tahamajs/RR / Docs /module_metadata /astro_quantizer.json
tahamajs's picture
download
raw
708 Bytes
{
"module_name": "astro_quantizer",
"class_name": "STEFakeQuantize",
"category": "compression",
"family": "compression",
"params": [
"in_features",
"out_features",
"bias",
"bits",
"symmetric",
"per_channel",
"model",
"use_ste"
],
"ops": [
"clamp",
"linear"
],
"docstring": "Straight-Through Estimator (STE) fake-quantization autograd function.\nClamps inputs to [qmin, qmax] in discrete integer space, de-quantizes back,\nand passes gradients through uninhibited where within clipping bounds.",
"rank": 16,
"d_state": 1,
"num_experts": 1,
"top_k": 1,
"sparsity_pct": 0.0,
"bits": 4,
"hash_val": 337753970793766220957567372448457046694
}

Xet Storage Details

Size:
708 Bytes
·
Xet hash:
633cd236142f512c2bac1d8bcd75122427a0d68a9b7d55c81c31985ecdef0276

Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.