The weights of the tokens don't seem to be applied

[Against-the-Wind]

In other_impls.py:
def token_weights(string, current_weight):
a = parse_parentheses(string)
out = []
for x in a:
weight = current_weight
if len(x) >= 2 and x[-1] == ")" and x[0] == "(":
x = x[1:-1]
xx = x.rfind(":")
weight *= 1.1
if xx > 0:
try:
weight = float(x[xx + 1 :])
x = x[:xx]
except:
pass
out += token_weights(x, weight)
else:
out += [(x, current_weight)]
return out
the weights of the tokens within the parentheses were mutiplied by 1.1.
But going through the whole repository, I don't seem to find where the weights were applied, reasons as follows:
In other_impls.py:
class ClipTokenWeightEncoder:
def encode_token_weights(self, token_weight_pairs):
tokens = list(map(lambda a: a[0], token_weight_pairs[0]))
out, pooled = self([tokens])
if pooled is not None:
first_pooled = pooled[0:1].cpu()
else:
first_pooled = pooled
output = [out[0:1]]
return torch.cat(output, dim=-2).cpu(), first_pooled
tokens = list(map(lambda a: a[0], token_weight_pairs[0])) means only passing the tokens to the variable tokens
And in sd3_infer.py:
def get_cond(self, prompt):
self.print("Encode prompt...")
tokens = self.tokenizer.tokenize_with_weights(prompt)
l_out, l_pooled = self.clip_l.model.encode_token_weights(tokens["l"])
g_out, g_pooled = self.clip_g.model.encode_token_weights(tokens["g"])
t5_out, t5_pooled = self.t5xxl.model.encode_token_weights(tokens["t5xxl"])
lg_out = torch.cat([l_out, g_out], dim=-1)
lg_out = torch.nn.functional.pad(lg_out, (0, 4096 - lg_out.shape[-1]))
return torch.cat([lg_out, t5_out], dim=-2), torch.cat(
(l_pooled, g_pooled), dim=-1
)
although using the function encode_token_weights, as mentioned before, the weights weren't used in the encoding process

[Against-the-Wind]

I did experiments with weights ranging from 0.0 to 10.0, and as I expected, there wasn't any significant change in the generated images. This further proves that the weights weren't used when generating images.