Click here to see the table of contents.
Note that this README is automatically generated - don't edit!
- Category: AI/ML optimization.
- CM GitHub repository: mlcommons@ck
- GitHub directory for this script: GitHub
- CM meta description for this script: _cm.json
- CM "database" tags to find this script: qaic,compile,model,model-compile,qaic-compile
- Output cached? True
cm pull repo mlcommons@ck
-
cm run script --tags=qaic,compile,model,model-compile,qaic-compile[,variations] [--input_flags]
-
cmr "qaic compile model model-compile qaic-compile[ variations]" [--input_flags]
Click here to expand this section.
import cmind
r = cmind.access({'action':'run'
'automation':'script',
'tags':'qaic,compile,model,model-compile,qaic-compile'
'out':'con',
...
(other input keys for this script)
...
})
if r['return']>0:
print (r['error'])
cmr "cm gui" --script="qaic,compile,model,model-compile,qaic-compile"
Use this online GUI to generate CM CMD.
cm docker script "qaic compile model model-compile qaic-compile[ variations]" [--input_flags]
-
No group (any variation can be selected)
Click here to expand this section.
_bert-99
- Environment variables:
- CM_COMPILE_BERT:
on
- CM_QAIC_MODEL_TO_CONVERT:
calibrate_bert_mlperf
- CM_QAIC_MODEL_COMPILER_PARAMS_BASE:
-aic-hw -aic-hw-version=2.0 -execute-nodes-in-fp16=Add,Div,Erf,Softmax -quantization-schema=symmetric_with_uint8 -quantization-precision=Int8 -quantization-precision-bias=Int32 -vvv -compile-only -onnx-define-symbol=batch_size,1 -onnx-define-symbol=seg_length,384 -multicast-weights -combine-inputs=false -combine-outputs=false
- CM_QAIC_MODEL_COMPILER_ARGS: ``
- CM_COMPILE_BERT:
- Workflow:
- Read "deps" on other CM scripts
- calibrate,qaic,_bert-99
- CM names:
--adr.['bert-profile', 'qaic-profile']...
- CM script: calibrate-model-for.qaic
- CM names:
- calibrate,qaic,_bert-99
- Read "deps" on other CM scripts
- Environment variables:
_bert-99,offline
- Environment variables:
- CM_QAIC_MODEL_COMPILER_ARGS:
-allocator-dealloc-delay=2 -size-split-granularity=1536 -vtcm-working-set-limit-ratio=1
- CM_QAIC_MODEL_COMPILER_ARGS_SUT:
-aic-num-cores=1 -mos=1 -ols=2
- CM_QAIC_MODEL_COMPILER_ARGS:
- Workflow:
- Environment variables:
_bert-99,offline,nsp.14
- Environment variables:
- CM_QAIC_MODEL_COMPILER_ARGS_SUT:
-aic-num-cores=1 -mos=1 -ols=3
- CM_QAIC_MODEL_COMPILER_ARGS_SUT:
- Workflow:
- Environment variables:
_bert-99,offline,nsp.16
- Environment variables:
- CM_QAIC_MODEL_COMPILER_ARGS_SUT:
-aic-num-cores=1 -mos=1 -ols=2
- CM_QAIC_MODEL_COMPILER_ARGS_SUT:
- Workflow:
- Environment variables:
_bert-99,server
- Environment variables:
- CM_QAIC_MODEL_COMPILER_ARGS:
-allocator-dealloc-delay=2 -size-split-granularity=1536 -vtcm-working-set-limit-ratio=1
- CM_QAIC_MODEL_COMPILER_ARGS_SUT:
-aic-num-cores=1 -mos=1 -ols=3
- CM_QAIC_MODEL_COMPILER_ARGS:
- Workflow:
- Environment variables:
_bert-99,server,nsp.14
- Environment variables:
- CM_QAIC_MODEL_COMPILER_ARGS_SUT:
-aic-num-cores=1 -mos=1 -ols=3
- CM_QAIC_MODEL_COMPILER_ARGS_SUT:
- Workflow:
- Environment variables:
_bert-99,singlestream
- Environment variables:
- CM_QAIC_MODEL_COMPILER_ARGS: ``
- CM_QAIC_MODEL_COMPILER_ARGS_SUT:
-aic-num-cores=8 -mos=8 -ols=1
- Workflow:
- Environment variables:
_bert-99,singlestream,nsp.14
- Environment variables:
- CM_QAIC_MODEL_COMPILER_ARGS_SUT:
-aic-num-cores=8 -mos=8 -ols=1
- CM_QAIC_MODEL_COMPILER_ARGS_SUT:
- Workflow:
- Environment variables:
_bert-99.9
- Environment variables:
- CM_COMPILE_BERT:
on
- CM_QAIC_MODEL_TO_CONVERT:
bert_mlperf
- CM_QAIC_MODEL_COMPILER_PARAMS_BASE:
-aic-hw -aic-hw-version=2.0 -convert-to-fp16 -vvv -compile-only -onnx-define-symbol=batch_size,1 -onnx-define-symbol=seg_length,384 -combine-inputs=false -combine-outputs=false
- CM_QAIC_MODEL_COMPILER_ARGS: ``
- CM_COMPILE_BERT:
- Workflow:
- Environment variables:
_bert-99.9,offline
- Environment variables:
- CM_QAIC_MODEL_COMPILER_ARGS_SUT:
-aic-num-cores=2 -mos=1 -ols=2
- CM_QAIC_MODEL_COMPILER_ARGS_SUT:
- Workflow:
- Environment variables:
_bert-99.9,offline,nsp.14
- Environment variables:
- CM_QAIC_MODEL_COMPILER_ARGS_SUT:
-aic-num-cores=2 -mos=1 -ols=2
- CM_QAIC_MODEL_COMPILER_ARGS_SUT:
- Workflow:
- Environment variables:
_bert-99.9,offline,nsp.16
- Environment variables:
- CM_QAIC_MODEL_COMPILER_ARGS_SUT:
-aic-num-cores=2 -mos=1 -ols=2
- CM_QAIC_MODEL_COMPILER_ARGS_SUT:
- Workflow:
- Environment variables:
_bert-99.9,server
- Environment variables:
- CM_QAIC_MODEL_COMPILER_ARGS_SUT:
-aic-num-cores=4
- CM_QAIC_MODEL_COMPILER_ARGS_SUT:
- Workflow:
- Environment variables:
_bert-99.9,server,nsp.14
- Environment variables:
- CM_QAIC_MODEL_COMPILER_ARGS_SUT:
-aic-num-cores=4
- CM_QAIC_MODEL_COMPILER_ARGS_SUT:
- Workflow:
- Environment variables:
_resnet50
- Environment variables:
- CM_COMPILE_RESNET:
on
- CM_QAIC_MODEL_TO_CONVERT:
compile_resnet50_tf
- CM_QAIC_MODEL_COMPILER_PARAMS_BASE:
-aic-hw -aic-hw-version=2.0 -quantization-schema=symmetric_with_uint8 -quantization-precision=Int8 -output-node-name=ArgMax -vvv -compile-only -use-producer-dma=1
- CM_COMPILE_RESNET:
- Workflow:
- Environment variables:
_resnet50,multistream
- Environment variables:
- CM_QAIC_MODEL_COMPILER_ARGS: ``
- CM_QAIC_MODEL_COMPILER_ARGS_SUT:
-aic-num-cores=4 -mos=1 -ols=1
- Workflow:
- Environment variables:
_resnet50,multistream,nsp.14
- Environment variables:
- CM_QAIC_MODEL_COMPILER_ARGS_SUT:
-aic-num-cores=4
- CM_QAIC_MODEL_COMPILER_ARGS_SUT:
- Workflow:
- Environment variables:
_resnet50,offline
- Environment variables:
- CM_QAIC_MODEL_COMPILER_ARGS:
-sdp-cluster-sizes=2,2 -multicast-weights
- CM_QAIC_MODEL_COMPILER_ARGS_SUT:
-aic-num-cores=4 -mos=1,2 -ols=4
- CM_QAIC_MODEL_COMPILER_ARGS:
- Workflow:
- Environment variables:
_resnet50,offline,nsp.14
- Environment variables:
- CM_QAIC_MODEL_COMPILER_ARGS_SUT:
-aic-num-cores=4 -mos=1,2 -ols=4
- CM_QAIC_MODEL_COMPILER_ARGS_SUT:
- Workflow:
- Environment variables:
_resnet50,server
- Workflow:
_resnet50,server,nsp.14
- Environment variables:
- CM_QAIC_MODEL_COMPILER_ARGS_SUT:
-aic-num-cores=4 -ols=4
- CM_QAIC_MODEL_COMPILER_ARGS:
-sdp-cluster-sizes=2,2 -mos=1,2 -multicast-weights
- CM_QAIC_MODEL_COMPILER_ARGS_SUT:
- Workflow:
- Environment variables:
_resnet50,server,nsp.16
- Environment variables:
- CM_QAIC_MODEL_COMPILER_ARGS_SUT:
-aic-num-cores=4 -ols=4
- CM_QAIC_MODEL_COMPILER_ARGS:
-sdp-cluster-sizes=4,4 -mos=1,4
- CM_QAIC_MODEL_COMPILER_ARGS_SUT:
- Workflow:
- Environment variables:
_resnet50,singlestream
- Environment variables:
- CM_QAIC_MODEL_COMPILER_ARGS:
-aic-num-of-instances=1
- CM_QAIC_MODEL_COMPILER_ARGS_SUT:
-aic-num-cores=8 -mos=1 -ols=1
- CM_QAIC_MODEL_COMPILER_ARGS:
- Workflow:
- Environment variables:
_resnet50,singlestream,nsp.14
- Environment variables:
- CM_QAIC_MODEL_COMPILER_ARGS_SUT:
-aic-num-cores=8 -mos=1 -ols=1
- CM_QAIC_MODEL_COMPILER_ARGS_SUT:
- Workflow:
- Environment variables:
_resnet50,tf
- Environment variables:
- CM_QAIC_MODEL_TO_CONVERT:
calibrate_resnet50_tf
- CM_QAIC_MODEL_TO_CONVERT:
- Workflow:
- Environment variables:
_retinanet
- Environment variables:
- CM_COMPILE_RETINANET:
on
- CM_QAIC_MODEL_TO_CONVERT:
calibrate_retinanet_no_nms_mlperf
- CM_QAIC_MODEL_COMPILER_ARGS:
-aic-enable-depth-first
- CM_QAIC_MODEL_COMPILER_PARAMS_BASE:
-aic-hw -aic-hw-version=2.0 -compile-only -enable-channelwise -onnx-define-symbol=batch_size,1 -node-precision-info=<<<CM_ML_MODEL_RETINANET_QAIC_NODE_PRECISION_INFO_FILE_PATH>>> -quantization-schema-constants=symmetric_with_uint8 -quantization-schema-activations=asymmetric -quantization-calibration=None
- CM_COMPILE_RETINANET:
- Workflow:
- Environment variables:
_retinanet,multistream
- Workflow:
_retinanet,nsp.14
- Workflow:
_retinanet,offline
- Environment variables:
- CM_QAIC_MODEL_COMPILER_ARGS_SUT:
-aic-num-cores=1 -mos=1 -ols=1
- CM_QAIC_MODEL_COMPILER_ARGS_SUT:
- Workflow:
- Environment variables:
_retinanet,offline,nsp.14
- Workflow:
_retinanet,server
- Workflow:
_retinanet,server,nsp.14
- Workflow:
_retinanet,singlestream
- Environment variables:
- CM_QAIC_MODEL_COMPILER_ARGS: ``
- CM_QAIC_MODEL_COMPILER_ARGS_SUT:
-aic-num-cores=8 -mos=1 -ols=1
- Workflow:
- Environment variables:
_retinanet,singlestream,nsp.14
- Environment variables:
- CM_QAIC_MODEL_COMPILER_ARGS_SUT:
-aic-num-cores=8 -mos=1 -ols=1
- CM_QAIC_MODEL_COMPILER_ARGS_SUT:
- Workflow:
- Environment variables:
-
Group "batch-size"
Click here to expand this section.
_bs.#
- Environment variables:
- CM_QAIC_MODEL_BATCH_SIZE:
#
- CM_QAIC_MODEL_BATCH_SIZE:
- Workflow:
- Environment variables:
_bs.1
- Environment variables:
- CM_QAIC_MODEL_BATCH_SIZE:
1
- CM_QAIC_MODEL_BATCH_SIZE:
- Workflow:
- Environment variables:
-
Group "calib-dataset-filter-size"
Click here to expand this section.
_filter-size.#
- Workflow:
-
Group "mlperf-scenario"
Click here to expand this section.
_multistream
- Workflow:
_offline
- Workflow:
_server
- Workflow:
_singlestream
(default)- Workflow:
-
Group "model-framework"
Click here to expand this section.
_tf
- Workflow:
-
Group "nsp"
Click here to expand this section.
_nsp.14
- Workflow:
_nsp.16
- Workflow:
_nsp.8
- Workflow:
_nsp.9
- Workflow:
-
Group "percentile-calibration"
Click here to expand this section.
_pc.#
- Environment variables:
- CM_QAIC_MODEL_COMPILER_PERCENTILE_CALIBRATION_VALUE:
#
- CM_QAIC_MODEL_COMPILER_QUANTIZATION_PARAMS:
-quantization-calibration=Percentile -percentile-calibration-value=<<<CM_QAIC_MODEL_COMPILER_PERCENTILE_CALIBRATION_VALUE>>>
- CM_QAIC_MODEL_COMPILER_PERCENTILE_CALIBRATION_VALUE:
- Workflow:
- Environment variables:
-
Group "quantization"
Click here to expand this section.
_no-quantized
- Environment variables:
- CM_QAIC_MODEL_QUANTIZATION:
no
- CM_QAIC_MODEL_QUANTIZATION:
- Workflow:
- Environment variables:
_quantized
(default)- Environment variables:
- CM_QAIC_MODEL_QUANTIZATION:
yes
- CM_QAIC_MODEL_QUANTIZATION:
- Workflow:
- Environment variables:
_quantized,_singlestream
Click here to expand this section.
--register=value
→CM_REGISTER_CACHE=value
Above CLI flags can be used in the Python CM API as follows:
r=cm.access({... , "register":...}
Click here to expand this section.
These keys can be updated via --env.KEY=VALUE
or env
dictionary in @input.json
or using script flags.
Click here to expand this section.
- Read "deps" on other CM scripts from meta
- detect,cpu
- CM script: detect-cpu
- get,qaic,apps,sdk
if (CM_REGISTER_CACHE != on)
- CM names:
--adr.['qaic-apps-sdk']...
- CM script: get-qaic-apps-sdk
- qaic,calibrate,_retinanet
if (CM_COMPILE_RETINANET == yes)
- CM names:
--adr.['retinanet-profile', 'qaic-profile']...
- CM script: calibrate-model-for.qaic
- qaic,calibrate,_resnet50
if (CM_COMPILE_RESNET == on) AND (CM_REGISTER_CACHE != on)
- CM names:
--adr.['resnet-profile', 'qaic-profile']...
- CM script: calibrate-model-for.qaic
- get,ml-model
- CM names:
--adr.['model-src']...
- CM script: get-ml-model-3d-unet-kits19
- CM script: get-ml-model-abtf-ssd-pytorch
- CM script: get-ml-model-bert-base-squad
- CM script: get-ml-model-bert-large-squad
- CM script: get-ml-model-dlrm-terabyte
- CM script: get-ml-model-efficientnet-lite
- CM script: get-ml-model-gptj
- CM script: get-ml-model-huggingface-zoo
- CM script: get-ml-model-llama2
- CM script: get-ml-model-mobilenet
- CM script: get-ml-model-neuralmagic-zoo
- CM script: get-ml-model-resnet50
- CM script: get-ml-model-retinanet
- CM script: get-ml-model-retinanet-nvidia
- CM script: get-ml-model-rnnt
- CM script: get-ml-model-stable-diffusion
- CM script: get-ml-model-tiny-resnet
- CM script: get-ml-model-using-imagenet-from-model-zoo
- CM names:
- detect,cpu
- Run "preprocess" function from customize.py
- Read "prehook_deps" on other CM scripts from meta
- Run native script if exists
- Read "posthook_deps" on other CM scripts from meta
- Run "postrocess" function from customize.py
- Read "post_deps" on other CM scripts from meta
cmr "qaic compile model model-compile qaic-compile[,variations]" [--input_flags] -j
CM_ML_MODEL_FILE_WITH_PATH
CM_QAIC_MODEL*
CM_ML_MODEL_FILE_WITH_PATH
CM_QAIC_MODEL_COMPILED_BINARY_WITH_PATH
CM_QAIC_MODEL_FINAL_COMPILATION_CMD