Coder Social home page Coder Social logo

interactive_e2e_speech_recognition's Issues

Loss descends unusually on reproduction

Thanks for your code!
But I the Loss descends unusually, so the recognition is also at low accuracy. In your result, the loss is descend rapidly.
I haven't modify the code yet.

My environment is:

  • PyTorch 1.4
  • torchaudio 0.4

The Loss log.

epoch 1: avg_loss 6.371933863713191
epoch 2: avg_loss 4.70939903992873
epoch 3: avg_loss 4.671203154783982
epoch 4: avg_loss 3.8700154561262865
epoch 5: avg_loss 3.468390097984901
epoch 6: avg_loss 3.273801638529851
epoch 7: avg_loss 3.261995315551758
epoch 8: avg_loss 3.233139918400691
epoch 9: avg_loss 3.1941104118640604
epoch 10: avg_loss 3.1285105301783633
epoch 11: avg_loss 3.14404559135437
epoch 12: avg_loss 3.095609554877648
epoch 13: avg_loss 3.068654280442458
epoch 14: avg_loss 3.057704448699951
epoch 15: avg_loss 3.0675538136408877
epoch 16: avg_loss 2.992294329863328
epoch 17: avg_loss 2.9800828236799974
epoch 18: avg_loss 2.9441260741307187
epoch 19: avg_loss 2.894284596809974
epoch 20: avg_loss 2.936523822637705
epoch 21: avg_loss 2.946532964706421
epoch 22: avg_loss 2.9159597066732554
epoch 23: avg_loss 2.8778565480158877
epoch 24: avg_loss 2.901293919636653
epoch 25: avg_loss 2.832904577255249
epoch 26: avg_loss 2.8665461356823263
epoch 27: avg_loss 2.870034859730647
epoch 28: avg_loss 2.8736569698040304
epoch 29: avg_loss 2.8818020820617676
epoch 30: avg_loss 2.940280767587515
epoch 31: avg_loss 2.8641006396367
epoch 32: avg_loss 2.755228482759916
epoch 33: avg_loss 2.7636474646054783
epoch 34: avg_loss 2.7970426632807803
epoch 35: avg_loss 2.8201447266798754
epoch 36: avg_loss 2.827310708852915
epoch 37: avg_loss 2.851408389898447
epoch 38: avg_loss 2.7742223739624023
epoch 39: avg_loss 2.797891195003803
epoch 40: avg_loss 2.7421194406656118
epoch 41: avg_loss 2.7057653023646426
epoch 42: avg_loss 2.82857293349046
epoch 43: avg_loss 2.75337945497953
epoch 44: avg_loss 2.8449047345381517
epoch 45: avg_loss 2.8148957582620473
epoch 46: avg_loss 2.7799972112362203
epoch 47: avg_loss 2.779747247695923
epoch 48: avg_loss 2.7723453778486986
epoch 49: avg_loss 2.7705439971043515
epoch 50: avg_loss 2.7763983469742994
epoch 51: avg_loss 2.716023793587318
epoch 52: avg_loss 2.7847886635706973
epoch 53: avg_loss 2.7497686789585996
epoch 54: avg_loss 2.7348319567166843
epoch 55: avg_loss 2.804599248445951
epoch 56: avg_loss 2.80349848820613
epoch 57: avg_loss 2.757105882351215
epoch 58: avg_loss 2.805212516051072
epoch 59: avg_loss 2.72385520201463
epoch 60: avg_loss 2.7668622640463023
epoch 61: avg_loss 2.702036655866183
epoch 62: avg_loss 2.8305543019221377
epoch 63: avg_loss 2.778572999514066
epoch 64: avg_loss 2.8520910923297587
epoch 65: avg_loss 2.7859388498159556
epoch 66: avg_loss 2.728653302559486
epoch 67: avg_loss 2.8242092774464536
epoch 68: avg_loss 2.773068721477802
epoch 69: avg_loss 2.7353643820835996
epoch 70: avg_loss 2.821800415332501
epoch 71: avg_loss 2.808179286810068
epoch 72: avg_loss 2.75307664504418
epoch 73: avg_loss 2.8359962976895847
epoch 74: avg_loss 2.736109733581543
epoch 75: avg_loss 2.8306180330423207
epoch 76: avg_loss 2.742473767353938
epoch 77: avg_loss 2.7637857107015757
epoch 78: avg_loss 2.7592280828035793
epoch 79: avg_loss 2.7490466007819543
epoch 80: avg_loss 2.828728043116056
epoch 81: avg_loss 2.917681437272292
epoch 82: avg_loss 2.8417651928388157
epoch 83: avg_loss 2.7903012128976674
epoch 84: avg_loss 2.6843167268312893
epoch 85: avg_loss 2.790263982919546
epoch 86: avg_loss 2.7775218303386984
epoch 87: avg_loss 2.7608486322256236
epoch 88: avg_loss 2.7123066095205455
epoch 89: avg_loss 2.7112955496861386
epoch 90: avg_loss 2.725948663858267
epoch 91: avg_loss 2.724459904890794
epoch 92: avg_loss 2.6767702469458947
epoch 93: avg_loss 2.7425847236926737
epoch 94: avg_loss 2.760955755527203
epoch 95: avg_loss 2.7504168106959415
epoch 96: avg_loss 2.6886834914867697
epoch 97: avg_loss 2.8154220581054688
epoch 98: avg_loss 2.736657527776865
epoch 99: avg_loss 2.725706320542556

Have you come across this phenomenon?
Any help will be appreciated!

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google ❤️ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.