constants.py

import os

# options
CORPUS_BY_DESC = {
    'RedPajama (LLaMA tokenizer), 1.4T tokens': 'v4_rpj_llama_s4',
    'Pile-train (LLaMA tokenizer), 380B tokens': 'v4_piletrain_llama',
    'C4-train (LLaMA tokenizer), 200B tokens': 'v4_c4train_llama',
    'Pile-val (LLaMA tokenizer), 390M tokens': 'v4_pileval_llama',
    'Pile-val (GPT-2 tokenizer), 380M tokens': 'v4_pileval_gpt2',
    'Dolma-sample (OLMo tokenizer), 8.0B tokens': 'v4_dolmasample_olmo',
}
CORPUS_DESCS = list(CORPUS_BY_DESC.keys())
ENGINE_BY_DESC = {
    'C++ (🚀🚀 Fast)': 'c++',
    'Python': 'python',
}
ENGINE_DESCS = list(ENGINE_BY_DESC.keys())
ENGINES = list(ENGINE_BY_DESC.values())

# engine
MAX_QUERY_CHARS = int(os.environ.get('MAX_QUERY_CHARS', 1000))
MAX_INPUT_DOC_TOKENS = int(os.environ.get('MAX_INPUT_DOC_TOKENS', 1000))
MAX_OUTPUT_DOC_TOKENS = int(os.environ.get('MAX_OUTPUT_DOC_TOKENS', 5000))
MAX_OUTPUT_NUM_DOCS = int(os.environ.get('MAX_OUTPUT_NUM_DOCS', 10)) # This number is also hard-coded in app.py
MAX_CNT_FOR_NTD = int(os.environ.get('MAX_CNT_FOR_NTD', 1000))
MAX_CLAUSE_FREQ_PER_SHARD = int(os.environ.get('MAX_CLAUSE_FREQ_FAST_APPROX_PER_SHARD', 50000))
MAX_DIFF_TOKENS = int(os.environ.get('MAX_DIFF_TOKENS', 100))
MAX_DIFF_BYTES = 2 * MAX_DIFF_TOKENS
MAX_CLAUSES_IN_CNF = int(os.environ.get('MAX_CLAUSES_IN_CNF', 4))
MAX_TERMS_IN_DISJ_CLAUSE = int(os.environ.get('MAX_TERMS_IN_DISJ_CLAUSE', 4))

# HF demo
API_URL = os.environ.get('API_URL', None)
DEFAULT_CONCURRENCY_LIMIT = os.environ.get('DEFAULT_CONCURRENCY_LIMIT', 10)
MAX_SIZE = os.environ.get('MAX_SIZE', 100)
MAX_THREADS = os.environ.get('MAX_THREADS', 40)
DEBUG = (os.environ.get('DEBUG', 'False') != 'False')