Abylay Ospan 50b236fbcc Add PyTorch and CUDA memory profiling systems ... Capture PyTorch execution traces and CUDA memory snapshots. Traces display detailed CPU and CUDA activity, including individual CUDA kernel calls. CUDA memory snapshots visualize all memory allocations, helping diagnose CUDA out-of-memory errors, investigate memory leaks, or understand GPU memory usage for educational purposes. Enable profiling with the --enable_profiling=True flag in speedrun.sh. See PROFILING.md for documentation and example visualizations.		2025-10-18 12:50:19 +00:00
..
base_eval.py	initial commit	2025-10-13 06:49:24 -07:00
base_loss.py	initial commit	2025-10-13 06:49:24 -07:00
base_train.py	Add PyTorch and CUDA memory profiling systems	2025-10-18 12:50:19 +00:00
chat_cli.py	initial commit	2025-10-13 06:49:24 -07:00
chat_eval.py	initial commit	2025-10-13 06:49:24 -07:00
chat_rl.py	initial commit	2025-10-13 06:49:24 -07:00
chat_sft.py	dont evaluate the sampling evals during SFT they are too slow. keep the multiple choice evals. delete unused imports	2025-10-15 16:42:23 +00:00
chat_web.py	also allow regenerating assistant message by clicking it, and make sure to feed good seed to generate	2025-10-16 01:28:37 +00:00
mid_train.py	fix bug in learning rate multiplier, it was ramping up instead of ramping down. see more in Issue #68. also add --dry_run option useful for experimentation	2025-10-15 16:35:04 +00:00
tok_eval.py	initial commit	2025-10-13 06:49:24 -07:00
tok_train.py	initial commit	2025-10-13 06:49:24 -07:00