I am unable to use my bf16 lora with flux-dev even though I am disabling go_fast parameter in the ui , and for the api call I gave it "false". still no change. why cant I do it? It gave me this error "Prediction failed. cannot access local variable 'weight_is_f8' where it is not associated with a value". here are the error logs .....
"Downloaded weights in 5.51s
2025-04-24 00:05:06.210 | INFO | fp8.lora_loading:convert_lora_weights:502 - Loading LoRA weights for /src/weights-cache/7d2104666de2bfa9
Warning - loading loras that fine-tune the text encoder is not supported at present, text encoder weights will be ignored
2025-04-24 00:05:06.711 | INFO | fp8.lora_loading:convert_lora_weights:523 - LoRA weights loaded
2025-04-24 00:05:06.712 | DEBUG | fp8.lora_loading:apply_lora_to_model_and_optionally_store_clones:610 - Extracting keys
2025-04-24 00:05:06.712 | DEBUG | fp8.lora_loading:apply_lora_to_model_and_optionally_store_clones:617 - Keys extracted
Applying LoRA: 0it [00:00, ?it/s]
Applying LoRA: 0it [00:00, ?it/s]
Traceback (most recent call last):
File "/root/.pyenv/versions/3.11.11/lib/python3.11/site-packages/cog/server/worker.py", line 352, in _predict
result = predict(**payload)
^^^^^^^^^^^^^^^^^^
File "/src/predict.py", line 539, in predict
model.handle_loras(lora_weights, lora_scale)
File "/src/bfl_predictor.py", line 108, in handle_loras
load_lora(model, lora_path, lora_scale, self.store_clones)
File "/root/.pyenv/versions/3.11.11/lib/python3.11/site-packages/torch/utils/_contextlib.py", line 116, in decorate_context
return func(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^
File "/src/fp8/lora_loading.py", line 545, in load_lora
apply_lora_to_model_and_optionally_store_clones(model, lora_weights, lora_scale, store_clones)
File "/root/.pyenv/versions/3.11.11/lib/python3.11/site-packages/torch/utils/_contextlib.py", line 116, in decorate_context
return func(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^
File "/src/fp8/lora_loading.py", line 668, in apply_lora_to_model_and_optionally_store_clones
if weight_is_f8:
^^^^^^^^^^^^
UnboundLocalError: cannot access local variable 'weight_is_f8' where it is not associated with a value"
I am unable to use my bf16 lora with flux-dev even though I am disabling go_fast parameter in the ui , and for the api call I gave it "false". still no change. why cant I do it? It gave me this error "Prediction failed. cannot access local variable 'weight_is_f8' where it is not associated with a value". here are the error logs .....
"Downloaded weights in 5.51s
2025-04-24 00:05:06.210 | INFO | fp8.lora_loading:convert_lora_weights:502 - Loading LoRA weights for /src/weights-cache/7d2104666de2bfa9
Warning - loading loras that fine-tune the text encoder is not supported at present, text encoder weights will be ignored
2025-04-24 00:05:06.711 | INFO | fp8.lora_loading:convert_lora_weights:523 - LoRA weights loaded
2025-04-24 00:05:06.712 | DEBUG | fp8.lora_loading:apply_lora_to_model_and_optionally_store_clones:610 - Extracting keys
2025-04-24 00:05:06.712 | DEBUG | fp8.lora_loading:apply_lora_to_model_and_optionally_store_clones:617 - Keys extracted
Applying LoRA: 0it [00:00, ?it/s]
Applying LoRA: 0it [00:00, ?it/s]
Traceback (most recent call last):
File "/root/.pyenv/versions/3.11.11/lib/python3.11/site-packages/cog/server/worker.py", line 352, in _predict
result = predict(**payload)
^^^^^^^^^^^^^^^^^^
File "/src/predict.py", line 539, in predict
model.handle_loras(lora_weights, lora_scale)
File "/src/bfl_predictor.py", line 108, in handle_loras
load_lora(model, lora_path, lora_scale, self.store_clones)
File "/root/.pyenv/versions/3.11.11/lib/python3.11/site-packages/torch/utils/_contextlib.py", line 116, in decorate_context
return func(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^
File "/src/fp8/lora_loading.py", line 545, in load_lora
apply_lora_to_model_and_optionally_store_clones(model, lora_weights, lora_scale, store_clones)
File "/root/.pyenv/versions/3.11.11/lib/python3.11/site-packages/torch/utils/_contextlib.py", line 116, in decorate_context
return func(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^
File "/src/fp8/lora_loading.py", line 668, in apply_lora_to_model_and_optionally_store_clones
if weight_is_f8:
^^^^^^^^^^^^
UnboundLocalError: cannot access local variable 'weight_is_f8' where it is not associated with a value"