parakeet-rnnt-1.1b / config.json
senstella's picture
Upload folder using huggingface_hub
292ccaa verified
raw
history blame
37.3 kB
{
"sample_rate": 16000,
"compute_eval_loss": false,
"log_prediction": true,
"rnnt_reduction": "mean_volume",
"skip_nan_grad": false,
"model_defaults": {
"enc_hidden": 1024,
"pred_hidden": 640,
"joint_hidden": 640
},
"train_ds": {
"manifest_filepath": "/disk1/NVIDIA/datasets/LibriSpeech_NeMo/librivox-train-all.json",
"sample_rate": 16000,
"batch_size": 16,
"shuffle": true,
"num_workers": 8,
"pin_memory": true,
"use_start_end_token": false,
"trim_silence": false,
"max_duration": 16.7,
"min_duration": 0.1,
"is_tarred": false,
"tarred_audio_filepaths": null,
"shuffle_n": 2048,
"bucketing_strategy": "fully_randomized",
"bucketing_batch_size": null
},
"validation_ds": {
"manifest_filepath": "/disk1/NVIDIA/datasets/LibriSpeech_NeMo/librivox-dev-clean.json",
"sample_rate": 16000,
"batch_size": 16,
"shuffle": false,
"use_start_end_token": false,
"num_workers": 8,
"pin_memory": true
},
"test_ds": {
"manifest_filepath": null,
"sample_rate": 16000,
"batch_size": 16,
"shuffle": false,
"use_start_end_token": false,
"num_workers": 8,
"pin_memory": true
},
"tokenizer": {
"dir": "/disk3/datasets/suno_asr_set3_tokenizer/tokenizer_spe_bpe_v1024/",
"type": "bpe",
"model_path": "nemo:7880ced2d9384574a93511f66fa41d40_tokenizer.model",
"vocab_path": "nemo:d7786fbc907f40f5a905d51dd63c6cd4_vocab.txt",
"spe_tokenizer_vocab": "nemo:ceccf81b4bfd4f448f63aa1aeb0d7e09_tokenizer.vocab"
},
"preprocessor": {
"_target_": "nemo.collections.asr.modules.AudioToMelSpectrogramPreprocessor",
"sample_rate": 16000,
"normalize": "per_feature",
"window_size": 0.025,
"window_stride": 0.01,
"window": "hann",
"features": 80,
"n_fft": 512,
"log": true,
"frame_splicing": 1,
"dither": 1e-05,
"pad_to": 0,
"pad_value": 0.0
},
"spec_augment": {
"_target_": "nemo.collections.asr.modules.SpectrogramAugmentation",
"freq_masks": 2,
"time_masks": 10,
"freq_width": 27,
"time_width": 0.05
},
"encoder": {
"_target_": "nemo.collections.asr.modules.ConformerEncoder",
"feat_in": 80,
"feat_out": -1,
"n_layers": 42,
"d_model": 1024,
"subsampling": "dw_striding",
"subsampling_factor": 8,
"subsampling_conv_channels": 256,
"causal_downsampling": false,
"reduction": null,
"reduction_position": null,
"reduction_factor": 1,
"ff_expansion_factor": 4,
"self_attention_model": "rel_pos",
"n_heads": 8,
"att_context_size": [
-1,
-1
],
"att_context_style": "regular",
"xscaling": true,
"untie_biases": true,
"pos_emb_max_len": 5000,
"conv_kernel_size": 9,
"conv_norm_type": "batch_norm",
"conv_context_size": null,
"dropout": 0.1,
"dropout_pre_encoder": 0.1,
"dropout_emb": 0.0,
"dropout_att": 0.1
},
"decoder": {
"_target_": "nemo.collections.asr.modules.RNNTDecoder",
"normalization_mode": null,
"random_state_sampling": false,
"blank_as_pad": true,
"prednet": {
"pred_hidden": 640,
"pred_rnn_layers": 2,
"t_max": null,
"dropout": 0.2
},
"vocab_size": 1024
},
"joint": {
"_target_": "nemo.collections.asr.modules.RNNTJoint",
"log_softmax": null,
"preserve_memory": false,
"fuse_loss_wer": true,
"fused_batch_size": 16,
"jointnet": {
"joint_hidden": 640,
"activation": "relu",
"dropout": 0.2,
"encoder_hidden": 1024,
"pred_hidden": 640
},
"num_classes": 1024,
"vocabulary": [
"<unk>",
"\u2581t",
"\u2581th",
"\u2581a",
"\u2581i",
"\u2581the",
"re",
"\u2581w",
"\u2581s",
"\u2581o",
"in",
"at",
"er",
"ou",
"nd",
"\u2581c",
"\u2581b",
"\u2581h",
"on",
"\u2581m",
"\u2581f",
"ing",
"\u2581to",
"en",
"\u2581p",
"\u2581and",
"\u2581d",
"es",
"or",
"an",
"ll",
"\u2581y",
"\u2581l",
"ed",
"\u2581of",
"\u2581in",
"it",
"is",
"\u2581you",
"\u2581that",
"ar",
"\u2581g",
"\u2581n",
"as",
"om",
"\u2581it",
"ic",
"ve",
"\u2581e",
"\u2581wh",
"\u2581be",
"us",
"le",
"al",
"ion",
"ow",
"\u2581we",
"\u2581re",
"\u2581is",
"ut",
"ot",
"ent",
"\u2581on",
"et",
"\u2581ha",
"ay",
"ct",
"\u2581he",
"id",
"\u2581for",
"\u2581st",
"ver",
"ly",
"ro",
"ig",
"\u2581so",
"ld",
"\u2581this",
"ke",
"\u2581u",
"se",
"all",
"st",
"ur",
"ce",
"ch",
"im",
"ith",
"\u2581as",
"\u2581k",
"\u2581an",
"\u2581was",
"\u2581j",
"\u2581with",
"ir",
"\u2581go",
"ra",
"\u2581do",
"\u2581have",
"\u2581li",
"\u2581sh",
"\u2581se",
"\u2581they",
"\u2581are",
"am",
"ht",
"\u2581but",
"ation",
"\u2581not",
"th",
"\u2581r",
"ally",
"ad",
"ust",
"\u2581or",
"\u2581com",
"ould",
"\u2581can",
"ill",
"\u2581ne",
"ight",
"\u2581ch",
"\u2581de",
"\u2581con",
"\u2581at",
"\u2581mo",
"ant",
"oo",
"il",
"\u2581me",
"\u2581what",
"\u2581there",
"ter",
"pe",
"\u2581ab",
"\u2581su",
"ere",
"ck",
"\u2581pro",
"\u2581al",
"\u2581fr",
"\u2581kn",
"\u2581all",
"ers",
"\u2581like",
"ge",
"\u2581ex",
"\u2581som",
"ul",
"\u2581your",
"\u2581v",
"pp",
"use",
"\u2581if",
"ess",
"ate",
"est",
"\u2581know",
"out",
"if",
"\u2581just",
"ment",
"qu",
"op",
"ain",
"\u2581one",
"ol",
"ri",
"art",
"very",
"\u2581wor",
"ive",
"ist",
"\u2581my",
"nt",
"ab",
"\u2581from",
"ort",
"\u2581ma",
"\u2581about",
"res",
"ity",
"\u2581out",
"\u2581bec",
"\u2581le",
"our",
"od",
"and",
"ink",
"ie",
"\u2581up",
"ind",
"os",
"un",
"ause",
"oug",
"um",
"\u2581some",
"\u2581int",
"\u2581by",
"\u2581pl",
"\u2581get",
"el",
"ard",
"\u2581when",
"\u2581don",
"her",
"\u2581will",
"\u2581us",
"\u2581would",
"ook",
"ies",
"ich",
"\u2581because",
"\u2581think",
"em",
"\u2581pe",
"\u2581his",
"ack",
"\u2581then",
"\u2581our",
"ide",
"\u2581tim",
"\u2581how",
"ven",
"\u2581tr",
"\u2581who",
"\u2581them",
"ure",
"\u2581ar",
"\u2581ye",
"\u2581more",
"\u2581going",
"ect",
"\u2581sa",
"\u2581cl",
"\u2581had",
"\u2581now",
"\u2581which",
"\u2581here",
"ous",
"\u2581their",
"\u2581tw",
"so",
"\u2581has",
"ud",
"\u2581co",
"\u2581ta",
"ound",
"\u2581were",
"ast",
"\u2581peop",
"ough",
"\u2581no",
"\u2581really",
"\u2581any",
"\u2581people",
"\u2581want",
"\u2581she",
"\u2581en",
"\u2581fa",
"\u2581te",
"ame",
"ine",
"\u2581qu",
"red",
"\u2581im",
"\u2581right",
"ther",
"\u2581act",
"\u2581thing",
"king",
"ose",
"\u2581ad",
"\u2581see",
"\u2581time",
"\u2581these",
"ci",
"one",
"\u2581say",
"\u2581also",
"\u2581fe",
"per",
"\u2581ag",
"\u2581man",
"ore",
"\u2581un",
"pt",
"\u2581her",
"\u2581look",
"ong",
"ice",
"\u2581very",
"ff",
"ions",
"\u2581comp",
"\u2581did",
"itt",
"\u2581well",
"\u2581other",
"iv",
"ase",
"ree",
"hing",
"\u2581lo",
"reat",
"\u2581cont",
"\u2581part",
"\u2581into",
"nder",
"\u2581been",
"are",
"\u2581am",
"ans",
"\u2581sp",
"\u2581two",
"ue",
"\u2581way",
"age",
"\u2581where",
"ite",
"\u2581dis",
"\u2581than",
"\u2581every",
"\u2581pr",
"\u2581po",
"ag",
"\u2581need",
"ach",
"iff",
"ence",
"pl",
"own",
"\u2581ac",
"ble",
"\u2581over",
"iz",
"\u2581work",
"\u2581res",
"\u2581make",
"\u2581could",
"\u2581off",
"ually",
"\u2581ro",
"\u2581back",
"able",
"ip",
"ry",
"\u2581him",
"\u2581cour",
"ber",
"\u2581pre",
"\u2581fir",
"\u2581spe",
"ap",
"ars",
"\u2581diff",
"ire",
"\u2581somet",
"\u2581imp",
"\u2581those",
"\u2581comm",
"ance",
"ick",
"\u2581even",
"ated",
"way",
"sel",
"\u2581let",
"\u2581br",
"ty",
"\u2581per",
"int",
"\u2581first",
"\u2581thr",
"\u2581under",
"ah",
"\u2581may",
"\u2581cou",
"\u2581new",
"ress",
"act",
"\u2581gr",
"ep",
"\u2581said",
"ations",
"\u2581good",
"ace",
"ass",
"\u2581does",
"orm",
"ish",
"\u2581af",
"ving",
"co",
"\u2581app",
"\u2581lot",
"\u2581things",
"\u2581tra",
"ittle",
"\u2581bl",
"\u2581little",
"\u2581mu",
"cess",
"fe",
"ome",
"\u2581inc",
"\u2581differe",
"ary",
"ical",
"\u2581only",
"ult",
"\u2581again",
"\u2581got",
"ens",
"\u2581gu",
"\u2581kind",
"\u2581much",
"ord",
"\u2581through",
"ition",
"ild",
"\u2581down",
"\u2581actually",
"\u2581something",
"ang",
"ru",
"ces",
"\u2581fl",
"ile",
"ater",
"\u2581ra",
"\u2581take",
"ict",
"ign",
"\u2581sc",
"vel",
"\u2581bet",
"\u2581tal",
"\u2581yeah",
"\u2581use",
"fore",
"\u2581bu",
"\u2581start",
"ory",
"be",
"\u2581day",
"wn",
"xt",
"ia",
"ak",
"\u2581after",
"\u2581should",
"\u2581fo",
"\u2581ho",
"\u2581hel",
"\u2581ind",
"\u2581uh",
"na",
"ial",
"other",
"\u2581ke",
"\u2581call",
"\u2581most",
"\u2581ok",
"\u2581different",
"\u2581em",
"ting",
"ple",
"\u2581being",
"\u2581bo",
"ning",
"\u2581too",
"ors",
"\u2581happ",
"ark",
"og",
"\u2581help",
"\u2581rem",
"du",
"ction",
"ood",
"\u2581ser",
"ether",
"ious",
"\u2581mean",
"\u2581many",
"\u2581court",
"\u2581bel",
"ade",
"\u2581la",
"ved",
"\u2581des",
"\u2581rec",
"\u2581jo",
"\u2581dec",
"ves",
"\u2581before",
"\u2581put",
"self",
"\u2581point",
"te",
"\u2581ev",
"form",
"ents",
"\u2581add",
"ody",
"thing",
"\u2581case",
"\u2581pers",
"\u2581cons",
"iss",
"\u2581three",
"oth",
"\u2581ph",
"\u2581come",
"\u2581find",
"\u2581why",
"ull",
"\u2581show",
"\u2581bas",
"\u2581great",
"ily",
"\u2581rel",
"\u2581sm",
"\u2581its",
"\u2581fact",
"\u2581pos",
"ool",
"ments",
"ise",
"nds",
"ys",
"\u2581try",
"ual",
"ful",
"erm",
"\u2581inter",
"ons",
"\u2581quest",
"\u2581sub",
"we",
"vers",
"\u2581supp",
"\u2581feel",
"\u2581same",
"ub",
"ates",
"urn",
"ert",
"\u2581inv",
"day",
"\u2581rep",
"igh",
"\u2581sy",
"\u2581inst",
"\u2581long",
"\u2581still",
"\u2581okay",
"ft",
"ific",
"atch",
"ought",
"ath",
"\u2581own",
"\u2581made",
"ix",
"ced",
"ks",
"lic",
"\u2581wr",
"de",
"\u2581cr",
"\u2581att",
"\u2581ob",
"\u2581world",
"\u2581sure",
"ward",
"\u2581bit",
"\u2581life",
"\u2581person",
"\u2581pres",
"ph",
"\u2581vide",
"\u2581reg",
"\u2581end",
"ject",
"ange",
"\u2581fin",
"ied",
"pect",
"\u2581didn",
"\u2581around",
"ian",
"\u2581car",
"ible",
"\u2581sim",
"ever",
"\u2581sch",
"ating",
"\u2581pol",
"\u2581set",
"\u2581oh",
"cy",
"\u2581real",
"\u2581import",
"\u2581count",
"\u2581um",
"\u2581next",
"cial",
"les",
"\u2581hu",
"\u2581acc",
"\u2581might",
"\u2581ent",
"\u2581doing",
"\u2581ins",
"\u2581gen",
"\u2581play",
"\u2581cle",
"\u2581another",
"ady",
"ular",
"ib",
"ways",
"ered",
"ility",
"ities",
"\u2581op",
"\u2581def",
"\u2581years",
"\u2581never",
"ower",
"ram",
"\u2581tell",
"\u2581sl",
"onna",
"ail",
"ren",
"ute",
"\u2581gonna",
"\u2581big",
"\u2581give",
"der",
"ount",
"\u2581ap",
"kes",
"\u2581state",
"\u2581cor",
"\u2581min",
"ically",
"\u2581mon",
"\u2581fam",
"\u2581important",
"\u2581always",
"\u2581high",
"\u2581four",
"\u2581gra",
"\u2581ca",
"\u2581stud",
"\u2581dist",
"\u2581talk",
"\u2581num",
"\u2581str",
"\u2581today",
"ract",
"\u2581while",
"ason",
"\u2581iss",
"\u2581sur",
"\u2581char",
"\u2581last",
"oy",
"ited",
"\u2581exper",
"\u2581place",
"\u2581tri",
"\u2581ear",
"\u2581belie",
"\u2581able",
"\u2581underst",
"\u2581che",
"\u2581both",
"ug",
"\u2581doesn",
"\u2581keep",
"\u2581happen",
"ings",
"iew",
"ather",
"\u2581ass",
"\u2581love",
"ative",
"av",
"\u2581yes",
"\u2581ele",
"\u2581year",
"\u2581such",
"\u2581video",
"ness",
"\u2581el",
"\u2581trans",
"\u2581five",
"\u2581produ",
"ave",
"erest",
"als",
"body",
"cus",
"\u2581found",
"atter",
"\u2581eff",
"\u2581god",
"\u2581used",
"llow",
"\u2581interest",
"\u2581question",
"hip",
"\u2581bus",
"\u2581ask",
"\u2581exam",
"\u2581prov",
"lud",
"\u2581form",
"\u2581law",
"ense",
"\u2581child",
"\u2581gl",
"ne",
"\u2581each",
"\u2581understand",
"\u2581care",
"stem",
"\u2581med",
"\u2581maybe",
"ably",
"\u2581det",
"\u2581coll",
"its",
"\u2581commun",
"\u2581hand",
"\u2581'",
"\u2581ref",
"\u2581lear",
"\u2581done",
"\u2581gener",
"vern",
"\u2581mr",
"ween",
"\u2581better",
"\u2581between",
"li",
"blem",
"\u2581system",
"ertain",
"\u2581school",
"\u2581eas",
"\u2581exp",
"\u2581war",
"ention",
"\u2581ty",
"\u2581govern",
"ues",
"\u2581problem",
"\u2581plan",
"ac",
"\u2581conf",
"\u2581course",
"ouse",
"\u2581mar",
"\u2581stand",
"\u2581sk",
"\u2581seco",
"uring",
"\u2581ed",
"\u2581mem",
"ros",
"cri",
"\u2581thought",
"cept",
"\u2581partic",
"\u2581test",
"olog",
"iness",
"\u2581far",
"led",
"\u2581col",
"\u2581looking",
"\u2581read",
"\u2581whether",
"\u2581word",
"me",
"\u2581once",
"ize",
"\u2581home",
"\u2581requ",
"gg",
"\u2581ide",
"\u2581thank",
"ures",
"\u2581called",
"\u2581cur",
"\u2581water",
"\u2581frie",
"\u2581side",
"\u2581best",
"\u2581number",
"oney",
"\u2581turn",
"ock",
"\u2581eng",
"\u2581top",
"\u2581open",
"ead",
"\u2581everything",
"\u2581term",
"\u2581prob",
"\u2581hard",
"\u2581fun",
"\u2581spec",
"\u2581dire",
"\u2581second",
"\u2581pa",
"\u2581build",
"\u2581run",
"\u2581sign",
"\u2581reason",
"\u2581inform",
"\u2581watch",
"ution",
"\u2581few",
"mo",
"\u2581hum",
"ision",
"\u2581ext",
"\u2581tog",
"\u2581conc",
"\u2581thous",
"\u2581thousand",
"\u2581support",
"\u2581together",
"\u2581six",
"ps",
"\u2581mark",
"ics",
"\u2581includ",
"ef",
"\u2581opp",
"ident",
"\u2581anything",
"\u2581met",
"\u2581bre",
"\u2581jud",
"\u2581away",
"\u2581old",
"\u2581prog",
"ten",
"\u2581book",
"\u2581says",
"\u2581seem",
"\u2581contin",
"\u2581process",
"\u2581sing",
"\u2581money",
"\u2581having",
"\u2581beg",
"\u2581comple",
"\u2581thir",
"\u2581using",
"\u2581ret",
"ger",
"\u2581head",
"\u2581cre",
"\u2581poss",
"enty",
"\u2581certain",
"\u2581clear",
"ines",
"\u2581wee",
"arch",
"\u2581inf",
"ont",
"\u2581sit",
"\u2581lead",
"alth",
"\u2581art",
"ross",
"\u2581pub",
"\u2581without",
"\u2581pret",
"\u2581getting",
"ient",
"\u2581z",
"\u2581wom",
"\u2581power",
"ational",
"ner",
"\u2581rest",
"\u2581believe",
"\u2581wa",
"\u2581aut",
"\u2581move",
"aim",
"\u2581sort",
"idence",
"\u2581creat",
"\u2581expl",
"\u2581name",
"\u2581went",
"\u2581eu",
"\u2581change",
"\u2581came",
"\u2581pay",
"ices",
"\u2581sin",
"\u2581pur",
"\u2581pass",
"\u2581whole",
"\u2581house",
"\u2581hund",
"\u2581hundred",
"\u2581pretty",
"\u2581trying",
"\u2581ple",
"\u2581allow",
"\u2581compan",
"\u2581government",
"\u2581small",
"\u2581light",
"\u2581bra",
"\u2581stu",
"aint",
"\u2581ah",
"\u2581prot",
"ets",
"\u2581cent",
"velop",
"\u2581family",
"\u2581business",
"ety",
"\u2581making",
"\u2581list",
"\u2581experi",
"eric",
"\u2581follow",
"ately",
"\u2581probably",
"\u2581appe",
"\u2581serv",
"\u2581val",
"\u2581leg",
"\u2581resp",
"\u2581develop",
"ready",
"\u2581already",
"\u2581sec",
"ell",
"\u2581saying",
"ash",
"\u2581hear",
"\u2581loc",
"\u2581adv",
"\u2581pri",
"ret",
"\u2581lar",
"\u2581beh",
"\u2581must",
"\u2581hon",
"\u2581means",
"ew",
"\u2581par",
"\u2581order",
"\u2581mom",
"gn",
"\u2581though",
"\u2581record",
"\u2581miss",
"\u2581dr",
"\u2581es",
"\u2581eight",
"\u2581ever",
"\u2581left",
"\u2581example",
"\u2581enough",
"osed",
"\u2581claim",
"ank",
"con",
"\u2581americ",
"\u2581information",
"\u2581arg",
"\u2581full",
"nce",
"\u2581consid",
"\u2581working",
"ature",
"\u2581",
"e",
"t",
"a",
"o",
"i",
"n",
"s",
"r",
"h",
"l",
"d",
"u",
"c",
"m",
"y",
"w",
"g",
"f",
"p",
"b",
"v",
"k",
"'",
"j",
"x",
"q",
"z"
]
},
"decoding": {
"strategy": "greedy_batch",
"greedy": {
"max_symbols": 10
},
"beam": {
"beam_size": 2,
"return_best_hypothesis": false,
"score_norm": true,
"tsd_max_sym_exp": 50,
"alsd_max_target_len": 2.0
}
},
"loss": {
"loss_name": "default",
"warprnnt_numba_kwargs": {
"fastemit_lambda": 0.0,
"clamp": -1.0
}
},
"variational_noise": {
"start_step": 0,
"std": 0.0
},
"optim": {
"name": "adamw",
"lr": 0.0025,
"betas": [
0.9,
0.98
],
"weight_decay": 0.001,
"sched": {
"name": "CosineAnnealing",
"warmup_steps": 15000,
"warmup_ratio": null,
"min_lr": 1e-06
}
},
"labels": [
"<unk>",
"\u2581t",
"\u2581th",
"\u2581a",
"\u2581i",
"\u2581the",
"re",
"\u2581w",
"\u2581s",
"\u2581o",
"in",
"at",
"er",
"ou",
"nd",
"\u2581c",
"\u2581b",
"\u2581h",
"on",
"\u2581m",
"\u2581f",
"ing",
"\u2581to",
"en",
"\u2581p",
"\u2581and",
"\u2581d",
"es",
"or",
"an",
"ll",
"\u2581y",
"\u2581l",
"ed",
"\u2581of",
"\u2581in",
"it",
"is",
"\u2581you",
"\u2581that",
"ar",
"\u2581g",
"\u2581n",
"as",
"om",
"\u2581it",
"ic",
"ve",
"\u2581e",
"\u2581wh",
"\u2581be",
"us",
"le",
"al",
"ion",
"ow",
"\u2581we",
"\u2581re",
"\u2581is",
"ut",
"ot",
"ent",
"\u2581on",
"et",
"\u2581ha",
"ay",
"ct",
"\u2581he",
"id",
"\u2581for",
"\u2581st",
"ver",
"ly",
"ro",
"ig",
"\u2581so",
"ld",
"\u2581this",
"ke",
"\u2581u",
"se",
"all",
"st",
"ur",
"ce",
"ch",
"im",
"ith",
"\u2581as",
"\u2581k",
"\u2581an",
"\u2581was",
"\u2581j",
"\u2581with",
"ir",
"\u2581go",
"ra",
"\u2581do",
"\u2581have",
"\u2581li",
"\u2581sh",
"\u2581se",
"\u2581they",
"\u2581are",
"am",
"ht",
"\u2581but",
"ation",
"\u2581not",
"th",
"\u2581r",
"ally",
"ad",
"ust",
"\u2581or",
"\u2581com",
"ould",
"\u2581can",
"ill",
"\u2581ne",
"ight",
"\u2581ch",
"\u2581de",
"\u2581con",
"\u2581at",
"\u2581mo",
"ant",
"oo",
"il",
"\u2581me",
"\u2581what",
"\u2581there",
"ter",
"pe",
"\u2581ab",
"\u2581su",
"ere",
"ck",
"\u2581pro",
"\u2581al",
"\u2581fr",
"\u2581kn",
"\u2581all",
"ers",
"\u2581like",
"ge",
"\u2581ex",
"\u2581som",
"ul",
"\u2581your",
"\u2581v",
"pp",
"use",
"\u2581if",
"ess",
"ate",
"est",
"\u2581know",
"out",
"if",
"\u2581just",
"ment",
"qu",
"op",
"ain",
"\u2581one",
"ol",
"ri",
"art",
"very",
"\u2581wor",
"ive",
"ist",
"\u2581my",
"nt",
"ab",
"\u2581from",
"ort",
"\u2581ma",
"\u2581about",
"res",
"ity",
"\u2581out",
"\u2581bec",
"\u2581le",
"our",
"od",
"and",
"ink",
"ie",
"\u2581up",
"ind",
"os",
"un",
"ause",
"oug",
"um",
"\u2581some",
"\u2581int",
"\u2581by",
"\u2581pl",
"\u2581get",
"el",
"ard",
"\u2581when",
"\u2581don",
"her",
"\u2581will",
"\u2581us",
"\u2581would",
"ook",
"ies",
"ich",
"\u2581because",
"\u2581think",
"em",
"\u2581pe",
"\u2581his",
"ack",
"\u2581then",
"\u2581our",
"ide",
"\u2581tim",
"\u2581how",
"ven",
"\u2581tr",
"\u2581who",
"\u2581them",
"ure",
"\u2581ar",
"\u2581ye",
"\u2581more",
"\u2581going",
"ect",
"\u2581sa",
"\u2581cl",
"\u2581had",
"\u2581now",
"\u2581which",
"\u2581here",
"ous",
"\u2581their",
"\u2581tw",
"so",
"\u2581has",
"ud",
"\u2581co",
"\u2581ta",
"ound",
"\u2581were",
"ast",
"\u2581peop",
"ough",
"\u2581no",
"\u2581really",
"\u2581any",
"\u2581people",
"\u2581want",
"\u2581she",
"\u2581en",
"\u2581fa",
"\u2581te",
"ame",
"ine",
"\u2581qu",
"red",
"\u2581im",
"\u2581right",
"ther",
"\u2581act",
"\u2581thing",
"king",
"ose",
"\u2581ad",
"\u2581see",
"\u2581time",
"\u2581these",
"ci",
"one",
"\u2581say",
"\u2581also",
"\u2581fe",
"per",
"\u2581ag",
"\u2581man",
"ore",
"\u2581un",
"pt",
"\u2581her",
"\u2581look",
"ong",
"ice",
"\u2581very",
"ff",
"ions",
"\u2581comp",
"\u2581did",
"itt",
"\u2581well",
"\u2581other",
"iv",
"ase",
"ree",
"hing",
"\u2581lo",
"reat",
"\u2581cont",
"\u2581part",
"\u2581into",
"nder",
"\u2581been",
"are",
"\u2581am",
"ans",
"\u2581sp",
"\u2581two",
"ue",
"\u2581way",
"age",
"\u2581where",
"ite",
"\u2581dis",
"\u2581than",
"\u2581every",
"\u2581pr",
"\u2581po",
"ag",
"\u2581need",
"ach",
"iff",
"ence",
"pl",
"own",
"\u2581ac",
"ble",
"\u2581over",
"iz",
"\u2581work",
"\u2581res",
"\u2581make",
"\u2581could",
"\u2581off",
"ually",
"\u2581ro",
"\u2581back",
"able",
"ip",
"ry",
"\u2581him",
"\u2581cour",
"ber",
"\u2581pre",
"\u2581fir",
"\u2581spe",
"ap",
"ars",
"\u2581diff",
"ire",
"\u2581somet",
"\u2581imp",
"\u2581those",
"\u2581comm",
"ance",
"ick",
"\u2581even",
"ated",
"way",
"sel",
"\u2581let",
"\u2581br",
"ty",
"\u2581per",
"int",
"\u2581first",
"\u2581thr",
"\u2581under",
"ah",
"\u2581may",
"\u2581cou",
"\u2581new",
"ress",
"act",
"\u2581gr",
"ep",
"\u2581said",
"ations",
"\u2581good",
"ace",
"ass",
"\u2581does",
"orm",
"ish",
"\u2581af",
"ving",
"co",
"\u2581app",
"\u2581lot",
"\u2581things",
"\u2581tra",
"ittle",
"\u2581bl",
"\u2581little",
"\u2581mu",
"cess",
"fe",
"ome",
"\u2581inc",
"\u2581differe",
"ary",
"ical",
"\u2581only",
"ult",
"\u2581again",
"\u2581got",
"ens",
"\u2581gu",
"\u2581kind",
"\u2581much",
"ord",
"\u2581through",
"ition",
"ild",
"\u2581down",
"\u2581actually",
"\u2581something",
"ang",
"ru",
"ces",
"\u2581fl",
"ile",
"ater",
"\u2581ra",
"\u2581take",
"ict",
"ign",
"\u2581sc",
"vel",
"\u2581bet",
"\u2581tal",
"\u2581yeah",
"\u2581use",
"fore",
"\u2581bu",
"\u2581start",
"ory",
"be",
"\u2581day",
"wn",
"xt",
"ia",
"ak",
"\u2581after",
"\u2581should",
"\u2581fo",
"\u2581ho",
"\u2581hel",
"\u2581ind",
"\u2581uh",
"na",
"ial",
"other",
"\u2581ke",
"\u2581call",
"\u2581most",
"\u2581ok",
"\u2581different",
"\u2581em",
"ting",
"ple",
"\u2581being",
"\u2581bo",
"ning",
"\u2581too",
"ors",
"\u2581happ",
"ark",
"og",
"\u2581help",
"\u2581rem",
"du",
"ction",
"ood",
"\u2581ser",
"ether",
"ious",
"\u2581mean",
"\u2581many",
"\u2581court",
"\u2581bel",
"ade",
"\u2581la",
"ved",
"\u2581des",
"\u2581rec",
"\u2581jo",
"\u2581dec",
"ves",
"\u2581before",
"\u2581put",
"self",
"\u2581point",
"te",
"\u2581ev",
"form",
"ents",
"\u2581add",
"ody",
"thing",
"\u2581case",
"\u2581pers",
"\u2581cons",
"iss",
"\u2581three",
"oth",
"\u2581ph",
"\u2581come",
"\u2581find",
"\u2581why",
"ull",
"\u2581show",
"\u2581bas",
"\u2581great",
"ily",
"\u2581rel",
"\u2581sm",
"\u2581its",
"\u2581fact",
"\u2581pos",
"ool",
"ments",
"ise",
"nds",
"ys",
"\u2581try",
"ual",
"ful",
"erm",
"\u2581inter",
"ons",
"\u2581quest",
"\u2581sub",
"we",
"vers",
"\u2581supp",
"\u2581feel",
"\u2581same",
"ub",
"ates",
"urn",
"ert",
"\u2581inv",
"day",
"\u2581rep",
"igh",
"\u2581sy",
"\u2581inst",
"\u2581long",
"\u2581still",
"\u2581okay",
"ft",
"ific",
"atch",
"ought",
"ath",
"\u2581own",
"\u2581made",
"ix",
"ced",
"ks",
"lic",
"\u2581wr",
"de",
"\u2581cr",
"\u2581att",
"\u2581ob",
"\u2581world",
"\u2581sure",
"ward",
"\u2581bit",
"\u2581life",
"\u2581person",
"\u2581pres",
"ph",
"\u2581vide",
"\u2581reg",
"\u2581end",
"ject",
"ange",
"\u2581fin",
"ied",
"pect",
"\u2581didn",
"\u2581around",
"ian",
"\u2581car",
"ible",
"\u2581sim",
"ever",
"\u2581sch",
"ating",
"\u2581pol",
"\u2581set",
"\u2581oh",
"cy",
"\u2581real",
"\u2581import",
"\u2581count",
"\u2581um",
"\u2581next",
"cial",
"les",
"\u2581hu",
"\u2581acc",
"\u2581might",
"\u2581ent",
"\u2581doing",
"\u2581ins",
"\u2581gen",
"\u2581play",
"\u2581cle",
"\u2581another",
"ady",
"ular",
"ib",
"ways",
"ered",
"ility",
"ities",
"\u2581op",
"\u2581def",
"\u2581years",
"\u2581never",
"ower",
"ram",
"\u2581tell",
"\u2581sl",
"onna",
"ail",
"ren",
"ute",
"\u2581gonna",
"\u2581big",
"\u2581give",
"der",
"ount",
"\u2581ap",
"kes",
"\u2581state",
"\u2581cor",
"\u2581min",
"ically",
"\u2581mon",
"\u2581fam",
"\u2581important",
"\u2581always",
"\u2581high",
"\u2581four",
"\u2581gra",
"\u2581ca",
"\u2581stud",
"\u2581dist",
"\u2581talk",
"\u2581num",
"\u2581str",
"\u2581today",
"ract",
"\u2581while",
"ason",
"\u2581iss",
"\u2581sur",
"\u2581char",
"\u2581last",
"oy",
"ited",
"\u2581exper",
"\u2581place",
"\u2581tri",
"\u2581ear",
"\u2581belie",
"\u2581able",
"\u2581underst",
"\u2581che",
"\u2581both",
"ug",
"\u2581doesn",
"\u2581keep",
"\u2581happen",
"ings",
"iew",
"ather",
"\u2581ass",
"\u2581love",
"ative",
"av",
"\u2581yes",
"\u2581ele",
"\u2581year",
"\u2581such",
"\u2581video",
"ness",
"\u2581el",
"\u2581trans",
"\u2581five",
"\u2581produ",
"ave",
"erest",
"als",
"body",
"cus",
"\u2581found",
"atter",
"\u2581eff",
"\u2581god",
"\u2581used",
"llow",
"\u2581interest",
"\u2581question",
"hip",
"\u2581bus",
"\u2581ask",
"\u2581exam",
"\u2581prov",
"lud",
"\u2581form",
"\u2581law",
"ense",
"\u2581child",
"\u2581gl",
"ne",
"\u2581each",
"\u2581understand",
"\u2581care",
"stem",
"\u2581med",
"\u2581maybe",
"ably",
"\u2581det",
"\u2581coll",
"its",
"\u2581commun",
"\u2581hand",
"\u2581'",
"\u2581ref",
"\u2581lear",
"\u2581done",
"\u2581gener",
"vern",
"\u2581mr",
"ween",
"\u2581better",
"\u2581between",
"li",
"blem",
"\u2581system",
"ertain",
"\u2581school",
"\u2581eas",
"\u2581exp",
"\u2581war",
"ention",
"\u2581ty",
"\u2581govern",
"ues",
"\u2581problem",
"\u2581plan",
"ac",
"\u2581conf",
"\u2581course",
"ouse",
"\u2581mar",
"\u2581stand",
"\u2581sk",
"\u2581seco",
"uring",
"\u2581ed",
"\u2581mem",
"ros",
"cri",
"\u2581thought",
"cept",
"\u2581partic",
"\u2581test",
"olog",
"iness",
"\u2581far",
"led",
"\u2581col",
"\u2581looking",
"\u2581read",
"\u2581whether",
"\u2581word",
"me",
"\u2581once",
"ize",
"\u2581home",
"\u2581requ",
"gg",
"\u2581ide",
"\u2581thank",
"ures",
"\u2581called",
"\u2581cur",
"\u2581water",
"\u2581frie",
"\u2581side",
"\u2581best",
"\u2581number",
"oney",
"\u2581turn",
"ock",
"\u2581eng",
"\u2581top",
"\u2581open",
"ead",
"\u2581everything",
"\u2581term",
"\u2581prob",
"\u2581hard",
"\u2581fun",
"\u2581spec",
"\u2581dire",
"\u2581second",
"\u2581pa",
"\u2581build",
"\u2581run",
"\u2581sign",
"\u2581reason",
"\u2581inform",
"\u2581watch",
"ution",
"\u2581few",
"mo",
"\u2581hum",
"ision",
"\u2581ext",
"\u2581tog",
"\u2581conc",
"\u2581thous",
"\u2581thousand",
"\u2581support",
"\u2581together",
"\u2581six",
"ps",
"\u2581mark",
"ics",
"\u2581includ",
"ef",
"\u2581opp",
"ident",
"\u2581anything",
"\u2581met",
"\u2581bre",
"\u2581jud",
"\u2581away",
"\u2581old",
"\u2581prog",
"ten",
"\u2581book",
"\u2581says",
"\u2581seem",
"\u2581contin",
"\u2581process",
"\u2581sing",
"\u2581money",
"\u2581having",
"\u2581beg",
"\u2581comple",
"\u2581thir",
"\u2581using",
"\u2581ret",
"ger",
"\u2581head",
"\u2581cre",
"\u2581poss",
"enty",
"\u2581certain",
"\u2581clear",
"ines",
"\u2581wee",
"arch",
"\u2581inf",
"ont",
"\u2581sit",
"\u2581lead",
"alth",
"\u2581art",
"ross",
"\u2581pub",
"\u2581without",
"\u2581pret",
"\u2581getting",
"ient",
"\u2581z",
"\u2581wom",
"\u2581power",
"ational",
"ner",
"\u2581rest",
"\u2581believe",
"\u2581wa",
"\u2581aut",
"\u2581move",
"aim",
"\u2581sort",
"idence",
"\u2581creat",
"\u2581expl",
"\u2581name",
"\u2581went",
"\u2581eu",
"\u2581change",
"\u2581came",
"\u2581pay",
"ices",
"\u2581sin",
"\u2581pur",
"\u2581pass",
"\u2581whole",
"\u2581house",
"\u2581hund",
"\u2581hundred",
"\u2581pretty",
"\u2581trying",
"\u2581ple",
"\u2581allow",
"\u2581compan",
"\u2581government",
"\u2581small",
"\u2581light",
"\u2581bra",
"\u2581stu",
"aint",
"\u2581ah",
"\u2581prot",
"ets",
"\u2581cent",
"velop",
"\u2581family",
"\u2581business",
"ety",
"\u2581making",
"\u2581list",
"\u2581experi",
"eric",
"\u2581follow",
"ately",
"\u2581probably",
"\u2581appe",
"\u2581serv",
"\u2581val",
"\u2581leg",
"\u2581resp",
"\u2581develop",
"ready",
"\u2581already",
"\u2581sec",
"ell",
"\u2581saying",
"ash",
"\u2581hear",
"\u2581loc",
"\u2581adv",
"\u2581pri",
"ret",
"\u2581lar",
"\u2581beh",
"\u2581must",
"\u2581hon",
"\u2581means",
"ew",
"\u2581par",
"\u2581order",
"\u2581mom",
"gn",
"\u2581though",
"\u2581record",
"\u2581miss",
"\u2581dr",
"\u2581es",
"\u2581eight",
"\u2581ever",
"\u2581left",
"\u2581example",
"\u2581enough",
"osed",
"\u2581claim",
"ank",
"con",
"\u2581americ",
"\u2581information",
"\u2581arg",
"\u2581full",
"nce",
"\u2581consid",
"\u2581working",
"ature",
"\u2581",
"e",
"t",
"a",
"o",
"i",
"n",
"s",
"r",
"h",
"l",
"d",
"u",
"c",
"m",
"y",
"w",
"g",
"f",
"p",
"b",
"v",
"k",
"'",
"j",
"x",
"q",
"z"
],
"target": "nemo.collections.asr.models.rnnt_bpe_models.EncDecRNNTBPEModel",
"nemo_version": "1.19.0rc0"
}