Python Chat Bot Tutorial - Chatbot with Deep Learning (Part 3)

[@mvaughn663]

for anyone following along in 2023 the function "tensorflow.reset_default_graph()" is deprecated instead "tensorflow.compat.v1.reset_default_graph()" should work

[@spiderhollow5658]

Thank you very much, I was wondering what to do about that

[@arbitraryguy606]

Thanks

[@alinecano3183]

Me salvaste!! X3

[@Anonymous51701]

Removing the batch_size argument improved my accuracy

[@royalstastista4757]

There is a minute correction in line 37 the *classes* is not defined yet , it should be *lables* and in line 29 code is ------words = [stemmer.stem(w.lower()) for w in words if w != "?"]words = [stemmer.stem(w.lower()) for w in words if w != "?"]------- VERY VERY VERY helpful tutorial , Love it

[@Epic-Sidemen-Moments]

Change to tags

[@laxmanjeergal3827]

number of epochs =." number of times it sees the data" the best explanation ever

[@remixowlz]

Thank you Tim! You made it very easy to understand especially not just with theories and definitions but you gave examples or referencing from the training data. Keep it up! :) I'm a newb at ML

[@리주민]

I recently found out that the person I was talking (texting) to at my uni was actually a chatbot. It was actually pretty good. Wonder if my professor is one too 😋

[@markcohen2533]

man got catfished by a machine

[@i.arcturus]

thank you tim! i tried other videos but none of them actually worked. my code would show a whole bunch of errors and i had no idea how to make sense of them. but this works! so thanks!

[@absolutelyproprietary6896]

Parse > Process > Convert > ✨magic✨ > result

[@dr_frankenmiller2607]

Great lesson Tim I be here again tomorrow for Part 4 Everybody stay healthy stay safe please

[@justhangingout9666]

Has this tutorial worked for you so far? I am having no problems until I type activate chatbot and it doesn't recognize the command activate. Any advice?

[@akhileshvvn]

great tutorials....now i am addicted to your channel....

[@floydian25]

Tim honestly this series is so confusing to me. Please go a little slow. Love your pygame series.

[@Daviesdev]

Bro we are on the same page here

[@Daviesdev]

I'm lost

[@jonathanoakey2778]

Dude just slow the video down to 0.5 if it's to fast for you, I for real watched this at x2

[@gulkbag3538]

This series is so awesome i am looking forward for the future vids

[@abhijeetjain2098]

faltu

[@dipanshusomeshwar]

If anyone finds LOOKUP ERROR!! Simply go to Python shell and type: >>>import nltk >>>nltk.download() and install tha punkt module from models...

[@drviber6865]

Hi, Try to run this command in the IDE it self for ex: import tensorflow import numpy import nltk nltk.download() and it will work I think u need to run the nltk.download for one time, all the best.

[@aravindnallajerla4614]

@@drviber6865 Hey Mr. Viber, I am getting the following error. Could you please help me out. Thanks in advance :) ValueError: setting an array element with a sequence Help me fix it.

[@conversationalai8320]

thanks for the tutorial. Nicely explained.

[@utkarshmudgal2879]

My model not working on heroku how to implement this model on heroku using flask?

[@krishshah8838]

doc=nltk.word_tokenize(doc) this line needs to be added before defining wrds in the second loop because to use stemmer first we need to tokenize words in every doc in doc_x,or elese wrds takes doc as string rather than list of words...it will rresolve the accuracy problem

[@gauravsarawgi6727]

where exactly....can u specify

[@Pl-vw4wd]

Awesome! Thanks for sharing that.

[@НатальяСоколова-я1г]

Would you reccomend some practical stuff (books, playlists, articles) on keras? intermediate level

[@harshvardhansharma2102]

Hi Tim if you could clear my doubt that will be of great help. Videos are awesome Tim...

[@drviber6865]

Hi great teacher Tim I had an error with tf which it doesnt work with the reset_default_graph, and the other issue is this one Traceback (most recent call last): File "sample_chatbot.py", line 54, in output_row[labels.index(docs_y[x])] = 1 ValueError: 'greeting' is not in list thanks .

[@anuragbhattacharjee7758]

Even am getting the same error...how did u solved it?

[@balasaiallagadda2791]

Yeah even I am getting the same error it says tensorflow has no attribute reset_default_graph When I use tensorflow.reset_default_graph()

[@balasaiallagadda2791]

Plz help me out

[@mcquilf2270]

import tensorflow as tf then replace tensorflow.reset_default_graph() with tf.compat.v1.reset_default_graph()

[@0568raju]

You are a good teacher. Next videos?

[@breadzeppelin2705]

For those of you getting very low accuracy, fix all the minor errors he points out, it should work

[@biswajitsatapathy1595]

tflearn is not working for tensorflow 2.0. so instead of downgrading to tensorflow1.14 is there any alternative to tflearn

[@gurpremsingh]

This reply is a bit late but, keras can be used. A guy named Kunal Duran made a video where he added onto this video series, you can check that out.

[@marioharoun5372]

I get error raise RuntimeError('Attempted to use a closed Session.') RuntimeError: Attempted to use a closed Session. i don't understand why. it is bc of the tensorflow I think

[@tony-todo]

Running the script as 11:28 does not work - it just stays, frozen cmd unable to do anything

[@doomguy606]

When I add new tags to the json files, the bot doesn't recognize. Am I missing something in the main.py file?

[@krishuynh3337]

it also happens to me

[@mr.jupiterekbghost1392]

it is giving me the error that activate is not recoganised as internal or external operable bach file

[@vasudevnair9670]

you only need to activate chatbot if youre using anaconda, I believe

[@alokranjan04]

Hi Tim I am getting the error ValueError: Cannot feed value of shape (8, 6) for Tensor 'TargetsData/Y:0', which has shape '(?, 46)'

[@ahmetsahin3388]

@Tech With Tim #TechWithTim File "Kuasar.py", line 37, in words = [stemmer.stem(w.lower()) for w in words if w not in "?"] File "Kuasar.py", line 37, in words = [stemmer.stem(w.lower()) for w in words if w not in "?"] TypeError: stem() missing 1 required positional argument: 'word'

[@alalz]

bit late, but did you check that you had brackets after the LancasterStemmer? around line 3 you should have: stemmer = LancasterStemmer()

[@mateuszbiskup4706]

ModuleNotFoundError: No module named 'tensorflow.contrib' To solve this issue we need to have tensorflow

[@skg2753]

I have also faced same issue, I have uninstalled latest version of tensorflow and installed 1.15 and it worked for me. I found that 1.15 has required files to run it(i.e. contrib)

[@dsnk3777]

please help me out, every time i try importing tensorflow on a conda environment it gives me an import error saying a dll file failed to load

[@ifredpingplay468]

download .dll and past it to system32

[@nabihassan7576]

7:34 i have a confusion that is softmax creates 6 neurons and each neuron represents a tag, then what if we have more than 6 tags

[@skullrazor]

It doesn't matter if u have more tags, (same as if there's one output) the output neurons always stays the same. In this case, its was an example of 6 neurons, but since the output layer contains len(output...), so basically the length of output array which is updated previously, it will have a variable number of neurons, so don't have to worry on tags. (btw, does tflearn work? it doesn't work for me, giving no module named "tensorflow.contrib")

[@nabihassan7576]

@@skullrazor you are right about neurons and it doesn't matter ichecked it. And my code is working perfectly with tensorflow 1.14 and python 3.6.8

[@sciesteam]

It give me error sys.stdout.flush() OSError [WinError 87] the parameter is incorrect

[@adityavikram5176]

hey... for this line output_row[labels.index(docs_y[x])]=1 iam getting error :- ValueError: 'greeting' is not in list can someone pls help?

[@ashishsapra835]

Exact same error

[@Adi02]

I'm using Vscode and when I run the code this error comes- ImportError: Could not find 'nvcuda.dll'. TensorFlow requires that this DLL be installed in a directory that is named in your %PATH% environment variable. Typically it is installed in 'C:\Windows\System32'. If it is not present, ensure that you have a CUDA-capable GPU with the correct driver installed. I believe it is got to do with the fact that I'm using an AMD GPU and Cuda is not compatible with all AMD GPU'S. Is there a solution to this?

[@anthonygonsalvis121]

I have a normal computer with NVIDIA GPU which is NOT CUDA-enabled and the program worked just fine for me (with one minor tweak to the latest TensorFlow I installed in mid-Nov-2020). I did NOT uninstall or changed to a lower version of TensorFlow. You just need "tensorflow.compat.v1.reset_default_graph()" Btw, for neural networks that do need a decent GPU, it's not just AMD GPU. Even if you have an NVIDIA GPU, it must be a CUDA-enabled GPU. Gaming computers are more likely to have CUDA-enabled NVIDIA GPUs. It's very important to check the H/W config very carefully when buying a new desktop or a laptop. As of mid-Nov-2020, NVIDIA RTX 2060 and higher models are CUDA-enabled. But that's still the "most basic" CUDA-enabled GPU you can get. Another KZbin AI/ML celebrity talks about having not one but two NVIDIA RTX 8000 GPUs (each retailing for ~$5,500 on Amazon) installed in his desktop/workstation! Regardless, I think for machine learning enthusiasts, it's best to have an upgradable desktop rather than a laptop whose GPU you can't upgrade. An alternative is to use any of the leading cloud service providers (AWS, Azure, GCP). I have heard that GCP offers a free tier that includes GPUs! (unlike AWS and Azure?). Many industrial-grade models apparently require the high-end NVIDIA Tesla GPUs.

[@shanboss8296]

Hi Tim, my accuracy is 0.05, what should I do to fix this?

[@justhangingout9666]

Increase the number of times you run the training, the 1000 he talked about. Has this tutorial worked for you so far? I am having no problems until I type activate chatbot and it doesn't recognize the command activate. Any advice?

[@kantulaev]

Ty, I also started to read keras book about ML, very interesting!

[@НатальяСоколова-я1г]

What a book?

[@naveinkumar3143]

Hey Tim can u explain me why do u use regreesion in this code this is a classifoiction problem right?

[@nabihassan7576]

did you find the answer

[@meryemamzil3519]

nice question

[@calebarulandu3068]

Hey Tim, is tflearn supported with tensorflow-gpu==2.0.0-beta1? When I try to import tflearn, it says that the contrib module wasn't found in tensorflow. Is there a tensorflow.keras equivalent for the model? I trying to use keras.Sequential(). So far I have: model = keras.Sequential([ keras.layers.Flatten(input_shape=[None, len(training[0])]), keras.layers.Dense(8), keras.layers.Dense(8), keras.layers.Dense(len(output[0]), activation="softmax"), ]) Also, when I try to run my code, there is a problem with nltk.word_tokenize and it says that punkt is not found? Are all of these errors due to deprecation?

[@calebarulandu3068]

I just found the solution to the nltk.word_tokenize but I'm still trying to figure out the model.

[@calebarulandu3068]

Nevermind. I just fixed the problem. It works properly now. I had to use np.reshape and the model: model = keras.Sequential([ keras.layers.Flatten(input_shape= (len(training[0]), )), keras.layers.Dense(8), keras.layers.Dense(8), keras.layers.Dense(len(output[0]), activation="softmax"), ]) Great Tutorial!

[@antonfranzen2233]

@@calebarulandu3068 Hey, i also got the error that says punkt is not found, how do i solve it??

[@calebarulandu3068]

@@antonfranzen2233 Essentially, you have to reshape the array using numpy so that the input and output shapes match. As for tflearn, try to use keras sequential. I had to make my own neural net design to accommodate for commands I couldn't find in the keras documentation. Try to play around with the arrays and neural net set up. If you need a more detailed step by step solution let me know, but you'll learn so much more playing with it and figuring it out on your own.

[@calebarulandu3068]

@@antonfranzen2233 I couldn't find a solution to the punkt problem. Keras or tensorflow docs is the way to go. I'm pretty sure that tflearn doesn't work with tensorflow anymore but it wouldn't hurt to double check.

[@richikpurkayastha8908]

can anyone please help me the code is showing invalid syntax net... in net= tflearn.regression(net) please help me

[@willjohnson4579]

could you paste your code and the error if you havent solved it?

[@richikpurkayastha8908]

@@willjohnson4579 sir i have solved it...

[@willjohnson4579]

@@richikpurkayastha8908 that's good news!

[@nabihassan7576]

Traceback (most recent call last): File "main.py", line 51, in output_row[labels.index(docs_y[x])] = 1 IndexError: list index out of range How do i fix this error. Pls help

[@hassanamraoui2794]

Hey man, I am having the same error. Did you fix it?

[@nabihassan7576]

@@hassanamraoui2794 yeah i forgot to define docs_y[ ] that's why I was having this error.

[@hassanamraoui2794]

Nabi hassan thanks for your help It works now. Have a good day

[@nabihassan7576]

@@hassanamraoui2794 glad to be of help

[@azizulhaqdharma5968]

Sorry but may I ask. If I have 11 tags in the intents do I still put the hidden layers as 8? Or do I change it to a number that's higher than the tags? Thanks

[@justhangingout9666]

You would keep them the same, if your code has a low accuracy rate then try increasing the 1000 number he talked about. Has this tutorial worked for you so far? I am having no problems until I type activate chatbot and it doesn't recognize the command activate. Any advice?

[@Daviesdev]

Hey Tim From what I'm seeing Pycharm or tensorflow doesn't recognise “tensorflow.reset_default_graph”

[@rameshsrivastavachandra]

try: tf.reset_default_graph

[@collinsmakamu4544]

tensorflow.compat.v1.reset_default_graph()

[@M45t3rJ4ck]

@@collinsmakamu4544 thanks bru!!

[@dlyq10]

Hi, first off, thanks Tim for an amazing tutorial series. I wanna ask, why do we ignore "?" in line 42, when we can just remove "?" from our list of words in line29 with words = [stemmer.stem(w.lower()) for w in words if w.isalpha()]? Thanks!

[@willjohnson4579]

doing this would remove numbers and currency symbols which are useful. For me i didnt lose accuracy when adding isalpha and even isalnum, but your mileage may vary

[@CDCBytes]

hello tim, it is a great tutorial, but I am getting an Value error : Cannot feed value of shape (8, 1, 46) for Tensor 'InputData/X:0', which has shape '(?, 1)'

[@spiderhollow5658]

I'm having the same issue

[@technoinfoworldwide2329]

I have no knowledge on machine learning coding but have concept what it can do .so may I watch your chat bot series?did you explain your coding and theoretical part so that I can easily understand the system?should I watch or I should learn tutorial on deep learning first?

[@fakesweta9467]

you must first learn the basic concept

[@unknowntraveller7046]

I am getting error like : The error you are encountering seems to be related to compatibility issues between TensorFlow and TFLearn libraries. The error message itself suggests that there is an issue with importing the is_sequence function from the nest module in TensorFlow. This is likely due to a mismatch between the versions of TensorFlow and TFLearn that you are using. what to do in discord no one is replying...

[@shayp9428]

lower your tensorflow version to 2.12.0 to match tflearn version 0.5.0

[@unknowntraveller7046]

@@shayp9428 let me check

[@tintinap]

Hi Tim, why it has to be 8 neurons and 2 hidden layers?

[@meryemamzil3519]

i have same question

[@meryemamzil3519]

i have same question

[@meryemamzil3519]

i have same question

[@TheChroniclez]

Hi @techwithtim . i got this problem where my training data intents.json only reads 1 data.. How do i solve this?i've already follow all the steps.

[@momeerrafaqat7243]

hey, everytime I run it my accuracy changes. and the max it went to is 0.4

[@nikhilambavaram6006]

Thats supposed to hapen

[@umamahfatima8571]

Hi, can anyone help me ? I am getting an error "Module tensorflow.contrib not found"

[@isaacmujuni5196]

You need to create a virtual environment. And then run tensor flow in that environment

[@umamahfatima8571]

I changed tensorflow's version from 2.0 to 1.14. Now it's working fine

[@altairlevega6613]

I received an error: ImportError: cannot import name 'is_sequence' from 'tensorflow.python.util.nest' What do I do?

[@gabrielkim4271]

tensorflow.reset_default_graph() i have a problem in this. curses is not supported on this machine (please install/reinstall curses for an optimal experience) Scipy not supported!

[@meesumraxa8618]

(from tensorflow.python.framework import ops) add this at the starting of code and change tensorflow.reset_default_graph() to ops.reset_default_graph()

[@josecarlosvegamoron6519]

module 'tensorflow' has no attribute 'scalar_summary' Please Help

[@jerrod7216]

Did you figure it out? I can't get past it

[@josecarlosvegamoron6519]

@@jerrod7216 tensorflow 1.15 Tflearn last versión Windows curses 1.0

[@jerrod7216]

Carlos Vega thank you but what is windows curses? I’m on Mac?

[@shubhankarkalekar4753]

AttributeError: module 'tensorflow' has no attribute 'reset_default_graph' help anyone plss

[@vichupb6223]

It is because you used a higher version of the Tensorflow , it could be fixed by replacing that line with -- - "tensorflow.compat.v1.reset_default_graph()"

[@venupunna8500]

hello Tim , Thanks for clear explaination , I cannot get the accuracy more than 23 % can you tell what might be the reason for this , I have tried by changing epochs and batch sizes but at most I get is 23

[@jenglong7826]

i get the same problem. Did u find any solutions?

[@CSEYOGESHWARANS]

bro are u active? i have one question about this proj

[@kaavijjj6575]

Thank you

[@priyanshitripathi1845]

Hey Tim it's giving me an error which says : tensorflow.reset_default_graph() AttributeError: module 'tensorflow' has no attribute 'reset_default_graph' Pls help me sort this out I'm stuck here for days now

[@elplaguister]

tensorflow.compat.v1.reset_default_graph()

[@retro9040]

i get a syntax error on net = tflearn.regression(net) any help? btw im using tensorflow 1.14 tflearn1.19 and python 3.7

[@prasheelchadotra7407]

Attribute errror list object has no attribute lower how could i solve it!

[@nikhilagarwal2249]

Hey Tim, why did you specifically take 45 neurons initially. I have been seeing many people use some particular amount of neurons at the first step. Is there a particular math calculation or any convention to decide these initial neurons? Please help me out here.

[@Robert-fp3rs]

@Nikhil Agarwal, i think in the video the number 45 is just an example, the number of neurons that you're talking about depends on the size or the length of ` training_data[0]`

[@nikhilagarwal2249]

@@Robert-fp3rs Thanks. That helped me. :)

[@sunnypatel137]

Hi @Tech With Tim I have use the same exact code as well as the json file, but still i am getting very less training accuracy(20%) where as you got accuracy somewhere around 90%! Can you point out what may have gone wrong with my model?

[@dance4technolife]

I got the same issue. I actually had 90+% of accuracy, and afterwards, with the same code when I added some small amounts to the JSON, accuracy went down to about 20%. Even with the old file. I don't know what the issue is all of a sudden, seems it went crazy on the same day!!!

[@krishshah8838]

doc=nltk.word_tokenize(doc) this line needs to be added before defining wrds in the second loop because to use stemmer first we need to tokenize words in every doc in doc_x,or elese wrds takes doc as string rather than list of words...it will rresolve the accuracy problem

[@asyrafzlkln1701]

@@krishshah8838 I already tried your solution. Still get the same accuracy 20% and loss is quite high as well. Any other way?

[@krishshah8838]

@@asyrafzlkln1701 it worked for me...ig I'll check n let you know

[@incongnitoh493]

w. lower() not work It show error list object has no attribute lower

[@SovonSingha]

why the hidden layers have only 8 neurons ? why it's not any other number ?

[@willjohnson4579]

probably a balance. I switched from 8 layer 8 layer 1,000 epoch 8 batch size way up to 16 layer 16 layer 10,000 epoch 16 batch size and averaged a 4% accuracy increase. Not worth the effort if you arent using GPU acceleration and going to 256x2 on 100,000 epoch and deploying it in a realistic scenario

[@nelsonbwino1524]

Hello Tim what an awesome video tutorial. Is there a way of implementing this model to a seq2seq rnn?

[@mayurmhamunkar6808]

can we check the accuracy and validation loss. how to do that can you please tell me?

[@tarirochagwiza1281]

Hie Time. I have this exact same code but when i run my main.py in cmd,I am getting this NameError: name 'wrds' is not defined on this line wrds = nltk.word_tokenize(wrds) ... where am I going wrong?

[@21fps50]

i am getting error of no module named "tensorflow.contrib"

[@tarirochagwiza1281]

@@21fps50 hi, I got that same error and its actually a tflearn error. See this page github.com/tensorflow/tensorflow/issues/30794 . so you need to downgrade your tensorflow to version 1.14.

[@pradyumankannan6802]

it should be wrds = nltk.word_tokenize(patterns) u r tokenizing wrds which is not yet defined

[@Alaska-mk4ok]

I am getting this error when I try to run the program: wrds = nltk.word_tokenize(pattern) NameError: name 'pattern' is not defined Can anyone help me?

[@TinyEarthling]

Try "patterns" instead

[@hakeemhairul2482]

This worked for me. type 'nltk.download('punkt')' under 'import nltk'

[@dawnbugay7094]

@@hakeemhairul2482 hi after that im experincing this error File "c:/Users/kishinki/Desktop/Python files/chatbot python/main.py", line 65, in net = tflearn.fully_connected(net, len(output[0]), activation="softmax") IndexError: index 0 is out of bounds for axis 0 with size 0 please help me

[@pythonocean7879]

tim will u plz make some interesting projects tuts plz.

[@deepakathirvel8023]

which library can be used instead of tf learn sir...

[@אופירברק-ע7ר]

Hi Tim, this is a great video I'm trying to make it. I have 17 types of tags so it shows me that the success rate is only 0.26%. So I added this line 4 more times - net = tflearn.fully_connected (net, 8) But it still stands at 0.26%...Do you have any idea what I can do?

[@justhangingout9666]

You want to increase the number of times you show the AI, that was the 1000 he talked about.

[@sainco3036]

Thanks.

[@johnstreety642]

would this chatbot work in google colab?

[@blazingwarlord6149]

Hey can someone tell me how to convert this tflearn code to keras code. Tflearn is not supported

[@Bruno-mi2rd]

Use keras.layers.Dense for a simple connected layer

[@ErgsYT]

It tells me tensorflow has no attribute reset_default_graph

[@csicee]

[Line 52] IndexError: list index out of range

[@adalbertoperezgarcia6137]

I have the same, how did you solve it?

[@csicee]

@@adalbertoperezgarcia6137 I didn't :/

[@shuvadipghosh374]

Please can you tell where can we find the source code to the chatbot

[@adityanambidi5129]

line 29, in words = [stemmer.stem(w.lower()) for w in words if w not in "?"] TypeError: stem() missing 1 required positional argument: 'word' What is wrong pls help

[@alalz]

bit late, but did you check that you had brackets on line 3? you should have: stemmer = LancasterStemmer()

[@runeeerdekens1256]

did i mess up? (im using pycharm) Traceback (most recent call last): File "main.py", line 6, in import tflearn ModuleNotFoundError: No module named 'tflearn'

[@pradyumankannan6802]

did u install tflearn

[@skullrazor]

pip install tflearn.. its probable that u installed tflearn outside of environment

[@balasaiallagadda2791]

Can I run this whole code in jupyter like does it work the same?

[@jdsguam]

My code crashed with: words =[stemmer.stem(w.lower())] for w in words if != in "?"] - red lines throughout the rest of the code.

[@M45t3rJ4ck]

should be: words = [stemmer.stem(w.lower()) for w in words if != "?"] drop the "]" after lower and the "in" before the ? hope it still helps

[@Old_SDC]

Be careful with Words = sorted(list(set(words))) On line 30. It’s lowercase not upper.

[@justhangingout9666]

Has this tutorial worked for you so far? I am having no problems until I type activate chatbot and it doesn't recognize the command activate. Any advice?

[@Old_SDC]

@@justhangingout9666 I’m pretty sure that originally when I followed the tutorial that it worked, but after a couple reloads it stopped working for me

[@bendrury1647]

line 61 or something but: net = tflearn.input_data(shape=[None, len(training[0]), 46]) to fix shape issue

[@justhangingout9666]

Has this tutorial worked for you so far? I am having no problems until I type activate chatbot and it doesn't recognize the command activate. Any advice?

[@bathob.motubudi]

Hi Tim, thank you so much for creating the series. I'm suing visual studio 2019 but I have created an environment for Python 3.6.8 and that is what I am using to follow your tutorial. The command model.save("model.tflearn") bombs out, I get an error that says ValueError: The passed save_path is not a valid checkpoint. Please assist in any way you can.

[@NoorJahanBK]

did you find a fix for this? im having the same issue

[@obaidullah214]

my acc is not going above 26%, is it because of differnt version of pip or something

[@Robert-fp3rs]

How do you save your model and talk to your chatbot anyway?

[@TechWithTim]

Keep watching

[@tsutsuihiroshi2949]

Do you have any reason why you use 8 neurons??? I'm completely new to this so I don't undestand... If you can explain more in detail on that, it will be great.

[@marcosiebecke7220]

I'm guessing it is an arbitrary choice based on number of inputs and outputs. Experiment is key for ML as far as I understood it. So try what happens when you have more topic and same number of neurons and/or more/less neurons and same number of inputs/outputs

[@krishanu7351]

u can simply use stopwords to remove unnecessary signs...

[@tristansun3274]

I spent 20 minutes trying to debug my code... I spelled n_epoch as n_epochs

[@sdolo222]

Hopefully you have a good IDE for debugging by now, I use Pycharm :)

[@xxMpEGxx]

Welcome to programming, this will happen multiple times in your life.

[@hardiktrikha1303]

In my work it comes inconsistent use of tabs and spaces in indentation

[@Bruno-mi2rd]

If you get this error you will have to erase everything and restart.

[@iamcarry]

if you still get that problem, just type Ctrl +shift + A then type "convert to spaces"

[@imoshikapalliyaguruge2822]

hey! Enjoyed your tutorial. Really appreciate your explanations. However, can you tell me how to fine tune responses of the bot?

[@abhijeetjain2098]

faltu...

[@harshvardhansharma2102]

output_row[labels.index(docs_y[x])]=1 can you please explain this ?

[@fakesweta9467]

same error can some one please help ValueError : 'greeting' is not in list

[@izmedmessiz]

@@fakesweta9467 check line 24, that should be docs_y.append(intent["tag"]).

[@franciszekjob8236]

same issue

[@aravindnallajerla4614]

Did anyone get error as following? ValueError: setting an array element with a sequence Help me fix it

[@flueepwrien6587]

there is problem with numpy dimensions, but I'm lookin for the answers

[@flueepwrien6587]

So after hour of trying to fix this issue, I finally fixed it! It is purely problem in your code. For me personaly, I misspelled words for wrds and the other way around. Check your code and ensure that it is exactly the same as in the video. Good luck.

[@flueepwrien6587]

Somehow I am getting accuracy of whopping 100% lmao

[@flueepwrien6587]

So after hour of trying to fix this issue, I finally fixed it! It is purely problem in your code. For me personaly, I misspelled words for wrds and the other way around. Check your code and ensure that it is exactly the same as in the video. Good luck.

[@israelaristide5506]

ValueError: ['hello'] is not in list im getting this error can anyone help??

[@lvlGhostlylvl]

Did you ever find a fix? For me it's greeting. The first item on the list, but when I check the debugger all tag's are listed in the labels.

[@sarthaknegi582]

I don't know why but somehow I am getting very low accuracy even after using the exact same data and same No. of hidden layers, Can anyone please help me?

[@justhangingout9666]

Has this tutorial worked for you so far? I am having no problems until I type activate chatbot and it doesn't recognize the command activate. Any advice?

[@saurabhchakravarti3133]

hey Guys, I am getting the below error. Can someone please help? File "C:\Users\---\---\---\---\---\--- ltk\stem\lancaster.py", line 290, in __getLastLetter for position in range(len(word)): TypeError: object of type 'int' has no len()

[@MananGandhi]

hi, i am getting a very low accuracy, around 20%. can someone help me

[@saif.warraich]

in case you are using the exactly same model as in video, you can try changing number of neurons in hidden layers , number of epoches or adding new layers,...

[@sebastiangutierrez9994]

Hello people i have a question Why did he choose 8 Neurons? I dont understand that. Thanks in advance for the answers

[@neologicalgamer3437]

It was mostly random. It's just a good number, not too high, and not too low. It suits this AI perfectly

[@mohmedbadr1947]

i got an accuracy of 0.214 what went wrong

[@MorDreadful]

much the same though Tensorflow and tflearn have alot of warnings. "You'll notice we had an accuracy of 99%". um. nope and changing 1000 to anything else makes no difference. I just wanna understand something for my final assignment report really so I can graduate. if it doesn't work for me great, as that also counts. :)

[@dance4technolife]

@@MorDreadful did you manage to change it? I got accuracy of 90+% at first and afterwards program became mad and I had to train it again and got low level of accuracy, and I didn't save the file of an old model

[@krishshah8838]

doc=nltk.word_tokenize(doc) this line needs to be added before defining wrds in the second loop because to use stemmer first we need to tokenize words in every doc in doc_x,or elese wrds takes doc as string rather than list of words...it will rresolve the accuracy problem

[@jenglong7826]

@@krishshah8838 by second loop do you mean "For x, doc in enumerate(docs_x):"?

[@jafarhasan4716]

Need tflearn code to be replaced by keras code