Upload folder using huggingface_hub

config.json

@@ -0,0 +1,18 @@
+{
+  "model_type": "custom-translation",
+  "architecture": "Keras Sequential",
+  "language": "en", 
+  "task": "translation",
+  "max_seq_length": 50,
+  "tokenizer": {
+    "type": "custom",
+    "file_1": "en_tokenizer.json",
+    "file_2": "fr_tokenizer.json"
+  },
+  "framework": "tensorflow",
+  "description": "This is a custom translation model built using TensorFlow Keras and integrated with a tokenizer for seamless text translation. It supports tokenization and text generation directly using a user-defined function.",
+  "custom_class_file": "translation_model.py",
+  "version": 1.0,
+  "author": "Shivam Ghuge",
+  "license": "apache-2.0"
+}

model_weight/model.weights.h5

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:82bf670e412232ab115881757ade0d35b95721444529dc042744c0ec63808062
+size 552959904

translate.py

@@ -0,0 +1,32 @@
+import json
+from tensorflow.keras.preprocessing.text import tokenizer_from_json
+from tensorflow.keras.preprocessing.sequence import pad_sequences
+from tensorflow.keras.models import load_model
+import numpy as np
+
+class TranslationModel:
+    def __init__(self, model_path,en_tokenizer,fr_tokenizer,max_len):
+        # Load the Keras model
+        self.model = load_model(model_path)
+
+        self.english_tokenizer = en_tokenizer
+
+        self.french_tokenizer = fr_tokenizer
+
+        # Define max sequence length
+        self.max_seq_length = max_len
+
+    def logits_index(self,text):
+        input_sequence = self.english_tokenizer.texts_to_sequences([text])
+        padded_input_sequence = pad_sequences(input_sequence, maxlen=self.max_seq_length, padding='post')
+        decoded_text = '<start>'
+
+        for i in range(self.max_seq_length):
+            target_sequence = self.french_tokenizer.texts_to_sequences([decoded_text])
+            padded_target_sequence = pad_sequences(target_sequence, maxlen=self.max_seq_length, padding='post')[:, :-1]
+
+            prediction = self.model([padded_input_sequence, padded_target_sequence])
+
+            idx = np.argmax(prediction[0, i, :])
+
+        return idx