wasi_nn_tensorflowlite.cpp: fix get_output return size

[yamt]

it should be byte size, not the number of (fp32) values.

i'm ambivalent about how to deal with the compatibility for the legacy wamr-specific "wasi_nn". for now, i avoided changing it. (so that existing tests using the legacy abi, namely test_tensorflow.c and test_tensorflow_quantized.c, pass as they are.) if we have any users who still want to use the legacy abi, i suppose they consider the compatibility is more important than the consistency with other backends.

cf. #4376

[yamt]

restrarting SGX 143
https://github.com/bytecodealliance/wasm-micro-runtime/actions/runs/15746783829/job/44384528925?pr=4390

[lum1n0us]

Not familiar with the LLM model, it seems there are assumptions about the format under different tensor->quantization.type. For me, it would be better to add comments to make it easier for readers to follow.

[lum1n0us]

if bh_memcpy_s() actually write tensor->bytes bytes, L403 should be *output_tensor_size = tensor->bytes;?

[yamt]

this ifdef block is for compatibility. (for the legacy "wasi_nn", it returns the number of floats)

[yamt]

i added a few comments

[lum1n0us]

if output_tensor_size is the number of fp32, L394 would be something like if (*output_tensor_size < tensor->bytes/sizeof(float)) when WASM_ENABLE_WASI_EPHEMERAL_NN == 0 ?

[yamt]

right. fixed. thank you.

[yamt]

Not familiar with the LLM model, it seems there are assumptions about the format under different tensor->quantization.type.

i'm not quite happy with these assumptions because it makes us incompatible with wasmedge.

For me, it would be better to add comments to make it easier for readers to follow.

totally agree.