mw66
mw66
This worked for me: change: config/train_shakespeare_char.py ``` gradient_accumulation_steps = 2 ``` then: ``` CUDA_VISIBLE_DEVICES=0,1 TORCHDYNAMO_DISABLE=1 TORCHDYNAMO_SUPPRESS_ERRORS=True \ torchrun --nproc_per_node gpu \ train.py config/train_shakespeare_char.py ```
try: https://github.com/karpathy/nanoGPT/issues/289#issuecomment-1726236977
Thanks. My question is more about what values can be used as special markers (without changing the original input text), e.g. when I prepare data, can I inject `-1` as...
> We might be able to get multiple tokens at a time with `idx_next = torch.multinomial(probs, num_samples=1)` being set to 2 or 5, etc, gotta test that. > This will...
Based on #435 it's possible to place the order, have you tried? BTW, is it possible to get the extended hour stock price? And what are the 43 stock list?
is it possible to get the extended hour stock price? Any code example? Thanks
[Parallel(n_jobs=-1)]: Using backend MultiprocessingBackend with 8 concurrent workers. hangs.
I tried: $ docker run -it qgc-linux-docker sh then: ``` $ ls -l /project total 4 drwxr-xr-x 2 root root 4096 Sep 12 00:49 build $ whoami user ``` so...
I build as root in that image, add `-u 0` ``` $ docker run -u 0 --rm -v ${PWD}:/project/source -v ${PWD}/build:/project/build qgc-linux-docker ``` seems working so far.
I think the real fix should be `chown user /project/build`