Chunking quantized model leads to unequal Chunks

[nighting0le01]

Chunking quantized model leads to unequal Chunks, say we have a ~153 MB model, it's gettting chunked to 153 and 2 kb ,

prog = _load_prog_from_mlmodel(model)

# Compute the incision point by bisecting the program based on weights size
op_idx, first_chunk_weights_size, total_weights_size = _get_op_idx_split_location(
    prog)
print(f"First  chunk size = {first_chunk_weights_size:.2f} MB") # 152.67 MB
print(f"Second chunk size = {total_weights_size - first_chunk_weights_size:.2f} MB") #0.42 MB
print(index=587/2720)
prog_chunk1 = _make_first_chunk_prog(f"index={op_idx}/{len(main_block.operations)") # 587/3000
prog_chunk2 = _make_second_chunk_prog(_load_prog_from_mlmodel(model), op_idx)

how can i chunk model with constant nodes (like in quantization). (might have trouble processing quantized consts)

[aseemw]

Chunking script is now being refactored to use coremltools.models.utils.bisect_model() API (#354)

Can you please try again with this coremltools API (coremltools==8.0b2)? And if the issue persists, can you please open an issue in coremltools Github, with the code to reproduce.

[nighting0le01]

@aseemw thanks i have opened it here:apple/coremltools#2320