Composite encoding with Grid

[diolatzis]

Composite encoding with Grid

[Tom94]

Hi Stavros, that's strange -- it should work. I quickly tried your config in our radiance caching code and it works fine on my machine.

For reference, here's the config we used in the paper and which I also confirmed working on latest tiny-cuda-nn:

{
    "loss": {
        "otype": "RelativeL2Luminance"
    },
    "optimizer": {
        "otype": "Ema",
        "decay": 0.99,
        "sensitivity": 0.1,
        "nested": {
            "otype": "ExponentialDecay",
            "decay_start": 4000,
            "decay_interval": 4000,
            "decay_base": 0.33,
            "nested": {
                "otype": "Adam",
                "learning_rate": 1e-2,
                "beta1": 0.9,
                "beta2": 0.99,
                "epsilon": 1e-15,
                "l2_reg": 1e-6
            }
        }
    },
    "encoding": {
        "otype": "Composite",
        "nested": [
            {
                "n_dims_to_encode": 3, // Position
                "otype": "HashGrid",
                "per_level_scale": 2.0,
                "log2_hashmap_size": 15,
                "base_resolution": 16,
                "n_levels": 16
            },
            {
                "n_dims_to_encode": 5, // Interesting conditionals
                "otype": "OneBlob",
                "n_bins": 4
            },
            {
                "n_dims_to_encode": 6, // Linear conditionals that should be identity encoded
                "otype": "Identity"
            }
        ]
    },
    "network": {
        "otype": "FullyFusedMLP",
        "activation": "ReLU",
        "output_activation": "None",
        "n_neurons": 64,
        "n_hidden_layers": 2
    }
}

To troubleshoot further:

• could you let me know which call is throwing the exception?
• are you using the config to instantiate a NetworkWithInputEncoding and are there any implementation details you can divulge about how the network/encoding is called?
• there might be something wrong with the input / output buffer sizes and the illegal memory access -- being undefined behavior -- only surfaces with some code paths.

Cheers!

[Tom94]

Closing on the assumption that the problem is solved by now. Please feel free to re-open and provide extra details if not.