Fix some bugs

config.json

@@ -1,56 +0,0 @@
-{
-  "architectures": [
-    "T5ForConditionalGeneration"
-  ],
-  "d_ff": 3072,
-  "d_kv": 64,
-  "d_model": 768,
-  "decoder_start_token_id": 0,
-  "dropout_rate": 0.1,
-  "eos_token_id": 1,
-  "feed_forward_proj": "relu",
-  "gradient_checkpointing": false,
-  "initializer_factor": 1.0,
-  "is_encoder_decoder": true,
-  "layer_norm_epsilon": 1e-06,
-  "model_type": "t5",
-  "n_positions": 512,
-  "num_decoder_layers": 12,
-  "num_heads": 12,
-  "num_layers": 12,
-  "output_past": true,
-  "pad_token_id": 0,
-  "relative_attention_num_buckets": 32,
-  "task_specific_params": {
-    "summarization": {
-      "early_stopping": true,
-      "length_penalty": 2.0,
-      "max_length": 200,
-      "min_length": 30,
-      "no_repeat_ngram_size": 3,
-      "num_beams": 4,
-      "prefix": "summarize: "
-    },
-    "translation_en_to_de": {
-      "early_stopping": true,
-      "max_length": 300,
-      "num_beams": 4,
-      "prefix": "translate English to German: "
-    },
-    "translation_en_to_fr": {
-      "early_stopping": true,
-      "max_length": 300,
-      "num_beams": 4,
-      "prefix": "translate English to French: "
-    },
-    "translation_en_to_ro": {
-      "early_stopping": true,
-      "max_length": 300,
-      "num_beams": 4,
-      "prefix": "translate English to Romanian: "
-    }
-  },
-  "transformers_version": "4.9.0.dev0",
-  "use_cache": true,
-  "vocab_size": 32128
-}

events.out.tfevents.1625682591.t1v-n-a0c138ef-w-0.124617.3.v2

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:5c069e81c193f5ba7a9c8cff114c5522e13d8efd16e2e8c055c880bf5010f334
-size 736165

flax_model.msgpack

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:12aea1d6f15b37764f5615dcb6d6bc6cc56e7d74cd3ce88cdd0469817b5a9c29
-size 891625348

src/{preparaing_recipe_nlg_dataset.py → create_dataset.py}

@@ -114,6 +114,7 @@ def main():
 
         return {
             "inputs": ner,
+            # "targets": f"title: {title} <section> ingredients: {ingredients} <section> directions: {steps}"
             "targets": f"title: {title} <section> ingredients: {ingredients} <section> directions: {steps}"
         }
 

src/run.sh

@@ -5,6 +5,7 @@ export LANG=C.UTF-8
 
 export OUTPUT_DIR=/home/m3hrdadfi/code/t5-recipe-generation
 export MODEL_NAME_OR_PATH=t5-base
+# export MODEL_NAME_OR_PATH=flax-community/t5-recipe-generation
 export NUM_BEAMS=3
 
 export TRAIN_FILE=/home/m3hrdadfi/code/data/train.csv

src/run_recipe_nlg_flax.py

@@ -21,6 +21,7 @@ Fine-tuning the library models for recipe-generation.
 import logging
 import os
 import random
+import re
 import sys
 import time
 from dataclasses import dataclass, field
@@ -375,7 +376,7 @@ def main():
             data_files["test"] = data_args.test_file
             extension = data_args.test_file.split(".")[-1]
 
-        print(data_files)
+        logger.info(data_files)
         dataset = load_dataset(
             extension,
             data_files=data_files,
@@ -551,10 +552,30 @@ def main():
     bleu = load_metric("sacrebleu")
     wer = load_metric("wer")
 
+    def skip_special_tokens_text(text):
+        new_text = []
+        for word in text.split():
+            word = word.strip()
+            if word:
+                if word not in special_tokens:
+                    new_text.append(word)
+        
+        return " ".join(new_text)
+    
+    def skip_special_tokens_texts(texts):
+        if isinstance(texts, list):
+            new_texts = [skip_special_tokens_text(text) for text in texts]
+        elif isinstance(texts, str):
+            new_texts = skip_special_tokens_text(texts)
+        else:
+            new_texts = []
+        
+        return new_texts
+
     def postprocess_text(preds, labels):
-        preds = [pred.strip() for pred in preds]
-        labels_bleu = [[label.strip()] for label in labels]
-        labels_wer = [label.strip() for label in labels]
+        preds = [skip_special_tokens_texts(pred.strip()) for pred in preds]
+        labels_bleu = [[skip_special_tokens_texts(label.strip())] for label in labels]
+        labels_wer = [skip_special_tokens_texts(label.strip()) for label in labels]
 
         return preds, [labels_bleu, labels_wer]
 
@@ -846,11 +867,6 @@ def main():
                         push_to_hub=training_args.push_to_hub,
                         commit_message=f"Saving weights and logs of step {cur_step}",
                     )
-                    tokenizer.save_pretrained(
-                        training_args.output_dir,
-                        push_to_hub=training_args.push_to_hub,
-                        commit_message=f"Saving tokenizer step {cur_step}",
-                    )
 
 
 if __name__ == "__main__":