Spaces:
Paused
Paused
updated with interface
Browse files
app.py
CHANGED
@@ -1,100 +1,65 @@
|
|
1 |
-
|
|
|
|
|
2 |
from transformers import AutoTokenizer, AutoModelForCausalLM
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
3 |
|
4 |
-
|
5 |
-
tokenizer = AutoTokenizer.from_pretrained("ai4bharat/Airavata")
|
6 |
-
model = AutoModelForCausalLM.from_pretrained("ai4bharat/Airavata")
|
7 |
-
|
8 |
-
def chat_interface(user_input, assistant_input):
|
9 |
-
# Concatenate the user and assistant inputs to simulate a chat conversation
|
10 |
-
chat_history = f"{assistant_input} User: {user_input}"
|
11 |
-
|
12 |
-
# Tokenize the chat history and generate the response
|
13 |
-
inputs = tokenizer(chat_history, return_tensors="pt", max_length=256, truncation=True)
|
14 |
-
outputs = model.generate(**inputs)
|
15 |
-
response = tokenizer.batch_decode(outputs, skip_special_tokens=True)[0]
|
16 |
-
|
17 |
-
return response, chat_history
|
18 |
-
|
19 |
-
# Define Gradio Chat Interface
|
20 |
-
iface = gr.ChatInterface(
|
21 |
-
chat_model=chat_interface,
|
22 |
-
title="GPT-2 Chat Interface",
|
23 |
-
inputs=["text", "text"],
|
24 |
-
outputs=["text", "text"],
|
25 |
-
)
|
26 |
-
|
27 |
-
# Launch Gradio Chat Interface
|
28 |
-
iface.launch()
|
29 |
-
|
30 |
-
|
31 |
-
|
32 |
-
# import torch
|
33 |
-
# from transformers import AutoTokenizer, AutoModelForCausalLM
|
34 |
-
# import gradio as gr
|
35 |
-
|
36 |
-
# device = "cuda" if torch.cuda.is_available() else "cpu"
|
37 |
-
|
38 |
-
|
39 |
-
# def create_prompt_with_chat_format(messages, bos="<s>", eos="</s>", add_bos=True):
|
40 |
-
# formatted_text = ""
|
41 |
-
# for message in messages:
|
42 |
-
# if message["role"] == "system":
|
43 |
-
# formatted_text += "<|system|>\n" + message["content"] + "\n"
|
44 |
-
# elif message["role"] == "user":
|
45 |
-
# formatted_text += "<|user|>\n" + message["content"] + "\n"
|
46 |
-
# elif message["role"] == "assistant":
|
47 |
-
# formatted_text += "<|assistant|>\n" + message["content"].strip() + eos + "\n"
|
48 |
-
# else:
|
49 |
-
# raise ValueError(
|
50 |
-
# "Tulu chat template only supports 'system', 'user' and 'assistant' roles. Invalid role: {}.".format(
|
51 |
-
# message["role"]
|
52 |
-
# )
|
53 |
-
# )
|
54 |
-
# formatted_text += "<|assistant|>\n"
|
55 |
-
# formatted_text = bos + formatted_text if add_bos else formatted_text
|
56 |
-
# return formatted_text
|
57 |
-
|
58 |
-
|
59 |
-
# def inference(input_prompts, model, tokenizer):
|
60 |
-
# input_prompts = [
|
61 |
-
# create_prompt_with_chat_format([{"role": "user", "content": input_prompt}], add_bos=False)
|
62 |
-
# for input_prompt in input_prompts
|
63 |
-
# ]
|
64 |
-
|
65 |
-
# encodings = tokenizer(input_prompts, padding=True, return_tensors="pt")
|
66 |
-
# encodings = encodings.to(device)
|
67 |
-
|
68 |
-
# with torch.inference_mode():
|
69 |
-
# outputs = model.generate(encodings.input_ids, do_sample=False, max_new_tokens=250)
|
70 |
-
|
71 |
-
# output_texts = tokenizer.batch_decode(outputs.detach(), skip_special_tokens=True)
|
72 |
-
|
73 |
-
# input_prompts = [
|
74 |
-
# tokenizer.decode(tokenizer.encode(input_prompt), skip_special_tokens=True) for input_prompt in input_prompts
|
75 |
-
# ]
|
76 |
-
# output_texts = [output_text[len(input_prompt) :] for input_prompt, output_text in zip(input_prompts, output_texts)]
|
77 |
-
# return output_texts
|
78 |
-
|
79 |
-
|
80 |
-
# model_name = "ai4bharat/Airavata"
|
81 |
-
|
82 |
-
# tokenizer = AutoTokenizer.from_pretrained(model_name, padding_side="left")
|
83 |
-
# tokenizer.pad_token = tokenizer.eos_token
|
84 |
-
# model = AutoModelForCausalLM.from_pretrained(model_name, torch_dtype=torch.bfloat16).to(device)
|
85 |
-
# print(f"Loading model: {model_name}")
|
86 |
-
|
87 |
-
# examples = [
|
88 |
-
# ["मैं अपने समय प्रबंधन कौशल को कैसे सुधार सकता हूँ? मुझे पांच बिंदु बताएं।"],
|
89 |
-
# ["मैं अपने समय प्रबंधन कौशल को कैसे सुधार सकता हूँ? मुझे पांच बिंदु बताएं और उनका वर्णन करें।"],
|
90 |
-
# ]
|
91 |
-
|
92 |
-
# def chat_interface(input_prompts):
|
93 |
-
# outputs = inference(input_prompts, model, tokenizer)
|
94 |
-
# return outputs
|
95 |
-
|
96 |
-
# gr.Interface(fn=chat_interface,
|
97 |
-
# inputs="text",
|
98 |
-
# outputs="text",
|
99 |
-
# examples=examples,
|
100 |
-
# title="CAMAI ChatBot").launch()
|
|
|
1 |
+
|
2 |
+
|
3 |
+
import torch
|
4 |
from transformers import AutoTokenizer, AutoModelForCausalLM
|
5 |
+
import gradio as gr
|
6 |
+
|
7 |
+
device = "cuda" if torch.cuda.is_available() else "cpu"
|
8 |
+
|
9 |
+
model_name = "ai4bharat/Airavata"
|
10 |
+
|
11 |
+
tokenizer = AutoTokenizer.from_pretrained(model_name, padding_side="left")
|
12 |
+
tokenizer.pad_token = tokenizer.eos_token
|
13 |
+
model = AutoModelForCausalLM.from_pretrained(model_name, torch_dtype=torch.bfloat16).to(device)
|
14 |
+
|
15 |
+
|
16 |
+
def create_prompt_with_chat_format(messages, bos="<s>", eos="</s>", add_bos=True):
|
17 |
+
formatted_text = ""
|
18 |
+
for message in messages:
|
19 |
+
if message["role"] == "system":
|
20 |
+
formatted_text += "<|system|>\n" + message["content"] + "\n"
|
21 |
+
elif message["role"] == "user":
|
22 |
+
formatted_text += "<|user|>\n" + message["content"] + "\n"
|
23 |
+
elif message["role"] == "assistant":
|
24 |
+
formatted_text += "<|assistant|>\n" + message["content"].strip() + eos + "\n"
|
25 |
+
else:
|
26 |
+
raise ValueError(
|
27 |
+
"Tulu chat template only supports 'system', 'user' and 'assistant' roles. Invalid role: {}.".format(
|
28 |
+
message["role"]
|
29 |
+
)
|
30 |
+
)
|
31 |
+
formatted_text += "<|assistant|>\n"
|
32 |
+
formatted_text = bos + formatted_text if add_bos else formatted_text
|
33 |
+
return formatted_text
|
34 |
+
|
35 |
+
|
36 |
+
def inference(input_prompts, model, tokenizer):
|
37 |
+
input_prompts = [
|
38 |
+
create_prompt_with_chat_format([{"role": "user", "content": input_prompt}], add_bos=False)
|
39 |
+
for input_prompt in input_prompts
|
40 |
+
]
|
41 |
+
|
42 |
+
encodings = tokenizer(input_prompts, padding=True, return_tensors="pt")
|
43 |
+
encodings = encodings.to(device)
|
44 |
+
|
45 |
+
with torch.inference_mode(): # Add missing import statement for torch.inference_mode()
|
46 |
+
outputs = model.generate(encodings.input_ids, do_sample=False, max_new_tokens=250)
|
47 |
+
|
48 |
+
output_texts = tokenizer.batch_decode(outputs.detach(), skip_special_tokens=True)
|
49 |
+
|
50 |
+
input_prompts = [
|
51 |
+
tokenizer.decode(tokenizer.encode(input_prompt), skip_special_tokens=True) for input_prompt in input_prompts
|
52 |
+
]
|
53 |
+
output_texts = [output_text[len(input_prompt) :] for input_prompt, output_text in zip(input_prompts, output_texts)]
|
54 |
+
return output_texts
|
55 |
+
|
56 |
+
|
57 |
+
def chat_interface(input_prompts):
|
58 |
+
outputs = inference(input_prompts, model, tokenizer)
|
59 |
+
return outputs
|
60 |
+
|
61 |
+
|
62 |
+
inputs = gr.inputs.Textbox(lines=2, label="User Input")
|
63 |
+
outputs = gr.outputs.Textbox(label="Assistant Response")
|
64 |
|
65 |
+
gr.Interface(fn=chat_interface, inputs=inputs, outputs=outputs, title="Chat Interface").launch()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|