256 lines
6.6 KiB
Python
256 lines
6.6 KiB
Python
import ctypes
|
|
|
|
from ctypes import (
|
|
c_int,
|
|
c_float,
|
|
c_double,
|
|
c_char_p,
|
|
c_void_p,
|
|
c_bool,
|
|
POINTER,
|
|
Structure,
|
|
)
|
|
|
|
import pathlib
|
|
|
|
# Load the library
|
|
libfile = pathlib.Path(__file__).parent / "libllama.so"
|
|
lib = ctypes.CDLL(str(libfile))
|
|
|
|
|
|
# C types
|
|
llama_context_p = c_void_p
|
|
|
|
|
|
llama_token = c_int
|
|
llama_token_p = POINTER(llama_token)
|
|
|
|
|
|
class llama_token_data(Structure):
|
|
_fields_ = [
|
|
("id", llama_token), # token id
|
|
("p", c_float), # probability of the token
|
|
("plog", c_float), # log probability of the token
|
|
]
|
|
|
|
|
|
llama_token_data_p = POINTER(llama_token_data)
|
|
|
|
|
|
class llama_context_params(Structure):
|
|
_fields_ = [
|
|
("n_ctx", c_int), # text context
|
|
("n_parts", c_int), # -1 for default
|
|
("seed", c_int), # RNG seed, 0 for random
|
|
("f16_kv", c_bool), # use fp16 for KV cache
|
|
(
|
|
"logits_all",
|
|
c_bool,
|
|
), # the llama_eval() call computes all logits, not just the last one
|
|
("vocab_only", c_bool), # only load the vocabulary, no weights
|
|
("use_mlock", c_bool), # force system to keep model in RAM
|
|
("embedding", c_bool), # embedding mode only
|
|
]
|
|
|
|
|
|
llama_context_params_p = POINTER(llama_context_params)
|
|
|
|
|
|
# Functions
|
|
|
|
|
|
def llama_context_default_params() -> llama_context_params:
|
|
params = lib.llama_context_default_params()
|
|
return params
|
|
|
|
|
|
lib.llama_context_default_params.argtypes = []
|
|
lib.llama_context_default_params.restype = llama_context_params
|
|
|
|
# Various functions for loading a ggml llama model.
|
|
# Allocate (almost) all memory needed for the model.
|
|
# Return NULL on failure
|
|
def llama_init_from_file(
|
|
path_model: bytes, params: llama_context_params
|
|
) -> llama_context_p:
|
|
return lib.llama_init_from_file(path_model, params)
|
|
|
|
|
|
lib.llama_init_from_file.argtypes = [c_char_p, llama_context_params]
|
|
lib.llama_init_from_file.restype = llama_context_p
|
|
|
|
# Frees all allocated memory
|
|
def llama_free(ctx: llama_context_p):
|
|
lib.llama_free(ctx)
|
|
|
|
|
|
lib.llama_free.argtypes = [llama_context_p]
|
|
lib.llama_free.restype = None
|
|
|
|
# TODO: not great API - very likely to change
|
|
# Returns 0 on success
|
|
def llama_model_quantize(
|
|
fname_inp: bytes, fname_out: bytes, itype: c_int, qk: c_int
|
|
) -> c_int:
|
|
return lib.llama_model_quantize(fname_inp, fname_out, itype, qk)
|
|
|
|
|
|
lib.llama_model_quantize.argtypes = [c_char_p, c_char_p, c_int, c_int]
|
|
lib.llama_model_quantize.restype = c_int
|
|
|
|
# Run the llama inference to obtain the logits and probabilities for the next token.
|
|
# tokens + n_tokens is the provided batch of new tokens to process
|
|
# n_past is the number of tokens to use from previous eval calls
|
|
# Returns 0 on success
|
|
def llama_eval(
|
|
ctx: llama_context_p,
|
|
tokens: llama_token_p,
|
|
n_tokens: c_int,
|
|
n_past: c_int,
|
|
n_threads: c_int,
|
|
) -> c_int:
|
|
return lib.llama_eval(ctx, tokens, n_tokens, n_past, n_threads)
|
|
|
|
|
|
lib.llama_eval.argtypes = [llama_context_p, llama_token_p, c_int, c_int, c_int]
|
|
lib.llama_eval.restype = c_int
|
|
|
|
|
|
# Convert the provided text into tokens.
|
|
# The tokens pointer must be large enough to hold the resulting tokens.
|
|
# Returns the number of tokens on success, no more than n_max_tokens
|
|
# Returns a negative number on failure - the number of tokens that would have been returned
|
|
# TODO: not sure if correct
|
|
def llama_tokenize(
|
|
ctx: llama_context_p,
|
|
text: bytes,
|
|
tokens: llama_token_p,
|
|
n_max_tokens: c_int,
|
|
add_bos: c_bool,
|
|
) -> c_int:
|
|
"""Convert the provided text into tokens.
|
|
The tokens pointer must be large enough to hold the resulting tokens.
|
|
Returns the number of tokens on success, no more than n_max_tokens
|
|
Returns a negative number on failure - the number of tokens that would have been returned
|
|
"""
|
|
return lib.llama_tokenize(ctx, text, tokens, n_max_tokens, add_bos)
|
|
|
|
|
|
lib.llama_tokenize.argtypes = [llama_context_p, c_char_p, llama_token_p, c_int, c_bool]
|
|
lib.llama_tokenize.restype = c_int
|
|
|
|
|
|
def llama_n_vocab(ctx: llama_context_p) -> c_int:
|
|
return lib.llama_n_vocab(ctx)
|
|
|
|
|
|
lib.llama_n_vocab.argtypes = [llama_context_p]
|
|
lib.llama_n_vocab.restype = c_int
|
|
|
|
|
|
def llama_n_ctx(ctx: llama_context_p) -> c_int:
|
|
return lib.llama_n_ctx(ctx)
|
|
|
|
|
|
lib.llama_n_ctx.argtypes = [llama_context_p]
|
|
lib.llama_n_ctx.restype = c_int
|
|
|
|
# Token logits obtained from the last call to llama_eval()
|
|
# The logits for the last token are stored in the last row
|
|
# Can be mutated in order to change the probabilities of the next token
|
|
# Rows: n_tokens
|
|
# Cols: n_vocab
|
|
def llama_get_logits(ctx: llama_context_p):
|
|
return lib.llama_get_logits(ctx)
|
|
|
|
|
|
lib.llama_get_logits.argtypes = [llama_context_p]
|
|
lib.llama_get_logits.restype = POINTER(c_float)
|
|
|
|
# Get the embeddings for the input
|
|
# shape: [n_embd] (1-dimensional)
|
|
def llama_get_embeddings(ctx: llama_context_p):
|
|
return lib.llama_get_embeddings(ctx)
|
|
|
|
lib.llama_get_embeddings.argtypes = [llama_context_p]
|
|
lib.llama_get_embeddings.restype = POINTER(c_float)
|
|
|
|
# Token Id -> String. Uses the vocabulary in the provided context
|
|
def llama_token_to_str(ctx: llama_context_p, token: int) -> bytes:
|
|
return lib.llama_token_to_str(ctx, token)
|
|
|
|
|
|
lib.llama_token_to_str.argtypes = [llama_context_p, llama_token]
|
|
lib.llama_token_to_str.restype = c_char_p
|
|
|
|
# Special tokens
|
|
|
|
def llama_token_bos() -> llama_token:
|
|
return lib.llama_token_bos()
|
|
|
|
|
|
lib.llama_token_bos.argtypes = []
|
|
lib.llama_token_bos.restype = llama_token
|
|
|
|
|
|
def llama_token_eos() -> llama_token:
|
|
return lib.llama_token_eos()
|
|
|
|
|
|
lib.llama_token_eos.argtypes = []
|
|
lib.llama_token_eos.restype = llama_token
|
|
|
|
|
|
# TODO: improve the last_n_tokens interface ?
|
|
def llama_sample_top_p_top_k(
|
|
ctx: llama_context_p,
|
|
last_n_tokens_data: llama_token_p,
|
|
last_n_tokens_size: c_int,
|
|
top_k: c_int,
|
|
top_p: c_double,
|
|
temp: c_double,
|
|
repeat_penalty: c_double,
|
|
) -> llama_token:
|
|
return lib.llama_sample_top_p_top_k(
|
|
ctx, last_n_tokens_data, last_n_tokens_size, top_k, top_p, temp, repeat_penalty
|
|
)
|
|
|
|
|
|
lib.llama_sample_top_p_top_k.argtypes = [
|
|
llama_context_p,
|
|
llama_token_p,
|
|
c_int,
|
|
c_int,
|
|
c_double,
|
|
c_double,
|
|
c_double,
|
|
]
|
|
lib.llama_sample_top_p_top_k.restype = llama_token
|
|
|
|
|
|
# Performance information
|
|
|
|
def llama_print_timings(ctx: llama_context_p):
|
|
lib.llama_print_timings(ctx)
|
|
|
|
|
|
lib.llama_print_timings.argtypes = [llama_context_p]
|
|
lib.llama_print_timings.restype = None
|
|
|
|
|
|
def llama_reset_timings(ctx: llama_context_p):
|
|
lib.llama_reset_timings(ctx)
|
|
|
|
|
|
lib.llama_reset_timings.argtypes = [llama_context_p]
|
|
lib.llama_reset_timings.restype = None
|
|
|
|
|
|
# Print system information
|
|
def llama_print_system_info() -> bytes:
|
|
"""Print system informaiton"""
|
|
return lib.llama_print_system_info()
|
|
|
|
|
|
lib.llama_print_system_info.argtypes = []
|
|
lib.llama_print_system_info.restype = c_char_p
|