Update README.md

README.md

@@ -5,9 +5,9 @@ widget:
 - text: "作为电子extra0的平台，京东绝对是领先者。如今的刘强extra1已经是身价过extra2的老板。"
 
 
-
 ---
 
+
 # Chinese T5 Version 1.1
 
 ## Model description
@@ -61,7 +61,7 @@ python3 preprocess.py --corpus_path corpora/cluecorpussmall.txt \
                       --vocab_path models/google_zh_with_sentinel_vocab.txt \
                       --dataset_path cluecorpussmall_t5-v1_1_seq128_dataset.pt \
                       --processes_num 32 --seq_length 128 \
-                      --dynamic_masking --target t5 
+                      --dynamic_masking --data_processor t5 
 ```
 
 ```
@@ -72,10 +72,7 @@ python3 pretrain.py --dataset_path cluecorpussmall_t5-v1_1_seq128_dataset.pt \
                     --world_size 8 --gpu_ranks 0 1 2 3 4 5 6 7 \
                     --total_steps 1000000 --save_checkpoint_steps 100000 --report_steps 50000 \
                     --learning_rate 1e-3 --batch_size 64 \
-                    --span_masking --span_geo_prob 0.3 --span_max_length 5 \
-                    --embedding word --relative_position_embedding --remove_embedding_layernorm --tgt_embedding word \
-                    --encoder transformer --mask fully_visible --layernorm_positioning pre \
-                    --feed_forward gated --decoder transformer --target t5
+                    --span_masking --span_geo_prob 0.3 --span_max_length 5
 ```
 
 Stage2:
@@ -85,7 +82,7 @@ python3 preprocess.py --corpus_path corpora/cluecorpussmall.txt \
                       --vocab_path models/google_zh_with_sentinel_vocab.txt \
                       --dataset_path cluecorpussmall_t5-v1_1_seq512_dataset.pt \
                       --processes_num 32 --seq_length 512 \
-                      --dynamic_masking --target t5 
+                      --dynamic_masking --data_processor t5 
 ```
 
 ```
@@ -97,10 +94,7 @@ python3 pretrain.py --dataset_path cluecorpussmall_t5-v1_1_seq512_dataset.pt \
                     --world_size 8 --gpu_ranks 0 1 2 3 4 5 6 7 \
                     --total_steps 250000 --save_checkpoint_steps 50000 --report_steps 10000 \
                     --learning_rate 5e-4 --batch_size 16 \
-                    --span_masking --span_geo_prob 0.3 --span_max_length 5 \
-                    --embedding word --relative_position_embedding --remove_embedding_layernorm --tgt_embedding word \
-                    --encoder transformer --mask fully_visible --layernorm_positioning pre \
-                    --feed_forward gated --decoder transformer --target t5
+                    --span_masking --span_geo_prob 0.3 --span_max_length 5
 ```
 
 Finally, we convert the pre-trained model into Huggingface's format: