[RLlib] Fix performance and functionality flaws in attention nets (via Trajectory view API). #11729
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
…ectory_view_api_attention_nets
…nto trajectory_view_api_attention_nets # Conflicts: # rllib/models/tf/attention_net.py # rllib/policy/view_requirement.py
…ectory_view_api_enable_by_default_for_all_simple
…ectory_view_api_attention_nets
…ectory_view_api_enable_by_default_for_all_simple
…ectory_view_api_attention_nets # Conflicts: # rllib/agents/ppo/ppo_tf_policy.py # rllib/evaluation/tests/test_trajectory_view_api.py # rllib/policy/tf_policy_template.py # rllib/utils/tf_ops.py
…on_nets # Conflicts: # rllib/agents/ppo/ppo_tf_policy.py # rllib/evaluation/collectors/simple_list_collector.py # rllib/evaluation/tests/test_trajectory_view_api.py # rllib/policy/dynamic_tf_policy.py # rllib/policy/policy.py # rllib/policy/sample_batch.py # rllib/policy/view_requirement.py
| def __init__(self, shift_before: int = 0): | ||
| self.shift_before = max(shift_before, 1) | ||
| def __init__(self, view_reqs): | ||
| self.shift_before = -min( |
There was a problem hiding this comment.
Might want to add comment to describe what this code does!
| def add_init_obs(self, episode_id: EpisodeID, agent_id: AgentID, | ||
| env_id: EnvID, init_obs: TensorType, | ||
| def add_init_obs(self, episode_id: EpisodeID, agent_index: int, | ||
| env_id: EnvID, t: int, init_obs: TensorType, | ||
| view_requirements: Dict[str, ViewRequirement]) -> None: | ||
| """Adds an initial observation (after reset) to the Agent's trajectory. |
There was a problem hiding this comment.
Change the description. It adds more than a single observation.
There was a problem hiding this comment.
No, it doesn't it's really just adds a single one. Same as it used to work w/ SampleBatchBuilder.
| / view_req.batch_repeat_value)) | ||
| repeat_count = (view_req.data_rel_pos_to - | ||
| view_req.data_rel_pos_from + 1) | ||
| data = np.asarray([ |
There was a problem hiding this comment.
Same as above, big confused. Add comments on what these lines of code do
There was a problem hiding this comment.
Provided an example.
| shift = view_req.data_rel_pos + obs_shift | ||
| # Shift is exactly 0: Use trajectory as is. | ||
| if shift == 0: | ||
| data = np_data[data_col][self.shift_before:] |
There was a problem hiding this comment.
Provided an example.
| @@ -208,6 +269,43 @@ def _build_buffers(self, single_row: Dict[str, TensorType]) -> None: | |||
| [np.zeros(shape=shape, dtype=dtype) | |||
| for _ in range(shift)] | |||
|
|
|||
| def _get_input_dict(self, view_reqs, abs_pos: int = -1) -> \ | |||
There was a problem hiding this comment.
Add description what this method does
| batch = SampleBatch(self.buffers) | ||
| assert SampleBatch.UNROLL_ID in batch.data | ||
| batch = SampleBatch( | ||
| self.buffers, _seq_lens=self.seq_lens, _dont_check_lens=True) |
There was a problem hiding this comment.
What is _dont_check_lens?
| self.count = 0 | ||
| if self.seq_lens is not None: | ||
| self.seq_lens = [] | ||
| return batch | ||
|
|
||
|
|
||
| class _PolicyCollectorGroup: |
There was a problem hiding this comment.
Probably add comments on what this is
rllib/policy/sample_batch.py
Outdated
| for i, seq_len in enumerate(self.seq_lens): | ||
| count += seq_len | ||
| if count >= end: | ||
| data["state_in_0"] = self.data["state_in_0"][state_start: |
There was a problem hiding this comment.
Add comment on what this does
| # Range of indices on time-axis, make sure to create | ||
| if view_req.data_rel_pos_from is not None: | ||
| ret[view_col] = np.zeros_like([[ | ||
| view_req.space.sample() |
There was a problem hiding this comment.
Same add comment here
| @@ -121,3 +125,14 @@ def rel_shift(x: TensorType) -> TensorType: | |||
| x = tf.reshape(x, x_size) | |||
|
|
|||
| return x | |||
|
|
|||
|
|
|||
| class PositionalEmbedding(tf.keras.layers.Layer): | |||
There was a problem hiding this comment.
Add comments on what this does (how it initializes embedding per position based on cos/sin something)
…ectory_view_api_attention_nets # Conflicts: # rllib/agents/trainer.py # rllib/evaluation/collectors/simple_list_collector.py # rllib/evaluation/tests/test_trajectory_view_api.py # rllib/models/tf/attention_net.py # rllib/policy/policy.py # rllib/policy/torch_policy_template.py # rllib/policy/view_requirement.py # src/ray/raylet/node_manager.cc
…ectory_view_api_attention_nets # Conflicts: # rllib/agents/trainer.py # rllib/evaluation/collectors/simple_list_collector.py # rllib/evaluation/tests/test_trajectory_view_api.py # rllib/models/tf/attention_net.py # rllib/policy/policy.py # rllib/policy/torch_policy_template.py # rllib/policy/view_requirement.py # src/ray/raylet/node_manager.cc
…ectory_view_api_attention_nets � Conflicts: � rllib/agents/ppo/appo_tf_policy.py � rllib/agents/ppo/ppo_torch_policy.py � rllib/agents/qmix/model.py � rllib/evaluation/collectors/simple_list_collector.py � rllib/evaluation/rollout_worker.py � rllib/evaluation/tests/test_trajectory_view_api.py � rllib/models/modelv2.py � rllib/policy/dynamic_tf_policy.py � rllib/policy/policy.py � rllib/policy/sample_batch.py � rllib/policy/view_requirement.py
…ectory_view_api_attention_nets
…ectory_view_api_attention_nets
…ectory_view_api_attention_nets
…ectory_view_api_attention_nets
|
Moved here: |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Why are these changes needed?
Related issue number
Checks
scripts/format.shto lint the changes in this PR.