Only expose libggml

Author: abetlen

llama_cpp/_ggml.py

@@ -5,139 +5,8 @@
 import os
 import pathlib
 
-import ctypes
-
 import llama_cpp._ctypes_extensions as ctypes_ext
 
-import numpy as np
-
-
 libggml_base_path = pathlib.Path(os.path.abspath(os.path.dirname(__file__))) / "lib"
 libggml = ctypes_ext.load_shared_library("ggml", libggml_base_path)
 
-ggml_function = ctypes_ext.ctypes_function_for_shared_library(libggml)
-
-
-# define GGML_MAX_DIMS           4
-GGML_MAX_DIMS = 4
-
-# define GGML_MAX_OP_PARAMS      64
-GGML_MAX_OP_PARAMS = 64
-
-# define GGML_MAX_SRC            10
-GGML_MAX_SRC = 10
-
-# define GGML_MAX_NAME           64
-GGML_MAX_NAME = 64
-
-
-# // n-dimensional tensor
-# struct ggml_tensor {
-#     enum ggml_type         type;
-#
-#     GGML_DEPRECATED(enum ggml_backend_type backend, "use the buffer type to find the storage location of the tensor");
-#
-#     struct ggml_backend_buffer * buffer;
-#
-#     int64_t ne[GGML_MAX_DIMS]; // number of elements
-#     size_t  nb[GGML_MAX_DIMS]; // stride in bytes:
-#                                // nb[0] = ggml_type_size(type)
-#                                // nb[1] = nb[0]   * (ne[0] / ggml_blck_size(type)) + padding
-#                                // nb[i] = nb[i-1] * ne[i-1]
-#
-#     // compute data
-#     enum ggml_op op;
-#
-#     // op params - allocated as int32_t for alignment
-#     int32_t op_params[GGML_MAX_OP_PARAMS / sizeof(int32_t)];
-#
-#     int32_t flags;
-#
-#     struct ggml_tensor * grad;
-#     struct ggml_tensor * src[GGML_MAX_SRC];
-#
-#     // source tensor and offset for views
-#     struct ggml_tensor * view_src;
-#     size_t               view_offs;
-#
-#     void * data;
-#
-#     char name[GGML_MAX_NAME];
-#
-#     void * extra; // extra things e.g. for ggml-cuda.cu
-#
-#     // char padding[4];
-# };
-class ggml_tensor(ctypes.Structure):
-    __fields__ = [
-        ("type", ctypes.c_int),
-        ("buffer", ctypes.c_void_p),
-        ("ne", ctypes.c_int64 * 8),
-        ("nb", ctypes.c_size_t * 8),
-        ("op", ctypes.c_int),
-        ("op_params", ctypes.c_int32 * 8),
-        ("flags", ctypes.c_int32),
-        ("grad", ctypes.c_void_p),
-        ("src", ctypes.c_void_p * 8),
-        ("view_src", ctypes.c_void_p),
-        ("view_offs", ctypes.c_size_t),
-        ("data", ctypes.c_void_p),
-        ("name", ctypes.c_char * 64),
-        ("extra", ctypes.c_void_p),
-    ]
-
-
-ggml_tensor_p = ctypes_ext.CtypesPointer[ggml_tensor]
-ggml_tensor_p_ctypes = ctypes.POINTER(ggml_tensor)
-
-
-# GGML_API GGML_CALL void ggml_backend_tensor_get(const struct ggml_tensor * tensor,       void * data, size_t offset, size_t size);
-@ggml_function(
-    "ggml_backend_tensor_get",
-    [ggml_tensor_p_ctypes, ctypes.c_void_p, ctypes.c_size_t, ctypes.c_size_t],
-    ctypes.c_void_p,
-)
-def ggml_backend_tensor_get(
-    tensor: ggml_tensor_p, data: ctypes.c_void_p, offset: int, size: int
-) -> None:
-    ...
-
-
-# GGML_API GGML_CALL size_t  ggml_nbytes      (const struct ggml_tensor * tensor);
-@ggml_function(
-    "ggml_nbytes",
-    [ggml_tensor_p_ctypes],
-    ctypes.c_size_t,
-)
-def ggml_nbytes(tensor: ggml_tensor_p) -> int:
-    ...
-
-
-# GGML_API GGML_CALL int64_t ggml_nelements   (const struct ggml_tensor * tensor);
-@ggml_function(
-    "ggml_nelements",
-    [ggml_tensor_p_ctypes],
-    ctypes.c_int64,
-)
-def ggml_nelements(tensor: ggml_tensor_p) -> int:
-    ...
-
-
-# GGML_API           int  ggml_n_dims       (const struct ggml_tensor * tensor); // returns 1 for scalars
-@ggml_function(
-    "ggml_n_dims",
-    [ggml_tensor_p_ctypes],
-    ctypes.c_int,
-)
-def ggml_n_dims(tensor: ggml_tensor_p) -> int:
-    ...
-
-
-def ggml_tensor_to_numpy(tensor: ggml_tensor_p):
-    nbytes = ggml_nbytes(tensor)
-    nelements = ggml_nelements(tensor)
-    data = np.empty(nelements, dtype=np.float32)
-    ggml_backend_tensor_get(
-        tensor, ctypes.cast(data.ctypes.data, ctypes.c_void_p), 0, nbytes
-    )
-    return data.reshape(tensor.contents.ne[: ggml_n_dims(tensor)])