mirror of
				https://github.com/ggml-org/llama.cpp.git
				synced 2025-10-30 08:42:00 +00:00 
			
		
		
		
	llava : replace ggml_cpy with ggml_cont
This commit is contained in:
		| @@ -618,7 +618,7 @@ static ggml_cgraph * clip_image_build_graph(clip_ctx * ctx, const clip_image_f32 | |||||||
|             KQV = ggml_reshape_4d(ctx0, KQV, d_head, num_positions, n_head, batch_size); |             KQV = ggml_reshape_4d(ctx0, KQV, d_head, num_positions, n_head, batch_size); | ||||||
|             KQV = ggml_cont(ctx0, ggml_permute(ctx0, KQV, 0, 2, 1, 3)); |             KQV = ggml_cont(ctx0, ggml_permute(ctx0, KQV, 0, 2, 1, 3)); | ||||||
|  |  | ||||||
|             cur = ggml_cpy(ctx0, KQV, ggml_new_tensor_3d(ctx0, GGML_TYPE_F32, hidden_size, num_positions, batch_size)); |             cur = ggml_cont_3d(ctx0, KQV, hidden_size, num_positions, batch_size); | ||||||
|         } |         } | ||||||
|  |  | ||||||
|         // attention output |         // attention output | ||||||
|   | |||||||
		Reference in New Issue
	
	Block a user
	 slaren
					slaren