evaluate epoch 1 model checkpoints loss and perplexity on 1% samples from the FineWeb-edu test set

Update README.md

make gpt2_data_loader DDP

add the plan for epoch 1 end of epoch research

update ignore

add cover to readme

disabling the torch.backends.cudnn.deterministic doesn't give a 2~3x speed boost

Delete ckpt_gpt2_epoch_1_500.pt

Delete ckpt_gpt2_epoch_1_0.pt

reorganize some logs

Update training_gpt2-FineWeb-smoke-test.log

fix a time stamp issue

Create cursor_gpt2_training_performance_analys.md

successfully trained the first epoch

epoch 1 first 12k batch success