update IPEX example QA to high version transformers (#1131)

Author: changwangss

examples/.config/model_params_pytorch.json

@@ -64,6 +64,15 @@
       "new_benchmark": false
     },
     "bert_large_ipex":{
+      "model_src_dir": "nlp/huggingface_models/question-answering/quantization/ptq_static/ipex",
+      "dataset_location": "",
+      "input_model": "",
+      "yaml": "conf.yaml",
+      "strategy": "basic",
+      "batchsize": 8,
+      "new_benchmark": false
+    },
+    "bert_large_1_10_ipex":{
       "model_src_dir": "nlp/huggingface_models/question-answering/quantization/ptq_static/ipex",
       "dataset_location": "/tf_dataset/pytorch/squad/dev-v1.1.json",
       "input_model": "/tf_dataset/pytorch/bert_large_ipex",

examples/pytorch/nlp/huggingface_models/question-answering/quantization/ptq_static/ipex/README.md

@@ -23,29 +23,55 @@ Follow [link](https://github.com/intel-innersource/frameworks.ai.models.intel-mo
 
 > Note: Intel® Extension for PyTorch* has PyTorch version requirement. Please check more detailed information via the URL below.
 
-- Install transformers and set tag to v3.0.2
+
+## Run
+If IPEX version higher than 1.12, please install transformers 4.19.0. We can use the model from huggingface model hub and squad dataset from datasets package, run script `run_qa.py` with command as following.
+
+- Install transformers
+```
+  pip install transformers == 4.19.0
+```
+- Command
+```
+  python run_qa.py 
+    --model_name_or_path bert-large-uncased-whole-word-masking-finetuned-squad \
+    --dataset_name squad \
+    --do_eval \
+    --max_seq_length 384 \
+    --doc_stride 128 \
+    --no_cuda \
+    --tune \
+    --output_dir ./savedresult 
+
+```
+
+
+```
+  bash run_tuning.sh --topology="bert_large_ipex"
+```
+```
+  bash run_benchmark.sh --topology="bert_large_ipex" --mode=benchmark
+```
+
+If IPEX verison is 1.10 or 1.11, please install transformers 3.0.2, prepare model, dataset and run script `run_qa_1_10.py` command as following.
+- install transformers
+```
+pip install transformers == 3.0.2
 ```
-  git clone https://github.com/huggingface/transformers.git
-  cd transformers
-  git checkout v3.0.2
-  pip install -e ./
-  cd ../
-  ```
 
 - Download dataset
   Please following this [link](https://github.com/huggingface/transformers/tree/v3.0.2/examples/question-answering) to get dev-v1.1.json
 
-- Downliad fine-tuned model
+- Download fine-tuned model
 ```
   mkdir bert_squad_model
   wget https://s3.amazonaws.com/models.huggingface.co/bert/bert-large-uncased-whole-word-masking-finetuned-squad-config.json -O bert_squad_model/config.json
   wget https://cdn.huggingface.co/bert-large-uncased-whole-word-masking-finetuned-squad-pytorch_model.bin  -O bert_squad_model/pytorch_model.bin
 ```
 
-
-## Run
+- Command
 ```
-  python run_qa.py 
+  python run_qa_1_10.py 
     --model_type bert 
     --model_name_or_path ./bert_squad_model/ #finetuned model
     --do_lower_case 
@@ -61,10 +87,11 @@ Follow [link](https://github.com/intel-innersource/frameworks.ai.models.intel-mo
     --int8_fp32
 ```
 
-## Quick start
 ```
-  bash run_tuning.sh --dataset_location=/path/to/dataset --input_model=/path/to/model 
+  bash run_tuning.sh --topology="bert_large_1_10_ipex" --dataset_location=/path/to/dataset --input_model=/path/to/model 
 ```
 ```
-  bash run_benchmark.sh --dataset_location=/path/to/dataset --input_model=/path/to/model --mode=benchmark
+  bash run_benchmark.sh --topology="bert_large_1_10_ipex" --dataset_location=/path/to/dataset --input_model=/path/to/model --mode=benchmark
 ```
+
+

examples/pytorch/nlp/huggingface_models/question-answering/quantization/ptq_static/ipex/requirements.txt

@@ -1,4 +1,4 @@
 accelerate
 datasets >= 1.8.0
-transformers == 3.0.2
+transformers == 4.19.0
 tensorboard

examples/pytorch/nlp/huggingface_models/question-answering/quantization/ptq_static/ipex/run_benchmark.sh

@@ -65,20 +65,35 @@ function run_benchmark {
         extra_cmd=$extra_cmd" --int8"
     fi
     echo $extra_cmd
-
-    python run_qa.py \
-        --model_type bert \
-        --model_name_or_path $input_model \
-        --do_lower_case \
-        --predict_file $dataset_location \
-        --tokenizer_name $tokenizer_name \
-        --do_eval \
-        --max_seq_length 384 \
-        --doc_stride 128 \
-        --no_cuda \
-        --output_dir $tuned_checkpoint \
-        $mode_cmd \
-        ${extra_cmd}
+    if [[ "${topology}" == "bert_large_ipex" ]]; then
+        model_name_or_path="bert-large-uncased-whole-word-masking-finetuned-squad"
+        python run_qa.py \
+            --model_name_or_path $model_name_or_path \
+            --dataset_name squad \
+            --do_eval \
+            --max_seq_length 384 \
+            --no_cuda \
+            --output_dir $tuned_checkpoint \
+            $mode_cmd \
+            ${extra_cmd}
+    fi
+    if [[ "${topology}" == "bert_large_1_10_ipex" ]]; then
+        pip install transformers==3.0.2
+        python run_qa_1_10.py \
+            --model_type bert \
+            --model_name_or_path $input_model \
+            --do_lower_case \
+            --predict_file $dataset_location \
+            --tokenizer_name $tokenizer_name \
+            --do_eval \
+            --max_seq_length 384 \
+            --doc_stride 128 \
+            --no_cuda \
+            --output_dir $tuned_checkpoint \
+            $mode_cmd \
+            ${extra_cmd}
+    fi
 }
 
-main "$@"
+
+main "$@"