File size: 4,552 Bytes
5d62acd |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 |
from __future__ import annotations
import os
from ctypes import (
c_bool,
c_char_p,
c_int,
c_uint8,
c_float,
c_void_p,
POINTER,
_Pointer, # type: ignore
Structure,
)
import pathlib
from typing import (
Union,
NewType,
Optional,
TYPE_CHECKING,
)
import llama_cpp.llama_cpp as llama_cpp
from llama_cpp._ctypes_extensions import (
load_shared_library,
ctypes_function_for_shared_library,
)
if TYPE_CHECKING:
from llama_cpp._ctypes_extensions import (
CtypesArray,
)
# Specify the base name of the shared library to load
_libllava_base_name = "llava"
_libllava_override_path = os.environ.get("LLAVA_CPP_LIB")
_libllava_base_path = pathlib.Path(os.path.abspath(os.path.dirname(__file__))) / "lib" if _libllava_override_path is None else pathlib.Path()
# Load the library
_libllava = load_shared_library(_libllava_base_name, _libllava_base_path)
ctypes_function = ctypes_function_for_shared_library(_libllava)
################################################
# llava.h
################################################
# struct clip_ctx;
clip_ctx_p = NewType("clip_ctx_p", int)
clip_ctx_p_ctypes = c_void_p
# struct llava_image_embed {
# float * embed;
# int n_image_pos;
# };
class llava_image_embed(Structure):
_fields_ = [
("embed", POINTER(c_float)),
("n_image_pos", c_int),
]
# /** sanity check for clip <-> llava embed size match */
# LLAVA_API bool llava_validate_embed_size(const llama_context * ctx_llama, const clip_ctx * ctx_clip);
@ctypes_function(
"llava_validate_embed_size",
[llama_cpp.llama_context_p_ctypes, clip_ctx_p_ctypes],
c_bool,
)
def llava_validate_embed_size(
ctx_llama: llama_cpp.llama_context_p, ctx_clip: clip_ctx_p, /
) -> bool:
...
# /** build an image embed from image file bytes */
# LLAVA_API struct llava_image_embed * llava_image_embed_make_with_bytes(struct clip_ctx * ctx_clip, int n_threads, const unsigned char * image_bytes, int image_bytes_length);
@ctypes_function(
"llava_image_embed_make_with_bytes",
[clip_ctx_p_ctypes, c_int, POINTER(c_uint8), c_int],
POINTER(llava_image_embed),
)
def llava_image_embed_make_with_bytes(
ctx_clip: clip_ctx_p,
n_threads: Union[c_int, int],
image_bytes: CtypesArray[c_uint8],
image_bytes_length: Union[c_int, int],
/,
) -> "_Pointer[llava_image_embed]":
...
# /** build an image embed from a path to an image filename */
# LLAVA_API struct llava_image_embed * llava_image_embed_make_with_filename(struct clip_ctx * ctx_clip, int n_threads, const char * image_path);
@ctypes_function(
"llava_image_embed_make_with_filename",
[clip_ctx_p_ctypes, c_int, c_char_p],
POINTER(llava_image_embed),
)
def llava_image_embed_make_with_filename(
ctx_clip: clip_ctx_p, n_threads: Union[c_int, int], image_path: bytes, /
) -> "_Pointer[llava_image_embed]":
...
# LLAVA_API void llava_image_embed_free(struct llava_image_embed * embed);
# /** free an embedding made with llava_image_embed_make_* */
@ctypes_function("llava_image_embed_free", [POINTER(llava_image_embed)], None)
def llava_image_embed_free(embed: "_Pointer[llava_image_embed]", /):
...
# /** write the image represented by embed into the llama context with batch size n_batch, starting at context pos n_past. on completion, n_past points to the next position in the context after the image embed. */
# LLAVA_API bool llava_eval_image_embed(struct llama_context * ctx_llama, const struct llava_image_embed * embed, int n_batch, int * n_past);
@ctypes_function(
"llava_eval_image_embed",
[
llama_cpp.llama_context_p_ctypes,
POINTER(llava_image_embed),
c_int,
POINTER(c_int),
],
c_bool,
)
def llava_eval_image_embed(
ctx_llama: llama_cpp.llama_context_p,
embed: "_Pointer[llava_image_embed]",
n_batch: Union[c_int, int],
n_past: "_Pointer[c_int]",
/,
) -> bool:
...
################################################
# clip.h
################################################
# /** load mmproj model */
# CLIP_API struct clip_ctx * clip_model_load (const char * fname, int verbosity);
@ctypes_function("clip_model_load", [c_char_p, c_int], clip_ctx_p_ctypes)
def clip_model_load(
fname: bytes, verbosity: Union[c_int, int], /
) -> Optional[clip_ctx_p]:
...
# /** free mmproj model */
# CLIP_API void clip_free(struct clip_ctx * ctx);
@ctypes_function("clip_free", [clip_ctx_p_ctypes], None)
def clip_free(ctx: clip_ctx_p, /):
...
|