webinfer 0.0.1

package/dist/index.d.ts

@@ -0,0 +1,22 @@
+/**
+ * WebInfer - High-performance LLM inference kernels for WebGPU
+ * "The cuDNN/FlashInfer of WebGPU"
+ */
+export { WebInferDevice, type DeviceInfo } from "./core/device.ts";
+export { Tensor, type DType } from "./core/tensor.ts";
+export { BufferPool } from "./core/buffer-pool.ts";
+export { KernelCache, type CacheStats } from "./jit/kernel-cache.ts";
+export { WGSLCompiler, type MatMulConfig } from "./jit/compiler.ts";
+export { matmul, matmulCPU, getMatMulCacheStats } from "./ops/matmul.ts";
+export { layerNorm, layerNormCPU, rmsNorm, rmsNormCPU, } from "./ops/normalization.ts";
+export { rope, ropeCPU, computeRoPEFrequencies, type RoPEConfig, } from "./ops/rope.ts";
+export { gelu, geluCPU, geluExactCPU, silu, siluCPU, relu, reluCPU, sigmoidCPU, } from "./ops/activations.ts";
+export { softmaxGPU, softmaxCPU, logSoftmaxCPU } from "./ops/softmax.ts";
+export { add, addCPU, mul, mulCPU, scale, scaleCPU, addScalarCPU, fmaCPU, } from "./ops/elementwise.ts";
+export { embedding, embeddingCPU, batchedEmbeddingCPU, } from "./ops/embedding.ts";
+export { transpose2D, transpose2DCPU, transposeCPU, reshapeCPU, permuteCPU, } from "./ops/reshape.ts";
+export { quantizeToInt8, quantizeToInt4, dequantizeInt8, dequantizeInt4, quantizationError, getMemorySavings, qmatmulInt8CPU, qmatmulInt4CPU, qmatmulInt8BlockCPU, estimateQMatMulFlops, estimateQMatMulBandwidth, type QuantConfig, type QuantizedTensor, } from "./quantization/index.ts";
+export { flashAttention, attentionCPU, type AttentionConfig, buildBlockSparseCSR, getSparsityRatio, estimateMemorySavings, type BlockSparseCSR, type AttentionPattern, buildCausalMask, getCausalSparsity, buildSlidingWindowMask, buildCausalSlidingWindowMask, getSlidingWindowSparsity, AttentionScheduler, type ChunkPlan, PagedKVCache, type PagedKVCacheConfig, type SequenceEntry, BlockManager, ContinuousBatchScheduler, type BlockManagerConfig, type AllocationPolicy, type AllocationRequest, } from "./attention/index.ts";
+export { topK, topKCPU, topKFilter, topPFilter, topPFilterCPU, sample, sampleCPU, sampleGreedy, sampleFromProbs, softmax, applyRepetitionPenalty, type SamplingConfig, } from "./sampling/index.ts";
+export { type ModelFormat, type SafetensorsDType, GGUFQuantType, GGUFMetadataValueType, type TensorInfo, type SafetensorsHeader, type ModelMetadata, type LoadedModel, type LoadOptions, parseSafetensorsHeader, loadSafetensors, loadSafetensorsFromUrl, isSafetensors, parseGGUFHeader, loadGGUF, loadGGUFFromUrl, loadGGUFTensor, isGGUF, loadModel, } from "./model/index.ts";
+export { type ModelConfig, type InferenceConfig, type GenerationConfig, type GenerationResult, type StreamToken, type FinishReason, type ForwardResult, DEFAULT_GENERATION_CONFIG, normalizeGenerationConfig, InferenceEngine, generate, generateStream, greedyDecode, sampleNextToken, } from "./inference/index.ts";