Repository navigation
Expand file tree
/
Copy pathllava_cpp.py
More file actions
240 lines (193 loc) · 7.29 KB
/
Copy pathllava_cpp.py
File metadata and controls
240 lines (193 loc) · 7.29 KB
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
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
from __future__ import annotations
import sys
import os
import ctypes
import functools
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 (
List,
Union,
NewType,
Optional,
TypeVar,
Callable,
Any,
TYPE_CHECKING,
Generic,
)
from typing_extensions import TypeAlias
import llama_cpp.llama_cpp as llama_cpp
# Load the library
def _load_shared_library(lib_base_name: str):
# Construct the paths to the possible shared library names
_base_path = pathlib.Path(os.path.abspath(os.path.dirname(__file__))) / "lib"
# Searching for the library in the current directory under the name "libllama" (default name
# for llamacpp) and "llama" (default name for this repo)
_lib_paths: List[pathlib.Path] = []
# Determine the file extension based on the platform
if sys.platform.startswith("linux"):
_lib_paths += [
_base_path / f"lib{lib_base_name}.so",
]
elif sys.platform == "darwin":
_lib_paths += [
_base_path / f"lib{lib_base_name}.so",
_base_path / f"lib{lib_base_name}.dylib",
]
elif sys.platform == "win32":
_lib_paths += [
_base_path / f"{lib_base_name}.dll",
_base_path / f"lib{lib_base_name}.dll",
]
else:
raise RuntimeError("Unsupported platform")
if "LLAVA_CPP_LIB" in os.environ:
lib_base_name = os.environ["LLAVA_CPP_LIB"]
_lib = pathlib.Path(lib_base_name)
_base_path = _lib.parent.resolve()
_lib_paths = [_lib.resolve()]
cdll_args = dict() # type: ignore
# Add the library directory to the DLL search path on Windows (if needed)
if sys.platform == "win32" and sys.version_info >= (3, 8):
os.add_dll_directory(str(_base_path))
if "CUDA_PATH" in os.environ:
os.add_dll_directory(os.path.join(os.environ["CUDA_PATH"], "bin"))
os.add_dll_directory(os.path.join(os.environ["CUDA_PATH"], "lib"))
cdll_args["winmode"] = ctypes.RTLD_GLOBAL
# Try to load the shared library, handling potential errors
for _lib_path in _lib_paths:
if _lib_path.exists():
try:
return ctypes.CDLL(str(_lib_path), **cdll_args) # type: ignore
except Exception as e:
raise RuntimeError(f"Failed to load shared library '{_lib_path}': {e}")
raise FileNotFoundError(
f"Shared library with base name '{lib_base_name}' not found"
)
# Specify the base name of the shared library to load
_libllava_base_name = "llava"
# Load the library
_libllava = _load_shared_library(_libllava_base_name)
# ctypes helper
if TYPE_CHECKING:
CtypesCData = TypeVar("CtypesCData", bound=ctypes._CData) # type: ignore
CtypesArray: TypeAlias = ctypes.Array[CtypesCData] # type: ignore
CtypesPointer: TypeAlias = ctypes._Pointer[CtypesCData] # type: ignore
CtypesVoidPointer: TypeAlias = ctypes.c_void_p
class CtypesRef(Generic[CtypesCData]):
pass
CtypesPointerOrRef: TypeAlias = Union[
CtypesPointer[CtypesCData], CtypesRef[CtypesCData]
]
CtypesFuncPointer: TypeAlias = ctypes._FuncPointer # type: ignore
F = TypeVar("F", bound=Callable[..., Any])
def ctypes_function_for_shared_library(lib: ctypes.CDLL):
def ctypes_function(
name: str, argtypes: List[Any], restype: Any, enabled: bool = True
):
def decorator(f: F) -> F:
if enabled:
func = getattr(lib, name)
func.argtypes = argtypes
func.restype = restype
functools.wraps(f)(func)
return func
else:
return f
return decorator
return ctypes_function
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): ...