{"record": {"aliases": ["distilbert-onnx-quantization-benchmark"], "artifacts": [{"id": "gh:parth-udawant/distilbert-onnx-quantization-benchmark", "kind": "repo", "url": "https://github.com/Parth-Udawant/distilbert-onnx-quantization-benchmark"}], "categories": ["General AI", "Language & Knowledge", "cs.AI"], "description": {"en": "An end-to-end ONNX quantization benchmark for a DistilBERT sentiment classifier, comparing PyTorch FP32, ONNX FP32, dynamic INT8, and static INT8 variants on accuracy, latency, and model size using a fixed 200-sample SST-2 validation set."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://github.com/Parth-Udawant/distilbert-onnx-quantization-benchmark", "source": "claire_radar", "source_url": "https://github.com/Parth-Udawant/distilbert-onnx-quantization-benchmark", "title": "DistilBERT ONNX Quantization Benchmark"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "key": "claire-radar:github:parth-udawant/distilbert-onnx-quantization-benchmark", "modality": null, "name": "DistilBERT ONNX Quantization Benchmark", "openness": {"code_license": null, "data_license": null, "evidence": [], "status": "unknown"}, "provenance": {"admission_policy_version": "2026-09-09.1", "confidence": "Medium", "crawl_bundle": "claire_radar_2026-10-14", "crawled_at": "2026-10-14T09:44:51Z", "data_status": "primary-source-candidate", "display_eligible": "false", "origin_record_id": "bm_distilbert-onnx-quantization-benchmark_2d54c553", "origin_source": "github", "recognition_confidence": "0.4", "record_sha256": "6b5193135430775ff4042f863c10ee8092e07325dbbd9dbba1f978c2e528baab", "relation": "unclear", "review_model": "deepseek-v4-pro", "review_state": "source-reviewed-exclude", "reviewed_at": "2026-09-08T16:57:34.057602Z", "source_url": "https://github.com/Parth-Udawant/distilbert-onnx-quantization-benchmark"}, "publisher": null, "released": "2026-09-06", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:github:parth-udawant/distilbert-onnx-quantization-benchmark", "source_url": "https://github.com/Parth-Udawant/distilbert-onnx-quantization-benchmark"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-github-parth-udawant-distilbert-onnx-quantization-benchmark", "source": "claire_radar", "source_benchmark_id": "github:parth-udawant/distilbert-onnx-quantization-benchmark", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": ["distilbert-onnx-quantization-benchmark"], "annotation": "Unknown", "applicationDomains": ["General AI"], "area": "Language & Knowledge", "attention": {"asOf": "2026-10-05", "attemptedAt": "2026-10-04T15:31:48.030916Z", "githubRepo": "https://github.com/Parth-Udawant/distilbert-onnx-quantization-benchmark", "githubScope": "benchmark_repo", "githubStars": 0, "hfDailySubmittedAt": null, "hfDataset": null, "hfDatasetDownloads": null, "hfDatasetLikes": null, "hfPaperUpvotes": null, "hfPaperUrl": null, "observedAt": "2026-10-04T15:31:48.030916Z", "signalStatus": {"githubStars": {"observedAt": "2026-10-04T15:31:48.030916Z", "state": "fresh"}, "hfDatasetDownloads": {"state": "not_applicable"}, "hfDatasetLikes": {"state": "not_applicable"}, "hfPaperUpvotes": {"state": "not_applicable"}}, "status": "fresh"}, "attentionForecast": {"confidence": "Low", "horizon": "7d", "reason": "The DistilBERT quantization comparison is a common educational and deployment topic, but its single-model scope is unlikely to generate significant early interest.", "score": 12}, "capabilities": [], "confidence": "Medium", "construction": "Unknown", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-09-09T01:08:06.842497Z", "inputHash": "ee12c567ddbe3c8ef0cf1ad0813598896ea7edcbb27a32f0617998ab5c408588", "model": "deepseek-v4-pro"}, "curation": {"admissionPolicyVersion": "2026-09-09.1", "canonicalNameEvidence": "# DistilBERT ONNX Quantization Benchmark End-to-end model optimization pipeline for a DistilBERT sentiment classifier: PyTorch to ONNX export, dynamic INT8 quantization, and static (calibration-based) INT8 quantization, with accuracy, latency, and size benchmarked at every step on a fixed evaluation set.", "canonicalNameSource": "official_readme", "decisionReason": "Primary-source review: this is a single DistilBERT quantization experiment on 200 existing SST-2 examples, not a new reusable benchmark family.", "evaluationEvidence": {"reuse": "Reproduce Run in order in Google Colab: 1. `baseline.py` - load model, establish FP32 accuracy/latency/size baseline 2. `onnx_export.py` - export to ONNX, verify correctness vs. PyTorch 3. `reshape_sizes.py` - fix a size-measurement bug (ONNX external data file) 4. `dynamic_quantization.py` - dynamic INT8 quantization + benchmark 5. `static_quantization.py` - static INT8 quantization + benchmark", "scoring": "| Variant | Accuracy | Latency (mean) | Latency (p95) | Size | |---|---|---|---|---| | PyTorch FP32 | 91.0% | 94.7ms | 207.2ms | 255.4MB | | ONNX Runtime FP32 | 91.0% | 43.8ms | 61.4ms | 256.2MB | | ONNX Dynamic INT8 | 90.0% | 22.4ms | 26.0ms | 64.7MB | | ONNX Static INT8 | 89.0% | 37.7ms | 68.4ms | 134.7MB |", "task": "End-to-end model optimization pipeline for a DistilBERT sentiment classifier: PyTorch to ONNX export, dynamic INT8 quantization, and static (calibration-based) INT8 quantization, with accuracy, latency, and size benchmarked at every step on a fixed evaluation set."}, "method": "primary-source-review-v1", "model": "deepseek-v4-pro", "reviewedAt": "2026-09-08T16:57:34.057602Z", "state": "source-reviewed-exclude"}, "dataStatus": "primary-source-candidate", "demo": false, "description": "An end-to-end ONNX quantization benchmark for a DistilBERT sentiment classifier, comparing PyTorch FP32, ONNX FP32, dynamic INT8, and static INT8 variants on accuracy, latency, and model size using a fixed 200-sample SST-2 validation set.", "discoverySources": [{"id": "github:parth-udawant/distilbert-onnx-quantization-benchmark", "type": "github", "url": "https://github.com/Parth-Udawant/distilbert-onnx-quantization-benchmark"}], "displayEligible": false, "domainCuration": {"method": "rules-v1", "state": "auto"}, "evaluationMode": "public_reusable", "evidence": {"reasonCodes": ["discovered via github", "benchmark term in abstract", "evaluation protocol evidence", "no explicit benchmark release evidence"], "snippet": "distilbert-onnx-quantization-benchmark End-to-end ONNX quantization benchmark for DistilBERT — FP32 vs dynamic vs static INT8, comparing accuracy, latency, and model size tradeoffs deep-learning dynamic-quantization huggingface huggingface-transformers machine-learning model-optimization nlp onnx onnx-models onnx-runtime onnxruntime pytorch quantization sentiment-analysis sst2 sst2-dataset # DistilBERT ONNX Quantization Benchmark End-to-end model optimization pipeline for a DistilBERT sentiment "}, "familyId": "bmf_0adcc6e6dae4", "firstSeenAt": "2026-09-07", "heat": null, "id": "bm_distilbert-onnx-quantization-benchmark_2d54c553", "indexedAt": "2026-09-07T06:22:04.880155Z", "industrySectors": [], "links": {"code": "https://github.com/Parth-Udawant/distilbert-onnx-quantization-benchmark"}, "metrics": [], "motivation": "distilbert-onnx-quantization-benchmark End-to-end ONNX quantization benchmark for DistilBERT — FP32 vs dynamic vs static INT8, comparing accuracy, latency, and model size tradeoffs deep-learning dynamic-quantization huggingface huggingface-transformers machine-learning model-optimization nlp onnx onnx-models onnx-runtime onnxruntime pytorch quantization sentiment-analysis sst2 sst2-dataset # DistilBERT ONNX Quantiza…", "name": "DistilBERT ONNX Quantization Benchmark", "oneLine": "An end-to-end ONNX quantization benchmark for a DistilBERT sentiment classifier, comparing PyTorch FP32, ONNX FP32, dynamic INT8, and static INT8 variants on accuracy, latency, and model size using a fixed 200-sample SST-2 validation set.", "paperTitle": "distilbert-onnx-quantization-benchmark", "primaryDomain": "General AI", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://github.com/Parth-Udawant/distilbert-onnx-quantization-benchmark", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-09-07T06:22:04.880155Z"}, "publications": [], "ranking": {"30d": {"confidence": "Low", "coverage": 0.55, "growth": {"components": {"githubStars": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}, "hfDatasetDownloads": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}, "hfPaperUpvotes": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}}, "confidence": "Low", "coverage": 0.0, "elapsedDays": 30, "historyDate": "2026-09-05", "rank": null, "score": null, "windowDays": 30}, "level": {"components": {"githubStars": {"normalization": "log-max", "normalized": 0.0, "value": 0}, "hfDatasetDownloads": {"normalization": "log-max", "normalized": null, "value": null}, "hfPaperUpvotes": {"normalization": "log-max", "normalized": null, "value": null}}, "confidence": "Low", "coverage": 0.55, "rank": 204, "score": 23}, "method": "attention-ranking-v13", "rank": 204, "score": 23}, "90d": {"confidence": "Low", "coverage": 0.55, "growth": {"components": {"githubStars": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}, "hfDatasetDownloads": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}, "hfPaperUpvotes": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}}, "confidence": "Low", "coverage": 0.0, "elapsedDays": null, "historyDate": null, "rank": null, "score": null, "windowDays": 90}, "level": {"components": {"githubStars": {"normalization": "log-max", "normalized": 0.0, "value": 0}, "hfDatasetDownloads": {"normalization": "log-max", "normalized": null, "value": null}, "hfPaperUpvotes": {"normalization": "log-max", "normalized": null, "value": null}}, "confidence": "Low", "coverage": 0.55, "rank": 468, "score": 23}, "method": "attention-ranking-v13", "rank": 468, "score": 23}}, "readiness": "Runnable", "recognitionConfidence": 0.4, "relation": "unclear", "releasedAt": "2026-09-06", "source": {"authors": ["Parth-Udawant"], "categories": [], "id": "github:parth-udawant/distilbert-onnx-quantization-benchmark", "publicSignals": {"githubStars": 0}, "title": "distilbert-onnx-quantization-benchmark", "type": "github", "url": "https://github.com/Parth-Udawant/distilbert-onnx-quantization-benchmark"}, "sourceUpdatedAt": "2026-09-06T17:06:18Z", "topics": ["cs.AI"], "venueAttempts": [], "whyItMatters": "The project offers a focused comparison of quantization strategies for a specific model, but it is a single-model optimization experiment without a distinct reusable evaluation protocol for other teams."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
