ML research notes
Public benchmarks, new capabilities, and the experiments that did not work.
Start with the software
Install Loggetta · Runtime · Kernels · Model and hardware requirements
Research update · 8 October 2026
The Unsloth comparison is now quoted as GPU time. Unsloth spends more of each step on the CPU than e4b does, so its wall-clock time grows on a slower host. The GPU-time ratio is the figure that does not depend on the host; the wall-clock ratio is quoted beside it, on its host. The 6 October figures below were wall readings from before e4b's current defaults.
| Workload | Result | Conditions |
|---|---|---|
| Qwen3-30B-A3B QLoRA | Unsloth spends 1.92× e4b's GPU time per step; wall-clock 2.80× | RTX 5090 on an AMD EPYC 7713 host, both frameworks on one stack, e4b at its defaults; comparable held-out loss; Unsloth peaked lower (24.27 vs 26.16 GB). Evidence |
| Packed 4K Qwen3 training | Unsloth spends 1.160× e4b's GPU time per step | RTX 5090, e4b at its defaults; wall-clock 1.472× on that host. Evidence |
| Wide decode steps (serving) | No measurable quality cost | Qwen3-30B-A3B int4 on one RTX 5090: one 64-row decode step against four 16-row pieces, within the 16-row arithmetic's own variation. Evidence |
Released the same day: experts4bit-qlora 0.50.0, where serve_paged sizes its slots to the GPU's memory and decodes each step as one graph by default (release notes), and Loggetta 0.3.1, a documentation update.
Download the source revisions, result records, and release status as JSON.
Research update · 6 October 2026
Bring your data, keep your adapters. New in Loggetta 0.3.0: local or Hub data, validation before model loading, reusable adapters, and a run report. Save and reload pass CPU tests; the full CUDA test is still to run.
| Workload | Result | Conditions |
|---|---|---|
| Qwen3-30B-A3B QLoRA | 2.352× vs Unsloth; 2.468× on a second host | RTX 5090, matched work and software, comparable held-out loss. Unsloth used 3.22 GB less peak VRAM. Evidence |
| Packed 4K Qwen3 training | 1.453× vs Unsloth, defaults as of 2026-10-06 | RTX 5090, torch 2.12.1, 28.23 vs 24.86 GB peak. Defaults have changed since. Evidence |
| Mixtral-8x7B QLoRA | 1.144× vs Unsloth | Same-stack RTX 5090 run. Comparable held-out loss; Unsloth used less VRAM and energy. Evidence |
The software stack matters. Packed training was slower on torch 2.8, and bucketed padding is not the cause. The decoded training route showed no measurable whole-step saving and stays opt-in. Read the test
Earlier research narratives
These essays preserve the work as it stood when written. Use the current software documentation for installation and support.
Receipts-driven engineeringhistorical
The methodology narrative behind grouped-nf4-gemm's blind-confirmed benchmarks: preregistration, OpenTimestamps stamping, and the published failures.
Published 2026-07-14 · last revised 2026-07-31 (dates from this site's history before the narrative was moved here) · current documentation: grouped-nf4-gemm
experts-4bit, the 2026 narrativehistorical
The earlier experts-4bit story: fused 4-bit experts, QLoRA and offload as they stood in July 2026.
Published 2026-07-09 · last revised 2026-07-31 (dates from this site's history before the narrative was moved here) · current documentation: experts4bit-qlora