* Fixed the LoRA handling bug in [`py/libs/xyplot.py:365-395`](py/libs/xyplot.py:365).
The original code used an if/else pattern that only processed one LoRA at a time:
```python
xy_values = x_value if self.x_type == "Lora" else y_value
```
This caused issues when both X and Y axes contained LoRAs - only the X axis LoRA was processed, and when only Y had LoRAs, the latent array wasn't populated correctly.
The fix now:
1. Creates an empty `lora_stack` list
2. Adds the X axis LoRA to the stack if `self.x_type == "Lora"`
3. Adds the Y axis LoRA to the stack if `self.y_type == "Lora"`
4. Appends any existing `plot_image_vars['lora_stack']` to the combined stack
5. Applies all LoRAs in sequence
This ensures both X and Y LoRAs are properly combined and applied when both axes contain LoRA values, fixing the `IndexError` in `rearrange_tensors()` that occurred due to mismatched latent array dimensions.
* Fixed: get_labels_and_sample loop structure (lines 588-623)
The IndexError persisted because the latent array length didn't match the expected grid dimensions. The issue is in the nested loop structure of get_labels_and_sample()
When only Y-axis has LoRA values (X is "None"), the nested loops don't generate the correct number of latents:
This results in 0 latents instead of len(y_values) latents, causing the IndexError in rearrange_tensors()
Replaced the nested loop structure with three explicit cases:
X-only variation (self.y_type == 'None'): Iterates over X values only
Y-only variation (self.x_type == 'None'): Iterates over Y values only
Both X and Y variation: Nested iteration over both axes
This ensures the correct number of latents are generated for all scenarios, fixing the IndexError in rearrange_tensors() that occurred when only Y-axis had values (like LoRAs).
* Fix for the LoRA label generation logic in [`py/libs/xyplot.py`](py/libs/xyplot.py:63-69). The changes made to the `define_variable()` method:
1. **Reduced model name truncation** from 30 to 25 characters to leave room for weight information
2. **Changed weight format** from `(0.50)` to ` w:0.50` for better visibility
3. **Only show weight when it differs from default** (1.0) - this ensures weight is displayed for non-default values
4. **Added bounds check** for `len(arr) > 3` before accessing `arr[3]` to prevent potential IndexError
Now when using the same LoRA with different weights (e.g., LoRA A at 0.5 and 1.0), each variation will have a distinct label in the axis, making it clear which weight is being applied in each column/row of the XY plot.
* Add submodule
* Set submodule branch to main
* Add PromptAwait node
* Fix loop has started for the second time, but the prompt word has not output new content
* Add input_1
* Change select widget to toolbar
* Modify some field names and displays
* Change select max-width
* Add output random seed in promptAwait
* Add lora weight to XY title axis. Trim lora desc.
* Add weights to lora names in xyPlots. Re-add because original lost in git merge mistake.
* Only add common label, if it's not an axis type.
* Remove bad comment.
---------
Co-authored-by: Mike Kinney <mike.kinney@valorepartners.com>
1. wildcard-replaced prompt will be returned in order rather than randomly
2. will return the amount of probilities and amount of probilities each option or wildcard
3. the prompt can be selected by offset argument, the order of probilities is fixed
4. even if the offset exceeds the total, it will not stop, but will always return to the last probility, requiring additional nodes to control.
ISO-8859-1 encoding can forcibly read any byte (it maps each byte directly to its corresponding character). This means it won't throw any decoding errors, but it will incorrectly interpret UTF-8 encoded Chinese characters as other characters, resulting in garbled text (mojibake).
ISO-8859-1编码可以强制读取任何字节（它会把每个字节都映射到对应的字符）
这意味着它不会抛出解码错误，但会把UTF-8编码的中文字符错误解释为其他字符
导致中文显示为乱码
- Update CLIP vision model URLs for IPAdapter and DynamiCrafter
- Improve error handling for model downloads with clearer error messages
- Add changelog entry for v1.2.8
kannkyo