Package index
-
ColEmbedding() - Distribution-aware Column-wise Embedding Module
-
Encoder() - Encoder
-
ICLearning() - Dataset-wise In-Context Learning Module
-
KVCache - Base Class for Key-Value Caches
-
KVCacheEntry - A Single Key-Value Cache Entry for an Attention Layer
-
NanoTabICLv2() - NanoTabICL v2 Model
-
QASSMaxMLP() - QASSMaxMLP – Query-Aware Scalable Softmax with MLPs
-
QuantileDistribution() - Probability distribution constructed from predicted quantiles
-
QuantileDistributionConfig - Configuration constants for Quantile Distribution
-
RotaryEmbedding() - Rotary Positional Embeddings Module
-
RowInteraction() - Row-Wise Feature Interaction Module
-
SSMax() - SSMax – Scalable Softmax with Learnable Per-Head Scaling
-
SSMaxMLP() - SSMaxMLP – Scalable Softmax with MLP-Computed Scaling
-
TabICLCache - Top-Level Cache Container for the Entire TabICL Model
-
TabICLv2() - TabICL: A Tabular In-Context Learning Foundation Model
-
async_copy_manager - Manages asynchronous GPU-to-CPU copies using CUDA streams
-
create_ssmax_layer() - Create an SSMax Layer by Type
-
disk_tensor - A tensor backed by a memory-mapped file on disk
-
enforce_monotonicity() - Enforce monotonicity of quantiles to fix crossing
-
flash_attn3_toggle() - Temporarily enable or disable Flash Attention 3
-
induced_self_attention_block() - Induced Self-Attention Block for efficient O(n) attention
-
inference_manager - Manages memory-efficient model inference by automatically batching inputs
-
kv_cache_concat() - Concatenate Multiple KVCache Objects
-
kv_cache_entry_concat() - Concatenate Multiple KVCacheEntry Objects
-
multihead_attention() - Enhanced multi-head attention with RoPE, scalable softmax, and KV caching
-
multihead_attention_block() - Attention block supporting RoPE, scalable softmax, and KV caching
-
one_hot_and_linear() - One-hot encoding combined with linear projection
-
pinned_buffer_pool - Pool of pinned CPU memory buffers for efficient GPU-to-CPU transfers
-
predict(<tab_icl_v2>)predict(<tab_icl_v2.regressor>)predict(<tab_icl_v2.classifier>)augment(<tab_icl_v2>) - Predict using
TabICL2 -
quantile_to_distribution() - Module wrapper for QuantileDistribution
-
set_transformer() - Stack of induced self-attention blocks.
-
skippable_linear() - Linear layer that handles inputs flagged with a skip value
-
tab_icl2() - Fit a TabICL2 model.
-
tab_icl2_config() - Configure TabICL2 model architecture
-
tab_icl2_control()inference_config() - Control TabICL2 inference execution
-
tabicl_cache_concat() - Concatenate Multiple TabICLCache Objects