tacit/nanochat
1
0
Fork 0
mirror of https://github.com/karpathy/nanochat.git synced 2025-12-06 04:12:13 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

move the lr schedule args up in base_train so they are tunable in configurator

Browse Source
This commit is contained in:
Andrej Karpathy 2025-10-24 13:27:31 +00:00
parent cc3636b01c
commit 81597cd616
1 changed files with 3 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

7
scripts/base_train.py
View File

@ -49,6 +49,9 @@ unembedding_lr = 0.004 # learning rate for the unembedding parameters (Adam)
weight_decay = 0.0 # weight decay for the embedding/unembedding parameters (Adam)
matrix_lr = 0.02 # learning rate for the matrix parameters (Muon)
grad_clip = 1.0 # gradient clipping value (0.0 = disabled)
warmup_ratio = 0.0 # ratio of iterations for LR warmup
warmdown_ratio = 0.2 # ratio of iterations for LR warmdown
final_lr_frac = 0.0 # final LR is this fraction of the initial LR
# Evaluation
eval_every = 250 # every how many steps to evaluate the model for val bpb
eval_tokens = 20*524288 # number of tokens to evaluate val loss on
@ -151,10 +154,6 @@ x, y = next(train_loader) # kick off load of the very first batch of data
# Set up hyperparameter schedulers
# Learning rate scheduler
# TODO: experiment with a short warmup for the AdamW params (expecting slight improvement)
warmup_ratio = 0.0 # ratio of iterations for LR warmup
warmdown_ratio = 0.2 # ratio of iterations for LR warmdown
final_lr_frac = 0.0 # final LR is this fraction of the initial LR
def get_lr_multiplier(it):
warmup_iters = round(warmup_ratio * num_iterations)
warmdown_iters = round(warmdown_ratio * num_iterations)