Skip to content

A LLM trained only on data from certain time periods to reduce modern bias

Python

Add files via upload

Hayk Grigorianc46dc1f40609
diff --git a/train_london.py b/train_london.py
new file mode 100644
index 0000000..805646d
--- /dev/null
+++ b/train_london.py
@@ -0,0 +1,27 @@
+out_dir = 'out_london'
+eval_interval = 100
+eval_iters = 20
+log_interval = 10
+
+always_save_checkpoint = False
+
+dataset = 'gpt_data_london'
+gradient_accumulation_steps = 2
+batch_size = 4
+block_size = 128
+
+n_layer = 4
+n_head = 4
+n_embd = 256
+dropout = 0.0
+
+learning_rate = 1e-3
+max_iters = 2000
+lr_decay_iters = 2000
+min_lr = 1e-4
+
+warmup_iters = 100
+device = 'cpu' # Change to 'cuda' if you have GPU, or 'mps' for Mac
+
+tokenizer_name = 'tokenizer_london'
+vocab_size = None