Commit History

19 days ago

  1. Composer final; probes off Running

    Verified signature
    15a87ca

20 days ago

  1. Composer fixes; clean 4-bit probe

    Verified signature
    d335996
  2. Quantization Composer, portable save, KV quant, OpenVINO/AutoRound/SINQ probe

    Verified signature
    ee6a75e
  3. 4-bit backend probe

    Verified signature
    c3e0722
  4. Explicit Load button; env-gated llama.cpp probe

    Verified signature
    1eca699
  5. one logical CPU per core, compressed-tensors float32 on CPU

    Verified signature
    04d4603
  6. THP allocator, core-pinning probe

    Verified signature
    0c38057
  7. bandwidth/hugepage/placement probe, forward-share log, mask-builder kwarg filter

    Verified signature
    c962244
  8. v4 mask-kwarg shim (Spark-X), speculators anchor-sampling detection, CPU draft-length advice

    Verified signature
    0f1956b
  9. normalized-aux drafters (T5Gemma), quantized drafter matmuls, visible generate retries, fp16 speculation advice

    Verified signature
    8922fe1
  10. dynamic_fp16 scheme, eager attention for eager-only remote code, enc-dec drafter start token, quantized draft models, spec timing

    Verified signature
    f9ed677
  11. Speculative: deferred commit for recurrent caches, enc-dec decoder discovery + width check, cumulative aux calibration, variant configs; v4 tied-weights shim; gated/native config guards; loader attempt logging

    Verified signature
    117a403
  12. Speculative: version-tolerant KV crop, real layer-skip probe (block output kinds), repo::variant drafters

    Verified signature
    cce4354
  13. Dynamic int8 in place, one Linear at a time (no deepcopy / full fp32 copy: fixes OOM on 2B models); compile-stable int4/int8 kernels

    Verified signature
    7ca9e8c
  14. Release compiled forward + block drafters on model eviction; raw-logit token probabilities; README

    Verified signature
    1c280ce
  15. QwenScope instruments (two-text comparison, positions, strength presets, original vs steered), speculative decoding, torch.compile, native int4/int8 kernels, sources

    Verified signature
    902a1a8
  16. Speculative decoding (draft, self, DFlash family), torch.compile, int8/int4 kernels

    Verified signature
    703682d
  17. Hardware probe, native int8/int4 CPU kernels

    Verified signature
    dba7466
  18. Log CPU model and native bf16 on a separate INIT line

    Verified signature
    5b43e68
  19. Decide auto dtype from native bf16 flags; log CPU model

    Verified signature
    970e1c9
  20. CPU decode: fp32 auto when bf16 is slower; quanto int8 on bf16 activations; flag CPU float8

    Verified signature
    0b41e30

22 days ago

  1. DFT/ALSO: setup lines to Space log, final row-weight summary, no perplexity for DFT

    Verified signature
    7218d58
  2. DFT training method + native ALSO optimizer (per-row DRO weights)

    Verified signature
    03b83f2
  3. Update Small LM Bistro

    Verified signature
    e762c2d
  4. SAR: drop stale prompts from other behaviours; manifold steering card warns when the layer does not drive the answer

    Verified signature
    5595019
  5. MANIFOLDS tab, SAR, revised SAE training, NLA reading discipline, tuning advisor, sources

    Verified signature
    c1edad6

23 days ago

  1. Rename NLA tab to NLA · SHOW & TELL

    Verified signature
    dcafe31

24 days ago

  1. Hover hints on checkbox labels

    Verified signature
    3ef05f8
  2. Oracle records note attached steering; README

    Verified signature
    3a84f79
  3. Activation Oracle reader in NLA · TALKING CURE

    Verified signature
    b2edced
  4. README: SAE · COANALYSIS, NLA · TALKING CURE, steering, NanoQuant, GPart/GAWS

    Verified signature
    249bdb9
  5. Steering panel: read contrasts from the unsteered model

    Verified signature
    a5bf697
  6. Inference-tab activation steering panel (CAA/ActAdd, directional ablation, α sweep); sink-safe contrasts; hints on file labels

    Verified signature
    c3dc21c
  7. NLA · TALKING CURE tab (Nano-NLA / kitft NLAs: read, scan, model diff, edit steering); SAE tab renamed SAE · COANALYSIS

    Verified signature
    85d4299
  8. Lab: SAE-COANALYST tab, NanoQuant, GPart/GAWS, steering registry

    Verified signature
    2eacce3

25 days ago

  1. Update README.md

    Verified signature
    cad5d38

26 days ago

  1. Auto training precision: keep bf16 for models whose fp32 weights would exceed 1/3 of RAM

    Verified signature
    061fd38
  2. Atelier audit: fix DPO ref-free/losses, CPT chunking, Decision scoring+calibration, Vision SFT, 8/4-bit AdamW, bespoke-model training, hot-swap; chat template presets; Hub model/dataset links

    Verified signature
    f1f0998
  3. Rename tab to ATELIER · TRAINING

    Verified signature
    b3884b3
  4. Support Universal programmatic architecture synthesis, on-the-fly safetensors conversion, and custom in-repo architectures (PoetLM, OpenPeerLLM, Aurora1.0, Aurora-80K, Axiom-1, circuit-sparsity)

    Verified signature
    b8c149e

27 days ago

  1. Support openai/circuit-sparsity and prune unlaunchable non-Transformers models (bistro/engine.py)

    Verified signature
    db4cf67
  2. Support openai/circuit-sparsity and prune unlaunchable non-Transformers models (bistro/arch_bridge.py)

    Verified signature
    f647a85
  3. Support openai/circuit-sparsity and prune unlaunchable non-Transformers models (bistro/config.py)

    Verified signature
    abab780
  4. Quantized checkpoints saved as safetensors (+legacy .pt/.bin loading); architecture bridge (native retry, in-repo code discovery, clear errors)

    Verified signature
    4de6884
  5. Private-repo adapter loading: HF token fields, hot-swap from Hub, checkpoint subfolder support

    Verified signature
    451f248
  6. Fix: native Muse2 support, Mamba/hybrid LoRA target compat (Falcon-H1), Granite-hybrid num_experts config, clearer load errors

    Verified signature
    eedcbc7
  7. Make message entry full-width above action buttons & optimize mobile layout

    Verified signature
    a151e61

28 days ago

  1. Fix base/custom model inference (Pragya-EXP2, Supra2-100M, cagliostro), CPU LoRA meta-tensor/FlexAttention issues & add sub-banner note

    Verified signature
    f23626a
  2. Merge branch #AlekseyCalvin/Small_LM_Bistro' into 'AlekseyCalvin/SLM_Bistro'

    50698e3
  3. Update README.md

    Verified signature
    224b982