SamplingConfig: Improve/extend docstrings, clearly explaining the parameters

This commit is contained in:
Dominik Jain 2023-12-14 18:52:20 +01:00
parent 1903a72ecb
commit 408d51f9de
2 changed files with 91 additions and 5 deletions

View File

@ -23,7 +23,7 @@ class ReplayBuffer:
:param size: the maximum size of replay buffer.
:param stack_num: the frame-stack sampling argument, should be greater than or
equal to 1. Default to 1 (no stacking).
:param ignore_obs_next: whether to store obs_next. Default to False.
:param ignore_obs_next: whether to not store obs_next. Default to False.
:param save_only_last_obs: only save the last obs/obs_next when it has a shape
of (timestep, ...) because of temporal stacking. Default to False.
:param sample_avail: the parameter indicating sampling only available index

View File

@ -6,29 +6,115 @@ from tianshou.utils.string import ToStringMixin
@dataclass
class SamplingConfig(ToStringMixin):
"""Sampling, epochs, parallelization, buffers, collectors, and batching."""
"""Configuration of sampling, epochs, parallelization, buffers, collectors, and batching."""
# TODO: What are the most reasonable defaults?
num_epochs: int = 100
"""
the number of epochs to run training for. An epoch is the outermost iteration level and each
epoch consists of a number of training steps and a test step, where each training step
* collects environment steps/transitions (collection step), adding them to the (replay)
buffer (see :attr:`step_per_collect`)
* performs one or more gradient updates (see :attr:`update_per_step`).
The number of training steps in each epoch is indirectly determined by
:attr:`step_per_epoch`: As many training steps will be performed as are required in
order to reach :attr:`step_per_epoch` total steps in the training environments.
Specifically, if the number of transitions collected per step is `c` (see
:attr:`step_per_collect`) and :attr:`step_per_epoch` is set to `s`, then the number
of training steps per epoch is `ceil(s / c)`.
Therefore, if `num_epochs = e`, the total number of environment steps taken during training
can be computed as `e * ceil(s / c) * c`.
"""
step_per_epoch: int = 30000
"""
the total number of environment steps to be made per epoch. See :attr:`num_epochs` for
an explanation of epoch semantics.
"""
batch_size: int = 64
"""for off-policy algorithms, this is the number of environment steps/transitions to sample
from the buffer for a gradient update; for on-policy algorithms, its use is algorithm-specific.
On-policy algorithms use the full buffer that was collected in the preceding collection step
but they may use this parameter to perform the gradient update using mini-batches of this size
(causing the gradient to be less accurate, a form of regularization).
"""
num_train_envs: int = -1
"""the number of training environments to use. If set to -1, use number of CPUs/threads."""
num_test_envs: int = 1
"""the number of test environments to use"""
buffer_size: int = 4096
"""the total size of the sample/replay buffer, in which environment steps (transitions) are
stored"""
step_per_collect: int = 2048
"""
the number of environment steps/transitions to collect in each collection step before the
network update within each training step.
Note that the exact number can be reached only if this is a multiple of the number of
training environments being used, as each training environment will produce the same
(non-zero) number of transitions.
Specifically, if this is set to `n` and `m` training environments are used, then the total
number of transitions collected per collection step is `ceil(n / m) * m =: c`.
See :attr:`num_epochs` for information on the total number of environment steps being
collected during training.
"""
repeat_per_collect: int | None = 10
"""
controls, within one gradient update step of an on-policy algorithm, the number of times an
actual gradient update is applied using the full collected dataset, i.e. if the parameter is
`n`, then the collected data shall be used five times to update the policy within the same
training step.
The parameter is ignored and may be set to None for off-policy and offline algorithms.
"""
update_per_step: float = 1.0
"""
Only used in off-policy algorithms.
How many gradient steps to perform per step in the environment (i.e., per sample added to the buffer).
for off-policy algorithms only: the number of gradient steps to perform per sample
collected (see :attr:`step_per_collect`).
Specifically, if this is set to `u` and the number of samples collected in the preceding
collection step is `n`, then `round(u * n)` gradient steps will be performed.
Note that for on-policy algorithms, only a single gradient update is usually performed,
because thereafter, the samples no longer reflect the behavior of the updated policy.
To change the number of gradient updates for an on-policy algorithm, use parameter
:attr:`repeat_per_collect` instead.
"""
start_timesteps: int = 0
"""
the number of environment steps to collect before the actual training loop begins
"""
start_timesteps_random: bool = False
# TODO can we set the parameters below intelligently? Perhaps based on env. representation?
"""
whether to use a random policy (instead of the initial or restored policy to be trained)
when collecting the initial :attr:`start_timesteps` environment steps before training
"""
replay_buffer_ignore_obs_next: bool = False
replay_buffer_save_only_last_obs: bool = False
"""if True, only the most recent frame is saved when appending to experiences rather than the
full stacked frames. This avoids duplicating observations in buffer memory. Set to False to
save stacked frames in full.
"""
replay_buffer_stack_num: int = 1
"""
the number of consecutive environment observations to stack and use as the observation input
to the agent for each time step. Setting this to a value greater than 1 can help agents learn
temporal aspects (e.g. velocities of moving objects for which only positions are observed).
"""
def __post_init__(self) -> None:
if self.num_train_envs == -1: