NCTCMumbai
commited on
Commit
·
aad5b37
1
Parent(s):
5131264
Update backend/query_llm.py
Browse files- backend/query_llm.py +10 -52
backend/query_llm.py
CHANGED
@@ -50,7 +50,7 @@ def format_prompt(message: str, api_kind: str):
|
|
50 |
raise ValueError("API is not supported")
|
51 |
|
52 |
|
53 |
-
def generate_hf(prompt: str, history: str, temperature: float = 0.9, max_new_tokens: int =
|
54 |
top_p: float = 0.95, repetition_penalty: float = 1.0) -> Generator[str, None, str]:
|
55 |
"""
|
56 |
Generate a sequence of tokens based on a given prompt and history using Mistral client.
|
@@ -70,57 +70,15 @@ def generate_hf(prompt: str, history: str, temperature: float = 0.9, max_new_tok
|
|
70 |
|
71 |
temperature = max(float(temperature), 1e-2) # Ensure temperature isn't too low
|
72 |
top_p = float(top_p)
|
73 |
-
|
74 |
-
|
75 |
-
|
76 |
-
|
77 |
-
|
78 |
-
|
79 |
-
|
80 |
-
|
81 |
-
|
82 |
-
minimum=0.0,
|
83 |
-
maximum=1.0,
|
84 |
-
step=0.05,
|
85 |
-
interactive=True,
|
86 |
-
info="Higher values produce more diverse outputs",
|
87 |
-
),
|
88 |
-
gr.Slider(
|
89 |
-
label="Max new tokens",
|
90 |
-
value=256,
|
91 |
-
minimum=0,
|
92 |
-
maximum=4048,
|
93 |
-
step=64,
|
94 |
-
interactive=True,
|
95 |
-
info="The maximum numbers of new tokens",
|
96 |
-
),
|
97 |
-
gr.Slider(
|
98 |
-
label="Top-p (nucleus sampling)",
|
99 |
-
value=0.90,
|
100 |
-
minimum=0.0,
|
101 |
-
maximum=1,
|
102 |
-
step=0.05,
|
103 |
-
interactive=True,
|
104 |
-
info="Higher values sample more low-probability tokens",
|
105 |
-
),
|
106 |
-
gr.Slider(
|
107 |
-
label="Repetition penalty",
|
108 |
-
value=1.2,
|
109 |
-
minimum=1.0,
|
110 |
-
maximum=2.0,
|
111 |
-
step=0.05,
|
112 |
-
interactive=True,
|
113 |
-
info="Penalize repeated tokens",
|
114 |
-
)
|
115 |
-
]
|
116 |
-
# generate_kwargs = {
|
117 |
-
# 'temperature': temperature,
|
118 |
-
# 'max_new_tokens': max_new_tokens,
|
119 |
-
# 'top_p': top_p,
|
120 |
-
# 'repetition_penalty': repetition_penalty,
|
121 |
-
# 'do_sample': True,
|
122 |
-
# 'seed': 42,
|
123 |
-
# }
|
124 |
|
125 |
formatted_prompt = format_prompt(prompt, "hf")
|
126 |
|
|
|
50 |
raise ValueError("API is not supported")
|
51 |
|
52 |
|
53 |
+
def generate_hf(prompt: str, history: str, temperature: float = 0.9, max_new_tokens: int = 4000,
|
54 |
top_p: float = 0.95, repetition_penalty: float = 1.0) -> Generator[str, None, str]:
|
55 |
"""
|
56 |
Generate a sequence of tokens based on a given prompt and history using Mistral client.
|
|
|
70 |
|
71 |
temperature = max(float(temperature), 1e-2) # Ensure temperature isn't too low
|
72 |
top_p = float(top_p)
|
73 |
+
|
74 |
+
generate_kwargs = {
|
75 |
+
'temperature': temperature,
|
76 |
+
'max_new_tokens': max_new_tokens,
|
77 |
+
'top_p': top_p,
|
78 |
+
'repetition_penalty': repetition_penalty,
|
79 |
+
'do_sample': True,
|
80 |
+
'seed': 42,
|
81 |
+
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
82 |
|
83 |
formatted_prompt = format_prompt(prompt, "hf")
|
84 |
|