rename to taco_lm

Author: Slash0BZ

README.md

@@ -1,4 +1,4 @@
-# TCS-Bert
+# TacoLM
 A variation of BERT that is aware of temporal common sense.
 
 ## Introduction
@@ -14,7 +14,7 @@ Here are some things you can do with this package out of the box.
 
 ### Train the main model 
 - Access and download `data/tmp_seq_data` at [Google Drive](https://drive.google.com/drive/folders/1kx5Vc8iFYorWHrxHndkUzOssdeOm8oYC?usp=sharing) (4.6 G)
-- run `sh train_tcs_bert.sh`
+- run `sh train_taco_lm.sh`
 
 The script is set to default parameters and will export the model to `models/`. You can configure differently by editing the script.
 
@@ -48,7 +48,7 @@ The intrinsic evaluation relies on pre-formatted data.
 #### MC-TACO Experiment
 See [MC-TACO](https://github.com/CogComp/MCTACO). 
 - use the augmented data under `data/mctaco-tcs`
-- use the transformer weights of `tcs_bert_epoch_2`
+- use the transformer weights of `taco_lm_epoch_2`
 
 ## Citation
 See the following paper: 

eval_hieve.sh

@@ -6,7 +6,7 @@ python examples/run_hieve.py \
   --do_train \
   --do_eval \
   --data_dir data/hieve \
-  --bert_model models/tcs_bert_epoch_2 \
+  --bert_model models/taco_lm_epoch_2 \
   --max_seq_length 128 \
   --train_batch_size 32 \
   --learning_rate 5e-5 \
@@ -20,7 +20,7 @@ python examples/run_hieve.py \
   --do_train \
   --do_eval \
   --data_dir data/hieve \
-  --bert_model models/tcs_bert_epoch_2 \
+  --bert_model models/taco_lm_epoch_2 \
   --max_seq_length 128 \
   --train_batch_size 32 \
   --learning_rate 5e-5 \
@@ -34,7 +34,7 @@ python examples/run_hieve.py \
   --do_train \
   --do_eval \
   --data_dir data/hieve \
-  --bert_model models/tcs_bert_epoch_2 \
+  --bert_model models/taco_lm_epoch_2 \
   --max_seq_length 128 \
   --train_batch_size 32 \
   --learning_rate 5e-5 \

eval_intrinsic.sh

@@ -32,7 +32,7 @@ python examples/run_tmp_seq_eval.py \
   --do_eval \
   --do_lower_case \
   --data_dir data/intrinsic/model.realnews.txt \
-  --bert_model models/tcs_bert_epoch_3  \
+  --bert_model models/taco_lm_epoch_3  \
   --max_seq_length 128 \
   --train_batch_size 32 \
   --learning_rate 2e-5 \
@@ -45,7 +45,7 @@ python examples/run_tmp_seq_eval.py \
   --do_eval \
   --do_lower_case \
   --data_dir data/intrinsic/model.udstmp.txt \
-  --bert_model models/tcs_bert_epoch_3 \
+  --bert_model models/taco_lm_epoch_3 \
   --max_seq_length 128 \
   --train_batch_size 32 \
   --learning_rate 2e-5 \

eval_timebank.sh

@@ -6,7 +6,7 @@ python examples/run_hieve.py \
   --do_train \
   --do_eval \
   --data_dir data/timebank \
-  --bert_model models/tcs_bert_epoch_2 \
+  --bert_model models/taco_lm_epoch_2 \
   --max_seq_length 128 \
   --train_batch_size 32 \
   --learning_rate 5e-5 \
@@ -20,7 +20,7 @@ python examples/run_hieve.py \
   --do_train \
   --do_eval \
   --data_dir data/timebank \
-  --bert_model models/tcs_bert_epoch_2 \
+  --bert_model models/taco_lm_epoch_2 \
   --max_seq_length 128 \
   --train_batch_size 32 \
   --learning_rate 5e-5 \
@@ -34,7 +34,7 @@ python examples/run_hieve.py \
   --do_train \
   --do_eval \
   --data_dir data/timebank \
-  --bert_model models/tcs_bert_epoch_2 \
+  --bert_model models/taco_lm_epoch_2 \
   --max_seq_length 128 \
   --train_batch_size 32 \
   --learning_rate 5e-5 \

train_tcs_bert.sh renamed to train_taco_lm.sh

@@ -11,4 +11,4 @@ python examples/run_tmp_seq.py \
   --train_batch_size 32 \
   --learning_rate 2e-5 \
   --num_train_epochs 3.0 \
-  --output_dir ./models/tcs_bert
+  --output_dir ./models/taco_lm