Upload flan_t5_xl_with_gpu.py

flan_t5_xl_with_gpu.py

@@ -0,0 +1,103 @@
+# -*- coding: utf-8 -*-
+"""Flan-t5-xl_with_GPU.ipynb
+
+Automatically generated by Colaboratory.
+
+Original file is located at
+    https://colab.research.google.com/drive/15P4GWaUNFBqJf5_I58DxSiusjPiHUeU6
+"""
+
+from IPython.display import HTML, display
+
+def set_css():
+  display(HTML('''
+  <style>
+    pre {
+      white-space: pre-wrap;
+    }
+  </style>
+  '''))
+get_ipython().events.register('pre_run_cell', set_css)
+
+import multiprocessing
+import torch
+
+cores = multiprocessing.cpu_count() # Count the number of cores in a computer
+cores
+
+!pip install -q transformers
+
+!pip install -U deep-translator
+
+from deep_translator import GoogleTranslator
+
+# Use any translator you like, in this example GoogleTranslator
+#translated = GoogleTranslator(source='hy', target='en').translate("Բարև, ո՞նց ես։")  # output -> Hello, how are you?
+
+!pip install accelerate bitsandbytes sentencepiece
+
+device = "cuda:0" if torch.cuda.is_available() else "cpu"
+device
+
+!pip install streamlit
+import streamlit as st
+
+x = st.slider('Select a value')
+st.write(x, 'squared is', x * x)
+
+!apt-get update -qq
+!apt-get install -y -qq git
+
+#!git clone https://huggingface.co/spaces/VGG1555/VGG1
+
+!pwd
+!ls -la /
+!cd /content/VGG1
+
+import torch
+from transformers import T5Tokenizer, T5ForConditionalGeneration
+
+tokenizer = T5Tokenizer.from_pretrained("google/flan-t5-xl")
+model = T5ForConditionalGeneration.from_pretrained("google/flan-t5-xl", device_map = "auto")
+
+# We are running FP32!
+
+my_text = "Summarize: \
+Science can ignite new discoveries for society, \
+Society has the tendency to refer to old, familiar ways of doing. \
+Chaos is also a part of our society, although increasingly often so.\
+Innovative ways lead to new growthin businesses and under certain conditions all of society participates."
+
+my_text = "Write an essay with 100 words about Quantum Physics and it's problems regarding our understanding of laws of Physics."
+
+my_text = "Q: Can Geoffrey Hinton have a conversation with George Washington? Give the rationale before answering."
+# my_text = "A short explanation of machine learning for medical applications."
+
+translated = "Q: Ի՞նչ է խնձորը։"
+my_text = GoogleTranslator(source='hy', target='en').translate(translated)
+
+# We tune it a little bit: SMILE:
+# Specify your downstream task: FP32
+
+torch.cuda.empty_cache()
+
+input_ids = tokenizer(my_text, return_tensors = "pt").input_ids.to("cuda")
+
+outputs = model.generate(input_ids,
+                         min_length = 20,
+                         max_new_tokens = 600,
+                         length_penalty = 1.6, # Set to values < 1.0 in order to encourage the model to generate shorter answers.
+                         num_beams = 6,
+                         no_repeat_ngram_size = 3,
+                         temperature = 0.9,
+                         top_k = 150, # default 50
+                         top_p = 0.92,
+                         repetition_penalty = 2.1)
+
+print(GoogleTranslator(source='en', target='hy').translate(tokenizer.decode(outputs[0], skip_special_tokens = True)))
+
+#!git clone https://huggingface.co/spaces/VGG1555/VGG1
+!cd /content/VGG1 && git status
+!cd /content/VGG1 && git add app.py
+!cd /content/VGG1 && git commit -m "Add application file"
+!cd /content/VGG1 && git push