search

louisoutin / rat_crypto_trader

Relation-Aware Transformer for Portfolio Policy Learning using Binance provider
20 stars 5 forks source link

launch_test #3

Open absolutProgrammer opened 2 years ago

absolutProgrammer commented 2 years ago

works completely incorrect. It has old usage of class DataMatrices and after changing it to a normal one (I have taken it from lanch_train and changed to date_start=ctx["test_range"]["start"] and date_end=ctx["test_range"]["end"]).

Now I got an error in rat/network/encoder_decoder.py file:

Traceback (most recent call last): File "/rat_crypto_trader_master/rat/main.py", line 202, in run_main() File "/rat_crypto_trader_master/rat/main.py", line 66, in run_main launch_test(ctx) File "/rat_crypto_trader_master/rat/main.py", line 181, in launch_test tst_portfolio_value, SR, CR, St_v, tst_pc_array, TO = test_net(DM, x_window_size, local_context_length, model, File "/rat_crypto_trader_master/rat/test.py", line 15, in test_net tst_long_term_w, tst_trg_y = test_online(DM, x_window_size, model, File "/rat_crypto_trader_master/rat/test.py", line 61, in test_online out = model.forward(tst_src, tst_currt_price, tst_previous_w, File "/rat_crypto_trader_master/rat/network/encoder_decoder.py", line 114, in forward out = torch.cat([decode_out, previous_w], 2) # [128,11,212] cat [128,11,1] -> [128,11,212+1] RuntimeError: torch.cat(): Sizes of tensors must match except in dimension 2. Got 1 and 0 in dimension 0 (The offending index is 1)

louisoutin commented 2 years ago

6 months I didn't touch that repo, I'm super busy I'll not have time to help you. However from what I remember, the code was running fine since this last commit.