train.py

# coding=utf-8
# Copyright 2021 The Google Research 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.

# Lint as: python3
"""Training script for Nerf."""
import functools
import gc
import time
from absl import app
from absl import flags
import flax
from flax.metrics import tensorboard
from flax.training import checkpoints
import jax
from jax import config
from jax import random
import jax.numpy as jnp
import numpy as np
# import wandb
from tqdm import tqdm

from nerf import datasets
from nerf import models
from nerf import utils
from nerf import clip_utils

FLAGS = flags.FLAGS

utils.define_flags()
config.parse_flags_with_absl()

# set up TPU for colab
import os
if "COLAB_TPU_ADDR" in os.environ:
    import jax.tools.colab_tpu
    jax.tools.colab_tpu.setup_tpu()
print(f"detected device: {jax.local_devices()}")


def train_step(model, clip_model, rng, state, batch, lr, step, K,):
    """One optimization step.

    Args:
        model: The linen model.
        rng: jnp.ndarray, random number generator.
        state: utils.TrainState, state of the model/optimizer.
        batch: dict, a mini-batch of data for training.
        lr: float, real-time learning rate.

    Returns:
        new_state: utils.TrainState, new training state.
        stats: list. [(loss, psnr), (loss_coarse, psnr_coarse)].
        rng: jnp.ndarray, updated random number generator.
    """
    rng, key_0, key_1 = random.split(rng, 3)

    def loss_fn(variables):
        rays = batch["rays"]
        ret = model.apply(variables, key_0, key_1, rays, FLAGS.randomized)
        if len(ret) not in (1, 2):
            raise ValueError(
                "ret should contain either 1 set of output (coarse only), or 2 sets"
                "of output (coarse as ret[0] and fine as ret[1]).")
        # The main prediction is always at the end of the ret list.
        rgb, unused_disp, unused_acc = ret[-1]
        loss = ((rgb - batch["pixels"][Ellipsis, :3]) ** 2).mean()
        psnr = utils.compute_psnr(loss)
        if len(ret) > 1:
            # If there are both coarse and fine predictions, we compute the loss for
            # the coarse prediction (ret[0]) as well.
            rgb_c, unused_disp_c, unused_acc_c = ret[0]
            loss_c = ((rgb_c - batch["pixels"][Ellipsis, :3]) ** 2).mean()
            psnr_c = utils.compute_psnr(loss_c)
        else:
            loss_c = 0.
            psnr_c = 0.

        def tree_sum_fn(fn):
            return jax.tree_util.tree_reduce(lambda x, y: x + fn(y),
                                             variables, initializer=0)

        weight_l2 = (tree_sum_fn(lambda z: jnp.sum(z ** 2)) /
                     tree_sum_fn(lambda z: jnp.prod(jnp.array(z.shape))))

        total_loss = loss + loss_c + FLAGS.weight_decay_mult * weight_l2
        stats = utils.Stats(loss=loss, psnr=psnr, loss_c=loss_c,
                            psnr_c=psnr_c, weight_l2=weight_l2)
        return total_loss, stats

    (_, stats), grad = (
        jax.value_and_grad(loss_fn, has_aux=True)(state.optimizer.target))
    stats = jax.lax.pmean(stats, axis_name="batch")
    
    # Clip the gradient by value.
    if FLAGS.grad_max_val > 0:
        clip_fn = lambda z: jnp.clip(z, -FLAGS.grad_max_val, FLAGS.grad_max_val)
        grad = jax.tree_util.tree_map(clip_fn, grad)

    # Clip the (possibly value-clipped) gradient by norm.
    if FLAGS.grad_max_norm > 0:
        grad_norm = jnp.sqrt(
            jax.tree_util.tree_reduce(
                lambda x, y: x + jnp.sum(y ** 2), grad, initializer=0))
        mult = jnp.minimum(1, FLAGS.grad_max_norm / (1e-7 + grad_norm))
        grad = jax.tree_util.tree_map(lambda z: mult * z, grad)

    return grad, stats, rng
    new_optimizer = state.optimizer.apply_gradient(grad, learning_rate =lr)
    new_state = state.replace(optimizer=new_optimizer)
    return new_state, stats, rng

def update_step(state, grad, lr):
    grad = jax.lax.pmean(grad, axis_name="batch")
    new_optimizer = state.optimizer.apply_gradient(grad, learning_rate=lr)
    new_state = state.replace(optimizer=new_optimizer)
    return new_state

def main(unused_argv):
    #wandb.init(project="hf-flax-clip-nerf", entity="wandb", sync_tensorboard=True)
    rng = random.PRNGKey(20200823)
    # Shift the numpy random seed by host_id() to shuffle data loaded by different
    # hosts.
    np.random.seed(20201473 + jax.host_id())

    if FLAGS.config is not None:
        utils.update_flags(FLAGS)
    if FLAGS.batch_size % jax.device_count() != 0:
        raise ValueError("Batch size must be divisible by the number of devices.")
    if FLAGS.train_dir is None:
        raise ValueError("train_dir must be set. None set now.")
    if FLAGS.data_dir is None:
        raise ValueError("data_dir must be set. None set now.")

    # setup CLIP model
    if FLAGS.use_semantic_loss:
        clip_model = clip_utils.init_CLIP(FLAGS.clip_output_dtype,
                                          FLAGS.clip_model_name)
        print(f'semantic loss ACTIVATED, CLIP is set up '
              f'(sc_loss_mult: {FLAGS.sc_loss_mult})')
    else:
        clip_model = None
        print('semantic loss DEACTIVATED, CLIP is set to None')
    
    dataset = datasets.get_dataset("train", FLAGS, clip_model)
    test_dataset = datasets.get_dataset("test", FLAGS, clip_model)

    # setup NeRF model
    rng, key = random.split(rng)
    model, variables = models.get_model(key, dataset.peek(), FLAGS)
    optimizer = flax.optim.Adam(FLAGS.lr_init).create(variables)
    state = utils.TrainState(optimizer=optimizer)
    del optimizer, variables
    learning_rate_fn = functools.partial(
        utils.learning_rate_decay,
        lr_init=FLAGS.lr_init,
        lr_final=FLAGS.lr_final,
        max_steps=FLAGS.max_steps,
        lr_delay_steps=FLAGS.lr_delay_steps,
        lr_delay_mult=FLAGS.lr_delay_mult)

    train_pstep = jax.pmap(
        functools.partial(train_step, model, clip_model),
        axis_name="batch",
        in_axes=(0, 0, 0, None, None, None),
        donate_argnums=(2,))
    
    update_pstep = jax.pmap(
        functools.partial(update_step,),
        axis_name="batch",
        in_axes=(0, 0, None),
        donate_argnums=(0,))

    def render_fn(variables, key_0, key_1, rays):
        return model.apply(variables, key_0, key_1, rays, FLAGS.randomized)

    render_pfn = jax.pmap(
        render_fn,
        in_axes=(None, None, None, 0),  # Only distribute the data input.
        donate_argnums=(3,),
        axis_name="batch")

    def render_fn_(variables, key_0, key_1, rays):
        return model.apply(variables, key_0, key_1, rays, False, True)

    render_pfn_ = jax.pmap(
        render_fn_,
        in_axes=(None, None, None, 0),  # Only distribute the data input.
        donate_argnums=(3,),
        axis_name="batch")

    # Compiling to the CPU because it's faster and more accurate.
    ssim_fn = jax.jit(
        functools.partial(utils.compute_ssim, max_val=1.), backend="cpu")

    if not utils.isdir(FLAGS.train_dir):
        utils.makedirs(FLAGS.train_dir)
    state = checkpoints.restore_checkpoint(FLAGS.train_dir, state)
    # Resume training a the step of the last checkpoint.
    init_step = state.optimizer.state.step + 1

    # for distributive training
    state = flax.jax_utils.replicate(state)
    if jax.host_id() == 0:
        summary_writer = tensorboard.SummaryWriter(FLAGS.train_dir)

    # Prefetch_buffer_size = 3 x batch_size
    pdataset = flax.jax_utils.prefetch_to_device(dataset, 3)
    n_local_devices = jax.local_device_count()
    rng = rng + jax.host_id()  # Make random seed separate across hosts.
    keys = random.split(rng, n_local_devices)  # For pmapping RNG keys.
    gc.disable()  # Disable automatic garbage collection for efficiency.
    stats_trace = []
    reset_timer = True

    # for semantic loss update
    sc_image = None
    sc_loss = 0.
    
    for step, batch in tqdm(zip(range(init_step, FLAGS.max_steps + 1), pdataset)):
        if reset_timer:
            t_loop_start = time.time()
            reset_timer = False
        lr = learning_rate_fn(step)

        grad, stats, keys = train_pstep(keys, state, batch, lr, step, FLAGS.sc_loss_every)
        
        if step%FLAGS.sc_loss_every == 0 and FLAGS.use_semantic_loss and step < FLAGS.stop_sc_loss:
            sc_batch = dataset.get_clip_data()
            if jax.local_device_count() > 1:
                sc_loss, sc_grad, sc_image = clip_utils.semantic_step_multi(render_pfn_, clip_model, keys[0], state, sc_batch, lr)
            else:
                sc_loss, sc_grad, sc_image = clip_utils.semantic_step_single(model, clip_model, keys[0], state, sc_batch, lr)

            leaves, treedef = jax.tree_flatten(grad)
            sc_leaves, _ = jax.tree_flatten(sc_grad)
            grad = treedef.unflatten(g+jnp.expand_dims(sc_g,0) for g, sc_g in zip(leaves, sc_leaves))
        
        state = update_pstep(state, grad, lr)

        if jax.host_id() == 0:
            stats_trace.append(stats)
        if step % FLAGS.gc_every == 0:
            gc.collect()

        # Log training summaries. This is put behind a host_id check because in
        # multi-host evaluation, all hosts need to run inference even though we
        # only use host 0 to record results.
        if jax.host_id() == 0:
            if step % FLAGS.print_every == 0:
                summary_writer.scalar("loss/train", stats.loss[0], step)
                summary_writer.scalar("sc_loss", sc_loss, step)
                summary_writer.scalar("psnr/train", stats.psnr[0], step)
                summary_writer.scalar("train_coarse/loss", stats.loss_c[0], step)
                summary_writer.scalar("train_coarse/psnr", stats.psnr_c[0], step)

                avg_loss = np.mean(np.concatenate([s.loss for s in stats_trace]))
                avg_psnr = np.mean(np.concatenate([s.psnr for s in stats_trace]))
                stats_trace = []
                summary_writer.scalar("train_avg/loss", avg_loss, step)
                summary_writer.scalar("train_avg/psnr", avg_psnr, step)

                steps_per_sec = FLAGS.print_every / (time.time() - t_loop_start)
                reset_timer = True
                rays_per_sec = FLAGS.batch_size * steps_per_sec
                summary_writer.scalar("stats/weight_l2", stats.weight_l2[0], step)
                summary_writer.scalar("stats/learning_rate", lr, step)
                summary_writer.scalar("iter_speed/train_steps_per_sec", steps_per_sec, step)
                summary_writer.scalar("iter_speed/train_rays_per_sec", rays_per_sec, step)
                precision = int(np.ceil(np.log10(FLAGS.max_steps))) + 1
                print(("{:" + "{:d}".format(precision) + "d}").format(step) +
                      f"/{FLAGS.max_steps:d}: " + f"i_loss={stats.loss[0]:0.4f}, " +
                      f"avg_loss={avg_loss:0.4f}, " +
                      f"weight_l2={stats.weight_l2[0]:0.2e}, " +
                      f"sc_loss={sc_loss:0.4f}, " +
                      f"lr={lr:0.2e}, {rays_per_sec:0.0f} rays/sec")
            if step % FLAGS.save_every == 0:
                state_to_save = jax.device_get(jax.tree_map(lambda x: x[0], state))
                checkpoints.save_checkpoint(
                    FLAGS.train_dir, state_to_save, int(step), keep=100)

        # Test-set evaluation.
        if FLAGS.render_every > 0 and step % FLAGS.render_every == 0:
            # We reuse the same random number generator from the optimization step
            # here on purpose so that the visualization matches what happened in
            # training.
            t_eval_start = time.time()
            eval_variables = jax.device_get(jax.tree_map(lambda x: x[0],
                                                         state)).optimizer.target
            test_case = next(test_dataset)
            pred_color, pred_disp, pred_acc = utils.render_image(
                functools.partial(render_pfn, eval_variables),
                test_case["rays"],
                keys[0],
                FLAGS.dataset == "llff",
                chunk=FLAGS.chunk)

            # Log eval summaries on host 0.
            if jax.host_id() == 0:
                psnr = utils.compute_psnr(
                    ((pred_color - test_case["pixels"]) ** 2).mean())
                ssim = ssim_fn(pred_color, test_case["pixels"])
                eval_time = time.time() - t_eval_start
                num_rays = jnp.prod(jnp.array(test_case["rays"].directions.shape[:-1]))
                rays_per_sec = num_rays / eval_time
                summary_writer.scalar("iter_speed/test_rays_per_sec", rays_per_sec, step)
                print(f"Eval {step}: {eval_time:0.3f}s., {rays_per_sec:0.0f} rays/sec")
                summary_writer.scalar("psnr/test", psnr, step)
                summary_writer.scalar("ssim/ssim", ssim, step)
                if sc_image is not None:
                    summary_writer .image("random_ray_image", sc_image, step)
                summary_writer.image("test_pred_color", pred_color, step)
                summary_writer.image("test_pred_disp", pred_disp, step)
                summary_writer.image("test_pred_acc", pred_acc, step)
                summary_writer.image("test_target", test_case["pixels"], step)

    if FLAGS.max_steps % FLAGS.save_every != 0:
        state = jax.device_get(jax.tree_map(lambda x: x[0], state))
        checkpoints.save_checkpoint(
            FLAGS.train_dir, state, int(FLAGS.max_steps), keep=100)


if __name__ == "__main__":
    app.run(main)