ML-Intern-lab
/

nanochat

Model card Files Files and versions

nanochat / scripts

127 kB

Ctrl+K

Ctrl+K

67 contributors

History: 143 commits

Andrej Karpathy

clean up fragile code

92d63d4 4 months ago

base_eval.py

10.3 kB
nuke huggingface and its giant dependency footprint from orbit. datasets still todo 4 months ago
base_train.py

32.3 kB
fix deprecated commit on torchao 4 months ago
chat_cli.py

3.92 kB
delete autocast, an unnecessary thorn in my side, manage dtypes directly 8 months ago
chat_eval.py

11.6 kB
delete spurious 'cute' bloat that doesn't matter to a serious research stack 4 months ago
chat_rl.py

16.9 kB
delete the whole report thing i think it was a bad idea it just bloats the code. doing this subtracts 540 LOC 4 months ago
chat_sft.py

24.7 kB
delete spurious 'cute' bloat that doesn't matter to a serious research stack 4 months ago
infer_bench.py

12.7 kB
clean up fragile code 4 months ago
tok_eval.py

11.1 kB
rename fwe to climbmix 4 months ago
tok_train.py

3.73 kB
fix token_bytes calculation to use raw token bytes instead of round-tripping through str, which corrupts tokens that are not valid standalone UTF-8 (e.g. bytes >= 0x80 became 3-byte U+FFFD). 193/32768 tokens were affected, so bpb of new runs will read ~0.05% higher than old runs. Thank you @ellenjxu #693 4 months ago