Relevance of NTv3 H100 GPU-hours per training stage (Boshar et al. 2025 Table 5) to "Select an execution workflow for large-scale diagnostic genomics"
- Use case
- Select an execution workflow for large-scale diagnostic genomics
- Assessed by
- NTv3 H100 GPU-hours per training stage (Boshar et al. 2025 Table 5)
- Relevance
- outside_scope
- Endpoint
- H100 GPU-hours per training stage (pre-training, length extension, post-training) for NTv3 8M, 100M and 650M (not shown as evidence)
Outside scope. Pre-training and post-training compute is spent once by the model's developers; a diagnostic team selecting an execution workflow uses the trained checkpoint, so these GPU-hours say nothing about the runtime, throughput or memory of running the model on its samples. Recorded so the figures are not read as execution cost. The authors' statements on inference hardware are kept as a claim (gfm-guide-20261014-claim-ntv3-inference-hardware), not a result.
Evidence
Source checking verifies the cited claim or transcription. It does not establish independent reproduction.
Evidence table
Inspect claims, sources and review details
Trace each statement to its source and review. A context-only reference supports the record generally; it does not verify an individual field. Source checking does not reproduce an experiment.
One row per statement and cited source. Multiple citations are not independent evaluations. Shared locators are labelled explicitly.
15 evidence rows matching the loaded filters
| Property and statement | Original source and location | Review and provenance |
|---|---|---|
Citation locations1 values
| A foundational model for joint sequence-function multi-species modeling at scale for long-range genomic prediction (NTv3), bioRxiv v1 gfm-guide-20261014-source-ntv3-v1: Table 5 and section 4.4.4 'Training Infrastructure' (page 34) Version: bioRxiv v1, posted 2025-12-25 (dated 2025-12-22 on the title page) | not individually reviewed No individual claim review recorded Audit detailsField: Source artifact SHA-256: Hash scope: SHA-256 of one retrieval of the bioRxiv PDF (2026-10-14T12:05:27Z). bioRxiv stamps the PDF on each download (XMP ModifyDate, MetadataDate, DocumentID and InstanceID, the Info ModDate and the trailer /ID), so later retrievals differ. The extractor's copy was not retained, so the pinned bytes could not be re-verified. The independent reviewer's retrieval on 2026-10-14 (15,535,038 bytes, 107 pages, ModDate D:20261014041002-07'00') had SHA-256 082ea9b096d634cad3fc568465ef887d2529060909777685261cfc0bb43fcbee; all 13 Table 5 cells and the quoted section 4.4.4 and Results sentences match in that copy. |
Constraints1 values
| A foundational model for joint sequence-function multi-species modeling at scale for long-range genomic prediction (NTv3), bioRxiv v1 gfm-guide-20261014-source-ntv3-v1: Table 5 and section 4.4.4 'Training Infrastructure' (page 34) Version: bioRxiv v1, posted 2025-12-25 (dated 2025-12-22 on the title page) | not individually reviewed No individual claim review recorded Audit detailsField: Source artifact SHA-256: Hash scope: SHA-256 of one retrieval of the bioRxiv PDF (2026-10-14T12:05:27Z). bioRxiv stamps the PDF on each download (XMP ModifyDate, MetadataDate, DocumentID and InstanceID, the Info ModDate and the trailer /ID), so later retrievals differ. The extractor's copy was not retained, so the pinned bytes could not be re-verified. The independent reviewer's retrieval on 2026-10-14 (15,535,038 bytes, 107 pages, ModDate D:20261014041002-07'00') had SHA-256 082ea9b096d634cad3fc568465ef887d2529060909777685261cfc0bb43fcbee; all 13 Table 5 cells and the quoted section 4.4.4 and Results sentences match in that copy. |
| Endpoint H100 GPU-hours per training stage (pre-training, length extension, post-training) for NTv3 8M, 100M and 650M (not shown as evidence) Context-only references | A foundational model for joint sequence-function multi-species modeling at scale for long-range genomic prediction (NTv3), bioRxiv v1 gfm-guide-20261014-source-ntv3-v1: Table 5 and section 4.4.4 'Training Infrastructure' (page 34) Version: bioRxiv v1, posted 2025-12-25 (dated 2025-12-22 on the title page) | not individually reviewed No individual claim review recorded Audit detailsField: Source artifact SHA-256: Hash scope: SHA-256 of one retrieval of the bioRxiv PDF (2026-10-14T12:05:27Z). bioRxiv stamps the PDF on each download (XMP ModifyDate, MetadataDate, DocumentID and InstanceID, the Info ModDate and the trailer /ID), so later retrievals differ. The extractor's copy was not retained, so the pinned bytes could not be re-verified. The independent reviewer's retrieval on 2026-10-14 (15,535,038 bytes, 107 pages, ModDate D:20261014041002-07'00') had SHA-256 082ea9b096d634cad3fc568465ef887d2529060909777685261cfc0bb43fcbee; all 13 Table 5 cells and the quoted section 4.4.4 and Results sentences match in that copy. |
| Claimed field links:assessed_by:gfm-guide-20261014-protocol-ntv3-training-gpu-hours Context-only references | A foundational model for joint sequence-function multi-species modeling at scale for long-range genomic prediction (NTv3), bioRxiv v1 gfm-guide-20261014-source-ntv3-v1: Table 5 and section 4.4.4 'Training Infrastructure' (page 34) Version: bioRxiv v1, posted 2025-12-25 (dated 2025-12-22 on the title page) | not individually reviewed No individual claim review recorded Audit detailsField: Source artifact SHA-256: Hash scope: SHA-256 of one retrieval of the bioRxiv PDF (2026-10-14T12:05:27Z). bioRxiv stamps the PDF on each download (XMP ModifyDate, MetadataDate, DocumentID and InstanceID, the Info ModDate and the trailer /ID), so later retrievals differ. The extractor's copy was not retained, so the pinned bytes could not be re-verified. The independent reviewer's retrieval on 2026-10-14 (15,535,038 bytes, 107 pages, ModDate D:20261014041002-07'00') had SHA-256 082ea9b096d634cad3fc568465ef887d2529060909777685261cfc0bb43fcbee; all 13 Table 5 cells and the quoted section 4.4.4 and Results sentences match in that copy. |
Limitations3 values
| A foundational model for joint sequence-function multi-species modeling at scale for long-range genomic prediction (NTv3), bioRxiv v1 gfm-guide-20261014-source-ntv3-v1: Table 5 and section 4.4.4 'Training Infrastructure' (page 34) Version: bioRxiv v1, posted 2025-12-25 (dated 2025-12-22 on the title page) | not individually reviewed No individual claim review recorded Audit detailsField: Source artifact SHA-256: Hash scope: SHA-256 of one retrieval of the bioRxiv PDF (2026-10-14T12:05:27Z). bioRxiv stamps the PDF on each download (XMP ModifyDate, MetadataDate, DocumentID and InstanceID, the Info ModDate and the trailer /ID), so later retrievals differ. The extractor's copy was not retained, so the pinned bytes could not be re-verified. The independent reviewer's retrieval on 2026-10-14 (15,535,038 bytes, 107 pages, ModDate D:20261014041002-07'00') had SHA-256 082ea9b096d634cad3fc568465ef887d2529060909777685261cfc0bb43fcbee; all 13 Table 5 cells and the quoted section 4.4.4 and Results sentences match in that copy. |
Pinned records3 values
| A foundational model for joint sequence-function multi-species modeling at scale for long-range genomic prediction (NTv3), bioRxiv v1 gfm-guide-20261014-source-ntv3-v1: Table 5 and section 4.4.4 'Training Infrastructure' (page 34) Version: bioRxiv v1, posted 2025-12-25 (dated 2025-12-22 on the title page) | not individually reviewed No individual claim review recorded Audit detailsField: Source artifact SHA-256: Hash scope: SHA-256 of one retrieval of the bioRxiv PDF (2026-10-14T12:05:27Z). bioRxiv stamps the PDF on each download (XMP ModifyDate, MetadataDate, DocumentID and InstanceID, the Info ModDate and the trailer /ID), so later retrievals differ. The extractor's copy was not retained, so the pinned bytes could not be re-verified. The independent reviewer's retrieval on 2026-10-14 (15,535,038 bytes, 107 pages, ModDate D:20261014041002-07'00') had SHA-256 082ea9b096d634cad3fc568465ef887d2529060909777685261cfc0bb43fcbee; all 13 Table 5 cells and the quoted section 4.4.4 and Results sentences match in that copy. |
| Rationale Outside scope. Pre-training and post-training compute is spent once by the model's developers; a diagnostic team selecting an execution workflow uses the trained checkpoint, so these GPU-hours say nothing about the runtime, throughput or memory of running the model on its samples. Recorded so the figures are not read as execution cost. The authors' statements on inference hardware are kept as a claim (gfm-guide-20261014-claim-ntv3-inference-hardware), not a result. Context-only references | A foundational model for joint sequence-function multi-species modeling at scale for long-range genomic prediction (NTv3), bioRxiv v1 gfm-guide-20261014-source-ntv3-v1: Table 5 and section 4.4.4 'Training Infrastructure' (page 34) Version: bioRxiv v1, posted 2025-12-25 (dated 2025-12-22 on the title page) | not individually reviewed No individual claim review recorded Audit detailsField: Source artifact SHA-256: Hash scope: SHA-256 of one retrieval of the bioRxiv PDF (2026-10-14T12:05:27Z). bioRxiv stamps the PDF on each download (XMP ModifyDate, MetadataDate, DocumentID and InstanceID, the Info ModDate and the trailer /ID), so later retrievals differ. The extractor's copy was not retained, so the pinned bytes could not be re-verified. The independent reviewer's retrieval on 2026-10-14 (15,535,038 bytes, 107 pages, ModDate D:20261014041002-07'00') had SHA-256 082ea9b096d634cad3fc568465ef887d2529060909777685261cfc0bb43fcbee; all 13 Table 5 cells and the quoted section 4.4.4 and Results sentences match in that copy. |
| Reason Add relevance judgements for the execution and compute protocols of the genomic foundation model guide sources batch (data/omics/gfm-guide-sources-20261014, issue #162), independently reviewed in PR #169. Context-only references | A foundational model for joint sequence-function multi-species modeling at scale for long-range genomic prediction (NTv3), bioRxiv v1 gfm-guide-20261014-source-ntv3-v1: Table 5 and section 4.4.4 'Training Infrastructure' (page 34) Version: bioRxiv v1, posted 2025-12-25 (dated 2025-12-22 on the title page) | not individually reviewed No individual claim review recorded Audit detailsField: Source artifact SHA-256: Hash scope: SHA-256 of one retrieval of the bioRxiv PDF (2026-10-14T12:05:27Z). bioRxiv stamps the PDF on each download (XMP ModifyDate, MetadataDate, DocumentID and InstanceID, the Info ModDate and the trailer /ID), so later retrievals differ. The extractor's copy was not retained, so the pinned bytes could not be re-verified. The independent reviewer's retrieval on 2026-10-14 (15,535,038 bytes, 107 pages, ModDate D:20261014041002-07'00') had SHA-256 082ea9b096d634cad3fc568465ef887d2529060909777685261cfc0bb43fcbee; all 13 Table 5 cells and the quoted section 4.4.4 and Results sentences match in that copy. |
| Relevance outside_scope Context-only references | A foundational model for joint sequence-function multi-species modeling at scale for long-range genomic prediction (NTv3), bioRxiv v1 gfm-guide-20261014-source-ntv3-v1: Table 5 and section 4.4.4 'Training Infrastructure' (page 34) Version: bioRxiv v1, posted 2025-12-25 (dated 2025-12-22 on the title page) | not individually reviewed No individual claim review recorded Audit detailsField: Source artifact SHA-256: Hash scope: SHA-256 of one retrieval of the bioRxiv PDF (2026-10-14T12:05:27Z). bioRxiv stamps the PDF on each download (XMP ModifyDate, MetadataDate, DocumentID and InstanceID, the Info ModDate and the trailer /ID), so later retrievals differ. The extractor's copy was not retained, so the pinned bytes could not be re-verified. The independent reviewer's retrieval on 2026-10-14 (15,535,038 bytes, 107 pages, ModDate D:20261014041002-07'00') had SHA-256 082ea9b096d634cad3fc568465ef887d2529060909777685261cfc0bb43fcbee; all 13 Table 5 cells and the quoted section 4.4.4 and Results sentences match in that copy. |
| Revision 1 Context-only references | A foundational model for joint sequence-function multi-species modeling at scale for long-range genomic prediction (NTv3), bioRxiv v1 gfm-guide-20261014-source-ntv3-v1: Table 5 and section 4.4.4 'Training Infrastructure' (page 34) Version: bioRxiv v1, posted 2025-12-25 (dated 2025-12-22 on the title page) | not individually reviewed No individual claim review recorded Audit detailsField: Source artifact SHA-256: Hash scope: SHA-256 of one retrieval of the bioRxiv PDF (2026-10-14T12:05:27Z). bioRxiv stamps the PDF on each download (XMP ModifyDate, MetadataDate, DocumentID and InstanceID, the Info ModDate and the trailer /ID), so later retrievals differ. The extractor's copy was not retained, so the pinned bytes could not be re-verified. The independent reviewer's retrieval on 2026-10-14 (15,535,038 bytes, 107 pages, ModDate D:20261014041002-07'00') had SHA-256 082ea9b096d634cad3fc568465ef887d2529060909777685261cfc0bb43fcbee; all 13 Table 5 cells and the quoted section 4.4.4 and Results sentences match in that copy. |
Sources and history
Release 2026-10-22-1b7cf027e4f6 · Record review: source checked
1 source record and release history
- A foundational model for joint sequence-function multi-species modeling at scale for long-range genomic prediction (NTv3), bioRxiv v1 · Original source · bioRxiv v1, posted 2025-12-25 (dated 2025-12-22 on the title page)
Technical metadata and extraction receipts
Stable ID: use-case-mapping-gfm-guide-20261014-ntv3-training-gpu-hours
- field
- links:assessed_by:gfm-guide-20261014-protocol-ntv3-training-gpu-hours
- value
- gfm-guide-20261014-protocol-ntv3-training-gpu-hours
- relevance
- outside_scope
- endpoint
- H100 GPU-hours per training stage (pre-training, length extension, post-training) for NTv3 8M, 100M and 650M (not shown as evidence)
- rationale
- Outside scope. Pre-training and post-training compute is spent once by the model's developers; a diagnostic team selecting an execution workflow uses the trained checkpoint, so these GPU-hours say nothing about the runtime, throughput or memory of running the model on its samples. Recorded so the figures are not read as execution cost. The authors' statements on inference hardware are kept as a claim (gfm-guide-20261014-claim-ntv3-inference-hardware), not a result.
- constraints
- Do not cite these GPU-hours as the cost of running NTv3 in a workflow.
- limitations
- Developer-reported.; Table 5 gives H100 GPU-hours for every stage, though the text says some training ran on Cloud TPU v6; whether TPU runs were converted is not stated.; Preprint (bioRxiv v1), not peer reviewed.
- revision
- 1
- reason
- Add relevance judgements for the execution and compute protocols of the genomic foundation model guide sources batch (data/omics/gfm-guide-sources-20261014, issue #162), independently reviewed in PR #169.
- citation locators
- source id: gfm-guide-20261014-source-ntv3-v1; locator: Table 5 and section 4.4.4 'Training Infrastructure' (page 34)
- source locator
- gfm-guide-20261014-source-ntv3-v1: Table 5 and section 4.4.4 'Training Infrastructure' (page 34)
- review
- method: scope-review; reviewer: claude; reviewer note: Claude (Opus 5.5) review agent review-169, independent of the agent that drafted these judgements for issue #162; no human review claimed; reviewed at: 2026-10-14T19:00:00Z; note: Independent relevance review of an outside-scope judgement: pretraining and post-training compute spent once by the developers does not measure running the model in a team's workflow. Grade, endpoint and limitations checked against the #138 protocol and result records reviewed in PR #150; the bioRxiv PDF could not be re-downloaded (HTTP rate limit), so Table 5 was not re-read by this reviewer. No measured evidence is shown. See docs/reviews/use-cases/diagnostic-genomics-model-execution-judgements-2026-10-14.md.
- pins
- record id: gfm-guide-20261014-protocol-ntv3-training-gpu-hours; fields: status; name; description; facets; source_ids; links; attributes; sha256: 70efa307f25be889a266294aa9e70da401a8962aca46b3a66d6145185e47b1a6; record id: gfm-guide-20261014-source-ntv3-v1; fields: status; attributes.artifact_sha256; sha256: 244bdf27c1468908b1caa1e8f07083398952e8318235a3c59db56ec09991e0b2; record id: use-case-diagnostic-genomics-model-execution; fields: status; name; description; facets; attributes.question; attributes.decision; attributes.inputs; attributes.output; attributes.exclusions; attributes.setting; attributes.clinical_scope; attributes.intended_users; sha256: 1c2a12431556cc6581f5aab019cf91576d2393387b88a8a2bea35ea2a3ee4a70