding(ctx, num_embeddings, embedding_dim) {}
-
-        virtual ggml::tensor *forward(ComputeContext *ctx, ggml::tensor *patches, int patches_per_row, ggml::tensor *text_input) = 0;
     };
 
     class Linear : public Block
@@ -1848,7 +1856,7 @@ namespace chatllm
               norm(ctx, hidden_size)
         {}
 
-        ggml::tensor *forward(ComputeContext *ctx, ggml::tensor *hidden_states, ggml::tensor *attention_output);
+        ggml::tensor *forward(ComputeContext *ctx, ggml::tensor *hidden_states, ggml::tensor *attention_output) override;
 
     public:
         Linear dense;