file_loader.cc expects to find converted weights files corresponding to falcon-7b model structure, but they don't exist in falcon-40b:

Done converting the weights...
Loading weight file word_embeddings_weight
Loading weight file layers_0_input_layernorm_weight
Could not open file: /tmp/mymodel/half-precision/layers_0_input_layernorm_weight
python: /tmp/pip-install-toyyxtv8/flexflow_209f7d4f52cb40178d7cc442e662a167/inference/file_loader.cc:347: void
load_from_file(DT*, size_t, std::string) [with DT = __half; size_t = long unsigned int; std::string = std::__cxx11::basic_string]: Assertion `in.good() && "incorrect weight file path"' failed.

Falcon-40b has no such weights:
https://huggingface.co/tiiuae/falcon-40b/blob/main/pytorch_model.bin.index.json
whereas Falcon-7b has them:
https://huggingface.co/tiiuae/falcon-7b/blob/main/pytorch_model.bin.index.json

So file_loader seems to expect both models to follow falcon-7b structure but falcon-40b is different.