🚀🚀 「大模型」2小时完全从0训练64M的小参数GPT！🌏 Train a 64M-parameter GPT from scratch in just 2h!
COMMITS
April 4, 2026
J
[update] readme
jingyaogong committed
April 2, 2026
J
[update] open causal
jingyaogong committed
J
[update] readme
jingyaogong committed
April 1, 2026
J
[update] readme
jingyaogong committed
March 31, 2026
J
[update] default model
jingyaogong committed
March 27, 2026
J
[update] random seed
jingyaogong committed
J
[update] fp16 inference
jingyaogong committed
March 26, 2026
J
[update] image
jingyaogong committed
J
[update] change default seq_len
jingyaogong committed
March 24, 2026
J
[update] minimind-3
jingyaogong committed
J
Merge pull request #698 from readlnh/master
jingyaogong committed
R
[fix] align log/save last-step check and ETA with 1-indexed step
readlnh committed
R
[fix] gradient accumulation step alignment
readlnh committed
February 6, 2026
J
[update] empty_think_ratio
jingyaogong committed
February 5, 2026
J
[update] empty_think_ratio
jingyaogong committed
J
[feat] data process
jingyaogong committed
January 30, 2026
J
[update] save interval
jingyaogong committed
J
[update] safe half
jingyaogong committed
January 18, 2026
J
[fix] data skip
jingyaogong committed
J
[update] shuffle data
jingyaogong committed
January 17, 2026
J
[fix] max length
jingyaogong committed
J
[update] pretrain load
jingyaogong committed
January 15, 2026
J
[update] align mask
jingyaogong committed
January 14, 2026
J
[update] align loss
jingyaogong committed
J
[fix] compile unpack
jingyaogong committed
J
[feat] add compile
jingyaogong committed
January 13, 2026
J
[update] prompt prefill
jingyaogong committed
January 7, 2026
J
[update] show speed
jingyaogong committed
J
[update] params log
jingyaogong committed
J
[update] mask log
jingyaogong committed