Hi,
I encountered an issue in the docs, with the wrong parameter name being included in docs/source/common_options.rst lines 523-528. Using version 0.10.4. This caused the follwing error for me:
usage: lora_combiner.py [-h] --model_path MODEL_PATH --lora_path LORA_PATH
--output_path OUTPUT_PATH [--is_rm]
[--ds.param_dtype {bf16,fp16}]
lora_combiner.py: error: unrecognized arguments: --param_dtype bf16
Is:
python -m openrlhf.cli.lora_combiner
--model_path meta-llama/Meta-Llama-3-8B
--lora_path ./checkpoint/llama3-8b-rm
--output_path ./checkpoint/llama-3-8b-rm-combined
--is_rm
--param_dtype bf16
Should be:
python -m openrlhf.cli.lora_combiner
--model_path meta-llama/Meta-Llama-3-8B
--lora_path ./checkpoint/llama3-8b-rm
--output_path ./checkpoint/llama-3-8b-rm-combined
--is_rm
--ds.param_dtype bf16
Hi,
I encountered an issue in the docs, with the wrong parameter name being included in docs/source/common_options.rst lines 523-528. Using version 0.10.4. This caused the follwing error for me:
Is:
python -m openrlhf.cli.lora_combiner
--model_path meta-llama/Meta-Llama-3-8B
--lora_path ./checkpoint/llama3-8b-rm
--output_path ./checkpoint/llama-3-8b-rm-combined
--is_rm
--param_dtype bf16
Should be:
python -m openrlhf.cli.lora_combiner
--model_path meta-llama/Meta-Llama-3-8B
--lora_path ./checkpoint/llama3-8b-rm
--output_path ./checkpoint/llama-3-8b-rm-combined
--is_rm
--ds.param_dtype bf16