-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathrun_init_full.sh
155 lines (134 loc) · 4.83 KB
/
run_init_full.sh
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
#!/bin/bash
FILE_NAME=$0
bbin=$1
if [ $# -ne 1 ] ; then
echo "bash ${0} bin"
exit 0
fi
DEVICE=3
echo "DEVICE : ${DEVICE}"
#RESIDUAL_CHANGE=2step_reduced0.75_LNatonce
TIMESTEP=1
while [ $TIMESTEP -le 1 ]
do
echo "timestep : ${TIMESTEP}"
#SEED=${RANDOM}
SEED=$(echo "${TIMESTEP}*100" | bc)
echo "SEED : ${SEED}"
TOTAL_NUM_UPDATES=30000
TRAIN_STEPS=10000000
VERSION=${TIMESTEP}
model_from=blx
#model_from=31.3
#model_from=33.3
#model_from=bl
#model_from=43.62
#model_from=43.81
#BART_PATH=./checkpoints/bart.large.cnn/model.pt
#BART_PATH=./checkpoints/bart.large/model.pt
#BART_PATH=./checkpoints/blx_exf_norm_bsz90_43.62/checkpoint88.pt
BART_PATH=./checkpoints/bart.large.xsum/model.pt
#BART_PATH=./checkpoints/blc_norm_xentstop_drop0.1_lr3.0e-05_wu500_bsz1_ufreq2/blc_xentstop_norm/checkpoint_best.pt
#BART_PATH=./checkpoints/bl_200_100_byR1stop/checkpoint.best_loss_6.50.pt
#BART_PATH=./checkpoints_low/bl_200_100_Ralphastop_r3f0.1_uniformnoise_drop0.1_lr3.0e-05_wu400_bsz16_ufreq3/bl_200_100_Ralphastop_r3f_33.3/checkpoint_best.pt
#BART_PATH=./checkpoints/blc_norm_xentstop_drop0.1_lr3.0e-05_wu500_bsz1_ufreq2/blc_xentstop_norm_43.81/checkpoint_best.pt
#MAX_TOKENS=2048
MAX_TOKENS=1024
FT_WARMUP_UPDATES=500
#LR=5e-04
#WEIGHT_DECAY=1.0e-04
#OPTIM=adabelief
#ADAM_EPS=1e-16
WEIGHT_DECAY=0.01
LR=3.0e-05
OPTIM=adam
ADAM_EPS=1e-8
FT_DROPOUT=0.1
FT_BSZ=128
FT_UPDATE_FRQ=32
FT_MAX_EPOCH=50
FT_PATIENCE=3
NOISE_TYPE=uniform
R3F_LAMBDA=0.1
STOP_CRITERIA=xentstop
#STOP_CRITERIA=Rstop
TRAIN_FILE=fairseq_cli/train.py
#TRAIN_FILE=fairseq_cli/train_CS.py
#TRAIN_FILE=fairseq_cli/train2.py
#TRAIN_FILE=fairseq_cli/train_midval.py
#TRAIN_FILE=fairseq_cli/train_midval_Rstop.py
#TRAIN_FILE=fairseq_cli/train_rouge_val_low.py
#TRAIN_FILE=fairseq_cli/train_R3F_CS_Rstoplow_midval.py
#TRAIN_FILE=fairseq_cli/train_rougeRL_Rval_midval.py
#TRAIN_FILE=fairseq_cli/train2.py
#TRAIN_FILE=fairseq_cli/train_rouge_val.py
#TRAIN_FILE=fairseq_cli/train_Rval_cnndmlow.py
#TRAIN_FILE=fairseq_cli/train_R3F_cossim_Rstop.py
#TRAIN_FILE=fairseq_cli/train_R3F_cossim_Rstoplow_Sbert.py
#TRAIN_FILE=fairseq_cli/train_Ralpha_val.py
#CRITERIONN=r3f
#CRITERIONN=r3f_wolog
#CRITERIONN=r3f_cossim0.3
#CRITERIONN=cossim0.3
CRITERIONN=norm
TRAIN_CRITERION=label_smoothed_cross_entropy
#TRAIN_CRITERION=label_smoothed_cross_entropy_r3f
#TRAIN_CRITERION=label_smoothed_cross_entropy_r3f_cossim
#TRAIN_CRITERION=label_smoothed_cross_entropy_cossim2
#TRAIN_CRITERION=semantic_similarity_loss
#ATTN_CHANGE=valrate
#RATE=0.25
FT_PATIENCE=3
#FT_PATIENCE=$(echo 5/${RATE} | bc)
#SAVE_DIR1=./checkpoints_low/xsumlow${VERSION}_${RESIDUAL_CHANGE}_seed${SEED}_${ATTN_CHANGE}${RATE}_from${model_from}_${CRITERIONN}_${OPTIM}_${STOP_CRITERIA}_drop${FT_DROPOUT}_lr${LR}_wu${FT_WARMUP_UPDATES}_bsz${FT_BSZ}_ufreq${FT_UPDATE_FRQ}
SAVE_DIR1=./checkpoints/xsum${VERSION}_${RESIDUAL_CHANGE}_seed${SEED}_${ATTN_CHANGE}${RATE}_from${model_from}_${CRITERIONN}_${OPTIM}_${STOP_CRITERIA}_dev${DEVICE}
if [ ! -d $SAVE_DIR1 ]; then
mkdir $SAVE_DIR1
fi
SAVE_DIR=${SAVE_DIR1}/from${model_from}_${STOP_CRITERIA}_${CRITERIONN}
if [ ! -d $SAVE_DIR ]; then
mkdir $SAVE_DIR
fi
# --noise-type ${NOISE_TYPE} --r3f-lambda ${R3F_LAMBDA} \
# --ddp-backend=no_c10d \
CUDA_VISIBLE_DEVICES=${DEVICE} python ${TRAIN_FILE} ${bbin} \
--memory-efficient-fp16 \
--fp16 \
--max-epoch ${FT_MAX_EPOCH} \
--max-update ${TRAIN_STEPS} \
--save-dir ${SAVE_DIR} \
--batch-size ${FT_BSZ} --seed ${SEED} \
--update-freq ${FT_UPDATE_FRQ} \
--keep-best-checkpoints 1 \
--no-last-checkpoints \
--keep-last-epochs 1 \
--patience ${FT_PATIENCE} \
--restore-file ${BART_PATH} \
--max-tokens ${MAX_TOKENS} \
--task translation \
--source-lang source --target-lang target \
--truncate-source \
--layernorm-embedding \
--share-all-embeddings \
--share-decoder-input-output-embed \
--reset-optimizer --reset-dataloader --reset-meters \
--required-batch-size-multiple 1 \
--arch bart_large \
--criterion ${TRAIN_CRITERION} \
--label-smoothing 0.1 \
--dropout ${FT_DROPOUT} --attention-dropout ${FT_DROPOUT} \
--weight-decay ${WEIGHT_DECAY} --optimizer ${OPTIM} --adam-betas "(0.9, 0.999)" --adam-eps ${ADAM_EPS} \
--clip-norm 0.1 \
--lr-scheduler polynomial_decay --lr $LR --total-num-update $TOTAL_NUM_UPDATES \
--warmup-updates ${FT_WARMUP_UPDATES} \
--skip-invalid-size-inputs-valid-test \
--find-unused-parameters --bpe gpt2;
wait $!
#SAMPLE_SIZE=7330
#CUDA_VISIBLE_DEVICES=${DEVICE} python ./xsum_ST/hypogen_Forunsup.py ${SAVE_DIR}/checkpoint_best.pt 0 '' ${SAMPLE_SIZE}
#CUDA_VISIBLE_DEVICES=${DEVICE} python ./xsum_ST/hypogen_fortest_cnndm.py ${SAVE_DIR}/checkpoint_best.pt 0 True
#CUDA_VISIBLE_DEVICES=${DEVICE} python ./xsum_ST/hypogen_Forunsup.py ${SAVE_DIR}/checkpoint_best.pt 0 True
CUDA_VISIBLE_DEVICES=${DEVICE} python ./xsum_ST/hypogen_ForFull_3bestckpts.py ${SAVE_DIR}/checkpoint_best.pt 0 True;
((TIMESTEP++))
done
python sms.py ${DEVICE}