search

abudesai / timeVAE

TimeVAE implementation in keras/tensorflow
MIT License
111 stars 23 forks source link

value error #5

Open uliuycy opened 5 months ago

uliuycy commented 5 months ago

ValueError: Layer encoder weight shape (3, 5, 200) is not compatible with provided weight shape (120, 200). Why is there such an error when using VAE dense

abudesai commented 5 months ago

Hi, I fixed the issue in the code. Please do git pull and re-run the script.

You can reach out to me again if you have any questions or still have issues running the code.

Best, Abu Desai

On Mon, Apr 15, 2024 at 2:21 AM liuchengcheng @.***> wrote:

ValueError: Layer encoder weight shape (3, 5, 200) is not compatible with provided weight shape (120, 200). Why is there such an error when using VAE dense

— Reply to this email directly, view it on GitHub https://github.com/abudesai/timeVAE/issues/5, or unsubscribe https://github.com/notifications/unsubscribe-auth/AHXIAM2JLHJAV2WP27YD5Q3Y5OLZHAVCNFSM6AAAAABGG6POBGVHI2DSMVQWIX3LMV43ASLTON2WKOZSGI2DGMJVGM3TIMA . You are receiving this because you are subscribed to this thread.Message ID: @.***>

uliuycy commented 5 months ago

Thanks for your reply.If I want to stack stocks data for generation, but it only has two dimensions, how can I obtain N, T, and D

abudesai commented 4 months ago

How many unique stocks (tickers) do you have? What is the value of T - how many time steps?

Normally, the synthetic data generation papers take a single series and generate windows from it.
For example, suppose you have a single series with 3 features and 1000 time steps in length. If you set window size to 100, then you get 1000 - 100 + 1 = 901 separate samples from the same series. Then your data will be of size [901, 100, 2]. Here N = 901, T= 100, and D = 2.