generate dummy model FIM

make_fim_model.py

import gguf

def decode_field(field: gguf.ReaderField):
    if field and field.types:
        main_type = field.types[0]

        if main_type == gguf.GGUFValueType.ARRAY:
            sub_type = field.types[-1]

            if sub_type == gguf.GGUFValueType.STRING:
                return [str(bytes(field.parts[idx]), encoding='utf-8') for idx in field.data]
            else:
                return [pv for idx in field.data for pv in field.parts[idx].tolist()]
        if main_type == gguf.GGUFValueType.STRING:
            return str(bytes(field.parts[-1]), encoding='utf-8')
        else:
            return field.parts[-1][0]

    return None

reader = gguf.GGUFReader('stories260K.gguf', 'r')

print(reader)

writer = gguf.GGUFWriter('stories260K-infill.gguf', arch='llama')
for field in reader.fields.values():
    if field.name == 'general.architecture': continue
    writer.add_key_value(field.name, decode_field(field), field.types[0])

writer.add_uint32("tokenizer.ggml.fim_pre_token_id", 500)
writer.add_uint32("tokenizer.ggml.fim_suf_token_id", 501)
writer.add_uint32("tokenizer.ggml.fim_mid_token_id", 502)
writer.add_uint32("tokenizer.ggml.fim_pad_token_id", 503)
writer.add_uint32("tokenizer.ggml.fim_rep_token_id", 504)
writer.add_uint32("tokenizer.ggml.fim_sep_token_id", 505)

for tensor in reader.tensors:
    writer.add_tensor_info(tensor.name, tensor.data.shape, tensor.data.dtype, tensor.data.nbytes, tensor.tensor_type)

writer.write_header_to_file()
writer.write_kv_data_to_file()
writer.write_ti_data_to_file()

for tensor in reader.tensors:
    writer.write_tensor_data(tensor.data)