Open kaustavha opened 6 years ago
@kaustavha Thanks for the feedback. Really appreciated.
Keep in mind that this is an open-source project, and you're welcome to make pull-requests.
I've fixed the typo in the requirements.txt
file.
I've been looking into trying to find a workaround for calling the constant model_accuracy()
function.
I didn't really think about calling that function later, and just decided to disable it after the competition ended. You can see that in line 317
in the contracts/Danku_demo.sol
contract.
The forward_pass2()
is still public, and should be able to take the network definition with weights, and a single data point (as defined in the input layer). The problem with this is that there's no trivial way to get the submitted models directly, as we would need to decode them in python. (they're public, but we don't have any proper getters for them in populus as far as I can see)
To understand which functions can't be called, just look at the assert()
call at the start of every function. If it has the following, the function won't run:
assert(contract_terminated == false);
I'll look into what I can do, but I'm also open to suggestings too.
Thanks for the highlighting info! :)
Are those for geth
? If so, I didn't want to include them because it's a seperate project, and it's installation instructions may change at anytime.
The linux system deps were for solc I believe, we used infura's API instead of setting up our own geth instance.
Hi there! Really interesting project. We unfortunately couldnt spend enough time on this before the submission period expired :( We also used google collaboratory and infura, which made the setup a lot easier.
Are there any plans on releasing a similar contract on a test-net for users to practice against?
Also had some requests:
Could you please fix the requirements.txt file? The new geopandas req needs an extra = sign:
Could we please get the python code or instructions on what function calls on the contract are still valid and which ones expire when? We noticed the
get_train_data
call didnt work at one point but works again now (maybe mistake on our end), andmodel_accuracy
doesnt work now:I'm guessing on recent commits here that only the
get_datapoint_pred
function would need to be changed to reference the local stashed file instead right?You can add solidity highlighting to show up on github by adding a
.gitattributes
file in the root with*.sol linguist-language=Solidity
. That would make reading the danku.sol contract much easier for people encountering it just now.On new linux instances (as in the case of using Googles collaboratory) we need a few more pre-requisites:
cc @ClairHu