fix static train in formula (#14826)

Author: liuhongen1234567

configs/rec/PP-FormuaNet/rec_pp_formulanet_l.yml

@@ -18,9 +18,10 @@ Global:
   rec_char_dict_path: &rec_char_dict_path ppocr/utils/dict/unimernet_tokenizer
   max_new_tokens: &max_new_tokens 1024
   input_size: &input_size [768, 768]
-  save_res_path: ./output/rec/predicts_unimernet_latexocr.txt
+  save_res_path: ./output/rec/predicts_pp_formulanet_l.txt
   allow_resize_largeImg: False
   start_ema: True
+  d2s_train_image_shape: [1,768,768]
 
 Optimizer:
   name: AdamW

configs/rec/PP-FormuaNet/rec_pp_formulanet_s.yml

@@ -18,9 +18,10 @@ Global:
   rec_char_dict_path: &rec_char_dict_path  ppocr/utils/dict/unimernet_tokenizer
   max_new_tokens: &max_new_tokens 1024
   input_size: &input_size [384, 384]
-  save_res_path: ./output/rec/predicts_unimernet_latexocr.txt
+  save_res_path: ./output/rec/predicts_pp_formulanet_s.txt
   allow_resize_largeImg: False
   start_ema: True
+  d2s_train_image_shape: [1,384,384]
 
 Optimizer:
   name: AdamW

configs/rec/rec_unimernet.yml

@@ -18,8 +18,9 @@ Global:
   rec_char_dict_path: &rec_char_dict_path ppocr/utils/dict/unimernet_tokenizer
   input_size: &input_size [192, 672]
   max_seq_len: &max_seq_len 1024
-  save_res_path: ./output/rec/predicts_unimernet_plus_config_latexocr.txt
+  save_res_path: ./output/rec/predicts_unimernet.txt
   allow_resize_largeImg: False
+  d2s_train_image_shape: [1,192,672]
 
 Optimizer:
   name: AdamW

ppocr/modeling/architectures/__init__.py

@@ -50,6 +50,9 @@ def apply_to_static(model, config, logger):
         "SVTR",
         "SVTR_HGNet",
         "LaTeXOCR",
+        "UniMERNet",
+        "PP-FormulaNet-S",
+        "PP-FormulaNet-L",
     ]
     if config["Architecture"]["algorithm"] in ["Distillation"]:
         algo = list(config["Architecture"]["Models"].values())[0]["algorithm"]
@@ -127,6 +130,16 @@ def apply_to_static(model, config, logger):
                 InputSpec(shape=[None, None], dtype="float32"),
             ]
         ]
+    elif algo in ["UniMERNet", "PP-FormulaNet-S", "PP-FormulaNet-L"]:
+        specs = [
+            [
+                InputSpec(
+                    [None] + config["Global"]["d2s_train_image_shape"], dtype="float32"
+                ),
+                InputSpec(shape=[None, None], dtype="float32"),
+                InputSpec(shape=[None, None], dtype="float32"),
+            ]
+        ]
     model = to_static(model, input_spec=specs)
     logger.info("Successfully to apply @to_static with specs: {}".format(specs))
     return model