Peak memory under 8-bit and 4-bit quantization relative to BF16 (Saadat and Fellay 2026 Figure 5)
GPU memory of each model at BF16, 8-bit and 4-bit weights (bitsandbytes), alongside embedding cosine similarity to BF16.
Overview
An explanatory profile has not yet been reviewed for this record.
Consult the linked sources for architecture or protocol details. Missing evidence is not evidence of a missing capability.
2 recorded evaluations, 2 metric rows. A comparison chart has not yet been validated for these results. The table retains the individual findings and their sources.
Results
Results are available, but no reviewed comparison panel is linked in this release.
All evaluations
2 evaluations · 2 results. Different protocols are not a single leaderboard.
Protocol: Peak memory under 8-bit and 4-bit quantization relative to BF16 (Saadat and Fellay 2026 Figure 5) · Dataset: Synthetic uniform random DNA sequences, seed 42 (Saadat and Fellay 2026)
Sorted by Peak memory (Embedding inference with 4-bit weights; the paper calls it parameter memory in Section 2.4 and peak memory in Figure 5) (lower is better). The best value in each column is highlighted. Decimals are rounded for display; each value links to the printed value and its source.
| Tested configuration | Peak memory (Embedding inference with 4-bit weights; the paper calls it parameter memory in Section 2.4 and peak memory in Figure 5) | Peak memory (Embedding inference with BF16 weights; the paper calls it parameter memory in Section 2.4 and peak memory in Figure 5) |
|---|---|---|
| NT 2.5B, 4-bit weights for embedding inference (Saadat and Fellay 2026)Evaluation | 5.6 | Not reported |
| NT 2.5B, BF16 weights for embedding inference (Saadat and Fellay 2026)Evaluation | Not reported | 6.7 |
All 2 result rows with coverage, uncertainty and sources
Filter evaluations
Applied filters: All linked evaluations
Source checking is not independent reproduction. Release 2026-10-22-1b7cf027e4f6.
Methods and evaluation design
Procedure, tasks and evaluated configurations
Recorded evaluations
Each evaluation records what was tested and under which conditions.
Baseline coverage
Reference methods help show what a model adds beyond simple controls. We track a null control and a conventional method for each protocol.
0 of 2 active baseline roles have published Rewire measurements in this release. Measurements on a selected protocol do not establish coverage of an entire suite.
No execution recipe linked to this protocol. Recipe availability does not establish a completed evaluation.
- External evaluations
- 2
Literature evidence is not a Rewire measurement. Executed but unpublished runs and private review status are not included.
Null control
Proposed control: requires review
Select a task-valid null control after reviewing inputs and metric
Protocol-specific applicability, permitted inputs, access, split, evaluator and execution requirements need review before implementation or execution.
This is a suggested selection rule, not a validated method or a measured score.
Conventional reference
Proposed control: requires review
Select an upstream conventional reference after reviewing the full protocol
Protocol-specific applicability, permitted inputs, access, split, evaluator and execution requirements need review before implementation or execution.
This is a suggested selection rule, not a validated method or a measured score.
Protocol coverage CSV (gzip) · Model evaluation matrix (gzip) · Source table (gzip) · Release and checksums (gzip)
Coverage is derived from release 2026-10-22-1b7cf027e4f6. Source citations describe the original records; they do not validate an unreviewed baseline proposal. No results have been generated by this audit.
Run instructions
No runnable recipe has been reviewed for this protocol. Dataset access, model requirements, licences and compute requirements must be checked against its sources before execution.
Strengths, limitations and unresolved questions
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.
0 evidence rows matching the loaded filters
| Property and statement | Original source and location | Review and provenance |
|---|
No evidence rows match these filters. Choose another scope or clear the search.
Sources and history
Release 2026-10-22-1b7cf027e4f6 · Record review: source checked
1 source record and release history
- Optimizing genomic language models for efficient training, fine-tuning, and inference (Saadat and Fellay, ICLR 2026 workshop) · Original source · OpenReview PDF headed 'Published as a workshop paper at ICLR 2026' (11 pages; PDF CreationDate 2026-03-09)
Technical metadata and extraction receipts
Stable ID: saadat-fellay-20261015-protocol-quantization-peak-memory
- areas
- dna-genomes
- contexts
- research
- protocol
- Load each model in BF16, 8-bit and 4-bit (bitsandbytes; a 'conservative strategy' for NT models), compute embeddings and report memory and cosine similarity of the embeddings to BF16.
- metric
- peak-memory
- unit
- gigabyte
- version
- ICLR 2026 workshop paper, Section 3.5
- limitations
- Section 2.4 calls the measure parameter memory; Figure 5 labels it peak memory. The printed NT 2.5B values (6.7 and 5.6 GB) also sit slightly above the plotted bars (about 6.6 and 5.45 GB). See source_discrepancy on the two NT 2.5B results.; Input length, batch size and input sequences are not stated; the link to the synthetic data follows the paper's two input categories and is not stated for this experiment.; The 'conservative strategy' for NT models is not defined.; Only the NT 2.5B BF16 and 4-bit values are printed; the other bars are plotted without values.
- source locator
- Section 2.4 item 5; Section 3.5 and Figure 5 (page 7)
- missing metadata
- replicates: reason: unreported
- review
- reviewer: claude; reviewer note: Claude review agent; no human review claimed. Independent Claude (Opus 5.5) review agent, separate from the extraction agent that built this batch. Automated review only. Transcription, not reproduction.; reviewed at: 2026-10-15T22:00:00Z; artifact sha256: 4bc2452110f280a7ba8baa4c7c28beac89769511019666bc77fc33af9b3bcd60; retrieval url: https://openreview.net/pdf?id=NUleWh1GFx; method: source-hash-verification; pdf-text-parse; ai-assisted-source-review; scope-review; method note: Descriptive fields, identity links and missing-metadata reasons checked against the PDF.; note: Limitation on the evidence concern corrected in review (no such concern exists); synthetic-data link kept as an inference the protocol discloses.
Related records
- uses data: Synthetic uniform random DNA sequences, seed 42 (Saadat and Fellay 2026)
- assessment: NT 2.5B 4-bit inference memory
- assessment: NT 2.5B BF16 inference memory