feat: refactor common data utilities of dtensor policy v2 #1710
Conversation
|
hemildesai
force-pushed
the
hemil/automodel-init-refactor
branch
from
4f66b8f to
81174e5
Compare
hemildesai
force-pushed
the
hemil/automodel-data-refactor
branch
from
0dfbb1e to
2959be7
Compare
|
Signed-off-by: Hemil Desai <hemild@nvidia.com>
hemildesai
force-pushed
the
hemil/automodel-data-refactor
branch
from
2959be7 to
be28e23
Compare
|
|
📝 WalkthroughWalkthroughIntroduces a new data processing module for automodel with dataclasses ( Changes
Sequence Diagram(s)sequenceDiagram
participant Iterator as Raw Iterator
participant MBIter as Microbatch Iterator
participant Proc as process_microbatch()
participant Tokenizer as Tokenizer
participant DPMesh as DP Mesh (all_reduce)
participant Inputs as ProcessedInputs
Iterator->>MBIter: raw BatchedDataDict
loop per item in iterator
MBIter->>Proc: call with BatchedDataDict
Proc->>Tokenizer: tokenize/prepare
alt Sequence Packing Enabled
Proc->>Proc: pack_sequences()
Proc->>Proc: get_flash_attention_kwargs()
end
alt Context Parallel (cp_size > 1)
Proc->>Proc: construct cp_buffers<br/>and seq_index
end
Proc->>Inputs: create ProcessedInputs
MBIter->>MBIter: wrap in ProcessedMicrobatch
MBIter->>Iterator: yield ProcessedMicrobatch
end
Iterator->>DPMesh: process_global_batch()<br/>extracts batch
DPMesh->>DPMesh: all_reduce(global_valid_seqs,<br/>global_valid_toks)
Estimated code review effort🎯 4 (Complex) | ⏱️ ~60 minutes Possibly related PRs
Suggested labels
Suggested reviewers
🚥 Pre-merge checks | ✅ 2 | ❌ 2❌ Failed checks (2 warnings)
✅ Passed checks (2 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing touches
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 5
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (1)
nemo_rl/models/policy/workers/dtensor_policy_worker_v2.py (1)
492-503: Avoid leaking metrics from dummy microbatches.
num_valid_samplesis only set for real microbatches; dummy iterations can reuse the prior value and append duplicate metrics. Explicitly zero it (or guard the append) for dummy batches.🐛 Suggested fix
- else: - loss *= 0 + else: + loss *= 0 + num_valid_samples = 0
🤖 Fix all issues with AI agents
In `@nemo_rl/models/automodel/data.py`:
- Around line 1-13: Update the file header year from 2025 to 2026 in the NVIDIA
copyright block at the top of nemo_rl/models/automodel/data.py; locate the
existing copyright comment starting with "# Copyright (c) 2025, NVIDIA
CORPORATION." and change "2025" to "2026" so the header reflects the current
year.
In `@nemo_rl/models/policy/workers/dtensor_policy_worker_v2.py`:
- Around line 639-651: The unpacking of packed logprobs uses input_ids.shape[0]
(which is 1 when packed) and thus allocates/iterates the wrong batch size;
instead read the original batch size from the batch metadata stored in lp_batch
(e.g., lp_batch.original_batch_size or lp_batch['original_batch_size']) and use
that value when allocating unpacked_logprobs and for the unpacking loop/range in
the unpacking logic (the code that creates unpacked_logprobs and iterates over
batch indices). Apply the same change to the other unpacking site referenced
(the block around the unpacking logic in the later section).
- Around line 969-973: In get_topk_logits, skip dummy microbatches by checking
batch_idx against iterator_len inside the loop over processed_iterator—after
obtaining batch_idx and processed_mb (and before using
lp_batch/processed_inputs/input_lengths), add a guard like if batch_idx >=
iterator_len: break (or continue if appropriate) so dummy microbatches do not
produce extra outputs; update any related logic in get_topk_logits to rely on
iterator_len instead of processing all entries from processed_iterator.
- Around line 866-870: In the score() loop inside dtensor_policy_worker_v2.py,
skip any dummy microbatches appended by sequence packing by checking the batch
index against iterator_len; inside the for batch_idx, processed_mb in
enumerate(processed_iterator) loop (which updates step and collects
all_rm_scores), add a guard like if batch_idx >= iterator_len: break or continue
so you do not process or append scores for padded/dummy batches (use the
existing batch_idx, iterator_len, processed_mb, all_rm_scores, step symbols to
locate the change).
In `@tests/unit/models/automodel/test_automodel_data.py`:
- Around line 70-125: Tests contain unused variables/fixtures (e.g.,
mb_iterator, dummy_iterator, mock_tokenizer, *args/**kwargs) that trigger Ruff
ARG001/ARG002/RUF059; update the tests (including functions like
test_dynamic_batching and usages around get_microbatch_iterator,
BatchedDataDict, make_microbatch_iterator_with_dynamic_shapes,
get_microbatch_iterator_dynamic_shapes_len) by renaming intentionally unused
locals/fixtures with a leading underscore (e.g., _mb_iterator, _dummy_iterator,
_mock_tokenizer or _args/_kwargs) or, where truly intentional, annotate the
binding with a trailing "# noqa" to suppress the linter; then re-run Ruff to
confirm all reported warnings (including the other listed test regions) are
resolved.
🧹 Nitpick comments (1)
nemo_rl/models/automodel/data.py (1)
217-220: Track the sequence‑packing workaround.
The TODO indicates a known workaround formin_seq_len; consider linking it to an issue or follow‑up ticket so it doesn’t linger.If you want, I can help draft a follow-up issue or propose a replacement implementation.
| # Copyright (c) 2025, NVIDIA CORPORATION. All rights reserved. | ||
| # | ||
| # 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. |
There was a problem hiding this comment.
Update the NVIDIA copyright year to 2026.
This is a new non-test file, so the header should reflect the current year.
🔧 Suggested fix
-# Copyright (c) 2025, NVIDIA CORPORATION. All rights reserved.
+# Copyright (c) 2026, NVIDIA CORPORATION. All rights reserved.As per coding guidelines, the NVIDIA header should include the current year.
📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
| # Copyright (c) 2025, NVIDIA CORPORATION. All rights reserved. | |
| # | |
| # 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. | |
| # Copyright (c) 2026, NVIDIA CORPORATION. All rights reserved. | |
| # | |
| # 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. |
🤖 Prompt for AI Agents
In `@nemo_rl/models/automodel/data.py` around lines 1 - 13, Update the file header
year from 2025 to 2026 in the NVIDIA copyright block at the top of
nemo_rl/models/automodel/data.py; locate the existing copyright comment starting
with "# Copyright (c) 2025, NVIDIA CORPORATION." and change "2025" to "2026" so
the header reflects the current year.
| lp_batch = processed_mb.data_dict | ||
| processed_inputs = processed_mb.processed_inputs | ||
|
|
||
| # Extract values from processed inputs | ||
| input_ids = processed_inputs.input_ids | ||
| attention_mask = processed_inputs.attention_mask | ||
| position_ids = processed_inputs.position_ids | ||
| flash_attn_kwargs = processed_inputs.flash_attn_kwargs | ||
| vlm_kwargs = processed_inputs.vlm_kwargs | ||
| cp_buffers = processed_inputs.cp_buffers | ||
| seq_index = processed_inputs.seq_index | ||
| seq_len = processed_inputs.seq_len | ||
|
|
There was a problem hiding this comment.
Use original batch size when unpacking packed logprobs.
With packing, input_ids.shape[0] is 1, so unpacked_logprobs only allocates a single row and drops the rest. Use the original batch size for the unpacking loop/shape.
🐛 Suggested fix
- processed_inputs = processed_mb.processed_inputs
+ processed_inputs = processed_mb.processed_inputs
+ original_batch_size = processed_mb.original_batch_size
...
- unpacked_logprobs = torch.zeros(
- (batch_size, seq_dim_size),
+ unpacked_logprobs = torch.zeros(
+ (original_batch_size, seq_dim_size),
dtype=token_logprobs.dtype,
device=token_logprobs.device,
)
cu_seqlens = flash_attn_kwargs.cu_seqlens_q
- for i in range(batch_size):
+ for i in range(original_batch_size):
start = cu_seqlens[i].item() + 1
end = cu_seqlens[i + 1].item()
seq_len_actual = input_lengths[i].item()Also applies to: 799-817
🤖 Prompt for AI Agents
In `@nemo_rl/models/policy/workers/dtensor_policy_worker_v2.py` around lines 639 -
651, The unpacking of packed logprobs uses input_ids.shape[0] (which is 1 when
packed) and thus allocates/iterates the wrong batch size; instead read the
original batch size from the batch metadata stored in lp_batch (e.g.,
lp_batch.original_batch_size or lp_batch['original_batch_size']) and use that
value when allocating unpacked_logprobs and for the unpacking loop/range in the
unpacking logic (the code that creates unpacked_logprobs and iterates over batch
indices). Apply the same change to the other unpacking site referenced (the
block around the unpacking logic in the later section).
| step = 0 | ||
| all_rm_scores = [] | ||
| for batch_idx, generate_batch in enumerate( | ||
| itertools.chain(mb_iterator, dummy_iterator) | ||
| ): | ||
| for batch_idx, processed_mb in enumerate(processed_iterator): | ||
| step += 1 | ||
| input_ids = generate_batch.get("input_ids").cuda() | ||
| input_lengths = generate_batch.get("input_lengths") | ||
| batch_size, seq_len = input_ids.shape | ||
| if self.enable_seq_packing: | ||
| input_ids, position_ids, _ = pack_sequences( | ||
| input_ids=input_ids, | ||
| input_lengths=input_lengths, | ||
| packed_sequence_size=[ | ||
| batch_size | ||
| ], # flash attention 2 expects flattened input | ||
| padding_value=self.tokenizer.eos_token_id, | ||
| return_attention_mask=False, | ||
| ) | ||
| seq_len = input_ids.shape[1] | ||
| attention_mask = None | ||
| flash_attn_kwargs = get_flash_attention_kwargs( | ||
| input_lengths=input_lengths, | ||
| ) | ||
| else: | ||
| # Create attention mask for right-padded data | ||
| post_attention_mask = torch.zeros( | ||
| (batch_size, seq_len), dtype=torch.bool, device=input_ids.device | ||
| ) | ||
| for i, length in enumerate(input_lengths): | ||
| # For right-padded sequence, set 1s at the beginning of the sequence | ||
| post_attention_mask[i, :length] = 1 | ||
| position_ids = torch.arange( | ||
| seq_len, device=input_ids.device | ||
| ).repeat(batch_size, 1) | ||
|
|
||
| attention_mask = torch.ones( | ||
| (batch_size, seq_len), | ||
| dtype=torch.bool, | ||
| device=input_ids.device, | ||
| ) | ||
| if self.cp_size > 1: | ||
| seq_index = torch.arange(seq_len, device=input_ids.device).repeat( | ||
| 1, 1 | ||
| ) | ||
| cp_buffers = [input_ids, position_ids, seq_index] | ||
| else: | ||
| cp_buffers = [] | ||
| seq_index = None | ||
| # Extract processed inputs |
There was a problem hiding this comment.
Skip dummy microbatches in score().
When sequence packing yields uneven batch counts, dummy batches are appended and currently produce extra scores. Break/continue once batch_idx >= iterator_len.
🐛 Suggested fix
- for batch_idx, processed_mb in enumerate(processed_iterator):
+ for batch_idx, processed_mb in enumerate(processed_iterator):
+ if batch_idx >= iterator_len:
+ break🧰 Tools
🪛 Ruff (0.14.13)
868-868: Loop control variable batch_idx not used within loop body
Rename unused batch_idx to _batch_idx
(B007)
🤖 Prompt for AI Agents
In `@nemo_rl/models/policy/workers/dtensor_policy_worker_v2.py` around lines 866 -
870, In the score() loop inside dtensor_policy_worker_v2.py, skip any dummy
microbatches appended by sequence packing by checking the batch index against
iterator_len; inside the for batch_idx, processed_mb in
enumerate(processed_iterator) loop (which updates step and collects
all_rm_scores), add a guard like if batch_idx >= iterator_len: break or continue
so you do not process or append scores for padded/dummy batches (use the
existing batch_idx, iterator_len, processed_mb, all_rm_scores, step symbols to
locate the change).
| for batch_idx, processed_mb in enumerate(processed_iterator): | ||
| # Extract data dict and processed inputs | ||
| lp_batch = processed_mb.data_dict | ||
| processed_inputs = processed_mb.processed_inputs | ||
| input_lengths = lp_batch.get("input_lengths") |
There was a problem hiding this comment.
Skip dummy microbatches in get_topk_logits().
Dummy batches will otherwise add extra outputs. Break/continue once batch_idx >= iterator_len.
🐛 Suggested fix
- for batch_idx, processed_mb in enumerate(processed_iterator):
+ for batch_idx, processed_mb in enumerate(processed_iterator):
+ if batch_idx >= iterator_len:
+ break🧰 Tools
🪛 Ruff (0.14.13)
969-969: Loop control variable batch_idx not used within loop body
Rename unused batch_idx to _batch_idx
(B007)
🤖 Prompt for AI Agents
In `@nemo_rl/models/policy/workers/dtensor_policy_worker_v2.py` around lines 969 -
973, In get_topk_logits, skip dummy microbatches by checking batch_idx against
iterator_len inside the loop over processed_iterator—after obtaining batch_idx
and processed_mb (and before using lp_batch/processed_inputs/input_lengths), add
a guard like if batch_idx >= iterator_len: break (or continue if appropriate) so
dummy microbatches do not produce extra outputs; update any related logic in
get_topk_logits to rely on iterator_len instead of processing all entries from
processed_iterator.
| mb_iterator, iterator_len, dummy_iterator = get_microbatch_iterator( | ||
| data=data, | ||
| cfg=cfg, | ||
| enable_seq_packing=enable_seq_packing, | ||
| mbs=mbs, | ||
| dp_mesh=mock_dp_mesh, | ||
| ) | ||
|
|
||
| # Verify iterator length | ||
| assert iterator_len == 4 # 16 / 4 = 4 | ||
|
|
||
| # Verify we can iterate through the data | ||
| batches = list(mb_iterator) | ||
| assert len(batches) == 4 | ||
| assert batches[0]["input_ids"].shape[0] == 4 | ||
|
|
||
| # Verify dummy iterator is empty | ||
| assert list(dummy_iterator) == [] | ||
|
|
||
| def test_dynamic_batching(self): | ||
| # Create test data | ||
| data = BatchedDataDict( | ||
| { | ||
| "input_ids": torch.randint(0, 1000, (8, 128)), | ||
| "sample_mask": torch.ones(8, dtype=torch.bool), | ||
| } | ||
| ) | ||
|
|
||
| # Mock the microbatch iterator methods | ||
| mock_iterator = iter([data, data, data]) | ||
| data.make_microbatch_iterator_with_dynamic_shapes = MagicMock( | ||
| return_value=mock_iterator | ||
| ) | ||
| data.get_microbatch_iterator_dynamic_shapes_len = MagicMock(return_value=3) | ||
|
|
||
| cfg = {"dynamic_batching": {"enabled": True}} | ||
| mbs = 4 | ||
| enable_seq_packing = False | ||
| mock_dp_mesh = MagicMock() | ||
|
|
||
| mb_iterator, iterator_len, dummy_iterator = get_microbatch_iterator( | ||
| data=data, | ||
| cfg=cfg, | ||
| enable_seq_packing=enable_seq_packing, | ||
| mbs=mbs, | ||
| dp_mesh=mock_dp_mesh, | ||
| ) | ||
|
|
||
| # Verify dynamic batching was used | ||
| assert iterator_len == 3 | ||
| data.make_microbatch_iterator_with_dynamic_shapes.assert_called_once() | ||
| data.get_microbatch_iterator_dynamic_shapes_len.assert_called_once() | ||
|
|
||
| # Verify dummy iterator is empty | ||
| assert list(dummy_iterator) == [] | ||
|
|
There was a problem hiding this comment.
🧩 Analysis chain
🏁 Script executed:
# First, let's explore the repository structure and locate the file
find . -name "test_automodel_data.py" -type fRepository: NVIDIA-NeMo/RL
Length of output: 110
🏁 Script executed:
# Check if Ruff is available and what version
ruff --versionRepository: NVIDIA-NeMo/RL
Length of output: 70
🏁 Script executed:
# Let's read the test file to understand the context
head -200 tests/unit/models/automodel/test_automodel_data.py | cat -nRepository: NVIDIA-NeMo/RL
Length of output: 7985
🏁 Script executed:
# Read lines 150-160 to see the context around lines 152-155
sed -n '145,165p' tests/unit/models/automodel/test_automodel_data.py | cat -nRepository: NVIDIA-NeMo/RL
Length of output: 859
🏁 Script executed:
# Read around lines 448-478
sed -n '440,485p' tests/unit/models/automodel/test_automodel_data.py | cat -nRepository: NVIDIA-NeMo/RL
Length of output: 2042
🏁 Script executed:
# Read around lines 967-969
sed -n '960,975p' tests/unit/models/automodel/test_automodel_data.py | cat -nRepository: NVIDIA-NeMo/RL
Length of output: 600
🏁 Script executed:
# Now run Ruff on the file to see the actual warnings
ruff check tests/unit/models/automodel/test_automodel_data.py --select=ARG001,ARG002,RUF059Repository: NVIDIA-NeMo/RL
Length of output: 11071
🏁 Script executed:
# Verify the line 967-969 reference mentioned in the review
sed -n '960,975p' tests/unit/models/automodel/test_automodel_data.py | cat -nRepository: NVIDIA-NeMo/RL
Length of output: 600
🏁 Script executed:
# Confirm the exact line numbers for lines 70 and 110 (test_regular_batching vs test_dynamic_batching)
sed -n '56,90p' tests/unit/models/automodel/test_automodel_data.py | cat -nRepository: NVIDIA-NeMo/RL
Length of output: 1347
🏁 Script executed:
# Get a better view of the exact errors and their line numbers for comprehensive coverage
ruff check tests/unit/models/automodel/test_automodel_data.py --select=ARG001,ARG002,RUF059 -v 2>&1 | grep -E "^|help:" | head -50Repository: NVIDIA-NeMo/RL
Length of output: 3053
🏁 Script executed:
# Let's get the total count and grouping of errors
ruff check tests/unit/models/automodel/test_automodel_data.py --select=ARG001,ARG002,RUF059 2>&1 | tail -1Repository: NVIDIA-NeMo/RL
Length of output: 142
Fix Ruff linter warnings for unused variables and parameters.
ARG001/ARG002/RUF059 will trigger on unused fixtures/locals (e.g., mb_iterator, mock_tokenizer, *args/**kwargs). Prefix with _ or add # noqa where intentional.
♻️ Example cleanup (apply similarly elsewhere)
- mb_iterator, iterator_len, dummy_iterator = get_microbatch_iterator(
+ _mb_iterator, iterator_len, dummy_iterator = get_microbatch_iterator(
data=data,
cfg=cfg,
enable_seq_packing=enable_seq_packing,
mbs=mbs,
dp_mesh=mock_dp_mesh,
)
...
- def side_effect(tensor, *args, **kwargs):
+ def side_effect(tensor, *_args, **_kwargs):
tensor[0] = 2 # Simulate max batch count
...
- def test_processed_microbatch_creation(self, mock_tokenizer):
+ def test_processed_microbatch_creation(self, _mock_tokenizer):
"""Test that ProcessedMicrobatch correctly stores all attributes."""Also applies to: 152-155, 448-478, 858, 742, 796, 841, 913, 985.
Please re-run Ruff after cleanup to confirm the warnings are resolved.
🧰 Tools
🪛 Ruff (0.14.13)
110-110: Unpacked variable mb_iterator is never used
Prefix it with an underscore or any other dummy variable pattern
(RUF059)
🤖 Prompt for AI Agents
In `@tests/unit/models/automodel/test_automodel_data.py` around lines 70 - 125,
Tests contain unused variables/fixtures (e.g., mb_iterator, dummy_iterator,
mock_tokenizer, *args/**kwargs) that trigger Ruff ARG001/ARG002/RUF059; update
the tests (including functions like test_dynamic_batching and usages around
get_microbatch_iterator, BatchedDataDict,
make_microbatch_iterator_with_dynamic_shapes,
get_microbatch_iterator_dynamic_shapes_len) by renaming intentionally unused
locals/fixtures with a leading underscore (e.g., _mb_iterator, _dummy_iterator,
_mock_tokenizer or _args/_kwargs) or, where truly intentional, annotate the
binding with a trailing "# noqa" to suppress the linter; then re-run Ruff to
confirm all reported warnings (including the other listed test regions) are
resolved.
|
| ) | ||
|
|
||
|
|
||
| def get_microbatch_iterator( |
There was a problem hiding this comment.
In the mcore data refactor PR, get_microbatch_iterator calls make_processed_microbatch_iterator directly, so the processed iterator is returned from this fn. Is that something we want to do here too? I made that decision because there's no instance where we'd want to call get_microbatch_iterator without make_processed_microbatch_iterator
There was a problem hiding this comment.
Ok yeah I can make that change here too.
|
terrykong
left a comment
terrykong
left a comment
There was a problem hiding this comment.
looks good to me, small comments. like the first PR can you run the dtensor v2 nightlies to make sure this change hasn't broken anything
| # For right-padded sequence, set 1s at the beginning of the sequence | ||
| post_attention_mask[i, :length] = 1 | ||
|
|
||
| # explicitly create position ids for the input, otherwise the sharding |
There was a problem hiding this comment.
this comment was dropped i think, could we add it back if needed?
another thing that i was reminded of @akoumpa mentioned something about position_ids being an issue if we're using sequence parallel. something about the sharding being incorrect but I don't recall the exact context and whether the guidance was to pass it or not. @hemildesai @akoumpa any idea on this?
| ).repeat(batch_size, 1) | ||
| flash_attn_kwargs = {} | ||
|
|
||
| # DTensor requires the casual attention kernel to hit, |
There was a problem hiding this comment.
another comment worth keeping?
| device=input_ids.device, | ||
| ) | ||
|
|
||
| # if there are multimodal kwargs, we don't need to add position_ids (computed internally) |
| def make_processed_microbatch_iterator( | ||
| raw_iterator: Iterator[BatchedDataDict[Any]], | ||
| tokenizer: AutoTokenizer, | ||
| enable_seq_packing: bool, |
There was a problem hiding this comment.
to mirror mcore, wdyt about just inferring this from the config like the mcore PR is?
cc @ashors1
Depends on #1709
Issues
#1589
Usage
# Add a code snippet demonstrating how to use thisBefore your PR is "Ready for review"
Pre checks:
Additional Information
Summary by CodeRabbit
New Features
Tests
Chores
✏️ Tip: You can customize this high-level summary in your review settings.