updated hypr for my gpu
This commit is contained in:
14
train.py
14
train.py
@@ -10,16 +10,16 @@ import data
|
|||||||
import log
|
import log
|
||||||
|
|
||||||
hypr = {
|
hypr = {
|
||||||
"embed_size": 256,
|
"embed_size": 512,
|
||||||
"n_heads": 4,
|
"n_heads": 8,
|
||||||
"n_blocks": 4,
|
"n_blocks": 6,
|
||||||
"block_size": 256,
|
"block_size": 256,
|
||||||
"batch_size": 16,
|
"batch_size": 16,
|
||||||
"starting_lr": 3e-4,
|
"starting_lr": 6e-4,
|
||||||
"minimum_lr": 3e-5,
|
"minimum_lr": 6e-5,
|
||||||
"warmup": 1_000,
|
"warmup": 1_000,
|
||||||
"steps": 5_000,
|
"steps": 20_000,
|
||||||
"encoding": "TinyLlama/TinyLlama-1.1B-Chat-v1.0",
|
"encoding": "gpt2",
|
||||||
"dataset": "HuggingFaceTB/smollm-corpus",
|
"dataset": "HuggingFaceTB/smollm-corpus",
|
||||||
"subset": "cosmopedia-v2",
|
"subset": "cosmopedia-v2",
|
||||||
}
|
}
|
||||||
|
|||||||
Reference in New Issue
Block a user