Alpamys
|
a4dfbb308c
|
feat(trainer): live TTS / BitNet / MoE-expert-quant trainers (v0.71.20)
Lift three v0.52.0 schema-only NotImplementedError stubs to real code.
- #131 TTS: TTSTrainerWrapper(SFTTrainerWrapper) — TTS fine-tune = next-token
CE over [text][audio-codec-token] chat; per-family emotion templating
(Orpheus/Oute) + codec special-token registration. Pre-encoded chat path
live-validated on SmolLM2-135M-Instruct; live-codec (data.format=audio)
hardware-gated per family.
- #134 BitNet: BitNetTrainerWrapper gated on onebitllms; export --format
bitnet|tq1_0 runs real llama.cpp TQ1_0 ternary GGUF export.
- #136 MoE: apply_moe_expert_quant swaps fused-MoE experts to bnb Linear4bit/
Linear8bitLt (pre-LoRA); train_router_only freezes experts (post-LoRA).
Live-validated on RTX 3050 (dequant err 0.0155).
Review fixes: H1 explicit Params4bit/Int8Params weight-carry; H2 quant
pre-LoRA / freeze post-LoRA + skip PEFT-wrapped modules; M4 device-aware
placement.
Tests 13807 -> 13874 (+69 in test_v07120.py, -2 lifted stubs in test_v0520.py).
|
2026-06-10 12:37:14 +05:00 |
|
Alpamys
|
2292e81c3f
|
feat(modality): v0.53.2 — lift Modality II stubs (distill + classifier + EBFT/GDPO + reasoning_effort)
Closes #132, #133, #135, #137. Records #71 ONNX QA (partial — tiny-gpt2
PASS, TinyLlama-1.1B blocked by host RAM during onnx.load post-process).
New trainer wrappers:
- DistillTrainerWrapper (soup_cli/trainer/distill.py) — student + frozen
teacher, KL/JS divergence kernels scaled by T**2, device-bridge for HF
Trainer auto-CUDA promotion, DataCollatorForSeq2Seq for variable-length
loss-masked rows, separate trust_remote_code resolution per model.
- ClassifierTrainerWrapper (soup_cli/trainer/classifier.py) — single/multi
label sequence classification, 1024-entry multi-label cap, label_names
string-to-int resolution. Routes classifier / reranker / cross_encoder.
Live loss kernels:
- apply_ebft_loss (structured / strided) + attach_ebft_compute_loss (SFT)
- apply_gdpo_loss (standard / length_normalized / margin) +
attach_gdpo_compute_loss (DPO). Both attach hooks idempotent.
Prompt-format wiring:
- apply_reasoning_effort_prefix injects gpt-oss
<|reasoning_effort|>{low,medium,high}<|/reasoning_effort|> header.
- build_assistant_only_labels(train_on_eot=True) keeps EOT/EOS unmasked.
Bugs surfaced + fixed during Wave 3 CPU smoke (regression guards in tests):
- Distill collator did not pad pre-tokenised labels (variable-length crash)
- Distill compute_loss device-mismatch when HF Trainer auto-promoted
student to CUDA while teacher stayed on CPU.
Tests: 7722 -> 7842 (+120 in test_v0532.py). 5 review agents run; every
CRITICAL/HIGH/MEDIUM/LOW finding fixed.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
|
2026-05-13 12:56:50 +05:00 |
|
Alpamys
|
df7f49feda
|
feat(v0.52.0): Modality II — TTS + Distillation + BitNet + EBFT-GDPO + MoE quant + reasoning_effort
7 schema-only Parts; live trainer / loss / export wiring deferred to v0.52.1
(mirrors v0.27.0 MII / v0.50.0 GRPO Plus / v0.51.0 hubs stub-then-live pattern).
- Part A: task='tts' + modality='audio_out' + 5 families (Orpheus/Sesame-CSM/
Llasa/Spark/Oute) + per-family emotion allowlist (Orpheus 8 / Oute 6) +
5 recipes (orpheus-tts-sft, sesame-csm-tts, llasa-tts, spark-tts, oute-tts).
- Part B: classifier / reranker / cross_encoder tasks + num_labels (bool-
before-int guard) + classifier_kind + label_names (dedup + null-byte + cap).
- Part C: task='distill' + teacher_model + distill_divergence (kl alias
canonicalises to forward_kl; Literal excludes alias) + distill_temperature
(math.isfinite + [0.05, 100.0] bounds).
- Part D: quantization='bitnet_1.58' (gated to non-MLX + text + task in
{sft, pretrain, dpo}) + Falcon-E BitNet recipe + soup export
--format bitnet/tq1_0 CLI stubs (yellow deferred-advisory panel, exit 0).
- Part E: EBFT (structured/strided + bounded ebft_temperature; SFT-only)
+ GDPO (standard/length_normalized/margin; DPO-family-only).
- Part F: moe_expert_quant (nf4/int8_rowwise) + train_router_only — both
require moe_lora=true (silent-no-op footgun rejection).
- Part G: reasoning_effort (low/medium/high) + train_on_eot — both gated to
the SFT-family task set (sft/pretrain/distill/classifier/reranker/
cross_encoder); rejected on DPO/GRPO/PPO/etc. with named offenders.
Review fixes (5 agents: python, security, code, tdd, verification-loop-manual):
- num_labels bool-before-int field_validator (security HIGH)
- reasoning_effort + train_on_eot SoupConfig task-gate (code HIGH)
- _validate_classifier_compat lazy-import early-return (code HIGH)
- Oute emotion allowlist via data-driven _FAMILY_EMOTIONS (python+code MED)
- _MAX_LEN -> _MAX_REASONING_EFFORT_LEN (python MED)
- validate_reasoning_effort wired via field_validator (security MED)
- distill_divergence Literal excludes "kl" alias (code MED)
- DIVERGENCES derived from _DIVERGENCE_ALIASES drift guard (python LOW)
- is_bitnet_model comment fixed to match impl (python LOW)
- sister-fn bool guards on every compat helper (python+security MED)
- TDD coverage gaps closed: EBFT variant oversize, GDPO full rejection matrix,
ebft_temperature explicit-exc table, TTS compat input guards, recipe
model-id null/whitespace, full reasoning_effort task-matrix.
Drift fixes:
- tests/test_onnx_tensorrt_export.py + tests/test_awq_gptq_export.py
SUPPORTED_FORMATS count bumped 5 -> 7 (bitnet + tq1_0 stubs).
- tests/test_recipes.py catalog_size assertion 106 -> 112 (5 TTS + Falcon-E).
Test count: 7184 -> 7456 (+272). 230 new tests in tests/test_v0520.py;
remainder from drift-fix parametrize expansions.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
|
2026-05-12 13:38:01 +05:00 |