Bibliographic record
Abstract
Highlights New / Updated PTQ Algorithms: Qronos support https://github.com/Xilinx/brevitas/pull/1311 Fixes / improvements to rotation equalization https://github.com/Xilinx/brevitas/pull/1310, https://github.com/Xilinx/brevitas/pull/1312 DDP-like bias correction for SDXL (experimental) https://github.com/Xilinx/brevitas/pull/1342 Improved layer support: Quantization of SDPA without FX https://github.com/Xilinx/brevitas/pull/1299 New export flows: Initial SHARK Export support: https://github.com/Xilinx/brevitas/pull/1300 Initial GGUF Export: https://github.com/Xilinx/brevitas/pull/1291 Improved examples: Qronos examples (paper) https://github.com/Xilinx/brevitas/pull/1311 "Benchmark" experiments for stable diffusion, imagenet https://github.com/Xilinx/brevitas/pull/1281 Post-training model expansion examples (paper) https://github.com/Xilinx/brevitas/pull/1355 Allow signed scales https://github.com/Xilinx/brevitas/pull/1308 QONNX export with dynamo=True https://github.com/Xilinx/brevitas/pull/1234 What's Changed Fix (setup): solve incompatibility between isort and yapf by @Giuseppe5 in https://github.com/Xilinx/brevitas/pull/1293 Feat (graph/hadamard): 152 had support by @Giuseppe5 in https://github.com/Xilinx/brevitas/pull/1295 feat (ex/llm): fully parametrise attention quantization by @nickfraser in https://github.com/Xilinx/brevitas/pull/1287 Feat (ex/llm): "auto" dtype by @pablomlago in https://github.com/Xilinx/brevitas/pull/1301 Setup: update transformers version by @Giuseppe5 in https://github.com/Xilinx/brevitas/pull/1304 Feat (brevitas_examples/llm): quant SDPA without FX by @Giuseppe5 in https://github.com/Xilinx/brevitas/pull/1299 Fix (brevitas_examples/llm): update README and yaml by @Giuseppe5 in https://github.com/Xilinx/brevitas/pull/1305 Setup: temporary pin pytest by @Giuseppe5 in https://github.com/Xilinx/brevitas/pull/1307 Feat (brevitas_examples/llm): configurable expansion step by @Giuseppe5 in https://github.com/Xilinx/brevitas/pull/1280 Versioning support in documentation by @Giuseppe5 in https://github.com/Xilinx/brevitas/pull/1298 Feat (graph/rotate): improve R2 region in SDPA by @Giuseppe5 in https://github.com/Xilinx/brevitas/pull/1310 Docs: improve docs build by @Giuseppe5 in https://github.com/Xilinx/brevitas/pull/1314 Fix (graph/rotation): rotation on subset of channels for SDPA by @Giuseppe5 in https://github.com/Xilinx/brevitas/pull/1312 Feat (brevitas_examples/llm): GGUF export by @Giuseppe5 in https://github.com/Xilinx/brevitas/pull/1291 Setup: bump torch version by @Giuseppe5 in https://github.com/Xilinx/brevitas/pull/1205 Fix (ex/imagenet): add forward pass in imagenet ptq example by @Giuseppe5 in https://github.com/Xilinx/brevitas/pull/1316 Feat (qronos): initial implementation of Qronos by @i-colbert in https://github.com/Xilinx/brevitas/pull/1311 Fix (graph/equalize): correct class check during rotation merging by @Giuseppe5 in https://github.com/Xilinx/brevitas/pull/1317 docs (core): Typo fix in docstring by @nickfraser in https://github.com/Xilinx/brevitas/pull/1318 Fix (llm): removing duplicate set_seed function by @i-colbert in https://github.com/Xilinx/brevitas/pull/1319 Fix (ex/common): save scales during optimization by @pablomlago in https://github.com/Xilinx/brevitas/pull/1313 Fix (llm): compatibility with non-uniform RMSNorm shapes by @i-colbert in https://github.com/Xilinx/brevitas/pull/1324 Fix (graph/gpxq): fix memory leak with weight_orig by @Giuseppe5 in https://github.com/Xilinx/brevitas/pull/1325 Shark LLM export by @Giuseppe5 in https://github.com/Xilinx/brevitas/pull/1300 Feat (scaling): rescaled min-max scaling and zero point by @i-colbert in https://github.com/Xilinx/brevitas/pull/1320 Fix (gguf): derived modify_tensors() returns generator by @i-colbert in https://github.com/Xilinx/brevitas/pull/1329 Fix (gpxq): device management of weight_orig for GPxQ by @i-colbert in https://github.com/Xilinx/brevitas/pull/1330 Fix (gguf): resolving zero point permutation issue with LlamaModel by @i-colbert in https://github.com/Xilinx/brevitas/pull/1332 Feat (examples): refactor imagenet and stable_diffusion entrypoints by @pablomlago in https://github.com/Xilinx/brevitas/pull/1281 Feat: skipping rotation optimization with load_checkpoint by @i-colbert in https://github.com/Xilinx/brevitas/pull/1331 Feat (core): Remove assumptions on positiveness of scales by @pablomlago in https://github.com/Xilinx/brevitas/pull/1308 Fix (export/qonnx): Add export support with dynamo=True by @nickfraser in https://github.com/Xilinx/brevitas/pull/1234 Fix (core/ops_ste): preserve dtype during clamp by @Giuseppe5 in https://github.com/Xilinx/brevitas/pull/1340 Fix (eq/rotation): find value if passed as kwarg by @nickfraser in https://github.com/Xilinx/brevitas/pull/1338 Feat (ex): tests Stable Diffusion and ImageNet by @pablomlago in https://github.com/Xilinx/brevitas/pull/1339 Feat (ex/sdxl): DDP-like bias correction for SDXL by @pablomlago in https://github.com/Xilinx/brevitas/pull/1342 Feat (graph): Minor refactoring layerwise_layer_handler by @pablomlago in https://github.com/Xilinx/brevitas/pull/1335 Fix (torch_utils): remove deprecated functions by @Giuseppe5 in https://github.com/Xilinx/brevitas/pull/1344 Fix (setup): test against latest 2.1 torch by @Giuseppe5 in https://github.com/Xilinx/brevitas/pull/1348 Rotation fix by @Giuseppe5 in https://github.com/Xilinx/brevitas/pull/1334 Docs (qronos): adding docs and configs by @i-colbert in https://github.com/Xilinx/brevitas/pull/1326 Feat: Added ONNX export to BNN-PYNQ example by @nickfraser in https://github.com/Xilinx/brevitas/pull/916 Fix (deps): set accelerate<1.10 by @nickfraser in https://github.com/Xilinx/brevitas/pull/1352 Fix (copyright): Fix some missing copyright headers by @nickfraser in https://github.com/Xilinx/brevitas/pull/1353 Fix (ex/llm/benchmark): import error by @nickfraser in https://github.com/Xilinx/brevitas/pull/1356 Setup: temporarily pin diffusers version by @Giuseppe5 in https://github.com/Xilinx/brevitas/pull/1360 Fix (core/scaling): handle edge cases with signed scale by @Giuseppe5 in https://github.com/Xilinx/brevitas/pull/1347 Fix(core/stochastic_round): adjust stochastic round device by @Giuseppe5 in https://github.com/Xilinx/brevitas/pull/1359 Feat (papers): expansion paper configs by @Giuseppe5 in https://github.com/Xilinx/brevitas/pull/1355 Fix (docs): correct link to docs in initial README by @Giuseppe5 in https://github.com/Xilinx/brevitas/pull/1361 Fix (setup.py): Update author and contact in setup.py by @nickfraser in https://github.com/Xilinx/brevitas/pull/1365 Docs (README): Update maximum PyTorch version by @nickfraser in https://github.com/Xilinx/brevitas/pull/1366 Docs (getting started): Typo fix in example by @nickfraser in https://github.com/Xilinx/brevitas/pull/1367 requirements: Updated PyTorch, python versions by @nickfraser in https://github.com/Xilinx/brevitas/pull/1370 Feat (core/scaling): Add option to restrict the output of (scale / threshold) by @nickfraser in https://github.com/Xilinx/brevitas/pull/1369 deps (ex) update accelerate version by @nickfraser in https://github.com/Xilinx/brevitas/pull/1371 Full Changelog: https://github.com/Xilinx/brevitas/compare/v0.12.0...v0.12.1
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.008 |
| Meta-epidemiology (narrow) | 0.005 | 0.002 |
| Meta-epidemiology (broad) | 0.002 | 0.002 |
| Bibliometrics | 0.003 | 0.003 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.005 | 0.005 |
| Open science | 0.006 | 0.003 |
| Research integrity | 0.003 | 0.004 |
| Insufficient payload (model declined to judge) | 0.364 | 0.398 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; the direct Gemma label and the distilled Codex classifier agree on what is shown here.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".