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

compute_metrics: pre-apply lowercasing + punctuation removal before filtering
empty references — a label like "?" passes r.strip() but becomes "" after
RemovePunctuation, causing jiwer to raise ValueError mid-evaluation.
Pass already-normalised strings directly to jiwer.cer / jiwer.wer so no
second transform is applied.

web_harvester: add commented-in registry entries for OpenSLR SLR106
(10k clean Guinea Pular utterances, 49 speakers, CC BY-SA 4.0) and SLR105
(Guinea radio corpus, Pular-tagged validation split). Both are uncommented
once the audio is uploaded to HF — best available Guinea Pular ASR data.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

notebooks/kaggle_master_trainer.ipynb

@@ -184,7 +184,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "# -- Cell 14: Data collator + CER metric --------------------------------------\nimport jiwer\nfrom dataclasses import dataclass\nfrom typing import Any, Dict, List\n\ntransform = jiwer.Compose([\n    jiwer.ToLowerCase(),\n    jiwer.RemoveMultipleSpaces(),\n    jiwer.Strip(),\n    jiwer.RemovePunctuation(),\n    jiwer.ReduceToListOfListOfWords(),\n])\n\n# CER transform (no word-split step needed)\n_cer_transform = jiwer.Compose([\n    jiwer.ToLowerCase(),\n    jiwer.RemoveMultipleSpaces(),\n    jiwer.Strip(),\n    jiwer.RemovePunctuation(),\n])\n\n\n@dataclass\nclass DataCollatorSpeechSeq2SeqWithPadding:\n    processor: Any\n\n    def __call__(self, features: List[Dict]) -> Dict:\n        import torch\n        input_feats  = [{'input_features': f['input_features']} for f in features]\n        batch        = self.processor.feature_extractor.pad(input_feats, return_tensors='pt')\n\n        # Leave features in fp32 -- AMP (fp16=True in TrainingArgs) handles casting\n\n        label_feats  = [{'input_ids': f['labels']} for f in features]\n        labels_batch = self.processor.tokenizer.pad(label_feats, return_tensors='pt')\n        labels       = labels_batch['input_ids'].masked_fill(\n            labels_batch.attention_mask.ne(1), -100\n        )\n        if (labels[:, 0] == self.processor.tokenizer.bos_token_id).all().item():\n            labels = labels[:, 1:]\n        batch['labels'] = labels\n        return batch\n\n\ndef compute_metrics(pred):\n    pred_ids  = pred.predictions\n    label_ids = pred.label_ids\n    label_ids[label_ids == -100] = processor.tokenizer.pad_token_id\n\n    pred_str  = processor.tokenizer.batch_decode(pred_ids,  skip_special_tokens=True)\n    label_str = processor.tokenizer.batch_decode(label_ids, skip_special_tokens=True)\n\n    # Filter out empty references (silent audio / all-padding labels) that\n    # cause jiwer to raise ValueError (\"non-empty list of strings\" required).\n    pairs = [(r, h) for r, h in zip(label_str, pred_str) if r.strip()]\n    if not pairs:\n        return {'cer': 0.0, 'wer': 0.0}\n    label_str, pred_str = zip(*pairs)\n\n    cer = jiwer.cer(\n        list(label_str), list(pred_str),\n        reference_transform=_cer_transform,\n        hypothesis_transform=_cer_transform,\n    )\n    wer = jiwer.wer(list(label_str), list(pred_str),\n                    hypothesis_transform=transform,\n                    reference_transform=transform)\n    return {'cer': round(cer, 4), 'wer': round(wer, 4)}\n\n\ncollator = DataCollatorSpeechSeq2SeqWithPadding(processor=processor)\nprint('Collator and WER metric ready')"
+    "# -- Cell 14: Data collator + CER metric --------------------------------------\nimport jiwer\nfrom dataclasses import dataclass\nfrom typing import Any, Dict, List\n\ntransform = jiwer.Compose([\n    jiwer.ToLowerCase(),\n    jiwer.RemoveMultipleSpaces(),\n    jiwer.Strip(),\n    jiwer.RemovePunctuation(),\n    jiwer.ReduceToListOfListOfWords(),\n])\n\n# CER transform (no word-split step needed)\n_cer_transform = jiwer.Compose([\n    jiwer.ToLowerCase(),\n    jiwer.RemoveMultipleSpaces(),\n    jiwer.Strip(),\n    jiwer.RemovePunctuation(),\n])\n\n\n@dataclass\nclass DataCollatorSpeechSeq2SeqWithPadding:\n    processor: Any\n\n    def __call__(self, features: List[Dict]) -> Dict:\n        import torch\n        input_feats  = [{'input_features': f['input_features']} for f in features]\n        batch        = self.processor.feature_extractor.pad(input_feats, return_tensors='pt')\n\n        # Leave features in fp32 -- AMP (fp16=True in TrainingArgs) handles casting\n\n        label_feats  = [{'input_ids': f['labels']} for f in features]\n        labels_batch = self.processor.tokenizer.pad(label_feats, return_tensors='pt')\n        labels       = labels_batch['input_ids'].masked_fill(\n            labels_batch.attention_mask.ne(1), -100\n        )\n        if (labels[:, 0] == self.processor.tokenizer.bos_token_id).all().item():\n            labels = labels[:, 1:]\n        batch['labels'] = labels\n        return batch\n\n\ndef _apply_jiwer_transform(texts, t):\n    \"\"\"Apply a jiwer Compose transform and return plain strings (not nested lists).\"\"\"\n    import re as _re\n    result = []\n    for s in texts:\n        s = s.lower()\n        s = _re.sub(r'[^\\w\\s]', '', s)   # RemovePunctuation equivalent\n        s = ' '.join(s.split())              # RemoveMultipleSpaces + Strip\n        result.append(s)\n    return result\n\n\ndef compute_metrics(pred):\n    pred_ids  = pred.predictions\n    label_ids = pred.label_ids\n    label_ids[label_ids == -100] = processor.tokenizer.pad_token_id\n\n    pred_str  = processor.tokenizer.batch_decode(pred_ids,  skip_special_tokens=True)\n    label_str = processor.tokenizer.batch_decode(label_ids, skip_special_tokens=True)\n\n    # Pre-normalise so we can filter empties AFTER the transform.\n    # A reference like \"?\" or \"1\" decodes to non-empty but becomes empty\n    # after punctuation/number removal -- jiwer crashes on empty references.\n    norm_ref  = _apply_jiwer_transform(label_str, _cer_transform)\n    norm_hyp  = _apply_jiwer_transform(pred_str,  _cer_transform)\n    pairs     = [(r, h) for r, h in zip(norm_ref, norm_hyp) if r.strip()]\n    if not pairs:\n        return {'cer': 0.0, 'wer': 0.0}\n    ref_clean, hyp_clean = zip(*pairs)\n\n    cer = jiwer.cer(list(ref_clean), list(hyp_clean))   # already normalised\n    wer = jiwer.wer(list(ref_clean), list(hyp_clean))   # already normalised\n    return {'cer': round(cer, 4), 'wer': round(wer, 4)}\n\n\ncollator = DataCollatorSpeechSeq2SeqWithPadding(processor=processor)\nprint('Collator and WER metric ready')"
    ]
   },
   {

src/data/web_harvester.py

@@ -69,6 +69,35 @@ HF_DATASET_REGISTRY = {
             "adlam":     True,
             "note":      "51 Adlam-script audio rows — converted to Latin before training",
         },
+        # OpenSLR SLR106 — West African Virtual Assistant ASR Corpus (Guinea, CC BY-SA 4.0)
+        # 10,083 clean utterances from 49 Guinea-native Pular speakers (read speech,
+        # multi-device, ages 5–76).  Best available clean Guinea Pular ASR data.
+        # Download manually from https://openslr.org/106/ and upload to HF before training.
+        # Uncomment once the dataset repo is populated.
+        # {
+        #     "repo":      "ous-sow/slr106-pular",
+        #     "config":    "default",
+        #     "split":     "train",
+        #     "audio_col": "audio",
+        #     "text_col":  "transcription",
+        #     "max":       10_000,
+        #     "license":   "cc-by-sa-4.0",
+        #     "note":      "OpenSLR SLR106 Guinea Pular — 10k clean utterances, 49 speakers",
+        # },
+        # OpenSLR SLR105 — West African Radio Corpus (Guinea, CC BY-SA 4.0)
+        # ~142 hours raw radio audio from 6 Guinea stations; Pular-tagged validation
+        # set of 300 clips.  Noisier than SLR106 but larger.
+        # Uncomment once uploaded to HF.
+        # {
+        #     "repo":      "ous-sow/slr105-pular",
+        #     "config":    "default",
+        #     "split":     "validation",
+        #     "audio_col": "audio",
+        #     "text_col":  "transcription",
+        #     "max":       300,
+        #     "license":   "cc-by-sa-4.0",
+        #     "note":      "OpenSLR SLR105 Guinea radio — Pular-tagged validation split",
+        # },
     ],
 }