File size: 5,631 Bytes
e29422a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
21d0902
3917612
0c66821
21d0902
 
 
 
e29422a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
from components.induce_personality import construct_big_five_words
from components.constant import (
    ACCESS,
    QUERY_REWRITING,
    RAG,
    PERSONALITY,
    PERSONALITY_LIST,
    REWRITE_PASSAGES,
    NUM_PASSAGES,
    DEVICE,
    RESPONSE_GENERATOR,
    TEMPLATE_PAYLOAD,
)
from components.prompt import SYSTEM_INSTRUCTION, RAG_INSTRUCTION, PERSONALITY_INSTRUCTION
import requests
import together


def generate_response_debugging(history):
    # outputs_text = "This is a test response"
    outputs_text = " ".join([item["content"] for item in history])
    history = history + [{"role": "assistant", "content": outputs_text}]
    return outputs_text, history


# REWRITER = "castorini/t5-base-canard"
def generate_response_together_api(history, max_tokens, client, model="meta-llama/Meta-Llama-3.1-8B-Instruct-Turbo"):
    together_request = {
        "model": model,
        "messages": history,
        "stream": False,
        "logprobs": False,
        "stop": ["<eos>", "<unk>", "<sep>", "<pad>", "<cls>", "<mask>"],
        "max_tokens": max_tokens,
    }
    response = client.chat.completions.create(**together_request)
    outputs_text = response.choices[0].message.content
    history = history + [{"role": "assistant", "content": outputs_text}]
    return outputs_text, history


def make_local_api_call(payload, api_url):
    try:
        # Send the POST request to the API
        response = requests.post(api_url, json=payload)

        # Check if the request was successful
        if response.status_code == 200:
            result = response.json()
            # Print the generated text
            return result.get("text", [""])[0]
            # if "logits" in result:
            #     print(f"Logits: {result['logits']}")
        else:
            # If there was an error, print the status code and message
            print(f"Error: {response.status_code}")
            print(response.text)

    except requests.exceptions.RequestException as e:
        print(f"Request failed: {e}")


def generate_response_local_api(history, terminator, max_tokens, api_url):
    payload = TEMPLATE_PAYLOAD.copy()
    payload.update(
        {
            "prompt": history,
            "max_tokens": max_tokens,
            "stop_token_ids": terminator,
        }
    )
    # Call the API to generate the response
    outputs_text = make_local_api_call(payload, api_url)

    if outputs_text:
        # Update history with the assistant's response
        history = history + [{"role": "assistant", "content": outputs_text}]
        return outputs_text, history
    else:
        print("Failed to generate a response.")
        return "Generation failed", history  # Return the original history in case of failure


def conversation_window(history, N=100):
    if len(history) > N:
        return history[2:]
    return history


def format_message_history(message, history):
    if not history:
        str_history = f"\n<user>: {message}\n<assistant>"
    else:
        # Query written
        str_history = (
            "".join(["".join(["\n<user>:" + item[0], "\n<assistant>:" + item[1]]) for item in history])
            + f"\n<user>: {message}\n<assistant>"
        )
    return str_history


def format_user_message(message, history):
    return history + [{"role": "user", "content": message}]


def format_context(message, history):
    return [{"role": "system", "content": message}] + history


def prepare_tokenizer(tokenizer):
    special_tokens = ["<eos>", "<unk>", "<sep>", "<pad>", "<cls>", "<mask>"]
    for token in special_tokens:
        if tokenizer.convert_tokens_to_ids(token) is None:
            tokenizer.add_tokens([token])

    if tokenizer.eos_token_id is None:
        tokenizer.eos_token_id = tokenizer.convert_tokens_to_ids("<eos>")
    terminators = [
        tokenizer.eos_token_id,
        # self.pipeline.tokenizer.convert_tokens_to_ids(""),
    ]
    return tokenizer, terminators


def gradio_to_huggingface_message(gradio_message):
    huggingface_message = []
    for user, bot in gradio_message:
        huggingface_message.append({"role": "user", "content": user})
        huggingface_message.append({"role": "assistant", "content": bot})
    return huggingface_message


def huggingface_to_gradio_message(huggingface_message):
    gradio_message = []
    store = []
    for utter in huggingface_message:
        if utter["role"] in ["user", "assistant"]:
            if utter["role"] == "assistant":
                store.append(utter["content"])
                gradio_message.append(store)
                store = []
            else:
                store.append(utter["content"])
    return gradio_message


def get_personality_instruction(personality):
    return PERSONALITY_INSTRUCTION.format(personality)


def get_system_instruction(rag=RAG, personality_list=None):
    if rag and personality_list:
        return (
            SYSTEM_INSTRUCTION
            + RAG_INSTRUCTION
            + get_personality_instruction(construct_big_five_words(personality_list))
        )
    elif personality_list:
        return SYSTEM_INSTRUCTION + get_personality_instruction(construct_big_five_words(personality_list))
    elif rag:
        return SYSTEM_INSTRUCTION + RAG_INSTRUCTION
    else:
        return SYSTEM_INSTRUCTION


def format_rag_context(rag_context):
    """
    rag_context [{"passage_id": clue_web, "passage_text": "abc"}, ...]
    """
    passage_context = "Context: \n"
    for passage_rank, info in enumerate(rag_context):
        passage_context += f"Passage ID: {info['passage_id']}, Text: {info['passage_text']}\n\n"
    return passage_context