-
|
I'm now slowly accumulating game records. |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment 2 replies
-
|
For me, 100Mil samples (not games) is enough to train a BERT model. |
Beta Was this translation helpful? Give feedback.
-
|
Thanks. Then how many Hanchans are approximately equivalent to 100M samples? |
Beta Was this translation helpful? Give feedback.
-
|
On average, you can extract just under 150 samples per Hanchan. Therefore, by simple calculation, it would take about 7,000 Hanchan to gather 1M samples. However, since the samples extracted from a single Hanchan are highly correlated with each other, it would be better to collect samples from as many Hanchans as possible and then randomly sample the training examples from them. |
Beta Was this translation helpful? Give feedback.
On average, you can extract just under 150 samples per Hanchan. Therefore, by simple calculation, it would take about 7,000 Hanchan to gather 1M samples. However, since the samples extracted from a single Hanchan are highly correlated with each other, it would be better to collect samples from as many Hanchans as possible and then randomly sample the training examples from them.