load_args_from_checkpoint fails if arguments are missing

[Mittagskogel]

Megatron-DeepSpeed/megatron/checkpointing.py

Lines 472 to 479 in 61d5d61

 # Args. 
 if not state_dict: 
     print_rank_0('Checkpoint not found to provide arguments, using provided arguments.') 
     return args 
  
 if 'args' not in state_dict: 
     print_rank_0('Checkpoint provided does not have arguments saved, using provided arguments.') 
     return args 

This function should always return two values, see default return in line 529. The function is not behaving as described in the warning messages.

Down the line, this causes a crash when trying to unpack two values in

Megatron-DeepSpeed/tools/checkpoint_loader_megatron.py

Line 58 in 61d5d61

 margs, checkpoint_args = load_args_from_checkpoint(margs) 

[s1gh]

Hello, have you resolved this issue? @Mittagskogel