From 5622400fabdc4e4701cd9bb2845b0a9ee1c76641 Mon Sep 17 00:00:00 2001 From: cTuning Date: Tue, 1 Aug 2023 13:38:14 +0000 Subject: [PATCH] Updated docs --- cm-mlops/script/get-cuda/README.md | 8 ++- .../get-ml-model-neuralmagic-zoo/README.md | 72 +++++++++++++++---- 2 files changed, 65 insertions(+), 15 deletions(-) diff --git a/cm-mlops/script/get-cuda/README.md b/cm-mlops/script/get-cuda/README.md index 3e58340338..dbd0d954e0 100644 --- a/cm-mlops/script/get-cuda/README.md +++ b/cm-mlops/script/get-cuda/README.md @@ -189,7 +189,13 @@ ___
Click here to expand this section. - 1. Read "deps" on other CM scripts from [meta](https://github.com/mlcommons/ck/tree/master/cm-mlops/script/get-cuda/_cm.json) + 1. ***Read "deps" on other CM scripts from [meta](https://github.com/mlcommons/ck/tree/master/cm-mlops/script/get-cuda/_cm.json)*** + * detect,os + - CM script: [detect-os](https://github.com/mlcommons/ck/tree/master/cm-mlops/script/detect-os) + * get,cl + * `if (CM_CUDA_FULL_TOOLKIT_INSTALL == yes AND CM_HOST_OS_TYPE == windows)` + * CM names: `--adr.['compiler']...` + - CM script: [get-cl](https://github.com/mlcommons/ck/tree/master/cm-mlops/script/get-cl) 1. ***Run "preprocess" function from [customize.py](https://github.com/mlcommons/ck/tree/master/cm-mlops/script/get-cuda/customize.py)*** 1. ***Read "prehook_deps" on other CM scripts from [meta](https://github.com/mlcommons/ck/tree/master/cm-mlops/script/get-cuda/_cm.json)*** * install,cuda,prebuilt diff --git a/cm-mlops/script/get-ml-model-neuralmagic-zoo/README.md b/cm-mlops/script/get-ml-model-neuralmagic-zoo/README.md index 3724e79756..9008ca6682 100644 --- a/cm-mlops/script/get-ml-model-neuralmagic-zoo/README.md +++ b/cm-mlops/script/get-ml-model-neuralmagic-zoo/README.md @@ -106,53 +106,97 @@ ___
Click here to expand this section. - * `_14layer_pruned50_quant-none-vnni` + * `_bert-base-pruned95_obs_quant-none` + - Aliases: `_model-stub.zoo:nlp/question_answering/bert-base/pytorch/huggingface/squad/pruned95_obs_quant-none` + - Environment variables: + - *CM_MODEL_ZOO_STUB*: `zoo:nlp/question_answering/bert-base/pytorch/huggingface/squad/pruned95_obs_quant-none` + - *CM_ML_MODEL_FULL_NAME*: `bert-base-pruned95_obs_quant-none-bert-99` + - *CM_ML_MODEL_STARTING_WEIGHTS_FILENAME*: `https://huggingface.co/bert-large-uncased` + - *CM_ML_MODEL_WEIGHT_TRANSFORMATIONS*: `quantization, unstructured pruning` + - *CM_ML_MODEL_WEIGHTS_DATA_TYPE*: `int8` + - *CM_ML_MODEL_INPUTS_DATA_TYPE*: `int64` + - *CM_ML_MODEL_RETRAINING*: `yes` + - Workflow: + * `_mobilebert-14layer_pruned50-none-vnni` + - Aliases: `_model-stub.zoo:nlp/question_answering/mobilebert-none/pytorch/huggingface/squad/14layer_pruned50-none-vnni` + - Environment variables: + - *CM_MODEL_ZOO_STUB*: `zoo:nlp/question_answering/mobilebert-none/pytorch/huggingface/squad/14layer_pruned50-none-vnni` + - *CM_ML_MODEL_FULL_NAME*: `mobilebert-14layer_pruned50-none-vnni-bert-99` + - *CM_ML_MODEL_STARTING_WEIGHTS_FILENAME*: `https://storage.googleapis.com/cloud-tpu-checkpoints/mobilebert/uncased_L-24_H-128_B-512_A-4_F-4_OPT.tar.gz` + - *CM_ML_MODEL_WEIGHT_TRANSFORMATIONS*: `unstructured pruning` + - *CM_ML_MODEL_WEIGHTS_DATA_TYPE*: `fp32` + - *CM_ML_MODEL_INPUTS_DATA_TYPE*: `fp32` + - *CM_ML_MODEL_RETRAINING*: `no` + - Workflow: + * `_mobilebert-14layer_pruned50_quant-none-vnni` - Aliases: `_model-stub.zoo:nlp/question_answering/mobilebert-none/pytorch/huggingface/squad/14layer_pruned50_quant-none-vnni` - Environment variables: - *CM_MODEL_ZOO_STUB*: `zoo:nlp/question_answering/mobilebert-none/pytorch/huggingface/squad/14layer_pruned50_quant-none-vnni` - - *CM_ML_MODEL_FULL_NAME*: `14layer_pruned50_quant-none-vnni` + - *CM_ML_MODEL_FULL_NAME*: `mobilebert-14layer_pruned50_quant-none-vnni-bert-99` - *CM_ML_MODEL_STARTING_WEIGHTS_FILENAME*: `https://storage.googleapis.com/cloud-tpu-checkpoints/mobilebert/uncased_L-24_H-128_B-512_A-4_F-4_OPT.tar.gz` - *CM_ML_MODEL_WEIGHT_TRANSFORMATIONS*: `quantization, unstructured pruning` - *CM_ML_MODEL_WEIGHTS_DATA_TYPE*: `int8` - *CM_ML_MODEL_INPUTS_DATA_TYPE*: `int64` - *CM_ML_MODEL_RETRAINING*: `yes` - Workflow: - * `_base_quant-none` + * `_mobilebert-base_quant-none` - Aliases: `_model-stub.zoo:nlp/question_answering/mobilebert-none/pytorch/huggingface/squad/base_quant-none` - Environment variables: - *CM_MODEL_ZOO_STUB*: `zoo:nlp/question_answering/mobilebert-none/pytorch/huggingface/squad/base_quant-none` - - *CM_ML_MODEL_FULL_NAME*: `base_quant-none` + - *CM_ML_MODEL_FULL_NAME*: `mobilebert-base_quant-none-bert-99` - *CM_ML_MODEL_STARTING_WEIGHTS_FILENAME*: `https://storage.googleapis.com/cloud-tpu-checkpoints/mobilebert/uncased_L-24_H-128_B-512_A-4_F-4_OPT.tar.gz` - *CM_ML_MODEL_WEIGHT_TRANSFORMATIONS*: `quantization, unstructured pruning` - *CM_ML_MODEL_WEIGHTS_DATA_TYPE*: `int8` - *CM_ML_MODEL_INPUTS_DATA_TYPE*: `int64` - *CM_ML_MODEL_RETRAINING*: `yes` - Workflow: + * `_mobilebert-none-base-none` + - Aliases: `_model-stub.zoo:nlp/question_answering/mobilebert-none/pytorch/huggingface/squad/base-none` + - Environment variables: + - *CM_MODEL_ZOO_STUB*: `zoo:nlp/question_answering/mobilebert-none/pytorch/huggingface/squad/base-none` + - *CM_ML_MODEL_FULL_NAME*: `mobilebert-none-base-none-bert-99` + - *CM_ML_MODEL_STARTING_WEIGHTS_FILENAME*: `https://storage.googleapis.com/cloud-tpu-checkpoints/mobilebert/uncased_L-24_H-128_B-512_A-4_F-4_OPT.tar.gz` + - *CM_ML_MODEL_WEIGHT_TRANSFORMATIONS*: `unstructured pruning` + - *CM_ML_MODEL_WEIGHTS_DATA_TYPE*: `fp32` + - *CM_ML_MODEL_INPUTS_DATA_TYPE*: `fp32` + - *CM_ML_MODEL_RETRAINING*: `no` + - Workflow: * `_model-stub.#` - Environment variables: - *CM_MODEL_ZOO_STUB*: `#` - Workflow: - * `_pruned95_obs_quant-none` - - Aliases: `_model-stub.zoo:nlp/question_answering/bert-base/pytorch/huggingface/squad/pruned95_obs_quant-none` + * `_obert-large-pruned95_quant-none-vnni` + - Aliases: `_model-stub.zoo:nlp/question_answering/obert-large/pytorch/huggingface/squad/pruned95_quant-none-vnni` - Environment variables: - - *CM_MODEL_ZOO_STUB*: `zoo:nlp/question_answering/bert-base/pytorch/huggingface/squad/pruned95_obs_quant-none` - - *CM_ML_MODEL_FULL_NAME*: `pruned95_obs_quant-none` + - *CM_MODEL_ZOO_STUB*: `zoo:nlp/question_answering/obert-large/pytorch/huggingface/squad/pruned95_quant-none-vnni` + - *CM_ML_MODEL_FULL_NAME*: `obert-large-pruned95_quant-none-vnni-bert-99` - *CM_ML_MODEL_STARTING_WEIGHTS_FILENAME*: `https://huggingface.co/bert-large-uncased` - *CM_ML_MODEL_WEIGHT_TRANSFORMATIONS*: `quantization, unstructured pruning` - *CM_ML_MODEL_WEIGHTS_DATA_TYPE*: `int8` - *CM_ML_MODEL_INPUTS_DATA_TYPE*: `int64` - *CM_ML_MODEL_RETRAINING*: `yes` - Workflow: - * `_pruned95_quant-none-vnni` - - Aliases: `_model-stub.zoo:nlp/question_answering/obert-large/pytorch/huggingface/squad/pruned95_quant-none-vnni` + * `_oberta-base-pruned90-quant-none` + - Aliases: `_model-stub.zoo:nlp/question_answering/oberta-base/pytorch/huggingface/squad/pruned90_quant-none` - Environment variables: - - *CM_MODEL_ZOO_STUB*: `zoo:nlp/question_answering/obert-large/pytorch/huggingface/squad/pruned95_quant-none-vnni` - - *CM_ML_MODEL_FULL_NAME*: `pruned95_quant-none-vnni` - - *CM_ML_MODEL_STARTING_WEIGHTS_FILENAME*: `https://huggingface.co/bert-large-uncased` + - *CM_MODEL_ZOO_STUB*: `zoo:nlp/question_answering/oberta-base/pytorch/huggingface/squad/pruned90_quant-none` + - *CM_ML_MODEL_FULL_NAME*: `oberta-base-pruned90-quant-none-bert-99` + - *CM_ML_MODEL_STARTING_WEIGHTS_FILENAME*: `https://huggingface.co/roberta-base` - *CM_ML_MODEL_WEIGHT_TRANSFORMATIONS*: `quantization, unstructured pruning` - *CM_ML_MODEL_WEIGHTS_DATA_TYPE*: `int8` - *CM_ML_MODEL_INPUTS_DATA_TYPE*: `int64` - - *CM_ML_MODEL_RETRAINING*: `yes` + - *CM_ML_MODEL_RETRAINING*: `no` + - Workflow: + * `_roberta-base-pruned85-quant-none` + - Aliases: `_model-stub.zoo:nlp/question_answering/roberta-base/pytorch/huggingface/squad/pruned85_quant-none` + - Environment variables: + - *CM_MODEL_ZOO_STUB*: `zoo:nlp/question_answering/roberta-base/pytorch/huggingface/squad/pruned85_quant-none` + - *CM_ML_MODEL_FULL_NAME*: `roberta-base-pruned85-quant-none-bert-99` + - *CM_ML_MODEL_STARTING_WEIGHTS_FILENAME*: `https://huggingface.co/roberta-base` + - *CM_ML_MODEL_WEIGHT_TRANSFORMATIONS*: `quantization, unstructured pruning` + - *CM_ML_MODEL_WEIGHTS_DATA_TYPE*: `int8` + - *CM_ML_MODEL_INPUTS_DATA_TYPE*: `int64` + - *CM_ML_MODEL_RETRAINING*: `no` - Workflow: