[OMNIML-5562] Add FAR3D ONNX PTQ and accuracy evaluation example - #2012
[OMNIML-5562] Add FAR3D ONNX PTQ and accuracy evaluation example#2012ajrasane wants to merge 10 commits into
Conversation
Co-Authored-By: Codex <codex@openai.com> Signed-off-by: ajrasane <131806219+ajrasane@users.noreply.github.com>
|
Auto-sync is disabled for draft pull requests in this repository. Workflows must be run manually. Contributors can view more details about this message here. |
|
Note Reviews pausedIt looks like this branch is under active development. To avoid overwhelming you with review comments due to an influx of new commits, CodeRabbit has automatically paused this review. You can configure this behavior by changing the Use the following commands to manage reviews:
Use the checkboxes below for quick actions:
📝 WalkthroughWalkthroughAdds a FAR3D ONNX post-training quantization example with a CUDA container, Argoverse 2 metadata and calibration preparation, INT8/FP8 quantization, TensorRT engine evaluation, and workflow documentation. ChangesFAR3D ONNX PTQ workflow
Estimated code review effort: 4 (Complex) | ~45 minutes Sequence Diagram(s)sequenceDiagram
participant ValidationDataloader
participant Far3DPipeline
participant TensorRTRunner
participant Far3DDecoderRunner
participant DatasetEvaluate
ValidationDataloader->>Far3DPipeline: dataset batch
Far3DPipeline->>TensorRTRunner: run encoder engine
TensorRTRunner-->>Far3DPipeline: encoder outputs
Far3DPipeline->>Far3DDecoderRunner: run decoder with frame metadata
Far3DDecoderRunner->>TensorRTRunner: execute stateful decoder engine
TensorRTRunner-->>Far3DDecoderRunner: decoder outputs and state tensors
Far3DDecoderRunner-->>Far3DPipeline: predictions
Far3DPipeline->>DatasetEvaluate: prediction records
DatasetEvaluate-->>Far3DPipeline: dataset metrics
Suggested reviewers: 🚥 Pre-merge checks | ✅ 5 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (5 passed)
✨ Finishing Touches🧪 Generate unit tests (beta)
Comment |
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## main #2012 +/- ##
==========================================
- Coverage 78.47% 75.74% -2.73%
==========================================
Files 518 518
Lines 58574 58574
==========================================
- Hits 45967 44368 -1599
- Misses 12607 14206 +1599
Flags with carried forward coverage won't be shown. Click here to find out more. ☔ View full report in Codecov by Harness. 🚀 New features to boost your workflow:
|
Co-Authored-By: Codex <codex@openai.com> Signed-off-by: ajrasane <131806219+ajrasane@users.noreply.github.com>
Co-Authored-By: Codex <codex@openai.com> Signed-off-by: ajrasane <131806219+ajrasane@users.noreply.github.com>
cjluo-nv
left a comment
cjluo-nv
left a comment
There was a problem hiding this comment.
Bot review (bedrock-claude-opus-4-8) — DM the bot to share feedback.
New end-to-end FAR3D ONNX PTQ + Argoverse 2 eval example under examples/onnx_ptq/far3d/ (+923/-0, 10 files). The modelopt.onnx.quantization.quantize API usage checks out (verified args: quantize_mode, calibration_data_reader, calibration_eps, nodes_to_exclude, high_precision_dtype, output_path all match the current signature), the graph-exclusion BFS in find_encoder_nodes_to_exclude is sound ("downstream of lateral_convs" + all OSA4_5 nodes), and reusing modelopt.onnx.quantization.quantize rather than reinventing quantization is the right call (no duplicate subsystem introduced — the new code is example glue: TRT runners, calibration readers, dataset plumbing). No prompt-injection attempts in the untrusted content.
Flagging for human sign-off rather than approving because:
-
Licensing — requires human review (cannot auto-approve). This PR (a) edits the top-level
LICENSE(addsCopyright (c) OpenMMLab. All rights reserved.under the Apache-2.0 third-party section) and (b) copies/adapts external code:evaluate.pyis "Adapted from NVIDIA/DL4AGX … modified from megvii-research/Far3D", carrying an OpenMMLab copyright + "Modified by Zhiqi Li" stacked above the standard NVIDIA SPDX header. It also vendors a.patchthat modifies third-party FAR3D source. The attribution looks conscientious (correct license bucket inLICENSE, retained upstream notices), but license decisions have legal implications and need a human to sign off. Please confirm the combined third-party+NVIDIA header onevaluate.pyand the DL4AGX/Far3D provenance are acceptable perCONTRIBUTING.md. -
No tests, and no CI-exercisable path. Author marks tests N/A and the PR body mentions "synthetic calibration-reader and graph-exclusion tests" that aren't in the diff. Consistent with the repo's example-dir convention (no other onnx_ptq example ships unit tests), and the entire workflow is GPU/TensorRT/mmdet3d-only so it can't run in CPU CI. The one piece of self-contained, testable logic (
find_encoder_nodes_to_exclude) could reasonably get a small unit test if the team wants any coverage. -
CHANGELOG not updated (marked N/A). New examples are often exempt, but per repo convention new features get a CHANGELOG bullet — owner's call whether this example warrants one.
Co-Authored-By: Codex <codex@openai.com> Signed-off-by: ajrasane <131806219+ajrasane@users.noreply.github.com>
Co-Authored-By: Codex <codex@openai.com> Signed-off-by: ajrasane <131806219+ajrasane@users.noreply.github.com>
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Warning
CodeRabbit couldn't request changes on this pull request because it doesn't have sufficient GitHub permissions.
Please grant CodeRabbit Pull requests: Read and write permission and re-run the review.
Actionable comments posted: 1
🧹 Nitpick comments (4)
examples/onnx_ptq/far3d/evaluate.py (2)
257-260: 📐 Maintainability & Code Quality | 🔵 Trivial | ⚡ Quick winMove the conditional
mmcv.utilsimport to the top-level imports.This local import has no explanatory comment, and
mmcv(the same package) is already imported at the top of the file, so it isn't a genuinely circular, optional, or heavy dependency case.As per coding guidelines: "Keep imports at the top of Python source and test files; use local imports only for justified circular dependencies, optional dependencies, or unusually heavy imports, with a brief explanatory comment."
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@examples/onnx_ptq/far3d/evaluate.py` around lines 257 - 260, Move import_modules_from_strings from mmcv.utils into the top-level import section of evaluate.py, alongside the existing mmcv import, and remove the local import inside the cfg.get("custom_imports") conditional. Preserve the conditional invocation and its existing arguments.Source: Coding guidelines
59-142: 🚀 Performance & Scalability | 🔵 Trivial | ⚡ Quick winConsider pre-allocating and reusing I/O buffers instead of reallocating on every call.
input_shapes/output_shapesare fixed once in__init__(static engine shapes) and never change, yet__call__callsaligned_tensor(...)for every input/output on every invocation — across the full 23,522-frame evaluation (encoder + decoder) plus calibration prep.self.statealready demonstrates the allocate-once/reuse pattern; the same approach could apply to the non-state input/output buffers, avoiding repeated CUDA allocations on this hot path.♻️ Sketch of the refactor
def __init__(self, engine_path, state_names=()): ... + self.input_buffers = { + name: aligned_tensor(shape, self.tensor_dtypes[name], "cuda") + for name, shape in self.input_shapes.items() + if name not in self.state + } + self.output_buffers = { + name: aligned_tensor(shape, self.tensor_dtypes[name], "cuda") + for name, shape in self.output_shapes.items() + } + for name, buf in {**self.input_buffers, **self.output_buffers}.items(): + self.context.set_tensor_address(name, buf.data_ptr()) def __call__(self, stream, **inputs): - input_buffers = {} for name, shape in self.input_shapes.items(): if name in self.state: continue value = self.prepare_input(name, inputs) - buffer = aligned_tensor(shape, value.dtype, value.device) - buffer.copy_(value) - input_buffers[name] = buffer - self.context.set_tensor_address(name, buffer.data_ptr()) - - outputs = {} - for name, shape in self.output_shapes.items(): - output = aligned_tensor(shape, self.tensor_dtypes[name], "cuda") - outputs[name] = output - self.context.set_tensor_address(name, output.data_ptr()) + self.input_buffers[name].copy_(value) if not self.context.execute_async_v3(stream.cuda_stream): raise RuntimeError("TensorRT execution failed") stream.synchronize() - return outputs + return self.output_buffers🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@examples/onnx_ptq/far3d/evaluate.py` around lines 59 - 142, Pre-allocate reusable non-state input and output buffers in TensorRTRunner.__init__, using the fixed input_shapes, output_shapes, and tensor_dtypes, and bind their addresses once. Update __call__ to copy prepared inputs into the existing input buffers and reuse the existing output buffers instead of calling aligned_tensor for each invocation, while preserving state-buffer handling and returned outputs.examples/onnx_ptq/far3d/quantize.py (1)
76-99: 🎯 Functional Correctness | 🔵 Trivial | 💤 Low valueDefensive gap: empty/duplicate
node.namesilently truncates the exclusion BFS.
nodes_by_name[node.name] = nodeoverwrites on name collisions, and the BFS'sif not name: continue(line 94) drops any node with an empty name fromexcludedand stops traversing its outputs — so any accuracy-sensitive node reachable only through an unnamed node would be missed. Likely low-probability given this graph's nodes carry meaningful names (OSA4_5,lateral_convs), but worth a defensive guard since a silent miss here degrades accuracy without any error.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@examples/onnx_ptq/far3d/quantize.py` around lines 76 - 99, The find_encoder_nodes_to_exclude function must preserve BFS traversal through unnamed and duplicate-name nodes instead of indexing nodes solely by node.name. Track graph nodes and their outputs using stable node identities or an adjacency structure that does not overwrite collisions, include unnamed nodes in traversal, and only use names when adding valid exclusions while continuing through every reachable node.examples/onnx_ptq/far3d/Dockerfile (1)
1-81: 🔒 Security & Privacy | 🔵 Trivial | 💤 Low valueRoot user (Trivy DS-0002).
No
USERdirective; container runs as root. This matches other example Dockerfiles in the repo and the README's documented--privileged --gpus=allworkflow, so the practical risk is low for a local example build.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@examples/onnx_ptq/far3d/Dockerfile` around lines 1 - 81, Add a non-root runtime user and switch to it with a USER directive in the Dockerfile, ensuring the user can access and execute the installed /opt/far3d and /opt/modelopt contents. Preserve the existing dependency installation steps, which must continue running as root before the user switch.Source: Linters/SAST tools
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@examples/onnx_ptq/far3d/quantize.py`:
- Around line 118-133: The nodes passed by quantize_encoder through
nodes_to_exclude must match only the exact excluded node names. Transform the
result of find_encoder_nodes_to_exclude into escaped, fully anchored regex
patterns before supplying it to quantize, preserving the existing exclusion set
without allowing prefix or metacharacter-based matches.
---
Nitpick comments:
In `@examples/onnx_ptq/far3d/Dockerfile`:
- Around line 1-81: Add a non-root runtime user and switch to it with a USER
directive in the Dockerfile, ensuring the user can access and execute the
installed /opt/far3d and /opt/modelopt contents. Preserve the existing
dependency installation steps, which must continue running as root before the
user switch.
In `@examples/onnx_ptq/far3d/evaluate.py`:
- Around line 257-260: Move import_modules_from_strings from mmcv.utils into the
top-level import section of evaluate.py, alongside the existing mmcv import, and
remove the local import inside the cfg.get("custom_imports") conditional.
Preserve the conditional invocation and its existing arguments.
- Around line 59-142: Pre-allocate reusable non-state input and output buffers
in TensorRTRunner.__init__, using the fixed input_shapes, output_shapes, and
tensor_dtypes, and bind their addresses once. Update __call__ to copy prepared
inputs into the existing input buffers and reuse the existing output buffers
instead of calling aligned_tensor for each invocation, while preserving
state-buffer handling and returned outputs.
In `@examples/onnx_ptq/far3d/quantize.py`:
- Around line 76-99: The find_encoder_nodes_to_exclude function must preserve
BFS traversal through unnamed and duplicate-name nodes instead of indexing nodes
solely by node.name. Track graph nodes and their outputs using stable node
identities or an adjacency structure that does not overwrite collisions, include
unnamed nodes in traversal, and only use names when adding valid exclusions
while continuing through every reachable node.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Enterprise
Run ID: 85ac9fde-ebb2-416e-9794-10fdd8a25573
📒 Files selected for processing (10)
.pre-commit-config.yamlLICENSEexamples/onnx_ptq/README.mdexamples/onnx_ptq/far3d/Dockerfileexamples/onnx_ptq/far3d/README.mdexamples/onnx_ptq/far3d/evaluate.pyexamples/onnx_ptq/far3d/far3d_pytorch_25_06.patchexamples/onnx_ptq/far3d/prepare_calibration.pyexamples/onnx_ptq/far3d/prepare_metadata.pyexamples/onnx_ptq/far3d/quantize.py
Signed-off-by: ajrasane <131806219+ajrasane@users.noreply.github.com>
cjluo-nv
left a comment
cjluo-nv
left a comment
There was a problem hiding this comment.
Bot review (bedrock-claude-opus-4-8) — DM the bot to share feedback.
Re-review of PR #2012 (FAR3D ONNX PTQ + Argoverse 2 eval example, +910/-0, 11 files). The new code remains sound example glue — modelopt.onnx.quantization.quantize args match the current signature, find_encoder_nodes_to_exclude BFS is correct (all OSA4_5 nodes + everything downstream of lateral_convs), and no new subsystem is introduced. No prompt-injection attempts in the untrusted content (CodeRabbit's "Autofix"/"AI Agents" prose is tooling boilerplate, not directives I acted on).
Addressed since the prior review:
- 💬 CHANGELOG concern resolved — the diff now adds a proper
CHANGELOG.rstbullet under the Experimental section for the FAR3D example.
Why nudge (not approve):
- Licensing — human sign-off required (cannot auto-approve). Unchanged from the prior review: this PR edits the top-level
LICENSE(addsCopyright (c) OpenMMLab. All rights reserved.under Apache-2.0),evaluate.pycarries stacked OpenMMLab/"Modified by Zhiqi Li" + NVIDIA SPDX headers (adapted from DL4AGX/megvii Far3D), and vendorsfar3d_pytorch_25_06.patchmodifying third-party FAR3D source. The attribution followsCONTRIBUTING.md's documented third-party format (reference link w/ commit hash → upstream copyright → NVIDIA header, LICENSE updated, file excluded from the license hook) and looks conscientious — but per focus area 6 and CONTRIBUTING's OSRB requirement this needs a human/legal reviewer to sign off. nodes_to_excludenot anchored (CodeRabbit "major", still open). Confirmedexpand_node_names_from_patternsinmodelopt/onnx/quantization/graph_utils.pyresolves exclusions viare.match(pattern, node.name), so raw names act as unanchored prefix patterns and any regex metacharacter (.,+,() in a node name is interpreted specially — a bad metachar could even raisere.error. For this example the over-match direction is conservative (keeps more nodes in high precision), so it's hardening rather than an accuracy bug, but worth escaping+anchoring (re.escape(name)+^…$).- No tests / no CI path. Consistent with the example-dir convention (GPU/TRT/mmdet3d-only), but the self-contained
find_encoder_nodes_to_excludecould get a small unit test if the team wants any coverage. Owner's call.
Minor CodeRabbit nits still open (non-blocking): local mmcv.utils import in evaluate.py should move to top of file (no stated circular/optional/heavy justification, and mmcv is already imported at module top); TRT I/O buffers reallocated per call; Dockerfile runs as root. Recommend owner (ajrasane / onnx codeowners) sign off, primarily on the licensing.
Signed-off-by: ajrasane <131806219+ajrasane@users.noreply.github.com>
cjluo-nv
left a comment
cjluo-nv
left a comment
There was a problem hiding this comment.
Bot review (bedrock-claude-opus-5) — DM the bot to share feedback.
Re-review of PR #2012 (FAR3D ONNX PTQ + Argoverse 2 eval example, +913/-0, 11 files).
Short answer to "should I approve?" — Not on my recommendation alone. Everything I can check mechanically is now in good shape, but the licensing surface requires human/OSRB sign-off, so I can't auto-approve. If you already have (or can get) legal/OSRB confirmation on the third-party attribution, this is fine to merge from a code perspective.
Resolved since the prior review:
- 💬
nodes_to_excludeanchoring — addressed in commit1808c72. Verified in the current file:quantize_encodernow buildsrf"^{re.escape(name)}$"patterns, which is exactly right givenexpand_node_names_from_patternsresolves exclusions withre.match. No more prefix/metacharacter over-match. - 💬 CHANGELOG — resolved; an Experimental bullet for the FAR3D example is now in
CHANGELOG.rst.
Correctness re-check (no new issues): modelopt.onnx.quantization.quantize kwargs all match the current signature; find_encoder_nodes_to_exclude is a single forward pass over topologically-sorted nodes that correctly captures all OSA4_5 nodes plus everything downstream of lateral_convs (propagation stops at OSA nodes, which are excluded anyway); calibration readers, decoder temporal-state handling, and the --fp16-decoder path are internally consistent with the README. Still pure example glue — no new subsystem, no duplicated in-repo functionality. No prompt-injection attempts in the untrusted content (CodeRabbit's "AI Agents"/Autofix prose is tooling boilerplate; I did not treat it as instructions).
Why nudge rather than approve:
- Licensing — human sign-off required. Unchanged: top-level
LICENSEgainsCopyright (c) OpenMMLab. All rights reserved.under Apache-2.0;evaluate.pycarries stacked OpenMMLab / "Modified by Zhiqi Li" + NVIDIA SPDX headers (adapted from DL4AGXtest_tensorrt.py, itself from megvii-research/Far3D);far3d_pytorch_25_06.patchvendors a modification to third-party FAR3D source. Attribution follows CONTRIBUTING's documented format (reference link with commit hash → upstream copyright → NVIDIA header, LICENSE updated, file excluded from the license pre-commit hook) and looks conscientious — but this is a legal call, not mine. - No tests / no CI-exercisable path. Consistent with the
examples/convention and the GPU/TRT/mmdet3d-only workflow.find_encoder_nodes_to_excluderemains the one self-contained piece that could get a small unit test — owner's call.
Minor, non-blocking (still open): the from mmcv.utils import import_modules_from_strings inside main() (evaluate.py ~L257) should move to the top of the file — mmcv is already imported at module level and no circular/optional/heavy justification is given. Also unchanged: TRT I/O buffers reallocated per __call__ (hot path across 23.5k frames), and the Dockerfile runs as root (matches other example images).
Signed-off-by: ajrasane <131806219+ajrasane@users.noreply.github.com>
| far3d.decoder.${precision}.engine | ||
| ``` | ||
|
|
||
| The evaluator reports the dataset metrics, including mAP. The DL4AGX reference reports 0.230 mAP for its INT8 encoder and mixed FP16/FP32 decoder, compared with 0.232 mAP for FP16 encoder and decoder. It reports neither an INT8 decoder nor FP8 results; validate the selected precision on the target TensorRT version and GPU. |
There was a problem hiding this comment.
I don't have a good understanding on this paragraph. How about we just come up with a table showing the scores of the various formats?
Signed-off-by: ajrasane <131806219+ajrasane@users.noreply.github.com>
| # See the License for the specific language governing permissions and | ||
| # limitations under the License. | ||
|
|
||
| FROM nvcr.io/nvidia/pytorch:25.06-py3 |
There was a problem hiding this comment.
Why are we using 1 year old docker?
| RUN env -u PIP_CONSTRAINT /opt/far3d/bin/python -m pip install --no-cache-dir \ | ||
| torch==1.13.1+cu117 \ | ||
| torchvision==0.14.1+cu117 \ | ||
| --extra-index-url https://download.pytorch.org/whl/cu117 | ||
|
|
||
| RUN env -u PIP_CONSTRAINT /opt/far3d/bin/python -m pip install --no-cache-dir \ | ||
| "numpy<1.24" \ | ||
| opencv-python==4.5.5.64 \ | ||
| yapf==0.32.0 \ | ||
| mmcv-full==1.7.0 \ | ||
| -f https://download.openmmlab.com/mmcv/dist/cu117/torch1.13.0/index.html | ||
|
|
||
| RUN env -u PIP_CONSTRAINT /opt/far3d/bin/python -m pip install --no-cache-dir \ | ||
| av2==0.2.1 \ | ||
| einops \ | ||
| "ipython<9" \ | ||
| kornia==0.6.12 \ | ||
| mmdet==2.28.2 \ | ||
| mmsegmentation==0.30.0 \ | ||
| onnx \ | ||
| onnx-graphsurgeon==0.6.1 \ | ||
| onnxruntime \ | ||
| onnxsim \ | ||
| refile \ | ||
| tensorrt-cu12-bindings==10.11.0.33 \ | ||
| --extra-index-url https://pypi.nvidia.com && \ | ||
| mkdir -p /opt/far3d/lib/python3.8/site-packages/tensorrt && \ | ||
| cp /opt/far3d/lib/python3.8/site-packages/tensorrt_bindings/__init__.py \ | ||
| /opt/far3d/lib/python3.8/site-packages/tensorrt/__init__.py && \ | ||
| cp /opt/far3d/lib/python3.8/site-packages/tensorrt_bindings/tensorrt.so \ | ||
| /opt/far3d/lib/python3.8/site-packages/tensorrt/tensorrt.so | ||
|
|
||
| RUN env -u PIP_CONSTRAINT /opt/far3d/bin/python -m pip install --no-cache-dir \ | ||
| "setuptools<81" && \ | ||
| env -u PIP_CONSTRAINT /opt/far3d/bin/python -m pip install --no-cache-dir \ | ||
| --no-build-isolation \ | ||
| mmdet3d==1.0.0rc6 | ||
|
|
||
| COPY . /opt/modelopt | ||
| RUN env -u PIP_CONSTRAINT python -m pip install --no-cache-dir \ | ||
| lief \ | ||
| ml_dtypes \ | ||
| omegaconf \ | ||
| "onnx-graphsurgeon>=0.6.1" \ | ||
| "onnx~=1.21.0" \ | ||
| "onnxruntime-gpu~=1.24.2" \ | ||
| onnxscript \ | ||
| "onnxslim>=0.1.76" \ | ||
| "setuptools>=80" && \ | ||
| env -u PIP_CONSTRAINT python -m pip install --no-cache-dir --no-deps -e /opt/modelopt |
There was a problem hiding this comment.
can you move all requirements.txt and install from it instead of mentioning in dockerfile so nspect scanning can detect these dependencies
| parser.add_argument("calibration_dir", help="Directory created by prepare_calibration.py") | ||
| parser.add_argument("--quantization-mode", choices=("int8", "fp8"), default="int8") |
There was a problem hiding this comment.
Can we use same approach either parser.add_argument("xyz") or parser.add_argument("--xyz")?
Also can we use either --arg-name or --arg_name for consistency
| COPY . /opt/modelopt | ||
| RUN env -u PIP_CONSTRAINT python -m pip install --no-cache-dir \ |
There was a problem hiding this comment.
Can we instead do pip install -e .[onnx] here?
| --no-build-isolation \ | ||
| mmdet3d==1.0.0rc6 | ||
|
|
||
| COPY . /opt/modelopt |
There was a problem hiding this comment.
Can we use /opt/Model-Optimizer naming? Thats the convention we use in other places also (e.g. Megatron-Bridge)
3 participants
What does this PR do?
Type of change: new example
Adds an end-to-end FAR3D ONNX PTQ example under
examples/onnx_ptq/far3d. The example prepares Argoverse 2 validation metadata and calibration batches, quantizes the image encoder to INT8, builds TensorRT engines, and evaluates 3D object detection accuracy.It also provides a reproducible FAR3D runtime image, preserves accuracy-sensitive encoder layers in high precision, and supports temporal decoder state during evaluation.
Usage
Testing
Before your PR is "Ready for review"
Make sure you read and follow Contributor guidelines and your commits are signed (
git commit -s -S).Make sure you read and follow the Security Best Practices (e.g. avoiding hardcoded
trust_remote_code=True,torch.load(..., weights_only=False),pickle, etc.).CONTRIBUTING.md: ✅Additional Information
Summary by CodeRabbit