From 52c85f847aba62ef2018162472007a167dc25622 Mon Sep 17 00:00:00 2001
From: ahotrod <44321615+ahotrod@users.noreply.github.com>
Date: Sun, 19 Apr 2020 10:49:24 -0700
Subject: [PATCH] Update README.md

---
 .../ahotrod/roberta_large_squad2/README.md    | 60 ++++++++-----------
 1 file changed, 24 insertions(+), 36 deletions(-)

diff --git a/model_cards/ahotrod/roberta_large_squad2/README.md b/model_cards/ahotrod/roberta_large_squad2/README.md
index 77a0c5841b..2fe38471d4 100644
--- a/model_cards/ahotrod/roberta_large_squad2/README.md
+++ b/model_cards/ahotrod/roberta_large_squad2/README.md
@@ -3,66 +3,54 @@
 ### with the following results:
 
 ```
-  "exact": 84.02257222269014,
-  "f1": 87.47063479332766,
+  "exact": 84.46896319380106,
+  "f1": 87.85388093408943,
   "total": 11873,
-  "HasAns_exact": 81.19095816464238,
-  "HasAns_f1": 88.0969714745582,
+  "HasAns_exact": 81.37651821862349,
+  "HasAns_f1": 88.1560607844881,
   "HasAns_total": 5928,
-  "NoAns_exact": 86.84608915054667,
-  "NoAns_f1": 86.84608915054667,
+  "NoAns_exact": 87.55256518082422,
+  "NoAns_f1": 87.55256518082422,
   "NoAns_total": 5945,
-  "best_exact": 84.02257222269014,
+  "best_exact": 84.46896319380106,
   "best_exact_thresh": 0.0,
-  "best_f1": 87.47063479332759,
+  "best_f1": 87.85388093408929,
   "best_f1_thresh": 0.0
 ```
 ### from script:
 ```
-python -m torch.distributed.launch --nproc_per_node=2 ${RUN_SQUAD_DIR}/run_squad.py \
+python ${EXAMPLES}/run_squad.py \
   --model_type roberta \
   --model_name_or_path roberta-large \
   --do_train \
-  --train_file ${SQUAD_DIR}/train-v2.0.json \
-  --predict_file ${SQUAD_DIR}/dev-v2.0.json \
+  --do_eval \
+  --train_file ${SQUAD}/train-v2.0.json \
+  --predict_file ${SQUAD}/dev-v2.0.json \
   --version_2_with_negative \
-  --num_train_epochs 2 \
-  --warmup_steps 328 \
-  --weight_decay 0.01 \
   --do_lower_case \
-  --learning_rate 1.5e-5 \
+  --num_train_epochs 3 \
+  --warmup_steps 1642 \
+  --weight_decay 0.01 \
+  --learning_rate 3e-5 \
+  --adam_epsilon 1e-6 \
   --max_seq_length 512 \
   --doc_stride 128 \
-  --save_steps 1000 \
-  --per_gpu_train_batch_size 1 \
-  --gradient_accumulation_steps 24 \
+  --per_gpu_train_batch_size 8 \
+  --gradient_accumulation_steps 6 \
+  --per_gpu_eval_batch_size 48 \
+  --threads 12 \
   --logging_steps 50 \
-  --threads 10 \
-  --overwrite_cache \
-  --overwrite_output_dir \
-  --output_dir ${MODEL_PATH}
-
-python ${RUN_SQUAD_DIR}/run_squad.py \
-  --model_type roberta \
-  --model_name_or_path ${MODEL_PATH} \
-  --do_eval \
-  --train_file ${SQUAD_DIR}/train-v2.0.json \
-  --predict_file ${SQUAD_DIR}/dev-v2.0.json \
-  --version_2_with_negative \
-  --do_lower_case \
-  --max_seq_length 512 \
-  --per_gpu_eval_batch_size 24 \
-  --eval_all_checkpoints \
+  --save_steps 2000 \
   --overwrite_output_dir \
   --output_dir ${MODEL_PATH}
 $@
 ```
 ### using the following system & software:
 ```
-OS/Platform: Linux-4.15.0-91-generic-x86_64-with-debian-buster-sid
-GPU/CPU: 2 x NVIDIA 1080Ti / Intel i7-8700
 Transformers: 2.7.0
 PyTorch: 1.4.0
 TensorFlow: 2.1.0
 Python: 3.7.7
+OS/Platform: Linux-5.3.0-46-generic-x86_64-with-debian-buster-sid
+CPU/GPU: Intel i9-9900K / NVIDIA Titan RTX 24GB
 ```