lewtun

.PHONY: install

#################################################################################
# GLOBALS                                                                       #
#################################################################################
SHELL=/bin/bash
CONDA_ACTIVATE=source $$(conda info --base)/etc/profile.d/conda.sh ; conda activate ; conda activate

#################################################################################
# COMMANDS                                                                      #

lewtun

lewtun

lewtun

lewtun

{"data": [{"title": "B00001WRSJ", "paragraphs": [{"qas": [{"question": "What is the tonal balance of these headphones?", "id": "d0781d13200014aa25860e44da9d5ea7", "answers": [{"text": "I have been a headphone fanatic for thirty years", "answer_start": 0}], "is_impossible": false}], "context": "I have been a headphone fanatic for thirty years and have owned and used a variety of headphones over those years, to include Stax SR-5, Sennheiser HD-424 and HD-580.  The Sony MDRV6 excells as the best value of any headphone that I've ever owned.  They are especially good at producing natural-sounding deep bass, and the overall octave-to-octave balance is excellent.  The sound quality is all in all comparable to other headphones that cost considerably more.The MDRV6 is especially well-suited for travel due to the collapsible design, and for noisy environments or for quiet environments such as a library where the sound emitted by open-back headphones would distract others.The MDRV6 is not quite as comfortable as some ot

lewtun

from datasets import load_dataset

def validate_datasets(reference_dataset, new_dataset):
    """Validate the column names and rows of the new dataset"""
    splits = list(reference_dataset.keys())
    for split in splits:
        ref_dset = reference_dataset[split]
        new_dset = new_dataset[split]
        # Check column names agree
        ref_cols = set(ref_dset.column_names)

lewtun

for question_type in ["How", "What", "Is"]:
    for question in (
        dfs["train"][dfs["train"].question.str.startswith(question_type)]
        .sample(n=3, random_state=42)['question']):
        print(question)

lewtun

from tqdm import tqdm
import torch

device = "cuda" if torch.cuda.is_available() else "cpu"

def chunks(list_of_elements, batch_size):
    """Yield successive batch-sized chunks from list_of_elements."""
    for i in range(0, len(list_of_elements), batch_size):
        yield list_of_elements[i : i + batch_size]

lewtun

if any(nd in n for nd in no_decay):
    params_without_wd.append(p)
else:
    params_with_wd.append(p)

lewtun

def get_grouped_params(model, no_decay=["bias", "LayerNorm.weight"]):
    params_with_wd, params_without_wd = [], []
    for n, p in model.named_parameters():
        if any(nd in n for nd in no_decay): 
            params_without_wd.append(p)
        else: 
            params_with_wd.append(p)
    return [{'params': params_with_wd, 'weight_decay': args.weight_decay},
            {'params': params_without_wd, 'weight_decay': 0.0}]