Garrett Pete
Fine tuning tutorial: https://www.youtube.com/watch?v=4iK-IuvatxI
Github: https://github.com/openai/gpt-2
Make sure to install python3, pip, and git using either apt-get or brew install.
Hope you learn some useful new stuff!
Garrett Pete
Fine tuning tutorial: https://www.youtube.com/watch?v=4iK-IuvatxI
Github: https://github.com/openai/gpt-2
Make sure to install python3, pip, and git using either apt-get or brew install.
Hope you learn some useful new stuff!
Comments are closed.
What specs are your laptop? I read your article and want to know how much of my resources I should allocate to this. Should I run it through a virtual machine or my home environment? Thanks
Hey, thanks for your video. Please could I ask for some advice. Im not a programmer but am really interested in ML. I've met all requirements but cant seem to reconfigure the generate_unconditional_samples.py to run the big model. Im not sure what Im doing wrong, I changed the default model to 1558M in the definitions and the path but I dont know the correct syntax
I would be interested if you could leave your email, or any kind of contact, so I can get in touch with you.
I'd like to hire you to do it for me, and tweak it a bit, for a purpose I have in my mind.
Thanks in advance
does it just write random stories or can you get it to rewrite passages?
Coincidentally I am at that same point with the same numpy issue once I pasted "python3 src/interactive_conditional_samples.py –top_k 40"
I did notice that when you posted this vid, you showed 3 python models. Now there are 4 models. The last one is a 1.558m file. Might that be the issue?
I have a question on this. If you just run the model like the way you do in the video does it develop in any way? Is it training input to input at all or just starting over each time?
Help
if i throw
pip install tqdm == 4.31.1
Result: Requirement already satisfied: tqdm == 4.31.1 in c: programdata miniconda3 lib site-packages (4.31.1)
but if I throw
python3 download_model.py 124M
result: from tqdm import tqdm
ModuleNotFoundError: No module named 'tqdm'
Hi, I jokingly decided to use GTP-2 and followed the instructions on this tutorial but I need help, on the last step where I put the sudo command for interactive conditional samples it keeps ending with an AttributeError: "module 'TensorFlow' has no attribute 'sort' ", I've installed the non GPU tensor flow through the command and don't know what I need to do to fix this, I'm using a old weak laptop with a 4th gen i3 if that's any help
Hi, I basically just followed your video and don't know anything about programing but after installing the sudo python3 download_model.py 124M command, I got stuck trying to run this command:
sudo python3 src/interactive_conditional_samples.py –top_k 40
The following appeared:
Traceback (most recent call last):
File "src/interactive_conditional_samples.py", line 6, in <module>
import numpy as np
ModuleNotFoundError: No module named 'numpy'
Would you know what I am supposed to do ?
UPDATE:
I have updated the TensorFlow and now this appears:
MacBook-Pro-de-Elie:gpt-2 elie$ sudo python3 src/interactive_conditional_samples.py –top_k 40
Password:
Traceback (most recent call last):
File "src/interactive_conditional_samples.py", line 7, in <module>
import tensorflow as tf
File "/Library/Frameworks/Python.framework/Versions/3.8/lib/python3.8/site-packages/tensorflow/__init__.py", line 41, in <module>
from tensorflow.python.tools import module_util as _module_util
File "/Library/Frameworks/Python.framework/Versions/3.8/lib/python3.8/site-packages/tensorflow/python/__init__.py", line 50, in <module>
from tensorflow.python import pywrap_tensorflow
ImportError: cannot import name 'pywrap_tensorflow' from partially initialized module 'tensorflow.python' (most likely due to a circular import) (/Library/Frameworks/Python.framework/Versions/3.8/lib/python3.8/site-packages/tensorflow/python/__init__.py)
I am Still stuck!!
soodoo
after running this command 'sudo python3 src/interactive_conditional_samples.py –top_k 40' I've got the following issue: Traceback (most recent call last):
File "src/interactive_conditional_samples.py", line 6, in <module>
import numpy as np
ModuleNotFoundError: No module named 'numpy'
muhammadsalimov@MUHAMMADs-MacBook-Pro gpt-2 %
What version of python do you have?
Why would you run git clone as sudo???
I got it to work with python 3.7 + tensorflow 1.15.2
Any ideas as to how to bring that generated text to a react front end?
Traceback (most recent call last):
File "src/interactive_conditional_samples.py", line 3, in <module>
import fire
ModuleNotFoundError: No module named 'fire'
What the heck
Thanks for this tutorial, Garrett. But like many others here, I am stuck with the "numpy" issue. I believe your tutorial is the most recent one on YT so I was hoping you could look into it. I wanna run GPT-2 so badly…..!!!
For anyone who is having pipeline issues, make sure you have Python 3.7 – 64 bit (not 32) installed and are using tensorflow 1.15. Tensorflow 2.x (anything after 2.0) does not include the same modules. Similarly, this will not work with the newest version of the 64 bit architecture – Python (3.9).
thanks
Does this work on Linux?
I got permission error is that because of my anaconda?
Successfully installed fire-0.3.1 regex-2017.4.5 requests-2.21.0 termcolor-1.1.0 tqdm-4.31.1 urllib3-1.24.3
(base) Dodjis-MacBook-Pro:gpt-2 dodjicohovi$ sudo python3 download_model.py 124M
Fetching checkpoint: 1.00kit [00:00, 479kit/s]
Fetching encoder.json: 1.04Mit [00:00, 6.96Mit/s]
Fetching hparams.json: 1.00kit [00:00, 501kit/s]
Fetching model.ckpt.data-00000-of-00001: 498Mit [00:21, 22.9Mit/s]
Fetching model.ckpt.index: 6.00kit [00:00, 2.79Mit/s]
Fetching model.ckpt.meta: 472kit [00:00, 5.09Mit/s]
Fetching vocab.bpe: 457kit [00:00, 3.69Mit/s]
(base) Dodjis-MacBook-Pro:gpt-2 dodjicohovi$ python3 download_model.py 355M
Traceback (most recent call last):
File "download_model.py", line 14, in <module>
os.makedirs(subdir)
File "/Users/dodjicohovi/anaconda3/lib/python3.7/os.py", line 221, in makedirs
mkdir(name, mode)
PermissionError: [Errno 13] Permission denied: 'models/355M'
I believe the version of the tf is not compatible. I don't know what I am doing
Traceback (most recent call last):
File "src/interactive_conditional_samples.py", line 7, in <module>
import tensorflow as tf
ModuleNotFoundError: No module named 'tensorflow'
(base) Dodjis-MacBook-Pro:gpt-2 dodjicohovi$ pip3 install tensorflow==1.12.0
ERROR: Could not find a version that satisfies the requirement tensorflow==1.12.0 (from versions: 1.13.0rc1, 1.13.0rc2, 1.13.1, 1.13.2, 1.14.0rc0, 1.14.0rc1, 1.14.0, 1.15.0rc0, 1.15.0rc1, 1.15.0rc2, 1.15.0rc3, 1.15.0, 1.15.2, 1.15.3, 2.0.0a0, 2.0.0b0, 2.0.0b1, 2.0.0rc0, 2.0.0rc1, 2.0.0rc2, 2.0.0, 2.0.1, 2.0.2, 2.1.0rc0, 2.1.0rc1, 2.1.0rc2, 2.1.0, 2.1.1, 2.2.0rc0, 2.2.0rc1, 2.2.0rc2, 2.2.0rc3, 2.2.0rc4, 2.2.0, 2.3.0rc0, 2.3.0rc1, 2.3.0rc2)
ERROR: No matching distribution found for tensorflow==1.12.0
(base) Dodjis-MacBook-Pro:gpt-2 dodjicohovi$ pip3 install tensorflow-gpu==1.12.0
ERROR: Could not find a version that satisfies the requirement tensorflow-gpu==1.12.0 (from versions: none)
ERROR: No matching distribution found for tensorflow-gpu==1.12.0
(base) Dodjis-MacBook-Pro:gpt-2 dodjicohovi$ sudo python3 src/interactive_conditional_samples.py –top_k 40
Traceback (most recent call last):
File "src/interactive_conditional_samples.py", line 9, in <module>
import model, sample, encoder
File "/Users/dodjicohovi/Desktop/AI-GPT-2/gpt-2/src/model.py", line 3, in <module>
from tensorflow.contrib.training import HParams
ModuleNotFoundError: No module named 'tensorflow.contrib'
(base) Dodjis-MacBook-Pro:gpt-2 dodjicohovi$
I enjoyed your detailed instructions etc now I am just curious to know when you are going to give instructions for GPT-3
lol sudo ! you don't get it dude ..
is this tutorial for windows 10 or another operativ system? if this is for beginners i should not look at the pro lessons 🙂
Hi Can I train my own model using 1558M or 775m data instead of using 117M data? and if it is possible, how can Ido that? thank u!
how do I install tensorflow?