Spaces:
Runtime error
Runtime error
Zwea Htet
commited on
Commit
•
e594eb9
1
Parent(s):
5e8fa58
update llama custom
Browse files- models/llamaCustom.py +69 -31
models/llamaCustom.py
CHANGED
@@ -58,34 +58,79 @@ def load_model(model_name: str):
|
|
58 |
return pipe
|
59 |
|
60 |
|
61 |
-
|
62 |
-
|
63 |
-
|
|
|
|
|
64 |
|
65 |
-
|
66 |
-
|
67 |
-
|
68 |
-
|
69 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
70 |
|
71 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
72 |
prompt_length = len(prompt)
|
73 |
-
response = self.pipeline(prompt, max_new_tokens=
|
74 |
|
75 |
# only return newly generated tokens
|
76 |
-
|
|
|
77 |
|
78 |
-
|
79 |
-
|
80 |
-
return {"name_of_model": self.llm_model_name}
|
81 |
|
82 |
-
|
83 |
-
|
84 |
-
|
|
|
|
|
|
|
85 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
86 |
|
87 |
class LlamaCustom:
|
|
|
|
|
|
|
|
|
|
|
88 |
def __init__(self, model_name: str) -> None:
|
|
|
|
|
|
|
|
|
|
|
89 |
self.vector_index = self.initialize_index(model_name=model_name)
|
90 |
|
91 |
@st.cache_resource
|
@@ -93,6 +138,7 @@ class LlamaCustom:
|
|
93 |
index_name = model_name.split("/")[-1]
|
94 |
|
95 |
file_path = f"./vectorStores/{index_name}"
|
|
|
96 |
if os.path.exists(path=file_path):
|
97 |
# rebuild storage context
|
98 |
storage_context = StorageContext.from_defaults(persist_dir=file_path)
|
@@ -105,23 +151,11 @@ class LlamaCustom:
|
|
105 |
# index = pickle.loads(file.readlines())
|
106 |
return index
|
107 |
else:
|
108 |
-
# define llm
|
109 |
-
prompt_helper = PromptHelper(
|
110 |
-
context_window=CONTEXT_WINDOW,
|
111 |
-
num_output=NUM_OUTPUT,
|
112 |
-
chunk_overlap_ratio=CHUNK_OVERLAP_RATION,
|
113 |
-
)
|
114 |
-
|
115 |
-
llm_predictor = LLMPredictor(llm=CustomLLM(llm_model_name=model_name))
|
116 |
-
service_context = ServiceContext.from_defaults(
|
117 |
-
llm_predictor=llm_predictor, prompt_helper=prompt_helper
|
118 |
-
)
|
119 |
-
|
120 |
# documents = prepare_data(r"./assets/regItems.json")
|
121 |
documents = SimpleDirectoryReader(input_dir="./assets/pdf").load_data()
|
122 |
|
123 |
index = GPTVectorStoreIndex.from_documents(
|
124 |
-
documents, service_context=service_context
|
125 |
)
|
126 |
|
127 |
# local write access
|
@@ -134,6 +168,10 @@ class LlamaCustom:
|
|
134 |
|
135 |
def get_response(self, query_str):
|
136 |
print("query_str: ", query_str)
|
137 |
-
query_engine = self.vector_index.as_query_engine()
|
|
|
|
|
|
|
138 |
response = query_engine.query(query_str)
|
|
|
139 |
return str(response)
|
|
|
58 |
return pipe
|
59 |
|
60 |
|
61 |
+
@st.cache_resource
|
62 |
+
def load_model(mode_name: str):
|
63 |
+
# llm_model_name = "bigscience/bloom-560m"
|
64 |
+
tokenizer = AutoTokenizer.from_pretrained(mode_name)
|
65 |
+
model = AutoModelForCausalLM.from_pretrained(mode_name, config="T5Config")
|
66 |
|
67 |
+
pipe = pipeline(
|
68 |
+
task="text-generation",
|
69 |
+
model=model,
|
70 |
+
tokenizer=tokenizer,
|
71 |
+
# device=0, # GPU device number
|
72 |
+
# max_length=512,
|
73 |
+
do_sample=True,
|
74 |
+
top_p=0.95,
|
75 |
+
top_k=50,
|
76 |
+
temperature=0.7,
|
77 |
+
)
|
78 |
|
79 |
+
return pipe
|
80 |
+
|
81 |
+
|
82 |
+
class OurLLM(CustomLLM):
|
83 |
+
def __init__(self, model_name: str, model_pipeline):
|
84 |
+
self.model_name = model_name
|
85 |
+
self.pipeline = model_pipeline
|
86 |
+
|
87 |
+
@property
|
88 |
+
def metadata(self) -> LLMMetadata:
|
89 |
+
"""Get LLM metadata."""
|
90 |
+
return LLMMetadata(
|
91 |
+
context_window=CONTEXT_WINDOW,
|
92 |
+
num_output=NUM_OUTPUT,
|
93 |
+
model_name=self.model_name,
|
94 |
+
)
|
95 |
+
|
96 |
+
def complete(self, prompt: str, **kwargs: Any) -> CompletionResponse:
|
97 |
prompt_length = len(prompt)
|
98 |
+
response = self.pipeline(prompt, max_new_tokens=NUM_OUTPUT)[0]["generated_text"]
|
99 |
|
100 |
# only return newly generated tokens
|
101 |
+
text = response[prompt_length:]
|
102 |
+
return CompletionResponse(text=text)
|
103 |
|
104 |
+
def stream_complete(self, prompt: str, **kwargs: Any) -> CompletionResponse:
|
105 |
+
raise NotImplementedError()
|
|
|
106 |
|
107 |
+
# def _call(self, prompt: str, stop: Optional[List[str]] = None) -> str:
|
108 |
+
# prompt_length = len(prompt)
|
109 |
+
# response = self.pipeline(prompt, max_new_tokens=525)[0]["generated_text"]
|
110 |
+
|
111 |
+
# # only return newly generated tokens
|
112 |
+
# return response[prompt_length:]
|
113 |
|
114 |
+
# @property
|
115 |
+
# def _identifying_params(self) -> Mapping[str, Any]:
|
116 |
+
# return {"name_of_model": self.model_name}
|
117 |
+
|
118 |
+
# @property
|
119 |
+
# def _llm_type(self) -> str:
|
120 |
+
# return "custom"
|
121 |
|
122 |
class LlamaCustom:
|
123 |
+
# define llm
|
124 |
+
# llm_predictor = LLMPredictor(llm=OurLLM())
|
125 |
+
# service_context = ServiceContext.from_defaults(
|
126 |
+
# llm_predictor=llm_predictor, prompt_helper=prompt_helper
|
127 |
+
# )
|
128 |
def __init__(self, model_name: str) -> None:
|
129 |
+
pipe = load_model(mode_name=model_name)
|
130 |
+
llm = OurLLM(model_name=model_name, model_pipeline=pipe)
|
131 |
+
self.service_context = ServiceContext.from_defaults(
|
132 |
+
llm=llm, prompt_helper=prompt_helper
|
133 |
+
)
|
134 |
self.vector_index = self.initialize_index(model_name=model_name)
|
135 |
|
136 |
@st.cache_resource
|
|
|
138 |
index_name = model_name.split("/")[-1]
|
139 |
|
140 |
file_path = f"./vectorStores/{index_name}"
|
141 |
+
|
142 |
if os.path.exists(path=file_path):
|
143 |
# rebuild storage context
|
144 |
storage_context = StorageContext.from_defaults(persist_dir=file_path)
|
|
|
151 |
# index = pickle.loads(file.readlines())
|
152 |
return index
|
153 |
else:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
154 |
# documents = prepare_data(r"./assets/regItems.json")
|
155 |
documents = SimpleDirectoryReader(input_dir="./assets/pdf").load_data()
|
156 |
|
157 |
index = GPTVectorStoreIndex.from_documents(
|
158 |
+
documents, service_context=self.service_context
|
159 |
)
|
160 |
|
161 |
# local write access
|
|
|
168 |
|
169 |
def get_response(self, query_str):
|
170 |
print("query_str: ", query_str)
|
171 |
+
# query_engine = self.vector_index.as_query_engine()
|
172 |
+
query_engine = self.vector_index.as_query_engine(
|
173 |
+
text_qa_template=text_qa_template, refine_template=refine_template
|
174 |
+
)
|
175 |
response = query_engine.query(query_str)
|
176 |
+
print("metadata: ", response.metadata)
|
177 |
return str(response)
|