Update README.md
Browse files
README.md
CHANGED
|
@@ -6,4 +6,76 @@ language:
|
|
| 6 |
- en
|
| 7 |
base_model:
|
| 8 |
- black-forest-labs/FLUX.1-dev
|
| 9 |
-
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 6 |
- en
|
| 7 |
base_model:
|
| 8 |
- black-forest-labs/FLUX.1-dev
|
| 9 |
+
---
|
| 10 |
+
#This prompt is from message 2. #The goal is to generate 100 messages per prompt.
|
| 11 |
+
|
| 12 |
+
prompt2 = "Vaping is risky"
|
| 13 |
+
|
| 14 |
+
#Below, we specify to use pytorch machine learning framework.
|
| 15 |
+
#You can also choose Tensorflow, but we use Pytorch here.
|
| 16 |
+
|
| 17 |
+
inputs = tokenizer(prompt2, return_tensors="pt")
|
| 18 |
+
---
|
| 19 |
+
#We generate 50 messages each time due to restrictions in Ram storage.
|
| 20 |
+
|
| 21 |
+
sample_outputs = bloom.generate(inputs["input_ids"],
|
| 22 |
+
temperature = 0.7,
|
| 23 |
+
max_new_tokens = 60,
|
| 24 |
+
do_sample=True,
|
| 25 |
+
top_k=40,
|
| 26 |
+
top_p=0.9,
|
| 27 |
+
num_return_sequences=50
|
| 28 |
+
)
|
| 29 |
+
|
| 30 |
+
print("Output:\n" + 100 * '-')
|
| 31 |
+
messages = []
|
| 32 |
+
for i, sample_output in enumerate(sample_outputs):
|
| 33 |
+
generated_messages = tokenizer.decode(sample_output, skip_special_tokens=True)
|
| 34 |
+
print("{}: {}".format(i, generated_messages))
|
| 35 |
+
messages.append(generated_messages)
|
| 36 |
+
|
| 37 |
+
print(messages)
|
| 38 |
+
---
|
| 39 |
+
#We save the AI-generated messages to google drive.
|
| 40 |
+
|
| 41 |
+
AI_messages = pd.DataFrame(messages, columns = ['tweet'])
|
| 42 |
+
AI_messages.to_csv('Vaping is risky1.csv', index = False)
|
| 43 |
+
---
|
| 44 |
+
#Then generate another 50 messages with prompt1 and then save to google drive.
|
| 45 |
+
|
| 46 |
+
AI_messages = pd.DataFrame(messages, columns = ['tweet'])
|
| 47 |
+
AI_messages.to_csv('Vaping is risky2.csv', index = False)
|
| 48 |
+
---
|
| 49 |
+
#This prompt is from message 3. #The goal is to generate 100 messages per prompt.
|
| 50 |
+
|
| 51 |
+
prompt3 = "Vapes and e-cigarettes increase your risk"
|
| 52 |
+
|
| 53 |
+
#Below, we specify to use pytorch machine learning framework.
|
| 54 |
+
#You can also choose Tensorflow, but we use Pytorch here.
|
| 55 |
+
|
| 56 |
+
inputs = tokenizer(prompt3, return_tensors="pt")
|
| 57 |
+
---
|
| 58 |
+
#We generate 50 messages each time due to restrictions in Ram storage.
|
| 59 |
+
|
| 60 |
+
sample_outputs = bloom.generate(inputs["input_ids"],
|
| 61 |
+
temperature = 0.7,
|
| 62 |
+
max_new_tokens = 60,
|
| 63 |
+
do_sample=True,
|
| 64 |
+
top_k=40,
|
| 65 |
+
top_p=0.9,
|
| 66 |
+
num_return_sequences=50
|
| 67 |
+
)
|
| 68 |
+
|
| 69 |
+
print("Output:\n" + 100 * '-')
|
| 70 |
+
messages = []
|
| 71 |
+
for i, sample_output in enumerate(sample_outputs):
|
| 72 |
+
generated_messages = tokenizer.decode(sample_output, skip_special_tokens=True)
|
| 73 |
+
print("{}: {}".format(i, generated_messages))
|
| 74 |
+
messages.append(generated_messages)
|
| 75 |
+
|
| 76 |
+
print(messages)
|
| 77 |
+
---
|
| 78 |
+
#We save the AI-generated messages to google drive.
|
| 79 |
+
|
| 80 |
+
AI_messages = pd.DataFrame(messages, columns = ['tweet'])
|
| 81 |
+
AI_messages.to_csv('Vapes and e-cigarettes increase your risk1.csv', index = False)
|