big_vision/configs/proj/paligemma/transfers/ocrvqa.py

# Copyright 2024 Big Vision Authors.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

# pylint: disable=line-too-long
r"""PaliGemma transfer to OCR-VQA, see (internal link) for details and notes.
"""

import big_vision.configs.common as bvcc
from big_vision.configs.proj.paligemma.transfers.common import combine_and_keep_train, combine_and_keep_eval, TOKENIZER


def training_data(res, *, final_split, text_len=32):
  """Creates training data config.

  See (internal link)
  You can add more arguments beside `res`, but give them good defaults.

  Args:
    res: The requested image resolution (eg 224).
    final_split: Train on all train+val data or train+val[20000:].
    text_len: sequence length.

  Returns:
    The ConfigDict for the input section.
  """
  # Note that the dataset is "unbatched" here, meaning each image_question pair
  # is one example. So while there are ~800k training examples, there's only
  # ~200k unique images, each one having on average 4 questions, and the
  # questions are highly regular:
  #   - What is the title of this book?
  #   - What type of book is this? OR What is the genre of this book?
  #   - Who wrote this book? OR Who is the author of this book?
  #   - Is this book related to [GENRE]? OR Is this a [GENRE] book? "yes"
  #   - Same but with answer "no"
  # So one obvious thing we could do in training is randomize the negative genre
  # question more using a custom pp op.
  c = bvcc.parse_arg('')  # Just make a configdict without extra import.
  c.data = dict(
      name='ocrvqa_id',
      split='train+val' if final_split else 'train + val[20_000:]',  # Val is 100k, we don't need that much!
  )
  c.pp = '|'.join([
      f'decode|resize({res}, antialias=True)|value_range(-1, 1)',
      'strfmt("answer en {question}", outkey="prefix")',
      'copy(inkey="answer", outkey="suffix")',
      combine_and_keep_train(text_len),
  ])
  return c


def add_eval(c, res, text_len=32, **kw):
  """OCR-VQA evaluators."""
  pp = '|'.join([
      f'decode|resize({res}, antialias=True)|value_range(-1, 1)',
      'strfmt("answer en {question}", outkey="prefix")',
      'copy(inkey="int_id", outkey="question_id")',
      combine_and_keep_eval(text_len, keep=('answer', 'question_id')),
  ])

  for freq, name, split in [
      (1/8, 'minitrain', 'train[:5120]'),  # To gauge memorization.
      (1/4, 'minival', 'val[:20_000]'),  # To tune hparams. SLOW!
      (1.0, 'eval', 'test'),  # Final number to report. Big => rare.
  ]:
    c.evals[f'ocrvqa/{name}'] = dict(
        type='proj.paligemma.transfers.vqa',
        to_lower=True,
        pred='decode', pred_kw={'max_decode_len': text_len},
        data={**training_data(res, final_split=True, text_len=text_len).data, 'split': split},
        log_percent=freq, skip_first=freq == 1, tokenizer=TOKENIZER, pp_fn=pp)
    c.evals[f'ocrvqa/{name}'].update(kw)


def add_eval_pplx(c, res, text_len=32):
  """Perplexity evaluator to test runs before implementing the real deal."""
  c_train = training_data(res, final_split=True, text_len=text_len)  # Use mostly same settings as training.
  for name, split in [
      ('minitrain', 'train[:5120]'),  # To gauge memorization.
      ('minival', 'val[:20_000]'),  # To tune hparams.
      ('eval', 'test'),  # To compute final publishable scores.
  ]:
    c.evals[f'ocrvqa/{name}/pplx'] = dict(
        type='proj.paligemma.perplexity', pred='logits',
        key='text', shift_labels=True,
        log_percent=0.1,  # Eval ~10x per run;
        data={**c_train.data, 'split': split},
        pp_fn=c_train.pp,
    )


def sweep_best(add, arg=None):
  """Train with best hyper-params."""
  c = bvcc.parse_arg(arg, final_split=False)
  add(**bvcc.arg(res=224, **c), lr=3e-6)
  add(**bvcc.arg(res=448, **c), lr=3e-6)
  add(**bvcc.arg(res=896, **c), lr=1e-5)


sweep = sweep_best  # Choose which sweep to run.


def get_config(arg=None):
  """Config for training."""
  c = bvcc.parse_arg(arg, mode='xm', res=224, final_split=False)

  c.input = training_data(c.res, final_split=c.final_split)

  # Instead of epochs, you can also use `total_examples` or `total_steps`.
  c.total_epochs = 3
  c.input.batch_size = 128
  c.optax_name = 'scale_by_adam'
  c.optax = dict(b2=0.999)
  c.lr = 3e-6
  c.wd = 0.0
  c.grad_clip_norm = 1.0
  c.label_smoothing = 0.0
  c.schedule = dict(decay_type='cosine', warmup_percent=0.05)

  # Add evaluators.
  c.evals = {}
  add_eval(c, c.res, batch_size=256)
  add_eval_pplx(c, c.res)

  # Model section.
  c.model_name = 'proj.paligemma.paligemma'
  c.model = {}
  c.model.img = dict(variant='So400m/14', pool_type='none', scan=True)
  c.model.llm = dict(vocab_size=256_000 + 1024 + 128, dropout=0.0)
  c.model_init = f'pt_{c.res}'

  # FSDP strategy.
  c.mesh = [('data', -1)]
  c.sharding_strategy = [('.*', 'fsdp(axis="data")')]
  c.sharding_rules = [('act_batch', ('data',))]

  # These probably do not need any change/tuning
  c.input.shuffle_buffer_size = 50_000
  c.log_training_steps = 50
  c.ckpt_steps = 1_000
  c.pp_modules = ['ops_general', 'ops_image', 'ops_text', 'proj.paligemma.ops']

  # Update configs for quicker local runs and avoid swapping.
  if c.mode in ('runlocal', 'mock'):
    c.input.shuffle_buffer_size = None
    for ev in c.evals.values():
      ev.data.split = ev.data.split.split('[')[0] + '[:16]'

  if c.mode == 'runlocal':
    c.log_training_steps = 1
    c.input.batch_size = 2

  c.seed = 0
  return c


def metrics(arg=None):  # pylint: disable=unused-argument
  m = ['training_loss']
  for split in ('eval', 'minival', 'minitrain'):
    m.append(f'ocrvqa/{split}/acc')
  return m