NicolaLancellotti commented on a change in pull request #8795: URL: https://github.com/apache/tvm/pull/8795#discussion_r696663975
########## File path: python/tvm/relay/backend/contrib/ethosu/vela_api.py ########## @@ -0,0 +1,313 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you under the Apache License, Version 2.0 (the +# "License"); you may not use this file except in compliance +# with the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, +# software distributed under the License is distributed on an +# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +# KIND, either express or implied. See the License for the +# specific language governing permissions and limitations +# under the License. +""" +This is an adapter module for conversions between TVM and Vela. +The following conversion APIs are added : + *Obtaining the best block config + *Compressing weights + *Packing biases +""" +import logging +import math +import numpy as np +from ethosu.vela import api as vapi + +from tvm.relay.backend.contrib.ethosu import util + +# pylint: disable=invalid-name +logger = logging.getLogger("Ethos-U") + +VELA_TO_NP_DTYPES = { + vapi.NpuDataType.UINT8: np.uint8, + vapi.NpuDataType.UINT16: np.uint16, + vapi.NpuDataType.INT8: np.int8, + vapi.NpuDataType.INT16: np.int16, + vapi.NpuDataType.INT32: np.int32, +} + +SCALE_BIAS_LENGTH = 10 + + +def get_optimal_block_config(npu_op, accel_type): + """ + "The NPU's unit of work is known as a block. It will fetch block(s) from Input + Feature Map (IFM) and a compute block for Output Feature Map (OFM). + Therefore, we need to pick an optimal block configuration considering bandwidth + to bring IFM blocks and the number of OFM block computes need to happen + to cover the OFM as indicated by the npu op. + + Parameters + ---------- + npu_op : ethosu.vela.api.NpuOperation + The NPU operation and its params + accel_type : ethosu.vela.api.NpuAccelerator + The NPU accelerator variant + Returns Review comment: ```suggestion Returns ``` There has to be a blank line before Returns in order for the doc to be built correctly. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: commits-unsubscr...@tvm.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org
