Splend1dchan/wav2vec2-large-lv60_t5lephone-small_nofreeze_bs16_forMINDS.en.all2

wav2vec2 -> t5lephone bs = 16 dropout = 0.3

performance : 29%

{ "architectures": [ "SpeechMixEEDT5" ], "decoder": { "nameorpath": "voidful/phonemebyt5", "addcrossattention": true, "architectures": [ "T5ForConditionalGeneration" ], "badwordsids": null, "bostokenid": null, "chunksizefeedforward": 0, "crossattentionhiddensize": null, "dff": 3584, "dkv": 64, "dmodel": 1472, "decoderstarttokenid": 0, "diversitypenalty": 0.0, "dosample": false, "dropoutrate": 0.1, "earlystopping": false, "encodernorepeatngramsize": 0, "eostokenid": 1, "feedforwardproj": "gated-gelu", "finetuningtask": null, "forcedbostokenid": null, "forcedeostokenid": null, "gradientcheckpointing": false, "id2label": { "0": "LABEL0", "1": "LABEL1" }, "initializerfactor": 1.0, "isdecoder": true, "isencoderdecoder": true, "label2id": { "LABEL0": 0, "LABEL1": 1 }, "layernormepsilon": 1e-06, "lengthpenalty": 1.0, "maxlength": 20, "minlength": 0, "modeltype": "t5", "norepeatngramsize": 0, "numbeamgroups": 1, "numbeams": 1, "numdecoderlayers": 4, "numheads": 6, "numlayers": 12, "numreturnsequences": 1, "outputattentions": false, "outputhiddenstates": false, "outputscores": false, "padtokenid": 0, "prefix": null, "problemtype": null, "prunedheads": {}, "relativeattentionmaxdistance": 128, "relativeattentionnumbuckets": 32, "removeinvalidvalues": false, "repetitionpenalty": 1.0, "returndict": true, "returndictingenerate": false, "septokenid": null, "taskspecificparams": null, "temperature": 1.0, "tieencoderdecoder": false, "tiewordembeddings": false, "tokenizerclass": "ByT5Tokenizer", "topk": 50, "topp": 1.0, "torchdtype": "float32", "torchscript": false, "transformersversion": "4.17.0", "typicalp": 1.0, "usebfloat16": false, "usecache": true, "vocabsize": 384 }, "encoder": { "nameorpath": "facebook/wav2vec2-large-lv60", "activationdropout": 0.1, "adapterkernelsize": 3, "adapterstride": 2, "addadapter": false, "addcrossattention": false, "applyspecaugment": true, "architectures": [ "Wav2Vec2ForPreTraining" ], "attentiondropout": 0.1, "badwordsids": null, "bostokenid": 1, "chunksizefeedforward": 0, "classifierprojsize": 256, "codevectordim": 768, "contrastivelogitstemperature": 0.1, "convbias": true, "convdim": [ 512, 512, 512, 512, 512, 512, 512 ], "convkernel": [ 10, 3, 3, 3, 3, 2, 2 ], "convstride": [ 5, 2, 2, 2, 2, 2, 2 ], "crossattentionhiddensize": null, "ctclossreduction": "sum", "ctczeroinfinity": false, "decoderstarttokenid": null, "diversitylossweight": 0.1, "diversitypenalty": 0.0, "dosample": false, "dostablelayernorm": true, "earlystopping": false, "encodernorepeatngramsize": 0, "eostokenid": 2, "featextractactivation": "gelu", "featextractdropout": 0.0, "featextractnorm": "layer", "featprojdropout": 0.1, "featquantizerdropout": 0.0, "finaldropout": 0.1, "finetuningtask": null, "forcedbostokenid": null, "forcedeostokenid": null, "gradientcheckpointing": false, "hiddenact": "gelu", "hiddendropout": 0.1, "hiddendropoutprob": 0.1, "hiddensize": 1024, "id2label": { "0": "LABEL0", "1": "LABEL1" }, "initializerrange": 0.02, "intermediatesize": 4096, "isdecoder": false, "isencoderdecoder": false, "label2id": { "LABEL0": 0, "LABEL1": 1 }, "layernormeps": 1e-05, "layerdrop": 0.0, "lengthpenalty": 1.0, "maskfeaturelength": 10, "maskfeatureminmasks": 0, "maskfeatureprob": 0.0, "masktimelength": 10, "masktimeminmasks": 2, "masktimeprob": 0.05, "maxlength": 20, "minlength": 0, "modeltype": "wav2vec2", "norepeatngramsize": 0, "numadapterlayers": 3, "numattentionheads": 16, "numbeamgroups": 1, "numbeams": 1, "numcodevectorgroups": 2, "numcodevectorspergroup": 320, "numconvposembeddinggroups": 16, "numconvposembeddings": 128, "numfeatextractlayers": 7, "numhiddenlayers": 24, "numnegatives": 100, "numreturnsequences": 1, "outputattentions": false, "outputhiddensize": 1024, "outputhiddenstates": false, "outputscores": false, "padtokenid": 0, "prefix": null, "problemtype": null, "projcodevectordim": 768, "prunedheads": {}, "removeinvalidvalues": false, "repetitionpenalty": 1.0, "returndict": true, "returndictingenerate": false, "septokenid": null, "taskspecificparams": null, "tdnndilation": [ 1, 2, 3, 1, 1 ], "tdnndim": [ 512, 512, 512, 512, 1500 ], "tdnnkernel": [ 5, 3, 3, 1, 1 ], "temperature": 1.0, "tieencoderdecoder": false, "tiewordembeddings": true, "tokenizerclass": null, "topk": 50, "topp": 1.0, "torchdtype": null, "torchscript": false, "transformersversion": "4.17.0", "typicalp": 1.0, "usebfloat16": false, "useweightedlayersum": false, "vocabsize": 32, "xvectoroutputdim": 512 }, "isencoderdecoder": true, "modeltype": "speechmix", "torchdtype": "float32", "transformersversion": null }