All authors
Claude Skills by wenmin-wu
github.com/wenmin-wu535 skills0 installs779 views
- Class Balanced Dataset MergeMerges multiple training datasets while keeping all positive examples and downsampling negatives to control class imbalance.Votes: 0GitHub stars: 61
- Cls Null Score Span RerankingReranks candidate answer spans by subtracting the CLS token's start+end logit sum as a null-answer baseline score.Votes: 0GitHub stars: 61
- Cp1252 Utf8 Encoding NormalizationResolves mixed cp1252/utf-8 encoding artifacts in text via round-trip encode/decode with custom error handlers and unidecode normalization.Votes: 0GitHub stars: 61
- Cross Model Candidate PoolMerge candidate pools from multiple independent seq2seq models before MBR selection to reduce shared failure modesVotes: 0GitHub stars: 61
- Custom Bpe TokenizerTrains a Byte-Pair Encoding tokenizer on the task corpus to capture domain-specific vocabulary, typos, and subword patterns.Votes: 0GitHub stars: 61
- Cyclical Lr TriangularCyclical learning rate (CLR) Keras callback that oscillates LR between base and max each batch for faster convergenceVotes: 0GitHub stars: 61
- Deberta ClassificationFine-tunes DeBERTa-v3 for text classification tasks. Use when building text classifiers, sentiment analysis, or multi-label classification on domain-specific data.Votes: 0GitHub stars: 61
- Domain Special Token EmbeddingAdd domain-specific categorical values as new special tokens, resize embeddings, and prepend them to input so the model learns domain-aware representationsVotes: 0GitHub stars: 61
- Domain Wordpiece TfidfTrains a WordPiece tokenizer on in-domain text, then feeds its subword token IDs into TF-IDF vectorization for domain-adapted sparse features.Votes: 0GitHub stars: 61
- Dropout Disabled InferenceExplicitly zero all dropout probabilities in transformer config at load time for fully deterministic inferenceVotes: 0GitHub stars: 61
- Dual Embedding ConcatConcatenate GloVe and FastText embedding matrices along feature axis for richer 600d word representationsVotes: 0GitHub stars: 61
- Dynamic Batch PaddingPads each batch to its actual max sequence length instead of the global max_len, reducing wasted computation.Votes: 0GitHub stars: 61
- Embedding Aware Punct NormalizationUse pretrained embedding vocab to decide which punctuation to keep, split, or remove before tokenizationVotes: 0GitHub stars: 61
- Embedding Coverage AnalysisMeasure pretrained embedding coverage over dataset vocab and return OOV words sorted by frequency for targeted preprocessingVotes: 0GitHub stars: 61
- Embedding Matrix Oov Gaussian InitInitialize out-of-vocabulary word embeddings with Gaussian noise matching the pretrained embedding distributionVotes: 0GitHub stars: 61
- Encoder Decoder Lr SplitUse separate learning rates for pretrained backbone (low) and randomly initialized classification head (high)Votes: 0GitHub stars: 61
- Fill Mask Entity ClassificationUses a fine-tuned masked language model to classify candidate entity spans by comparing fill-mask probabilities for positive vs negative marker tokens.Votes: 0GitHub stars: 61
- Gpu Tfidf Cosine MatchingGPU-accelerated TF-IDF vectorization via RAPIDS cuML with chunked cosine similarity for large-scale text matchingVotes: 0GitHub stars: 61
- Gradient CheckpointingTrades compute for memory by recomputing intermediate activations during backprop instead of storing them, reducing memory from O(n) to O(sqrt(n)).Votes: 0GitHub stars: 61
- Head Tail TruncationPreserves both the start and end of long text sequences when truncating for transformer input limits.Votes: 0GitHub stars: 61
- Hierarchical Label EncodingConcatenate multi-level categorical fields into a compound label (Category:Subcategory) for flat multiclass classificationVotes: 0GitHub stars: 61
- Identity Weighted Bce LossWeight BCE loss by identity subgroup membership to debias predictions — upweight samples where identity conflicts with labelVotes: 0GitHub stars: 61
- In Task PretrainingFurther pretrains a transformer with masked language modeling on the target task's own text before fine-tuning.Votes: 0GitHub stars: 61
- Jaccard Dedup Prediction FilterDeduplicates extracted entity predictions by filtering out candidates whose Jaccard word-overlap with already-accepted labels exceeds a threshold.Votes: 0GitHub stars: 61
- Jaccard Fbeta Ner MetricComputes micro F-beta for NER where true positives are determined by Jaccard word-overlap >= threshold rather than exact string match.Votes: 0GitHub stars: 61
- Japanese Transliteration NormalizationConverts Japanese scripts (Hiragana, Katakana, Kanji) to romanized ASCII using pykakasi for cross-script entity matching.Votes: 0GitHub stars: 61
- Joint Span Answer Type HeadDual-head transformer model that jointly predicts start/end span logits from sequence output and answer type from pooled CLS output.Votes: 0GitHub stars: 61
- Kendall Tau Ordering MetricEvaluates predicted sequence ordering quality using Kendall Tau correlation via efficient O(n log n) inversion counting.Votes: 0GitHub stars: 61
- Keystroke Essay ReconstructionReplay keystroke activity logs (Input/Replace/Paste/Remove/Move) against a string buffer to reconstruct the evolving essay textVotes: 0GitHub stars: 61
- Label Smoothing BinaryApplies label smoothing to binary cross-entropy loss to reduce overconfidence and improve generalization in text classification.Votes: 0GitHub stars: 61
- Last Hidden States ConcatConcatenate the last two transformer hidden states along the feature dimension before the task head for richer token representationsVotes: 0GitHub stars: 61
- Layerwise Lr DecayApplies different learning rates to transformer encoder vs task-specific head, with no weight decay on bias and LayerNorm.Votes: 0GitHub stars: 61
- Lda Topic ModelingLatent Dirichlet Allocation on CountVectorizer bag-of-words to discover latent topics with per-document topic distributions for feature engineering or EDAVotes: 0GitHub stars: 61
- Learned Attention PoolingReplace mean pooling with a trainable attention network (Linear-Tanh-Linear-Softmax) that learns token importance weights over transformer hidden statesVotes: 0GitHub stars: 61
- Length Sorted BatchingSort texts by length before batching with dynamic padding to minimize wasted padding tokens and speed up transformer inferenceVotes: 0GitHub stars: 61
- Lstm Over Transformer LayersFeeds CLS token embeddings from each transformer layer into a BiLSTM to learn an optimal combination across layer depth.Votes: 0GitHub stars: 61
- Map At K MetricCustom MAP@K evaluation metric — scores top-K classification predictions with reciprocal rank weighting for HuggingFace TrainerVotes: 0GitHub stars: 61
- Masked Bce Span LossBinary cross-entropy loss with mask to ignore special and padding tokens in token-level span classificationVotes: 0GitHub stars: 61
- Mbr Decoding RerankingMinimum Bayes Risk decoding — select the candidate with highest average chrF++ agreement against all others in the poolVotes: 0GitHub stars: 61
- Mean Max Concat PoolingConcatenates token-level mean pooling and max pooling from the last hidden state for a richer sequence representation.Votes: 0GitHub stars: 61
- Mean Pooling With MaskComputes attention-mask-weighted mean of token embeddings, excluding padding tokens from the average.Votes: 0GitHub stars: 61
- Minmax Normalized Ensemble BlendMin-max normalizes each model's predictions to [0,1] before averaging, ensuring equal contribution regardless of score distribution scale.Votes: 0GitHub stars: 61
- Multi Metric String SimilarityComputes multiple complementary string similarity scores (Gestalt, Levenshtein, Jaro-Winkler, LCS) per field pair as features for entity matching classifiers.Votes: 0GitHub stars: 61
- Multi Retriever Union EnsembleRun multiple independent retrieve-rerank pipelines and union-merge their predicted IDs per query via explode-groupby-uniqueVotes: 0GitHub stars: 61
- Multi Temperature Candidate SamplingGenerate diverse translation candidates by running nucleus sampling at multiple temperatures then pooling for MBR selectionVotes: 0GitHub stars: 61
- Multisample DropoutApplies multiple dropout masks to the same hidden state and averages predictions for regularization and variance reduction.Votes: 0GitHub stars: 61
- Negative Sample DownsamplingDownsamples documents with no entity labels while keeping all positive samples, balancing class distribution in NER training without discarding entity-bearing examples.Votes: 0GitHub stars: 61
- Neutral Short Text FallbackOverride model span predictions with full text for neutral sentiment or very short inputs where sub-span extraction is unreliableVotes: 0GitHub stars: 61
- O Class Threshold SuppressionThresholds the O-class (non-entity) softmax probability in NER: if below threshold, overrides with the best non-O class to boost entity recall.Votes: 0GitHub stars: 61
- Onnx Mixed Precision ExportExports a HuggingFace transformer to ONNX with dynamic axes, then auto-converts to BF16 mixed precision for 30-200% GPU inference speedup with 2x memory reduction.Votes: 0GitHub stars: 61