if object query is random, as shown in code, how to evaluate to get a steady result?

[RedBlack888]

if object query is random, as shown in code, how to evaluate to get a steady result when inference?
self.query_embed = nn.Embedding(num_queries, hidden_dim)

hs = self.transformer(self.input_proj(src), mask, self.query_embed.weight, pos[-1])[0]

Best.

[AjibolaPy]

It's learnable. I think you'll use the learnt weight on inference