ayo/ipex-llm
1
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions 1

LLM: fix and update related license in llama pybinding (#8250)

Browse source
This commit is contained in:
binbin Deng 2023-06-01 17:09:15 +08:00 committed by GitHub
parent 141febec1f
commit 3a9aa23835
3 changed files with 212 additions and 96 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

127
python/llm/src/bigdl/llm/ggml/model/llama/llama.py
View file

@ -14,6 +14,33 @@
# limitations under the License. # limitations under the License.
# #
# ===========================================================================
#
# This file is adapted from
# https://github.com/abetlen/llama-cpp-python/blob/main/llama_cpp/llama.py
#
# MIT License
#
# Copyright (c) 2023 Andrei Betlen
#
# Permission is hereby granted, free of charge, to any person obtaining a copy
# of this software and associated documentation files (the "Software"), to deal
# in the Software without restriction, including without limitation the rights
# to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
# copies of the Software, and to permit persons to whom the Software is
# furnished to do so, subject to the following conditions:
#
# The above copyright notice and this permission notice shall be included in all
# copies or substantial portions of the Software.
#
# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
# AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
# OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
# SOFTWARE.
# This would makes sure Python is aware there is more than one sub-package within bigdl, # This would makes sure Python is aware there is more than one sub-package within bigdl,
# physically located elsewhere. # physically located elsewhere.
# Otherwise there would be module not found error in non-pip's setting as Python would # Otherwise there would be module not found error in non-pip's setting as Python would
@ -27,7 +54,7 @@ import math
import multiprocessing import multiprocessing
from typing import List, Optional, Union, Generator, Sequence, Iterator, Deque, Tuple from typing import List, Optional, Union, Generator, Sequence, Iterator, Deque, Tuple
from collections import deque, OrderedDict from collections import deque, OrderedDict
from bigdl.llm.utils.common import invalidInputError
from . import llama_cpp from . import llama_cpp
from .llama_types import * from .llama_types import *
@ -61,8 +88,7 @@ class LlamaCache:
def __getitem__(self, key: Sequence[int]) -> "LlamaState": def __getitem__(self, key: Sequence[int]) -> "LlamaState":
key = tuple(key) key = tuple(key)
_key = self._find_longest_prefix_key(key) _key = self._find_longest_prefix_key(key)
if _key is None: invalidInputError(_key is not None, "Key not found.")
raise KeyError(f"Key not found")
value = self.cache_state[_key] value = self.cache_state[_key]
self.cache_state.move_to_end(_key) self.cache_state.move_to_end(_key)
return value return value
@ -122,7 +148,8 @@ class Llama:
Args: Args:
model_path: Path to the model. model_path: Path to the model.
n_ctx: Maximum context size. n_ctx: Maximum context size.
n_parts: Number of parts to split the model into. If -1, the number of parts is automatically determined. n_parts: Number of parts to split the model into. If -1, the number of parts
is automatically determined.
seed: Random seed. 0 for random. seed: Random seed. 0 for random.
f16_kv: Use half-precision for key/value cache. f16_kv: Use half-precision for key/value cache.
logits_all: Return logits for all tokens, not just the last token. logits_all: Return logits for all tokens, not just the last token.
@ -130,10 +157,12 @@ class Llama:
use_mmap: Use mmap if possible. use_mmap: Use mmap if possible.
use_mlock: Force the system to keep the model in RAM. use_mlock: Force the system to keep the model in RAM.
embedding: Embedding mode only. embedding: Embedding mode only.
n_threads: Number of threads to use. If None, the number of threads is automatically determined. n_threads: Number of threads to use. If None, the number of threads is
automatically determined.
n_batch: Maximum number of prompt tokens to batch together when calling llama_eval. n_batch: Maximum number of prompt tokens to batch together when calling llama_eval.
last_n_tokens_size: Maximum number of tokens to keep in the last_n_tokens deque. last_n_tokens_size: Maximum number of tokens to keep in the last_n_tokens deque.
lora_base: Optional path to base model, useful if using a quantized base model and you want to apply LoRA to an f16 model. lora_base: Optional path to base model, useful if using a quantized base model and
you want to apply LoRA to an f16 model.
lora_path: Path to a LoRA file to apply to the model. lora_path: Path to a LoRA file to apply to the model.
verbose: Print verbose output to stderr. verbose: Print verbose output to stderr.
@ -169,18 +198,17 @@ class Llama:
self.lora_base = lora_base self.lora_base = lora_base
self.lora_path = lora_path self.lora_path = lora_path
### DEPRECATED ### # DEPRECATED
self.n_parts = n_parts self.n_parts = n_parts
### DEPRECATED ### # DEPRECATED
if not os.path.exists(model_path): invalidInputError(os.path.exists(model_path), f"Model path does not exist: {model_path}.")
raise ValueError(f"Model path does not exist: {model_path}")
self.ctx = llama_cpp.llama_init_from_file( self.ctx = llama_cpp.llama_init_from_file(
self.model_path.encode("utf-8"), self.params self.model_path.encode("utf-8"), self.params
) )
assert self.ctx is not None invalidInputError(self.ctx is not None, "The attribute `ctx` of `Llama` object is None.")
if self.lora_path: if self.lora_path:
if llama_cpp.llama_apply_lora_from_file( if llama_cpp.llama_apply_lora_from_file(
@ -191,9 +219,9 @@ class Llama:
else llama_cpp.c_char_p(0), else llama_cpp.c_char_p(0),
llama_cpp.c_int(self.n_threads), llama_cpp.c_int(self.n_threads),
): ):
raise RuntimeError( invalidInputError(False,
f"Failed to apply LoRA from lora path: {self.lora_path} to base path: {self.lora_base}" "Failed to apply LoRA from lora path: "
) f"{self.lora_path} to base path: {self.lora_base}")
if self.verbose: if self.verbose:
print(llama_cpp.llama_print_system_info().decode("utf-8"), file=sys.stderr) print(llama_cpp.llama_print_system_info().decode("utf-8"), file=sys.stderr)
@ -233,7 +261,7 @@ class Llama:
Returns: Returns:
A list of tokens. A list of tokens.
""" """
assert self.ctx is not None invalidInputError(self.ctx is not None, "The attribute `ctx` of `Llama` object is None.")
n_ctx = llama_cpp.llama_n_ctx(self.ctx) n_ctx = llama_cpp.llama_n_ctx(self.ctx)
tokens = (llama_cpp.llama_token * int(n_ctx))() tokens = (llama_cpp.llama_token * int(n_ctx))()
n_tokens = llama_cpp.llama_tokenize( n_tokens = llama_cpp.llama_tokenize(
@ -253,10 +281,8 @@ class Llama:
llama_cpp.c_int(n_tokens), llama_cpp.c_int(n_tokens),
llama_cpp.c_bool(add_bos), llama_cpp.c_bool(add_bos),
) )
if n_tokens < 0: invalidInputError(n_tokens >= 0,
raise RuntimeError( f'Failed to tokenize: text="{text}" n_tokens={n_tokens}')
f'Failed to tokenize: text="{text}" n_tokens={n_tokens}'
)
return list(tokens[:n_tokens]) return list(tokens[:n_tokens])
def detokenize(self, tokens: List[int]) -> bytes: def detokenize(self, tokens: List[int]) -> bytes:
@ -268,7 +294,7 @@ class Llama:
Returns: Returns:
The detokenized string. The detokenized string.
""" """
assert self.ctx is not None invalidInputError(self.ctx is not None, "The attribute `ctx` of `Llama` object is None.")
output = b"" output = b""
for token in tokens: for token in tokens:
output += llama_cpp.llama_token_to_str( output += llama_cpp.llama_token_to_str(
@ -295,7 +321,7 @@ class Llama:
Args: Args:
tokens: The list of tokens to evaluate. tokens: The list of tokens to evaluate.
""" """
assert self.ctx is not None invalidInputError(self.ctx is not None, "The attribute `ctx` of `Llama` object is None.")
n_ctx = int(llama_cpp.llama_n_ctx(self.ctx)) n_ctx = int(llama_cpp.llama_n_ctx(self.ctx))
for i in range(0, len(tokens), self.n_batch): for i in range(0, len(tokens), self.n_batch):
batch = tokens[i: min(len(tokens), i + self.n_batch)] batch = tokens[i: min(len(tokens), i + self.n_batch)]
@ -308,8 +334,7 @@ class Llama:
n_past=llama_cpp.c_int(n_past), n_past=llama_cpp.c_int(n_past),
n_threads=llama_cpp.c_int(self.n_threads), n_threads=llama_cpp.c_int(self.n_threads),
) )
if int(return_code) != 0: invalidInputError(int(return_code) == 0, f"llama_eval returned {return_code}.")
raise RuntimeError(f"llama_eval returned {return_code}")
# Save tokens # Save tokens
self.eval_tokens.extend(batch) self.eval_tokens.extend(batch)
# Save logits # Save logits
@ -338,8 +363,9 @@ class Llama:
mirostat_eta: llama_cpp.c_float, mirostat_eta: llama_cpp.c_float,
penalize_nl: bool = True, penalize_nl: bool = True,
): ):
assert self.ctx is not None invalidInputError(self.ctx is not None, "The attribute `ctx` of `Llama` object is None.")
assert len(self.eval_logits) > 0 invalidInputError(len(self.eval_logits) > 0,
"The attribute `eval_logits` of `Llama` object is None.")
n_vocab = self.n_vocab() n_vocab = self.n_vocab()
n_ctx = self.n_ctx() n_ctx = self.n_ctx()
top_k = llama_cpp.c_int(n_vocab) if top_k.value <= 0 else top_k top_k = llama_cpp.c_int(n_vocab) if top_k.value <= 0 else top_k
@ -467,7 +493,7 @@ class Llama:
Returns: Returns:
The sampled token. The sampled token.
""" """
assert self.ctx is not None invalidInputError(self.ctx is not None, "The attribute `ctx` of `Llama` object is None.")
last_n_tokens_data = [llama_cpp.llama_token(0)] * max( last_n_tokens_data = [llama_cpp.llama_token(0)] * max(
0, self.last_n_tokens_size - len(self.eval_tokens) 0, self.last_n_tokens_size - len(self.eval_tokens)
) + list(self.eval_tokens)[-self.last_n_tokens_size:] ) + list(self.eval_tokens)[-self.last_n_tokens_size:]
@ -509,7 +535,8 @@ class Llama:
Examples: Examples:
>>> llama = Llama("models/ggml-7b.bin") >>> llama = Llama("models/ggml-7b.bin")
>>> tokens = llama.tokenize(b"Hello, world!") >>> tokens = llama.tokenize(b"Hello, world!")
>>> for token in llama.generate(tokens, top_k=40, top_p=0.95, temp=1.0, repeat_penalty=1.1): >>> for token in llama.generate(tokens, top_k=40, top_p=0.95,
>>> temp=1.0, repeat_penalty=1.1):
... print(llama.detokenize([token])) ... print(llama.detokenize([token]))
Args: Args:
@ -523,7 +550,7 @@ class Llama:
Yields: Yields:
The generated tokens. The generated tokens.
""" """
assert self.ctx is not None invalidInputError(self.ctx is not None, "The attribute `ctx` of `Llama` object is None.")
if reset and len(self.eval_tokens) > 0: if reset and len(self.eval_tokens) > 0:
longest_prefix = 0 longest_prefix = 0
@ -577,13 +604,11 @@ class Llama:
Returns: Returns:
An embedding object. An embedding object.
""" """
assert self.ctx is not None invalidInputError(self.ctx is not None, "The attribute `ctx` of `Llama` object is None.")
model_name: str = model if model is not None else self.model_path model_name: str = model if model is not None else self.model_path
if self.params.embedding == False: invalidInputError(self.params.embedding,
raise RuntimeError( "Llama model must be created with embedding=True to call this method.")
"Llama model must be created with embedding=True to call this method"
)
if self.verbose: if self.verbose:
llama_cpp.llama_reset_timings(self.ctx) llama_cpp.llama_reset_timings(self.ctx)
@ -657,7 +682,7 @@ class Llama:
mirostat_eta: float = 0.1, mirostat_eta: float = 0.1,
model: Optional[str] = None, model: Optional[str] = None,
) -> Union[Iterator[Completion], Iterator[CompletionChunk]]: ) -> Union[Iterator[Completion], Iterator[CompletionChunk]]:
assert self.ctx is not None invalidInputError(self.ctx is not None, "The attribute `ctx` of `Llama` object is None.")
completion_id: str = f"cmpl-{str(uuid.uuid4())}" completion_id: str = f"cmpl-{str(uuid.uuid4())}"
created: int = int(time.time()) created: int = int(time.time())
completion_tokens: List[int] = [] completion_tokens: List[int] = []
@ -673,10 +698,9 @@ class Llama:
if self.verbose: if self.verbose:
llama_cpp.llama_reset_timings(self.ctx) llama_cpp.llama_reset_timings(self.ctx)
if len(prompt_tokens) + max_tokens > int(llama_cpp.llama_n_ctx(self.ctx)): invalidInputError(len(prompt_tokens) + max_tokens <= int(llama_cpp.llama_n_ctx(self.ctx)),
raise ValueError( "Requested tokens exceed context window of "
f"Requested tokens exceed context window of {llama_cpp.llama_n_ctx(self.ctx)}" f"{llama_cpp.llama_n_ctx(self.ctx)}.")
)
if stop != []: if stop != []:
stop_sequences = [s.encode("utf-8") for s in stop] stop_sequences = [s.encode("utf-8") for s in stop]
@ -684,9 +708,8 @@ class Llama:
stop_sequences = [] stop_sequences = []
if logprobs is not None and self.params.logits_all is False: if logprobs is not None and self.params.logits_all is False:
raise ValueError( invalidInputError(False,
"logprobs is not supported for models created with logits_all=False" "logprobs is not supported for models created with logits_all=False")
)
if self.cache: if self.cache:
try: try:
@ -1294,9 +1317,9 @@ class Llama:
n_threads=self.n_threads, n_threads=self.n_threads,
lora_base=self.lora_base, lora_base=self.lora_base,
lora_path=self.lora_path, lora_path=self.lora_path,
### DEPRECATED ### # DEPRECATED
n_parts=self.n_parts, n_parts=self.n_parts,
### DEPRECATED ### # DEPRECATED
) )
def __setstate__(self, state): def __setstate__(self, state):
@ -1321,12 +1344,11 @@ class Llama:
) )
def save_state(self) -> LlamaState: def save_state(self) -> LlamaState:
assert self.ctx is not None invalidInputError(self.ctx is not None, "The attribute `ctx` of `Llama` object is None.")
state_size = llama_cpp.llama_get_state_size(self.ctx) state_size = llama_cpp.llama_get_state_size(self.ctx)
llama_state = (llama_cpp.c_uint8 * int(state_size))() llama_state = (llama_cpp.c_uint8 * int(state_size))()
n_bytes = llama_cpp.llama_copy_state_data(self.ctx, llama_state) n_bytes = llama_cpp.llama_copy_state_data(self.ctx, llama_state)
if int(n_bytes) > int(state_size): invalidInputError(int(n_bytes) <= int(state_size), "Failed to copy llama state data.")
raise RuntimeError("Failed to copy llama state data")
llama_state_compact = (llama_cpp.c_uint8 * int(n_bytes))() llama_state_compact = (llama_cpp.c_uint8 * int(n_bytes))()
llama_cpp.ctypes.memmove(llama_state_compact, llama_state, int(n_bytes)) llama_cpp.ctypes.memmove(llama_state_compact, llama_state, int(n_bytes))
if self.verbose: if self.verbose:
@ -1342,26 +1364,27 @@ class Llama:
) )
def load_state(self, state: LlamaState) -> None: def load_state(self, state: LlamaState) -> None:
assert self.ctx is not None invalidInputError(self.ctx is not None, "The attribute `ctx` of `Llama` object is None.")
self.eval_tokens = state.eval_tokens.copy() self.eval_tokens = state.eval_tokens.copy()
self.eval_logits = state.eval_logits.copy() self.eval_logits = state.eval_logits.copy()
state_size = state.llama_state_size state_size = state.llama_state_size
if llama_cpp.llama_set_state_data(self.ctx, state.llama_state) != state_size: invalidInputError(llama_cpp.llama_set_state_data(self.ctx,
raise RuntimeError("Failed to set llama state data") state.llama_state) == state_size,
"Failed to set llama state data.")
def n_ctx(self) -> int: def n_ctx(self) -> int:
"""Return the context window size.""" """Return the context window size."""
assert self.ctx is not None invalidInputError(self.ctx is not None, "The attribute `ctx` of `Llama` object is None.")
return llama_cpp.llama_n_ctx(self.ctx) return llama_cpp.llama_n_ctx(self.ctx)
def n_embd(self) -> int: def n_embd(self) -> int:
"""Return the embedding size.""" """Return the embedding size."""
assert self.ctx is not None invalidInputError(self.ctx is not None, "The attribute `ctx` of `Llama` object is None.")
return llama_cpp.llama_n_embd(self.ctx) return llama_cpp.llama_n_embd(self.ctx)
def n_vocab(self) -> int: def n_vocab(self) -> int:
"""Return the vocabulary size.""" """Return the vocabulary size."""
assert self.ctx is not None invalidInputError(self.ctx is not None, "The attribute `ctx` of `Llama` object is None.")
return llama_cpp.llama_n_vocab(self.ctx) return llama_cpp.llama_n_vocab(self.ctx)
@staticmethod @staticmethod

143
python/llm/src/bigdl/llm/ggml/model/llama/llama_cpp.py
View file

@ -13,6 +13,32 @@
# See the License for the specific language governing permissions and # See the License for the specific language governing permissions and
# limitations under the License. # limitations under the License.
# #
# ===========================================================================
#
# This file is adapted from
# https://github.com/abetlen/llama-cpp-python/blob/main/llama_cpp/llama_cpp.py
#
# MIT License
#
# Copyright (c) 2023 Andrei Betlen
#
# Permission is hereby granted, free of charge, to any person obtaining a copy
# of this software and associated documentation files (the "Software"), to deal
# in the Software without restriction, including without limitation the rights
# to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
# copies of the Software, and to permit persons to whom the Software is
# furnished to do so, subject to the following conditions:
#
# The above copyright notice and this permission notice shall be included in all
# copies or substantial portions of the Software.
#
# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
# AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
# OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
# SOFTWARE.
# This would makes sure Python is aware there is more than one sub-package within bigdl, # This would makes sure Python is aware there is more than one sub-package within bigdl,
# physically located elsewhere. # physically located elsewhere.
@ -36,6 +62,7 @@ from ctypes import (
c_size_t, c_size_t,
) )
import pathlib import pathlib
from bigdl.llm.utils.common import invalidInputError
# Load the library # Load the library
@ -48,7 +75,7 @@ def _load_shared_library(lib_base_name: str):
elif sys.platform == "win32": elif sys.platform == "win32":
lib_ext = ".dll" lib_ext = ".dll"
else: else:
raise RuntimeError("Unsupported platform") invalidInputError(False, "Unsupported platform.")
# Construct the paths to the possible shared library names # Construct the paths to the possible shared library names
_base_path = pathlib.Path(__file__).parent.parent.parent.parent.resolve() _base_path = pathlib.Path(__file__).parent.parent.parent.parent.resolve()
@ -81,11 +108,9 @@ def _load_shared_library(lib_base_name: str):
try: try:
return ctypes.CDLL(str(_lib_path), **cdll_args) return ctypes.CDLL(str(_lib_path), **cdll_args)
except Exception as e: except Exception as e:
raise RuntimeError(f"Failed to load shared library '{_lib_path}': {e}") invalidInputError(False, f"Failed to load shared library '{_lib_path}': {e}.")
raise FileNotFoundError( invalidInputError(False, f"Shared library with base name '{lib_base_name}' not found.")
f"Shared library with base name '{lib_base_name}' not found"
)
# Specify the base name of the shared library to load # Specify the base name of the shared library to load
@ -300,7 +325,8 @@ _lib.llama_free.restype = None
# TODO: not great API - very likely to change # TODO: not great API - very likely to change
# Returns 0 on success # Returns 0 on success
# nthread - how many threads to use. If <=0, will use std::thread::hardware_concurrency(), else the number given # nthread - how many threads to use. If <=0, will use std::thread::hardware_concurrency(),
# else the number given
# LLAMA_API int llama_model_quantize( # LLAMA_API int llama_model_quantize(
# const char * fname_inp, # const char * fname_inp,
# const char * fname_out, # const char * fname_out,
@ -399,7 +425,8 @@ _lib.llama_set_state_data.restype = c_size_t
# Save/load session file # Save/load session file
# LLAMA_API bool llama_load_session_file(struct llama_context * ctx, const char * path_session, llama_token * tokens_out, size_t n_token_capacity, size_t * n_token_count_out); # LLAMA_API bool llama_load_session_file(struct llama_context * ctx, const char * path_session,
# llama_token * tokens_out, size_t n_token_capacity, size_t * n_token_count_out);
def llama_load_session_file( def llama_load_session_file(
ctx: llama_context_p, ctx: llama_context_p,
path_session: bytes, path_session: bytes,
@ -422,7 +449,8 @@ _lib.llama_load_session_file.argtypes = [
_lib.llama_load_session_file.restype = c_size_t _lib.llama_load_session_file.restype = c_size_t
# LLAMA_API bool llama_save_session_file(struct llama_context * ctx, const char * path_session, const llama_token * tokens, size_t n_token_count); # LLAMA_API bool llama_save_session_file(struct llama_context * ctx, const char * path_session,
# const llama_token * tokens, size_t n_token_count);
def llama_save_session_file( def llama_save_session_file(
ctx: llama_context_p, ctx: llama_context_p,
path_session: bytes, path_session: bytes,
@ -601,8 +629,10 @@ _lib.llama_init_candidates.argtypes = [
_lib.llama_init_candidates.restype = None _lib.llama_init_candidates.restype = None
# @details Repetition penalty described in CTRL academic paper https://arxiv.org/abs/1909.05858, with negative logit fix. # @details Repetition penalty described in CTRL academic paper https://arxiv.org/abs/1909.05858,
# LLAMA_API void llama_sample_repetition_penalty(struct llama_context * ctx, llama_token_data_array * candidates, const llama_token * last_tokens, size_t last_tokens_size, float penalty); # with negative logit fix.
# LLAMA_API void llama_sample_repetition_penalty(struct llama_context * ctx, llama_token_data_array
# * candidates, const llama_token * last_tokens, size_t last_tokens_size, float penalty);
def llama_sample_repetition_penalty( def llama_sample_repetition_penalty(
ctx: llama_context_p, ctx: llama_context_p,
candidates, # type: _Pointer[llama_token_data_array] candidates, # type: _Pointer[llama_token_data_array]
@ -625,8 +655,11 @@ _lib.llama_sample_repetition_penalty.argtypes = [
_lib.llama_sample_repetition_penalty.restype = None _lib.llama_sample_repetition_penalty.restype = None
# @details Frequency and presence penalties described in OpenAI API https://platform.openai.com/docs/api-reference/parameter-details. # @details Frequency and presence penalties described in OpenAI API
# LLAMA_API void llama_sample_frequency_and_presence_penalties(struct llama_context * ctx, llama_token_data_array * candidates, const llama_token * last_tokens, size_t last_tokens_size, float alpha_frequency, float alpha_presence); # https://platform.openai.com/docs/api-reference/parameter-details.
# LLAMA_API void llama_sample_frequency_and_presence_penalties(struct llama_context * ctx,
# llama_token_data_array * candidates, const llama_token * last_tokens,
# size_t last_tokens_size, float alpha_frequency, float alpha_presence);
def llama_sample_frequency_and_presence_penalties( def llama_sample_frequency_and_presence_penalties(
ctx: llama_context_p, ctx: llama_context_p,
candidates, # type: _Pointer[llama_token_data_array] candidates, # type: _Pointer[llama_token_data_array]
@ -656,8 +689,10 @@ _lib.llama_sample_frequency_and_presence_penalties.argtypes = [
_lib.llama_sample_frequency_and_presence_penalties.restype = None _lib.llama_sample_frequency_and_presence_penalties.restype = None
# @details Sorts candidate tokens by their logits in descending order and calculate probabilities based on logits. # @details Sorts candidate tokens by their logits in descending order and calculate probabilities
# LLAMA_API void llama_sample_softmax(struct llama_context * ctx, llama_token_data_array * candidates); # based on logits.
# LLAMA_API void llama_sample_softmax(struct llama_context * ctx,
# llama_token_data_array * candidates);
def llama_sample_softmax( def llama_sample_softmax(
ctx: llama_context_p, candidates # type: _Pointer[llama_token_data] ctx: llama_context_p, candidates # type: _Pointer[llama_token_data]
): ):
@ -671,8 +706,10 @@ _lib.llama_sample_softmax.argtypes = [
_lib.llama_sample_softmax.restype = None _lib.llama_sample_softmax.restype = None
# @details Top-K sampling described in academic paper "The Curious Case of Neural Text Degeneration" https://arxiv.org/abs/1904.09751 # @details Top-K sampling described in academic paper
# LLAMA_API void llama_sample_top_k(struct llama_context * ctx, llama_token_data_array * candidates, int k, size_t min_keep); # "The Curious Case of Neural Text Degeneration" https://arxiv.org/abs/1904.09751
# LLAMA_API void llama_sample_top_k(struct llama_context * ctx,
# llama_token_data_array * candidates, int k, size_t min_keep);
def llama_sample_top_k( def llama_sample_top_k(
ctx: llama_context_p, ctx: llama_context_p,
candidates, # type: _Pointer[llama_token_data_array] candidates, # type: _Pointer[llama_token_data_array]
@ -691,8 +728,10 @@ _lib.llama_sample_top_k.argtypes = [
_lib.llama_sample_top_k.restype = None _lib.llama_sample_top_k.restype = None
# @details Nucleus sampling described in academic paper "The Curious Case of Neural Text Degeneration" https://arxiv.org/abs/1904.09751 # @details Nucleus sampling described in academic paper
# LLAMA_API void llama_sample_top_p(struct llama_context * ctx, llama_token_data_array * candidates, float p, size_t min_keep); # "The Curious Case of Neural Text Degeneration" https://arxiv.org/abs/1904.09751
# LLAMA_API void llama_sample_top_p(struct llama_context * ctx,
# llama_token_data_array * candidates, float p, size_t min_keep);
def llama_sample_top_p( def llama_sample_top_p(
ctx: llama_context_p, ctx: llama_context_p,
candidates, # type: _Pointer[llama_token_data_array] candidates, # type: _Pointer[llama_token_data_array]
@ -711,8 +750,10 @@ _lib.llama_sample_top_p.argtypes = [
_lib.llama_sample_top_p.restype = None _lib.llama_sample_top_p.restype = None
# @details Tail Free Sampling described in https://www.trentonbricken.com/Tail-Free-Sampling/. # @details Tail Free Sampling described in
# LLAMA_API void llama_sample_tail_free(struct llama_context * ctx, llama_token_data_array * candidates, float z, size_t min_keep); # https://www.trentonbricken.com/Tail-Free-Sampling/.
# LLAMA_API void llama_sample_tail_free(struct llama_context * ctx,
# llama_token_data_array * candidates, float z, size_t min_keep);
def llama_sample_tail_free( def llama_sample_tail_free(
ctx: llama_context_p, ctx: llama_context_p,
candidates, # type: _Pointer[llama_token_data_array] candidates, # type: _Pointer[llama_token_data_array]
@ -731,8 +772,10 @@ _lib.llama_sample_tail_free.argtypes = [
_lib.llama_sample_tail_free.restype = None _lib.llama_sample_tail_free.restype = None
# @details Locally Typical Sampling implementation described in the paper https://arxiv.org/abs/2202.00666. # @details Locally Typical Sampling implementation described in the paper
# LLAMA_API void llama_sample_typical(struct llama_context * ctx, llama_token_data_array * candidates, float p, size_t min_keep); # https://arxiv.org/abs/2202.00666.
# LLAMA_API void llama_sample_typical(struct llama_context * ctx,
# llama_token_data_array * candidates, float p, size_t min_keep);
def llama_sample_typical( def llama_sample_typical(
ctx: llama_context_p, ctx: llama_context_p,
candidates, # type: _Pointer[llama_token_data_array] candidates, # type: _Pointer[llama_token_data_array]
@ -751,7 +794,8 @@ _lib.llama_sample_typical.argtypes = [
_lib.llama_sample_typical.restype = None _lib.llama_sample_typical.restype = None
# LLAMA_API void llama_sample_temperature(struct llama_context * ctx, llama_token_data_array * candidates, float temp); # LLAMA_API void llama_sample_temperature(struct llama_context * ctx,
# llama_token_data_array * candidates, float temp);
def llama_sample_temperature( def llama_sample_temperature(
ctx: llama_context_p, ctx: llama_context_p,
candidates, # type: _Pointer[llama_token_data_array] candidates, # type: _Pointer[llama_token_data_array]
@ -768,13 +812,25 @@ _lib.llama_sample_temperature.argtypes = [
_lib.llama_sample_temperature.restype = None _lib.llama_sample_temperature.restype = None
# @details Mirostat 1.0 algorithm described in the paper https://arxiv.org/abs/2007.14966. Uses tokens instead of words. # @details Mirostat 1.0 algorithm described in the paper https://arxiv.org/abs/2007.14966.
# @param candidates A vector of `llama_token_data` containing the candidate tokens, their probabilities (p), and log-odds (logit) for the current position in the generated text. # Uses tokens instead of words.
# @param tau The target cross-entropy (or surprise) value you want to achieve for the generated text. A higher value corresponds to more surprising or less predictable text, while a lower value corresponds to less surprising or more predictable text. # @param candidates A vector of `llama_token_data` containing the candidate tokens,
# @param eta The learning rate used to update `mu` based on the error between the target and observed surprisal of the sampled word. A larger learning rate will cause `mu` to be updated more quickly, while a smaller learning rate will result in slower updates. # their probabilities (p), and log-odds (logit) for the current position in the generated text.
# @param m The number of tokens considered in the estimation of `s_hat`. This is an arbitrary value that is used to calculate `s_hat`, which in turn helps to calculate the value of `k`. In the paper, they use `m = 100`, but you can experiment with different values to see how it affects the performance of the algorithm. # @param tau The target cross-entropy (or surprise) value you want to achieve for the generated
# @param mu Maximum cross-entropy. This value is initialized to be twice the target cross-entropy (`2 * tau`) and is updated in the algorithm based on the error between the target and observed surprisal. # text. A higher value corresponds to more surprising or less predictable text, while a lower
# LLAMA_API llama_token llama_sample_token_mirostat(struct llama_context * ctx, llama_token_data_array * candidates, float tau, float eta, int m, float * mu); # value corresponds to less surprising or more predictable text.
# @param eta The learning rate used to update `mu` based on the error between the target and
# observed surprisal of the sampled word. A larger learning rate will cause `mu` to be
# updated more quickly, while a smaller learning rate will result in slower updates.
# @param m The number of tokens considered in the estimation of `s_hat`. This is an arbitrary value
# that is used to calculate `s_hat`, which in turn helps to calculate the value of `k`.
# In the paper, they use `m = 100`, but you can experiment with different values to see
# how it affects the performance of the algorithm.
# @param mu Maximum cross-entropy. This value is initialized to be twice the target cross-entropy
# (`2 * tau`) and is updated in the algorithm based on the error between the target and
# observed surprisal.
# LLAMA_API llama_token llama_sample_token_mirostat(struct llama_context * ctx,
# llama_token_data_array * candidates, float tau, float eta, int m, float * mu);
def llama_sample_token_mirostat( def llama_sample_token_mirostat(
ctx: llama_context_p, ctx: llama_context_p,
candidates, # type: _Pointer[llama_token_data_array] candidates, # type: _Pointer[llama_token_data_array]
@ -797,12 +853,21 @@ _lib.llama_sample_token_mirostat.argtypes = [
_lib.llama_sample_token_mirostat.restype = llama_token _lib.llama_sample_token_mirostat.restype = llama_token
# @details Mirostat 2.0 algorithm described in the paper https://arxiv.org/abs/2007.14966. Uses tokens instead of words. # @details Mirostat 2.0 algorithm described in the paper https://arxiv.org/abs/2007.14966.
# @param candidates A vector of `llama_token_data` containing the candidate tokens, their probabilities (p), and log-odds (logit) for the current position in the generated text. # Uses tokens instead of words.
# @param tau The target cross-entropy (or surprise) value you want to achieve for the generated text. A higher value corresponds to more surprising or less predictable text, while a lower value corresponds to less surprising or more predictable text. # @param candidates A vector of `llama_token_data` containing the candidate tokens,
# @param eta The learning rate used to update `mu` based on the error between the target and observed surprisal of the sampled word. A larger learning rate will cause `mu` to be updated more quickly, while a smaller learning rate will result in slower updates. # their probabilities (p), and log-odds (logit) for the current position in the generated text.
# @param mu Maximum cross-entropy. This value is initialized to be twice the target cross-entropy (`2 * tau`) and is updated in the algorithm based on the error between the target and observed surprisal. # @param tau The target cross-entropy (or surprise) value you want to achieve for the generated
# LLAMA_API llama_token llama_sample_token_mirostat_v2(struct llama_context * ctx, llama_token_data_array * candidates, float tau, float eta, float * mu); # text. A higher value corresponds to more surprising or less predictable text, while a lower value
# corresponds to less surprising or more predictable text.
# @param eta The learning rate used to update `mu` based on the error between the target and
# observed surprisal of the sampled word. A larger learning rate will cause `mu` to be
# updated more quickly, while a smaller learning rate will result in slower updates.
# @param mu Maximum cross-entropy. This value is initialized to be twice the target cross-entropy
# (`2 * tau`) and is updated in the algorithm based on the error between the target
# and observed surprisal.
# LLAMA_API llama_token llama_sample_token_mirostat_v2(struct llama_context * ctx,
# llama_token_data_array * candidates, float tau, float eta, float * mu);
def llama_sample_token_mirostat_v2( def llama_sample_token_mirostat_v2(
ctx: llama_context_p, ctx: llama_context_p,
candidates, # type: _Pointer[llama_token_data_array] candidates, # type: _Pointer[llama_token_data_array]
@ -824,7 +889,8 @@ _lib.llama_sample_token_mirostat_v2.restype = llama_token
# @details Selects the token with the highest probability. # @details Selects the token with the highest probability.
# LLAMA_API llama_token llama_sample_token_greedy(struct llama_context * ctx, llama_token_data_array * candidates); # LLAMA_API llama_token llama_sample_token_greedy(struct llama_context * ctx,
# llama_token_data_array * candidates);
def llama_sample_token_greedy( def llama_sample_token_greedy(
ctx: llama_context_p, ctx: llama_context_p,
candidates, # type: _Pointer[llama_token_data_array] candidates, # type: _Pointer[llama_token_data_array]
@ -840,7 +906,8 @@ _lib.llama_sample_token_greedy.restype = llama_token
# @details Randomly selects a token from the candidates based on their probabilities. # @details Randomly selects a token from the candidates based on their probabilities.
# LLAMA_API llama_token llama_sample_token(struct llama_context * ctx, llama_token_data_array * candidates); # LLAMA_API llama_token llama_sample_token(struct llama_context * ctx,
# llama_token_data_array * candidates);
def llama_sample_token( def llama_sample_token(
ctx: llama_context_p, ctx: llama_context_p,
candidates, # type: _Pointer[llama_token_data_array] candidates, # type: _Pointer[llama_token_data_array]

26
python/llm/src/bigdl/llm/ggml/model/llama/llama_types.py
View file

@ -13,6 +13,32 @@
# See the License for the specific language governing permissions and # See the License for the specific language governing permissions and
# limitations under the License. # limitations under the License.
# #
# ===========================================================================
#
# This file is adapted from
# https://github.com/abetlen/llama-cpp-python/blob/main/llama_cpp/llama_types.py
#
# MIT License
#
# Copyright (c) 2023 Andrei Betlen
#
# Permission is hereby granted, free of charge, to any person obtaining a copy
# of this software and associated documentation files (the "Software"), to deal
# in the Software without restriction, including without limitation the rights
# to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
# copies of the Software, and to permit persons to whom the Software is
# furnished to do so, subject to the following conditions:
#
# The above copyright notice and this permission notice shall be included in all
# copies or substantial portions of the Software.
#
# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
# AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
# OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
# SOFTWARE.
# This would makes sure Python is aware there is more than one sub-package within bigdl, # This would makes sure Python is aware there is more than one sub-package within bigdl,
# physically located elsewhere. # physically located elsewhere.