Flux nodes work fine in workflow from localhost but causes error when same workflow used in API

### Expected Behavior

Image created

### Actual Behavior

Error:

model weight dtype torch.float8_e4m3fn, manual cast: torch.float16
model_type FLOW
Using xformers attention in VAE
Using xformers attention in VAE
Requested to load FluxClipModel_
Loading 1 new model
Requested to load Flux
Loading 1 new model
100%|██████████████████████████████████████████████████████████████████████████████| 6/6 [01:32<00:00, 15.39s/it]
Using xformers attention in VAE
Using xformers attention in VAE
Requested to load AutoencoderKL
Loading 1 new model
!!! Exception during processing!!! Given groups=1, weight of size [4, 4, 1, 1], expected input[1, 16, 160, 90] to have 4 channels, but got 16 channels instead
Traceback (most recent call last):
File "C:\inetpub\wwwroot\ComfyUI\ComfyUI\execution.py", line 152, in recursive_execute
output_data, output_ui = get_output_data(obj, input_data_all)
File "C:\inetpub\wwwroot\ComfyUI\ComfyUI\execution.py", line 82, in get_output_data
return_values = map_node_over_list(obj, input_data_all, obj.FUNCTION, allow_interrupt=True)
File "C:\inetpub\wwwroot\ComfyUI\ComfyUI\execution.py", line 75, in map_node_over_list
results.append(getattr(obj, func)(**slice_dict(input_data_all, i)))
File "C:\inetpub\wwwroot\ComfyUI\ComfyUI\nodes.py", line 270, in decode
return (vae.decode(samples["samples"]), )
File "C:\inetpub\wwwroot\ComfyUI\ComfyUI\comfy\sd.py", line 322, in decode
pixel_samples[x:x+batch_number] = self.process_output(self.first_stage_model.decode(samples).to(self.output_device).float())
File "C:\inetpub\wwwroot\ComfyUI\ComfyUI\comfy\ldm\models\autoencoder.py", line 199, in decode
dec = self.post_quant_conv(z)
File "C:\inetpub\wwwroot\ComfyUI\python_embeded\lib\site-packages\torch\nn\modules\module.py", line 1553, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
File "C:\inetpub\wwwroot\ComfyUI\python_embeded\lib\site-packages\torch\nn\modules\module.py", line 1562, in _call_impl
return forward_call(*args, **kwargs)
File "C:\inetpub\wwwroot\ComfyUI\ComfyUI\comfy\ops.py", line 93, in forward
return super().forward(*args, **kwargs)
File "C:\inetpub\wwwroot\ComfyUI\python_embeded\lib\site-packages\torch\nn\modules\conv.py", line 458, in forward
return self._conv_forward(input, self.weight, self.bias)
File "C:\inetpub\wwwroot\ComfyUI\python_embeded\lib\site-packages\torch\nn\modules\conv.py", line 454, in _conv_forward
return F.conv2d(input, weight, bias, self.stride,
RuntimeError: Given groups=1, weight of size [4, 4, 1, 1], expected input[1, 16, 160, 90] to have 4 channels, but got 16 channels instead

Prompt executed in 194.95 seconds

### Steps to Reproduce

Run workflow...
[Flux full checkpoint API test.json](https://github.com/user-attachments/files/16606291/Flux.full.checkpoint.API.test.json)

### Debug Logs

```powershell
got prompt
[rgthree] Using rgthree's optimized recursive execution.
model weight dtype torch.float8_e4m3fn, manual cast: torch.float16
model_type FLOW
Using xformers attention in VAE
Using xformers attention in VAE
Requested to load FluxClipModel_
Loading 1 new model
Requested to load FluxClipModel_
Loading 1 new model
clip missing: ['text_projection.weight']
Requested to load Flux
Loading 1 new model
100%|██████████████████████████████████████████████████████████████████████████████| 6/6 [01:37<00:00, 16.30s/it]
Using xformers attention in VAE
Using xformers attention in VAE
Requested to load AutoencodingEngine
Loading 1 new model
Prompt executed in 255.51 seconds
got prompt
[rgthree] Using rgthree's optimized recursive execution.
100%|██████████████████████████████████████████████████████████████████████████████| 6/6 [01:37<00:00, 16.31s/it]
Requested to load AutoencodingEngine
Loading 1 new model
Prompt executed in 106.29 seconds
got prompt
[rgthree] Using rgthree's optimized recursive execution.
model weight dtype torch.float8_e4m3fn, manual cast: torch.float16
model_type FLOW
Using xformers attention in VAE
Using xformers attention in VAE
Requested to load FluxClipModel_
Loading 1 new model
Requested to load Flux
Loading 1 new model
100%|██████████████████████████████████████████████████████████████████████████████| 6/6 [01:32<00:00, 15.39s/it]
Using xformers attention in VAE
Using xformers attention in VAE
Requested to load AutoencoderKL
Loading 1 new model
!!! Exception during processing!!! Given groups=1, weight of size [4, 4, 1, 1], expected input[1, 16, 160, 90] to have 4 channels, but got 16 channels instead
Traceback (most recent call last):
File "C:\inetpub\wwwroot\ComfyUI\ComfyUI\execution.py", line 152, in recursive_execute
output_data, output_ui = get_output_data(obj, input_data_all)
File "C:\inetpub\wwwroot\ComfyUI\ComfyUI\execution.py", line 82, in get_output_data
return_values = map_node_over_list(obj, input_data_all, obj.FUNCTION, allow_interrupt=True)
File "C:\inetpub\wwwroot\ComfyUI\ComfyUI\execution.py", line 75, in map_node_over_list
results.append(getattr(obj, func)(**slice_dict(input_data_all, i)))
File "C:\inetpub\wwwroot\ComfyUI\ComfyUI\nodes.py", line 270, in decode
return (vae.decode(samples["samples"]), )
File "C:\inetpub\wwwroot\ComfyUI\ComfyUI\comfy\sd.py", line 322, in decode
pixel_samples[x:x+batch_number] = self.process_output(self.first_stage_model.decode(samples).to(self.output_device).float())
File "C:\inetpub\wwwroot\ComfyUI\ComfyUI\comfy\ldm\models\autoencoder.py", line 199, in decode
dec = self.post_quant_conv(z)
File "C:\inetpub\wwwroot\ComfyUI\python_embeded\lib\site-packages\torch\nn\modules\module.py", line 1553, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
File "C:\inetpub\wwwroot\ComfyUI\python_embeded\lib\site-packages\torch\nn\modules\module.py", line 1562, in _call_impl
return forward_call(*args, **kwargs)
File "C:\inetpub\wwwroot\ComfyUI\ComfyUI\comfy\ops.py", line 93, in forward
return super().forward(*args, **kwargs)
File "C:\inetpub\wwwroot\ComfyUI\python_embeded\lib\site-packages\torch\nn\modules\conv.py", line 458, in forward
return self._conv_forward(input, self.weight, self.bias)
File "C:\inetpub\wwwroot\ComfyUI\python_embeded\lib\site-packages\torch\nn\modules\conv.py", line 454, in _conv_forward
return F.conv2d(input, weight, bias, self.stride,
RuntimeError: Given groups=1, weight of size [4, 4, 1, 1], expected input[1, 16, 160, 90] to have 4 channels, but got 16 channels instead

Prompt executed in 194.95 seconds
```

### Other

To confirm, I can run this workflow myself in the web interface without any issues. But when I copy the workflow to an API I get the above error. Some how the vae does not appear to be matching I believe.. 16 channels instead of 4.. I don't understand how this works in the web interface but not through the API.