generated from ShadowVR/AI_botter
166 lines
6.6 KiB
JavaScript
166 lines
6.6 KiB
JavaScript
/**
|
|
Copyright (c) 2014 Jameson Little
|
|
|
|
Permission is hereby granted, free of charge, to any person obtaining a copy
|
|
of this software and associated documentation files (the "Software"), to deal
|
|
in the Software without restriction, including without limitation the rights
|
|
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
|
|
copies of the Software, and to permit persons to whom the Software is
|
|
furnished to do so, subject to the following conditions:
|
|
|
|
The above copyright notice and this permission notice shall be included in
|
|
all copies or substantial portions of the Software.
|
|
|
|
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
|
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
|
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
|
|
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
|
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
|
|
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
|
|
THE SOFTWARE.
|
|
*/
|
|
const lookup = [];
|
|
const revLookup = [];
|
|
const code = "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+/";
|
|
for (var i = 0, len = code.length; i < len; ++i) {
|
|
lookup[i] = code[i];
|
|
revLookup[code.charCodeAt(i)] = i;
|
|
}
|
|
// Support decoding URL-safe base64 strings, as Node.js does.
|
|
// See: https://en.wikipedia.org/wiki/Base64#URL_applications
|
|
revLookup["-".charCodeAt(0)] = 62;
|
|
revLookup["_".charCodeAt(0)] = 63;
|
|
function tripletToBase64(num) {
|
|
return (lookup[(num >> 18) & 0x3f] +
|
|
lookup[(num >> 12) & 0x3f] +
|
|
lookup[(num >> 6) & 0x3f] +
|
|
lookup[num & 0x3f]);
|
|
}
|
|
function encodeChunk(uint8, start, end) {
|
|
var tmp;
|
|
var output = [];
|
|
for (var i = start; i < end; i += 3) {
|
|
tmp =
|
|
((uint8[i] << 16) & 0xff0000) +
|
|
((uint8[i + 1] << 8) & 0xff00) +
|
|
(uint8[i + 2] & 0xff);
|
|
output.push(tripletToBase64(tmp));
|
|
}
|
|
return output.join("");
|
|
}
|
|
function fromByteArray(uint8) {
|
|
var tmp;
|
|
var len = uint8.length;
|
|
var extraBytes = len % 3; // if we have 1 byte left, pad 2 bytes
|
|
var parts = [];
|
|
var maxChunkLength = 16383; // must be multiple of 3
|
|
// go through the array every three bytes, we'll deal with trailing stuff later
|
|
for (var i = 0, len2 = len - extraBytes; i < len2; i += maxChunkLength) {
|
|
parts.push(encodeChunk(uint8, i, i + maxChunkLength > len2 ? len2 : i + maxChunkLength));
|
|
}
|
|
// pad the end with zeros, but make sure to not forget the extra bytes
|
|
if (extraBytes === 1) {
|
|
tmp = uint8[len - 1];
|
|
parts.push(lookup[tmp >> 2] + lookup[(tmp << 4) & 0x3f] + "==");
|
|
}
|
|
else if (extraBytes === 2) {
|
|
tmp = (uint8[len - 2] << 8) + uint8[len - 1];
|
|
parts.push(lookup[tmp >> 10] +
|
|
lookup[(tmp >> 4) & 0x3f] +
|
|
lookup[(tmp << 2) & 0x3f] +
|
|
"=");
|
|
}
|
|
return parts.join("");
|
|
}
|
|
function is_printable(u) {
|
|
// printable ascii characters according to python
|
|
return !(u <= 31 || (u >= 127 && u <= 160) || u == 173);
|
|
}
|
|
function data_gym_to_mergeable_bpe_ranks(vocal_bpe_contents, encoder_json_contents) {
|
|
const rank_to_intbyte = Array.from({ length: 2 ** 8 }, (_, i) => i).filter((i) => is_printable(i) && String.fromCharCode(i) !== " ");
|
|
const data_gym_byte_to_byte = rank_to_intbyte.reduce((memo, item) => {
|
|
memo[String.fromCharCode(item)] = item;
|
|
return memo;
|
|
}, {});
|
|
let n = 0;
|
|
for (let b = 0; b < 2 ** 8; b++) {
|
|
if (!rank_to_intbyte.includes(b)) {
|
|
rank_to_intbyte.push(b);
|
|
data_gym_byte_to_byte[String.fromCharCode(2 ** 8 + n)] = b;
|
|
n += 1;
|
|
}
|
|
}
|
|
if (rank_to_intbyte.length !== 2 ** 8) {
|
|
throw new Error("rank_to_intbyte.length must be 2**8");
|
|
}
|
|
// vocab_bpe contains the merges along with associated ranks
|
|
const bpe_merges = vocal_bpe_contents
|
|
.split("\n")
|
|
.slice(1, -1)
|
|
.map((merge_str) => merge_str.split(" "));
|
|
function decode_data_gym(value) {
|
|
return value.split("").map((b) => data_gym_byte_to_byte[b]);
|
|
}
|
|
// add the single byte tokens
|
|
const bpe_ranks = Object.fromEntries(rank_to_intbyte.map((b, i) => [b, i]));
|
|
// add the merged tokens
|
|
n = rank_to_intbyte.length;
|
|
for (const [first, second] of bpe_merges) {
|
|
bpe_ranks[[...decode_data_gym(first), ...decode_data_gym(second)].join(",")] = n;
|
|
n += 1;
|
|
}
|
|
// check that the encoder file matches the merges file
|
|
// this sanity check is important since tiktoken assumes that ranks are ordered the same
|
|
// as merge priority
|
|
const encoder_json = JSON.parse(encoder_json_contents);
|
|
const encoder_json_loaded = Object.fromEntries(Object.entries(encoder_json).map(([k, v]) => [
|
|
decode_data_gym(k).join(","),
|
|
v,
|
|
]));
|
|
// drop these two special tokens if present, since they're not mergeable bpe tokens
|
|
delete encoder_json_loaded[decode_data_gym("<|endoftext|>").join(",")];
|
|
delete encoder_json_loaded[decode_data_gym("<|startoftext|>").join(",")];
|
|
function normalize_map(items) {
|
|
return JSON.stringify(Object.keys(items)
|
|
.sort()
|
|
.map((key) => [key, items[key]]));
|
|
}
|
|
if (normalize_map(bpe_ranks) !== normalize_map(encoder_json_loaded)) {
|
|
throw new Error("bpe_ranks !== encoder_json_loaded");
|
|
}
|
|
return bpe_ranks;
|
|
}
|
|
function dump_tiktoken_bpe(bpe_ranks) {
|
|
return (Object.entries(bpe_ranks)
|
|
.sort((a, b) => a[1] - b[1])
|
|
.map(([token_str, rank]) => [
|
|
fromByteArray(token_str.split(",").map((i) => Number.parseInt(i, 10))),
|
|
rank,
|
|
].join(" "))
|
|
.join("\n") + "\n");
|
|
}
|
|
export async function load(registry, customFetch) {
|
|
const ofetch = customFetch
|
|
? customFetch
|
|
: (url) => fetch(url).then((r) => r.text());
|
|
if ("data_gym_to_mergeable_bpe_ranks" in registry) {
|
|
const [vocab_bpe, encoder_json] = await Promise.all([
|
|
ofetch(registry.data_gym_to_mergeable_bpe_ranks.vocab_bpe_file),
|
|
ofetch(registry.data_gym_to_mergeable_bpe_ranks.encoder_json_file),
|
|
]);
|
|
return {
|
|
explicit_n_vocab: registry.explicit_n_vocab,
|
|
pat_str: registry.pat_str,
|
|
special_tokens: registry.special_tokens,
|
|
bpe_ranks: dump_tiktoken_bpe(data_gym_to_mergeable_bpe_ranks(vocab_bpe, encoder_json)),
|
|
};
|
|
}
|
|
else {
|
|
return {
|
|
explicit_n_vocab: registry.explicit_n_vocab,
|
|
pat_str: registry.pat_str,
|
|
special_tokens: registry.special_tokens,
|
|
bpe_ranks: await ofetch(registry.load_tiktoken_bpe),
|
|
};
|
|
}
|
|
}
|