Update README.md
Browse files
README.md
CHANGED
@@ -40,6 +40,10 @@ https://docs.qq.com/doc/DQ1Vzd3VCTllFaXBv?_t=1685979317852&u=e7c714671e694797a04
|
|
40 |
|
41 |
如果觉得画面内容**不够丰富**,你可以尝试**细致地描述**,使画面更加**贴近你的想象**。
|
42 |
|
|
|
|
|
|
|
|
|
43 |
我一般在效果不符合预期时使用clip2。
|
44 |
|
45 |
**随你喜好使用lora!**
|
@@ -78,6 +82,10 @@ If the face appears to be falling apart, and you want to **improve the quality o
|
|
78 |
|
79 |
If you feel that the content of the picture is **not rich enough**, You can try to **describe in detail** to make the picture more **closely to your imagination.**
|
80 |
|
|
|
|
|
|
|
|
|
81 |
I usually use clip2 when the results don't meet expectations.
|
82 |
|
83 |
**Use lora as you like!**
|
|
|
40 |
|
41 |
如果觉得画面内容**不够丰富**,你可以尝试**细致地描述**,使画面更加**贴近你的想象**。
|
42 |
|
43 |
+
**提示词的权重以及顺序会影响它在画面里的重要程度。**
|
44 |
+
|
45 |
+
**如果有无法作出反应的提示词,请按以下顺序排查问题:同义词(同一概念的不同描述),提示词冲突(正面和负面),模型问题(看其他模型能否对同样的提示词作出反应),**embedding(我并没有使用它的习惯,但考虑到它的原理,我将它放上来作为参考)。
|
46 |
+
|
47 |
我一般在效果不符合预期时使用clip2。
|
48 |
|
49 |
**随你喜好使用lora!**
|
|
|
82 |
|
83 |
If you feel that the content of the picture is **not rich enough**, You can try to **describe in detail** to make the picture more **closely to your imagination.**
|
84 |
|
85 |
+
**The weight of the prompt word and the order in which it is used affects how important it is in the picture.**
|
86 |
+
|
87 |
+
**If there are prompt words that you cannot respond to, please rank the problems in the following order: synonyms (different descriptions of the same concept), prompt word conflicts (positive and negative), model problems (see if other models can respond to the same prompt words),** embedding (I am not in the habit of using it, but considering its rationale, I put it up as a reference).
|
88 |
+
|
89 |
I usually use clip2 when the results don't meet expectations.
|
90 |
|
91 |
**Use lora as you like!**
|