NanoSocrates/Project_Model/Libs/BPE/Classes/NanoSocraTrainer.py

249 lines
7.1 KiB
Python
Raw Normal View History

2025-09-30 13:33:40 +02:00
from collections import deque
2025-10-01 12:21:42 +02:00
import datetime
2025-09-30 13:33:40 +02:00
from pathlib import Path
import re
2025-10-01 12:21:42 +02:00
from ..Classes import (
NanoSocratesBPE,
NanoSocratesChunker,
NanoSocratesSplitter,
NanoSocratesBatchMemoryBPE,
)
2025-09-30 13:33:40 +02:00
from ..Enums import TokenType
2025-10-01 12:21:42 +02:00
from ..Utils import (
special_regex_maker,
iterator_with_checks,
save_nanos_vocabulary,
load_nanos_vocabulary,
save_json,
load_json,
)
2025-09-30 13:33:40 +02:00
class NanoSocraTrainer:
def __init__(
self,
max_vocabulary: int,
special_vocabulary: list[str],
chunk_size: int,
merge_treshold: int = 0,
max_iterations: int = 0,
2025-10-01 12:21:42 +02:00
print_after_iterations: int = 1,
2025-09-30 13:33:40 +02:00
) -> None:
# Bytes
BYTE_RESERVED_TOKENS = 256
SPECIAL_RESERVED_TOKENS = len(special_vocabulary)
RESERVED_TOKENS = BYTE_RESERVED_TOKENS + SPECIAL_RESERVED_TOKENS
self.__max_vocabulary = max_vocabulary - RESERVED_TOKENS
self.__max_iterations = max_iterations
self.__chunk_size = chunk_size
self.__merge_treshold = merge_treshold
self.__special_token_regex = special_regex_maker(special_vocabulary)
2025-09-30 23:58:31 +02:00
self.__print_after_iterations = print_after_iterations
2025-09-30 13:33:40 +02:00
def trainBPE(
2025-10-01 12:21:42 +02:00
self,
path: Path,
cache_dir: Path,
bpe: NanoSocratesBPE | None = None,
resume_from_iter: int = 0,
2025-09-30 13:33:40 +02:00
) -> NanoSocratesBPE:
if not path.is_file():
raise FileNotFoundError()
if not cache_dir.is_dir():
raise NotADirectoryError()
if bpe is None:
bpe = NanoSocratesBPE()
BPE = bpe
if BPE.vocabulary_size > self.__max_vocabulary:
return BPE
exit = False
cached = False
current_iteration = 0
2025-10-01 12:21:42 +02:00
input_path = path
2025-09-30 13:33:40 +02:00
2025-10-01 12:21:42 +02:00
NEXT_ITERATION = resume_from_iter + 1 if resume_from_iter != 0 else 0
2025-09-30 13:33:40 +02:00
2025-10-01 12:21:42 +02:00
PATH_GEN = self.__switch_paths(path, cache_dir, NEXT_ITERATION)
MEMORY_PATH_GEN = self.__switch_memory(cache_dir, resume_from_iter)
2025-09-30 13:33:40 +02:00
2025-10-01 12:21:42 +02:00
if resume_from_iter != 0:
cached = True
current_iteration = resume_from_iter
input_path = next(PATH_GEN)
# UGLY: fixes a bug immediately, unfortunately
_, _ = next(MEMORY_PATH_GEN)
_, voc_cache_path = next(MEMORY_PATH_GEN)
vocabulary = load_nanos_vocabulary(voc_cache_path)
BPE = NanoSocratesBPE(vocabulary)
2025-09-30 13:33:40 +02:00
2025-10-01 12:21:42 +02:00
while not exit:
2025-09-30 13:33:40 +02:00
out_path = next(PATH_GEN)
2025-10-01 12:21:42 +02:00
internal_cache_path, vocabulary_cache = next(MEMORY_PATH_GEN)
2025-09-30 13:33:40 +02:00
current_iteration = self.__increment_counter(current_iteration)
LAST_VOC_SIZE = BPE.vocabulary_size
FILE = open(out_path, "w")
2025-09-30 23:58:31 +02:00
last_memory = None
2025-10-01 12:21:42 +02:00
2025-09-30 23:58:31 +02:00
for _, memory, output in self.__round_train(input_path, BPE, cached):
last_memory = memory
2025-09-30 13:33:40 +02:00
FILE.write(output)
FILE.close()
2025-10-01 12:21:42 +02:00
internal_cache = {
"finished_iter": current_iteration,
"read_from": f"{input_path}",
"wrote_to": f"{out_path}",
"at": datetime.datetime.now(datetime.timezone.utc).strftime(
"%Y-%m-%d %H:%M:%S.%f"
)[:-3],
}
VOCABULARY = BPE.vocabulary
save_json(internal_cache, internal_cache_path)
save_nanos_vocabulary(VOCABULARY, vocabulary_cache)
2025-09-30 13:33:40 +02:00
cached = True
input_path = out_path
NEW_VOC_SIZE = BPE.vocabulary_size
2025-09-30 23:58:31 +02:00
if current_iteration % self.__print_after_iterations == 0:
2025-10-01 12:21:42 +02:00
2025-09-30 23:58:31 +02:00
DELIMITER = "==============="
2025-10-01 12:21:42 +02:00
DEBUG = "\n".join(
[
DELIMITER,
f"ITERATION: {current_iteration}",
DELIMITER,
f"\tVocabulary size: {BPE.vocabulary_size}\n",
f"\tFrequencies:\n{last_memory.frequencies}\n", # type: ignore (pretty sure it's not None)
f"\tvocabulary:\n{BPE.vocabulary}",
DELIMITER,
"",
]
)
2025-09-30 23:58:31 +02:00
print(DEBUG)
2025-09-30 13:33:40 +02:00
if LAST_VOC_SIZE == NEW_VOC_SIZE:
exit = True
continue
if current_iteration == self.__max_iterations:
exit = True
continue
if BPE.vocabulary_size == self.__max_vocabulary:
exit = True
continue
return BPE
2025-10-01 12:21:42 +02:00
def __round_train(self, path: Path, bpe: NanoSocratesBPE, cached: bool):
2025-09-30 13:33:40 +02:00
CHUNKER = NanoSocratesChunker(self.__chunk_size, self.__special_token_regex)
SPLITTER = NanoSocratesSplitter(self.__special_token_regex)
BPE = bpe
memory = NanoSocratesBatchMemoryBPE({}, self.__merge_treshold)
CHUNKER_GENERATOR = iterator_with_checks(CHUNKER.chunk(path))
for chunk, last_chunk in CHUNKER_GENERATOR:
2025-10-01 12:21:42 +02:00
PIECE_GENERATOR = iterator_with_checks(SPLITTER.split_text(chunk))
2025-09-30 13:33:40 +02:00
for piece, last_piece in PIECE_GENERATOR:
LAST_BATCH = last_chunk and last_piece
PIECE, TOKEN_TYPE = piece
if TOKEN_TYPE != TokenType.BPE:
_, _, out = BPE.fit([], memory, LAST_BATCH)
yield (BPE, memory, PIECE)
continue
PIECE_DATA = self.__make_list_ids(PIECE, cached)
_, _, out = BPE.fit(PIECE_DATA, memory, LAST_BATCH)
OUT_STRING = f"{out}"
yield (BPE, memory, OUT_STRING)
def __increment_counter(self, counter: int):
# What if overflows???
try:
counter += 1
except:
print("Integer overflow")
counter = 1
return counter
def __make_list_ids(self, corpus: str, cached: bool):
if not cached:
2025-09-30 23:58:31 +02:00
return list(corpus.encode("utf-8"))
2025-09-30 13:33:40 +02:00
2025-10-01 12:21:42 +02:00
REDUCED_CORPUS_LEN = len(corpus) - 1
2025-09-30 13:33:40 +02:00
# Skip these cars "[" "]"
INTS = corpus[1:REDUCED_CORPUS_LEN]
2025-10-01 12:21:42 +02:00
INT_LIST = list(map(int, INTS.split(",")))
2025-09-30 13:33:40 +02:00
return INT_LIST
2025-10-01 12:21:42 +02:00
def __switch_paths(self, path: Path, cache_path: Path, initial_iteration: int):
2025-09-30 13:33:40 +02:00
2025-10-01 12:21:42 +02:00
CORPUS_TMP_1 = cache_path / "corpus-tmp1.txt"
CORPUS_TMP_2 = cache_path / "corpus-tmp2.txt"
2025-09-30 13:33:40 +02:00
switch = True
2025-10-01 12:21:42 +02:00
if initial_iteration % 2 == 1:
switch = False
del initial_iteration
2025-09-30 13:33:40 +02:00
while True:
if switch:
2025-10-01 12:21:42 +02:00
yield CORPUS_TMP_1
2025-09-30 13:33:40 +02:00
else:
2025-10-01 12:21:42 +02:00
yield CORPUS_TMP_2
2025-09-30 13:33:40 +02:00
switch = not switch
2025-10-01 12:21:42 +02:00
def __switch_memory(self, cache_path: Path, initial_iteration: int):
INTERNAL_TMP_1 = cache_path / "internal-tmp1.json"
INTERNAL_TMP_2 = cache_path / "internal-tmp2.json"
VOCAB_TMP_1 = cache_path / "voc-tmp1.json"
VOCAB_TMP_2 = cache_path / "voc-tmp2.json"
switch = False
if initial_iteration % 2 == 1:
switch = True
del initial_iteration
while True:
if switch:
yield (INTERNAL_TMP_1, VOCAB_TMP_1)
else:
yield (INTERNAL_TMP_2, VOCAB_TMP_2)
switch = not switch