@huggingface/tasks
Version:
List of ML tasks for huggingface.co/tasks
191 lines (190 loc) • 11.5 kB
JavaScript
;
Object.defineProperty(exports, "__esModule", { value: true });
exports.GGMLQuantizationType = exports.GGUF_QUANT_ORDER = exports.GGUF_QUANT_RE_GLOBAL = exports.GGUF_QUANT_RE = exports.GGMLFileQuantizationType = void 0;
exports.parseGGUFQuantLabel = parseGGUFQuantLabel;
exports.findNearestQuantType = findNearestQuantType;
// This list is copied from gguf/types.ts, but will all types available (for backward compatibility)
// NOT to be confused with GGMLQuantizationType, a FileQuantization can contain multiple GGMLQuantizationType
// For example, Q4_K_M model can contains Q4_K and Q6_K tensors
var GGMLFileQuantizationType;
(function (GGMLFileQuantizationType) {
GGMLFileQuantizationType[GGMLFileQuantizationType["F32"] = 0] = "F32";
GGMLFileQuantizationType[GGMLFileQuantizationType["F16"] = 1] = "F16";
GGMLFileQuantizationType[GGMLFileQuantizationType["Q4_0"] = 2] = "Q4_0";
GGMLFileQuantizationType[GGMLFileQuantizationType["Q4_1"] = 3] = "Q4_1";
GGMLFileQuantizationType[GGMLFileQuantizationType["Q4_1_SOME_F16"] = 4] = "Q4_1_SOME_F16";
GGMLFileQuantizationType[GGMLFileQuantizationType["Q4_2"] = 5] = "Q4_2";
GGMLFileQuantizationType[GGMLFileQuantizationType["Q4_3"] = 6] = "Q4_3";
GGMLFileQuantizationType[GGMLFileQuantizationType["Q8_0"] = 7] = "Q8_0";
GGMLFileQuantizationType[GGMLFileQuantizationType["Q5_0"] = 8] = "Q5_0";
GGMLFileQuantizationType[GGMLFileQuantizationType["Q5_1"] = 9] = "Q5_1";
GGMLFileQuantizationType[GGMLFileQuantizationType["Q2_K"] = 10] = "Q2_K";
GGMLFileQuantizationType[GGMLFileQuantizationType["Q3_K_S"] = 11] = "Q3_K_S";
GGMLFileQuantizationType[GGMLFileQuantizationType["Q3_K_M"] = 12] = "Q3_K_M";
GGMLFileQuantizationType[GGMLFileQuantizationType["Q3_K_L"] = 13] = "Q3_K_L";
GGMLFileQuantizationType[GGMLFileQuantizationType["Q4_K_S"] = 14] = "Q4_K_S";
GGMLFileQuantizationType[GGMLFileQuantizationType["Q4_K_M"] = 15] = "Q4_K_M";
GGMLFileQuantizationType[GGMLFileQuantizationType["Q5_K_S"] = 16] = "Q5_K_S";
GGMLFileQuantizationType[GGMLFileQuantizationType["Q5_K_M"] = 17] = "Q5_K_M";
GGMLFileQuantizationType[GGMLFileQuantizationType["Q6_K"] = 18] = "Q6_K";
GGMLFileQuantizationType[GGMLFileQuantizationType["IQ2_XXS"] = 19] = "IQ2_XXS";
GGMLFileQuantizationType[GGMLFileQuantizationType["IQ2_XS"] = 20] = "IQ2_XS";
GGMLFileQuantizationType[GGMLFileQuantizationType["Q2_K_S"] = 21] = "Q2_K_S";
GGMLFileQuantizationType[GGMLFileQuantizationType["IQ3_XS"] = 22] = "IQ3_XS";
GGMLFileQuantizationType[GGMLFileQuantizationType["IQ3_XXS"] = 23] = "IQ3_XXS";
GGMLFileQuantizationType[GGMLFileQuantizationType["IQ1_S"] = 24] = "IQ1_S";
GGMLFileQuantizationType[GGMLFileQuantizationType["IQ4_NL"] = 25] = "IQ4_NL";
GGMLFileQuantizationType[GGMLFileQuantizationType["IQ3_S"] = 26] = "IQ3_S";
GGMLFileQuantizationType[GGMLFileQuantizationType["IQ3_M"] = 27] = "IQ3_M";
GGMLFileQuantizationType[GGMLFileQuantizationType["IQ2_S"] = 28] = "IQ2_S";
GGMLFileQuantizationType[GGMLFileQuantizationType["IQ2_M"] = 29] = "IQ2_M";
GGMLFileQuantizationType[GGMLFileQuantizationType["IQ4_XS"] = 30] = "IQ4_XS";
GGMLFileQuantizationType[GGMLFileQuantizationType["IQ1_M"] = 31] = "IQ1_M";
GGMLFileQuantizationType[GGMLFileQuantizationType["BF16"] = 32] = "BF16";
GGMLFileQuantizationType[GGMLFileQuantizationType["Q4_0_4_4"] = 33] = "Q4_0_4_4";
GGMLFileQuantizationType[GGMLFileQuantizationType["Q4_0_4_8"] = 34] = "Q4_0_4_8";
GGMLFileQuantizationType[GGMLFileQuantizationType["Q4_0_8_8"] = 35] = "Q4_0_8_8";
GGMLFileQuantizationType[GGMLFileQuantizationType["TQ1_0"] = 36] = "TQ1_0";
GGMLFileQuantizationType[GGMLFileQuantizationType["TQ2_0"] = 37] = "TQ2_0";
GGMLFileQuantizationType[GGMLFileQuantizationType["MXFP4_MOE"] = 38] = "MXFP4_MOE";
// custom quants used by unsloth
// they are not officially a scheme enum value in GGUF, but only here for naming
GGMLFileQuantizationType[GGMLFileQuantizationType["Q2_K_XL"] = 1000] = "Q2_K_XL";
GGMLFileQuantizationType[GGMLFileQuantizationType["Q3_K_XL"] = 1001] = "Q3_K_XL";
GGMLFileQuantizationType[GGMLFileQuantizationType["Q4_K_XL"] = 1002] = "Q4_K_XL";
GGMLFileQuantizationType[GGMLFileQuantizationType["Q5_K_XL"] = 1003] = "Q5_K_XL";
GGMLFileQuantizationType[GGMLFileQuantizationType["Q6_K_XL"] = 1004] = "Q6_K_XL";
GGMLFileQuantizationType[GGMLFileQuantizationType["Q8_K_XL"] = 1005] = "Q8_K_XL";
})(GGMLFileQuantizationType || (exports.GGMLFileQuantizationType = GGMLFileQuantizationType = {}));
const ggufQuants = Object.values(GGMLFileQuantizationType).filter((v) => typeof v === "string");
exports.GGUF_QUANT_RE = new RegExp(`(?<quant>${ggufQuants.join("|")})` + "(_(?<sizeVariation>[A-Z]+))?");
exports.GGUF_QUANT_RE_GLOBAL = new RegExp(exports.GGUF_QUANT_RE, "g");
function parseGGUFQuantLabel(fname) {
const quantLabel = fname.toUpperCase().match(exports.GGUF_QUANT_RE_GLOBAL)?.at(-1); // if there is multiple quant substrings in a name, we prefer the last one
return quantLabel;
}
// order of quantization, from biggest to smallest
// this list must be in sync with the order in GGMLFileQuantizationType
// the gguf.spec.ts tests are using verify if the order is correct
exports.GGUF_QUANT_ORDER = [
GGMLFileQuantizationType.F32,
GGMLFileQuantizationType.BF16,
GGMLFileQuantizationType.F16,
GGMLFileQuantizationType.Q8_K_XL,
GGMLFileQuantizationType.Q8_0,
// 6-bit quantizations
GGMLFileQuantizationType.Q6_K_XL,
GGMLFileQuantizationType.Q6_K,
// 5-bit quantizations
GGMLFileQuantizationType.Q5_K_XL,
GGMLFileQuantizationType.Q5_K_M,
GGMLFileQuantizationType.Q5_K_S,
GGMLFileQuantizationType.Q5_0,
GGMLFileQuantizationType.Q5_1,
// 4-bit quantizations
GGMLFileQuantizationType.Q4_K_XL,
GGMLFileQuantizationType.Q4_K_M,
GGMLFileQuantizationType.Q4_K_S,
GGMLFileQuantizationType.IQ4_NL,
GGMLFileQuantizationType.IQ4_XS,
GGMLFileQuantizationType.Q4_0_4_4,
GGMLFileQuantizationType.Q4_0_4_8,
GGMLFileQuantizationType.Q4_0_8_8,
GGMLFileQuantizationType.Q4_1_SOME_F16,
GGMLFileQuantizationType.Q4_0,
GGMLFileQuantizationType.Q4_1,
GGMLFileQuantizationType.Q4_2,
GGMLFileQuantizationType.Q4_3,
GGMLFileQuantizationType.MXFP4_MOE,
// 3-bit quantizations
GGMLFileQuantizationType.Q3_K_XL,
GGMLFileQuantizationType.Q3_K_L,
GGMLFileQuantizationType.Q3_K_M,
GGMLFileQuantizationType.Q3_K_S,
GGMLFileQuantizationType.IQ3_M,
GGMLFileQuantizationType.IQ3_S,
GGMLFileQuantizationType.IQ3_XS,
GGMLFileQuantizationType.IQ3_XXS,
// 2-bit quantizations
GGMLFileQuantizationType.Q2_K_XL,
GGMLFileQuantizationType.Q2_K,
GGMLFileQuantizationType.Q2_K_S,
GGMLFileQuantizationType.IQ2_M,
GGMLFileQuantizationType.IQ2_S,
GGMLFileQuantizationType.IQ2_XS,
GGMLFileQuantizationType.IQ2_XXS,
// 1-bit quantizations
GGMLFileQuantizationType.IQ1_S,
GGMLFileQuantizationType.IQ1_M,
GGMLFileQuantizationType.TQ1_0,
GGMLFileQuantizationType.TQ2_0,
];
// This function finds the nearest quantization type that is less than or equal to the given quantization type.
// It returns undefined if no such quantization type is found.
function findNearestQuantType(quant, availableQuants) {
// Create a map for quick index lookup from the defined order
const orderMap = new Map();
exports.GGUF_QUANT_ORDER.forEach((q, index) => {
orderMap.set(q, index);
});
const targetIndex = orderMap.get(quant) ?? 0; // the 0 case should never happen
// Filter the available quantizations to include only those defined in the order map,
// then sort them according to the GGUF_QUANT_ORDER (from largest/index 0 to smallest/highest index).
const sortedAvailable = availableQuants
.filter((q) => orderMap.has(q))
.sort((a, b) => (orderMap.get(a) ?? Infinity) - (orderMap.get(b) ?? Infinity));
// If no valid quantizations are available after filtering
if (sortedAvailable.length === 0) {
return undefined;
}
// Iterate through the sorted available quantizations (largest to smallest).
// Find the first one whose order index is >= the target index.
// This means finding the largest quantization that is smaller than or equal to the target.
for (const availableQuant of sortedAvailable) {
// We know the key exists due to the filter above.
const availableIndex = orderMap.get(availableQuant) ?? 0;
if (availableIndex >= targetIndex) {
return availableQuant;
}
}
// If the loop completes, it means all available quantizations are larger (have a smaller index)
// than the target quantization. In this case, return the "smallest" available quantization,
// which is the last element in the sorted list (highest index among available).
return sortedAvailable[sortedAvailable.length - 1];
}
// This list is only used to calculate the size of the model, NOT to be confused with the quantization FILE type
var GGMLQuantizationType;
(function (GGMLQuantizationType) {
GGMLQuantizationType[GGMLQuantizationType["F32"] = 0] = "F32";
GGMLQuantizationType[GGMLQuantizationType["F16"] = 1] = "F16";
GGMLQuantizationType[GGMLQuantizationType["Q4_0"] = 2] = "Q4_0";
GGMLQuantizationType[GGMLQuantizationType["Q4_1"] = 3] = "Q4_1";
GGMLQuantizationType[GGMLQuantizationType["Q5_0"] = 6] = "Q5_0";
GGMLQuantizationType[GGMLQuantizationType["Q5_1"] = 7] = "Q5_1";
GGMLQuantizationType[GGMLQuantizationType["Q8_0"] = 8] = "Q8_0";
GGMLQuantizationType[GGMLQuantizationType["Q8_1"] = 9] = "Q8_1";
GGMLQuantizationType[GGMLQuantizationType["Q2_K"] = 10] = "Q2_K";
GGMLQuantizationType[GGMLQuantizationType["Q3_K"] = 11] = "Q3_K";
GGMLQuantizationType[GGMLQuantizationType["Q4_K"] = 12] = "Q4_K";
GGMLQuantizationType[GGMLQuantizationType["Q5_K"] = 13] = "Q5_K";
GGMLQuantizationType[GGMLQuantizationType["Q6_K"] = 14] = "Q6_K";
GGMLQuantizationType[GGMLQuantizationType["Q8_K"] = 15] = "Q8_K";
GGMLQuantizationType[GGMLQuantizationType["IQ2_XXS"] = 16] = "IQ2_XXS";
GGMLQuantizationType[GGMLQuantizationType["IQ2_XS"] = 17] = "IQ2_XS";
GGMLQuantizationType[GGMLQuantizationType["IQ3_XXS"] = 18] = "IQ3_XXS";
GGMLQuantizationType[GGMLQuantizationType["IQ1_S"] = 19] = "IQ1_S";
GGMLQuantizationType[GGMLQuantizationType["IQ4_NL"] = 20] = "IQ4_NL";
GGMLQuantizationType[GGMLQuantizationType["IQ3_S"] = 21] = "IQ3_S";
GGMLQuantizationType[GGMLQuantizationType["IQ2_S"] = 22] = "IQ2_S";
GGMLQuantizationType[GGMLQuantizationType["IQ4_XS"] = 23] = "IQ4_XS";
GGMLQuantizationType[GGMLQuantizationType["I8"] = 24] = "I8";
GGMLQuantizationType[GGMLQuantizationType["I16"] = 25] = "I16";
GGMLQuantizationType[GGMLQuantizationType["I32"] = 26] = "I32";
GGMLQuantizationType[GGMLQuantizationType["I64"] = 27] = "I64";
GGMLQuantizationType[GGMLQuantizationType["F64"] = 28] = "F64";
GGMLQuantizationType[GGMLQuantizationType["IQ1_M"] = 29] = "IQ1_M";
GGMLQuantizationType[GGMLQuantizationType["BF16"] = 30] = "BF16";
GGMLQuantizationType[GGMLQuantizationType["TQ1_0"] = 34] = "TQ1_0";
GGMLQuantizationType[GGMLQuantizationType["TQ2_0"] = 35] = "TQ2_0";
GGMLQuantizationType[GGMLQuantizationType["MXFP4"] = 39] = "MXFP4";
})(GGMLQuantizationType || (exports.GGMLQuantizationType = GGMLQuantizationType = {}));