Commit History

Phrasebook: multi-alias + French source keys, plus misses-script stub
5815492
Running

jefffffff9 Claude Opus 4.7 commited on

Forbid parenthetical glosses in LLM replies
757e833

jefffffff9 Claude Opus 4.7 commited on

Stage 4: split translate/reply UI + CPU-safe TTS + reply-not-translate prompt
9e99c2c

jefffffff9 Claude Opus 4.7 commited on

Per-turn telemetry + RAG few-shot on phrasebook miss
064d08b

jefffffff9 commited on

Remove stale Qwen references from minimal-baseline docstrings
cc82bd8

jefffffff9 commited on

Ground-zero Stages 1–3: dialect anchors + phrasebook short-circuit + Aya-Expanse
d0e28fa

jefffffff9 commited on

space(ground-zero): point Space at app_minimal.py
26659d8

Broulaye Doumbia commited on

step 2: add app_minimal.py — zero-shot baseline for rebuild.
34edd5f

Broulaye Doumbia commited on

fix read me
da3a060

Broulaye Doumbia commited on

restructuring
76aac1b

Broulaye Doumbia commited on

push docs and script
cc8b90c

Broulaye Doumbia commited on

Add torchcodec install for datasets 4.x audio decoding
bb78cbf

jefffffff9 Claude Opus 4.7 commited on

Prepare training stack for RunPod: env-aware notebook + bootstrap script
9049ef3

jefffffff9 Claude Sonnet 4.6 commited on

Align Whisper default to turbo-v3 + add document upload to Knowledge Base tab
cc50efb

jefffffff9 Claude Sonnet 4.6 commited on

Fix WhisperProcessor import in reload + upgrade base to large-v3-turbo
c33a061

jefffffff9 Claude Sonnet 4.6 commited on

Fix mel-bin mismatch: load per-language processor from fine-tuned checkpoint
7fae91b

jefffffff9 Claude Sonnet 4.6 commited on

Fix jiwer crash on post-normalisation empty refs; register SLR106/105 datasets
6682858

jefffffff9 Claude Sonnet 4.6 commited on

Fix SyntaxError in Cell 17: unterminated f-string literal
58f431a

jefffffff9 Claude Sonnet 4.6 commited on

Fix compute_metrics crash on empty eval references in Fula training
3632a23

jefffffff9 Claude Sonnet 4.6 commited on

Fix Self-Teaching language detection: parse code from dropdown label
33c3a5a

jefffffff9 Claude Sonnet 4.6 commited on

Add Adlam/Pular Fula integration: transliterator + 3 new datasets + normalisation pipeline
ced078c

jefffffff9 Claude Sonnet 4.6 commited on

Fix language mixing: per-language prompts + Mali Bambara / Guinea Pular context
40cf84d

jefffffff9 Claude Sonnet 4.6 commited on

Fix model loading stuck forever + unhandled TTS crash in conversation mode
618eab5

Broulaye Doumbia commited on

Fix: add trust_remote_code=True for datasets 3.x compatibility
71bb3bc

Broulaye Doumbia commited on

Fix 'Model loading' stuck state: block in _do_asr until Whisper is ready
1958814

jefffffff9 Claude Sonnet 4.6 commited on

Fix conversation mode timeout: two-stage pipeline + faster LLM
8d7d9d8

jefffffff9 Claude Sonnet 4.6 commited on

Add real conversational memory + live learning to Conversation Mode
ad902c6

jefffffff9 Claude Sonnet 4.6 commited on

Fix slow build: strip runtime-irrelevant heavy packages from requirements.txt
bfe5b59

jefffffff9 Claude Sonnet 4.6 commited on

Phase 3: Voice-to-Voice S2S pipeline — F5-TTS, LLM brain, CER metric
8952fff

jefffffff9 Claude Sonnet 4.6 commited on

Fix Cell 16 ValueError: load model fp32 so AMP gradient scaler works
cd017e2

jefffffff9 Claude Sonnet 4.6 commited on

Fix Self-Teaching tab: float sliders, deduplication, Kaggle API fallback
24b1617

jefffffff9 Claude Sonnet 4.6 commited on

Fix Kaggle trigger 401: pass credentials as env vars instead of KAGGLE_CONFIG_DIR
670a1d1

jefffffff9 Claude Sonnet 4.6 commited on

Fix kaggle trigger: use kaggle binary not python -m kaggle
7f169d8

jefffffff9 commited on

Add kaggle to requirements.txt for Space trigger
317a499

jefffffff9 commited on

chore: force Space restart
be96d03

jefffffff9 commited on

Fix Kaggle trigger: use kaggle CLI push instead of fake /run endpoint
3d43385

jefffffff9 Claude Sonnet 4.6 commited on

Fix Wikipedia and HF harvest buttons
a5737ac

jefffffff9 Claude Sonnet 4.6 commited on

Fix: point Space to app.py (was app_lab.py)
520ca12

jefffffff9 commited on

Add mass training + internet self-teaching features
eddcaea

jefffffff9 Claude Sonnet 4.6 commited on

Align Space app with Kaggle notebook training pipeline
98eece8

jefffffff9 Claude Sonnet 4.6 commited on

Fix eval_loss nan + add generation_config to Seq2SeqTrainer
9f07d46

jefffffff9 Claude Sonnet 4.6 commited on

Fix GenerationConfig deprecation warning in cells 13 and 16
1f2993e

jefffffff9 Claude Sonnet 4.6 commited on

Auto-cap MAX_STEPS based on dataset size after Cell 12 builds train_ds
73a3f3a

jefffffff9 Claude Sonnet 4.6 commited on

Fix Cell 15: detect eval_strategy vs evaluation_strategy at runtime
1c462f3

jefffffff9 Claude Sonnet 4.6 commited on

Proactive notebook audit: fix 5 issues before they cause errors
aef60bc

jefffffff9 Claude Sonnet 4.6 commited on

Simplify Cell 2: no restart needed since PEFT is removed
8e41441

jefffffff9 Claude Sonnet 4.6 commited on

Fix Cell 16 TypeError: replace PEFT LoRA with freeze/unfreeze fine-tuning
6dc0dfc

jefffffff9 Claude Sonnet 4.6 commited on

Fix Cell 16 TypeError: target encoder-only LoRA to avoid PEFT/transformers 5.x conflict
c37ce59

jefffffff9 Claude Sonnet 4.6 commited on

Fix Cell 14: cast input_features to fp16 in collator
3ed28bd

jefffffff9 Claude Sonnet 4.6 commited on

Fix Cell 16: remove tokenizer arg from Seq2SeqTrainer (renamed in transformers 5.x)
829fd3d

jefffffff9 Claude Sonnet 4.6 commited on