"attempt to index upvalue 't2c' (a nil value)"

[stalagmite7]

I'm trying to convert a torch model, but it breaks with this error:

th torch2caffe/torch2caffe.lua --input "/root/crnn/model/crnn_demo/crnn_demo_model.t7"  64 1 32 100
	Opts: %s	{
	  input_dims = {
		64,
		1,
		32,
		100
	  },
	  format = "lua",
	  input = "/root/crnn/model/crnn_demo/crnn_demo_model.t7",
	  caffemodel = "",
	  preprocessing = "",
	  prototxt = "",
	  verify = "",
	  input_tensor = ""
	}	
	Parsed opts: %s	{
	  input_dims = {
		64,
		1,
		32,
		100
	  },
	  inputs = {
		{
		  name = "data",
		  input_dims = {
			64,
			1,
			32,
			100
		  }
		}
	  },
	  input = "/root/crnn/model/crnn_demo/crnn_demo_model.t7",
	  format = "lua",
	  preprocessing = "",
	  verify = "",
	  prototxt = "",
	  caffemodel = ""
	}	
	Running with model: 	
	{
	  parameters : 
		{
		  1 : CudaTensor - size: 64x1x3x3
		  2 : CudaTensor - size: 64
		  3 : CudaTensor - size: 128x64x3x3
		  4 : CudaTensor - size: 128
		  5 : CudaTensor - size: 256x128x3x3
		  6 : CudaTensor - size: 256
		  7 : CudaTensor - size: 256
		  8 : CudaTensor - size: 256
		  9 : CudaTensor - size: 256x256x3x3
		  10 : CudaTensor - size: 256
		  11 : CudaTensor - size: 512x256x3x3
		  12 : CudaTensor - size: 512
		  13 : CudaTensor - size: 512
		  14 : CudaTensor - size: 512
		  15 : CudaTensor - size: 512x512x3x3
		  16 : CudaTensor - size: 512
		  17 : CudaTensor - size: 512x512x2x2
		  18 : CudaTensor - size: 512
		  19 : CudaTensor - size: 512
		  20 : CudaTensor - size: 512
		  21 : CudaTensor - size: 1024x512
		  22 : CudaTensor - size: 1024
		  23 : CudaTensor - size: 1024x256
		  24 : CudaTensor - size: 1024
		  25 : CudaTensor - size: 1024x512
		  26 : CudaTensor - size: 1024
		  27 : CudaTensor - size: 1024x256
		  28 : CudaTensor - size: 1024
		  29 : CudaTensor - size: 256x256
		  30 : CudaTensor - size: 256
		  31 : CudaTensor - size: 256x256
		  32 : CudaTensor - size: 256
		  33 : CudaTensor - size: 1024x256
		  34 : CudaTensor - size: 1024
		  35 : CudaTensor - size: 1024x256
		  36 : CudaTensor - size: 1024
		  37 : CudaTensor - size: 1024x256
		  38 : CudaTensor - size: 1024
		  39 : CudaTensor - size: 1024x256
		  40 : CudaTensor - size: 1024
		  41 : CudaTensor - size: 37x256
		  42 : CudaTensor - size: 37
		  43 : CudaTensor - size: 37x256
		  44 : CudaTensor - size: 37
		}
	  bnVars : 
		{
		  1 : CudaTensor - size: 256
		  2 : CudaTensor - size: 256
		  3 : CudaTensor - size: 512
		  4 : CudaTensor - size: 512
		  5 : CudaTensor - size: 512
		  6 : CudaTensor - size: 512
		}
	}
	/root/torch/install/bin/luajit: ./torch2caffe/lib.lua:158: attempt to index upvalue 't2c' (a nil value)
	stack traceback:
		./torch2caffe/lib.lua:158: in function 'convert'
		./torch2caffe/lib.lua:168: in function 'main'
		torch2caffe/torch2caffe.lua:23: in main chunk
		[C]: in function 'dofile'
		/root/torch/install/lib/luarocks/rocks/trepl/scm-1/bin/th:145: in main chunk
		[C]: at 0x00406670

Seems like in torch2caffe/lib.lua, the lib_py.py that is added as a requirement in line 19 is a nil value in function M.convert(), which I don't understand, because if this is a scope issue, isn't declaring it as a local variable at the beginning of the function going to ensure that it is seen in the rest of the body?