lambdaofgod

pl_en_translation

translation -> english model -> backtranslation

Opus Marian model works Helsinki-NLP/opus-mt-pl-en

you can

lambdaofgod

chatgpt_experiment

ChatGPT API monthly cost estimate

Assume using ChatGPT 30 days a month, with 100 queries per day.

Mean tokens per query is upper limit (we estimated it using a bunch of queries), in practice this would be lower, especially because it can be bounded by asking ChatGPT to only answer with fixed number of sentences

api_cost_per_token = 2e-6

lambdaofgod

from rust_functions import tokenize_python_code

def test_simple_input():
    example_code = """
    def foo():
        return x + 1
"""
    expected_tokens = ["def", "foo", "(", ")", "return", "x", "+", "1"]
    assert tokenize_rust(example_code.strip()) == expected_tokens

lambdaofgod

"""
Query:
Write a gradio app that shows results of searching in a list of texts:
- tokenizing texts with nltk
- using rank_bm25 library
- displaying results as dataframe under the search box

Response:
Here is an example of a gradio app that uses the rank_bm25 library to search through a list of texts, tokenizes the texts and queries using nltk, and shows the results as a dataframe under the search box:
"""

lambdaofgod

sudo apt-get install -y libgl1-mesa-glx libegl1-mesa libxrandr2 libxrandr2 libxss1 libxcursor1 libxcomposite1 libasound2 libxi6 libxtst6 icu-devtools  libicu-dev
wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh
bash Miniconda3-latest-Linux-x86_64.sh

sudo apt-get install npm
sudo npm i elasticdump

curl -fsSL https://download.docker.com/linux/ubuntu/gpg | sudo apt-key add -
sudo add-apt-repository \
   "deb [arch=amd64] https://download.docker.com/linux/ubuntu \

lambdaofgod

import ktrain
zsl = ktrain.text.ZeroShotClassifier('huggingface/prunebert-base-uncased-6-finepruned-w-distil-mnli')

lambdaofgod

def build_segmentation_model(
        input_shape,
        n_classes,
        base_block_size=BASE_BLOCK_SIZE,
        base_dropout_rate=BASE_DROPOUT_RATE,
        activation=ACTIVATION
    ):
    # Build U-Net segmentation_model
    inputs = layers.Input(input_shape)
    s = layers.Lambda(lambda x: x - 0.5) (inputs)

lambdaofgod

apt-get install vim git wget

lambdaofgod

# Example NeoMutt config file for the index-color feature.

# Entire index line
color index white black '.*'
# Author name, %A %a %F %L %n
# Give the author column a dark grey background
color index_author default color234 '.*'
# Highlight a particular from (~f)
color index_author brightyellow color234 '~fRay Charles'
# Message flags, %S %Z

lambdaofgod

import torch
import ot
from sklearn import metrics


roberta = torch.hub.load('pytorch/fairseq', 'roberta.large')
roberta.eval()  # disable dropout (or leave in train mode to finetune)


def get_roberta_features(text):