The failed case in my hand: the resultant shard files were small in size, 1MB each more or less (chunk size were set at 64,64,64). Doing the same without --translate works fine (where the source were a traditional Precomputed with 0 voxel_offset).
Since the sharded format doesn't intrinsically use voxel_offset to my understanding, one strategy to consider for fixing could be just do the transfer with 0 offset and then just patch the info file? - This doubles as an existing workaround for this issue.
The failed case in my hand: the resultant shard files were small in size, 1MB each more or less (chunk size were set at 64,64,64). Doing the same without
--translateworks fine (where the source were a traditional Precomputed with 0 voxel_offset).Since the sharded format doesn't intrinsically use voxel_offset to my understanding, one strategy to consider for fixing could be just do the transfer with 0 offset and then just patch the info file? - This doubles as an existing workaround for this issue.