Octopus Command Line
Usage and Arguments
octopus -h
Octopus Command Line Interface (CLI) octopus_cli support two types of inputs:
-tor--text: Write you input text directly on the command line. The generation will display directly on the terminal.
-for--input_file: import the text from file. The generation will saved on the JSON format file.
octopus_cli has the following optional arguments.
usage: octopus [-h] -x PREFIX [-t TEXT] [-f INPUT_FILE] [-m SEARCH_METHOD]
[-s SEQ_LENGTH] [-o MAX_OUTPUTS] [-b NUM_BEAMS]
[-n NO_REPEAT_NGRAM_SIZE] [-k TOP_K] [-p TOP_P] [-c CACHE_DIR]
[-l LOGGING_FILE] [-bs BATCH_SIZE]
Octopus Command Line Interface (CLI)
options:
-h, --help show this help message and exit
-x PREFIX, --prefix PREFIX
task prefix [diacritize, correct_grammar, paraphrase,
,answer_question,generate_question,summarize,
generate_title,
translitrate_ar2en,translitrate_en2ar]
-t TEXT, --text TEXT translate the input text into Arabic
-f INPUT_FILE, --input_file INPUT_FILE
path of input file
-m SEARCH_METHOD, --search_method SEARCH_METHOD
Turjuman translation search method should be one of
the follows ['greedy', 'beam', 'sampling'], default
value is beam search
-s SEQ_LENGTH, --seq_length SEQ_LENGTH
The maximum sequence length value, default value is
300
-o MAX_OUTPUTS, --max_outputs MAX_OUTPUTS
The maxmuim of the output tanslations, default value
is 1
-b NUM_BEAMS, --num_beams NUM_BEAMS
Number of beams, default value is 5
-n NO_REPEAT_NGRAM_SIZE, --no_repeat_ngram_size NO_REPEAT_NGRAM_SIZE
Number of n-gram that doesn't appears twice, default
value is 2
-k TOP_K, --top_k TOP_K
Sample from top K likely next words instead of all
words, default value is 50
-p TOP_P, --top_p TOP_P
Sample from the smallest set whose cumulative
probability mass exceeds p for next words, default
value is 0.95
-c CACHE_DIR, --cache_dir CACHE_DIR
The cache directory path, default value is
octopus_cache directory
-l LOGGING_FILE, --logging_file LOGGING_FILE
The logging file path, default value is None
-bs BATCH_SIZE, --batch_size BATCH_SIZE
The maximum number of source examples utilized in one
iteration
Google Colab Link
You can find the full examples on the Google Colab on the following link https://colab.research.google.com/github/UBC-NLP/octopus/blob/main/examples/octopus_cli.ipynb