Manuel de Prada Corral manueldeprada

Cached CI for HF proposal: Record‑and‑Bake HTTP Cache

We’re seeing repeated CI failures in a fresh container when tests make live HTTP calls. Example from today, 5 or so rerun failures on:

tests/models/pix2struct/test_image_processing_pix2struct.py::Pix2StructImageProcessingTest::test_expected_patches
tests/models/pix2struct/test_image_processing_pix2struct.py::Pix2StructImageProcessingTest::test_call_vqa

First, it is you need to create a conda or venv environment. This mamba.yml file contains the packages:

name: QAFactEval
channels:
- conda-forge
dependencies:
- python=3.8.18=hd12c33a_0_cpython
- spacy=2.2.4 
- spacy-model-en_core_web_sm=2.2.5

	#!/usr/bin/python3

	import re
	import signal
	import dbus
	from gi.repository import GLib
	from dbus.mainloop.glib import DBusGMainLoop

	import gi
	gi.require_version('Gst', '1.0')

	from decoders import inject_supervitamined_decoders, StochasticBeamSearchDecoder, FakeTransformer
	from transformers import T5ForConditionalGeneration, T5Tokenizer
	import torch


	# pip install decoders

	# this demonstration uses a fake toy transformer (https://manueldeprada.com/blog/posts/toy-probabilistic-transformer/)
	# to test the correctness of the stochastic beam search implementation
	def test_fake_transformer():

	from transformers import PreTrainedModel, PretrainedConfig, PreTrainedTokenizer, BatchEncoding
	from transformers.modeling_outputs import Seq2SeqLMOutput
	import torch


	class FakeTransformerConfig(PretrainedConfig):
	model_type = "FakeTransformer"

	def __init__(self, vocab_size=4, **kwargs):
	super().__init__(pad_token_id=-1, eos_token_id=3, bos_token_id=0, **kwargs)