z1311 / Character-Level-Text-Generation

Character Level Text Generation with RNN, LSTM and GRU. Implemented using PyTorch and trained with Marvel Cinematic Universe(MCU) Dialogue Corpus.

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Character Level Text Generation

Character Level Text Generation with RNN, LSTM and GRU. The language models are trained with Marvel Cinematic Universe(MCU) Dialogue Corpus dataset.

Generate Sample Text

Flags

-m or --modelname: model name from the list [rnn, lstm, gru] (required)
-i or --initialstr: initial string. Default: Avengers Assemble (optional)
-n or --charlen: output characters length. Default: 200 (optional)

$ python3 sample.py -m gru
$ python3 sample.py -m gru -i "Infinity Stones" -n 300

Sample Output

Small Suggestion: Don't try to make sense out of the generated sample outputs.

AVENGERS ASSEMBLE --
SSLYS) shifting, over here!
They're coming for them.
There's seen in your six.
That she sisters and get me some fight for the government
is gonna be the best case scenario.
You let them go, Nat.
We're getting out of here.
As much as I hate the weapons.
And then he comes behind the energy denied.

I can do this all day --
and she seemed the best and burn, boy!
Shouldn't tell me about the car.
When you're all the bells, it looks.
You didn't do you see your bag corpee to control it?
And that's the others are being future of your letten.
This is what we were thanky."
I'm gonna miss you. It's the one.

I'm INEVITABLE! --
I was kind of hoping when I'm allowed to tell you where you're gonna like the Infinity Stone that could drown in the ocean
needs to get information on the detention of wrended in New York side.
I can get you better. That's right.
How does the FBI changen is the only last different personnel things...
and the light destruction every sequence a volunteer something I won't be.

Some Indian(Tollywood) Famous Dialogues

I believe in war not in morality --
than the monster in two minutes,
I think we learn that the lab back for years to come.
There's the dessers back in plenty of time.
You're already gonna make the lead, 80 heads existed.
Oh, what's the most lies.
Anyway, she really need my eyes existed permon diles for your loss!

Don't trouble the trouble, if you trouble the trouble, trouble troubles you.I am not the trouble. I am the truth. --
and listed the exact moment you have to do to close my men.
And he was an angel...
very pudding discipting 3,000 feet,
she's dead.
What the hell?
Not be a good man too.
You busy out!
No, this must have heard about the code.
But I'm really long down good aim you should download a VPN

Convergence Graph

RNN, LSTM and GRU trained over 500 epochs and their losses. loss_graph

About

Character Level Text Generation with RNN, LSTM and GRU. Implemented using PyTorch and trained with Marvel Cinematic Universe(MCU) Dialogue Corpus.


Languages

Language:Jupyter Notebook 97.9%Language:Python 2.1%