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
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383 | class WatsonxLLM(CustomLLM):
"""
IBM watsonx.ai large language models.
Example:
`pip install llama-index-llms-ibm`
```python
from llama_index.llms.ibm import WatsonxLLM
watsonx_llm = WatsonxLLM(
model_id="google/flan-ul2",
url="https://us-south.ml.cloud.ibm.com",
apikey="*****",
project_id="*****",
)
```
"""
model_id: Optional[str] = Field(
default=None, description="Type of model to use.", allow_mutation=False
)
deployment_id: Optional[str] = Field(
default=None, description="Id of deployed model to use.", allow_mutation=False
)
temperature: Optional[float] = Field(
default=None,
description="The temperature to use for sampling.",
)
max_new_tokens: Optional[int] = Field(
default=None,
description="The maximum number of tokens to generate.",
)
additional_params: Optional[Dict[str, Any]] = Field(
default_factory=None,
description="Additional generation params for the watsonx.ai models.",
)
project_id: Optional[str] = Field(
default=None,
description="ID of the Watson Studio project.",
allow_mutation=False,
)
space_id: Optional[str] = Field(
default=None, description="ID of the Watson Studio space.", allow_mutation=False
)
url: Optional[SecretStr] = Field(
default=None,
description="Url to Watson Machine Learning or CPD instance",
allow_mutation=False,
)
apikey: Optional[SecretStr] = Field(
default=None,
description="Apikey to Watson Machine Learning or CPD instance",
allow_mutation=False,
)
token: Optional[SecretStr] = Field(
default=None, description="Token to CPD instance", allow_mutation=False
)
password: Optional[SecretStr] = Field(
default=None, description="Password to CPD instance", allow_mutation=False
)
username: Optional[SecretStr] = Field(
default=None, description="Username to CPD instance", allow_mutation=False
)
instance_id: Optional[SecretStr] = Field(
default=None, description="Instance_id of CPD instance", allow_mutation=False
)
version: Optional[SecretStr] = Field(
default=None, description="Version of CPD instance", allow_mutation=False
)
verify: Union[str, bool, None] = Field(
default=None,
description="""
User can pass as verify one of following:
the path to a CA_BUNDLE file
the path of directory with certificates of trusted CAs
True - default path to truststore will be taken
False - no verification will be made
""",
allow_mutation=False,
)
validate_model: bool = Field(
default=True, description="Model id validation", allow_mutation=False
)
_model: ModelInference = PrivateAttr()
_client: Optional[APIClient] = PrivateAttr()
_model_info: Optional[Dict[str, Any]] = PrivateAttr()
_deployment_info: Optional[Dict[str, Any]] = PrivateAttr()
_context_window: Optional[int] = PrivateAttr()
_text_generation_params: Dict[str, Any] | None = PrivateAttr()
def __init__(
self,
model_id: Optional[str] = None,
deployment_id: Optional[str] = None,
temperature: Optional[float] = None,
max_new_tokens: Optional[int] = None,
additional_params: Optional[Dict[str, Any]] = None,
project_id: Optional[str] = None,
space_id: Optional[str] = None,
url: Optional[str] = None,
apikey: Optional[str] = None,
token: Optional[str] = None,
password: Optional[str] = None,
username: Optional[str] = None,
instance_id: Optional[str] = None,
version: Optional[str] = None,
verify: Union[str, bool, None] = None,
api_client: Optional[APIClient] = None,
validate_model: bool = True,
callback_manager: Optional[CallbackManager] = None,
**kwargs: Any,
) -> None:
"""
Initialize LLM and watsonx.ai ModelInference.
"""
callback_manager = callback_manager or CallbackManager([])
additional_params = additional_params or {}
self._context_window = kwargs.get("context_window")
creds = (
resolve_watsonx_credentials(
url=url,
apikey=apikey,
token=token,
username=username,
password=password,
instance_id=instance_id,
)
if not isinstance(api_client, APIClient)
else {}
)
super().__init__(
model_id=model_id,
deployment_id=deployment_id,
temperature=temperature,
max_new_tokens=max_new_tokens,
additional_params=additional_params,
project_id=project_id,
space_id=space_id,
url=creds.get("url"),
apikey=creds.get("apikey"),
token=creds.get("token"),
password=creds.get("password"),
username=creds.get("username"),
instance_id=creds.get("instance_id"),
version=version,
verify=verify,
_client=api_client,
validate_model=validate_model,
callback_manager=callback_manager,
**kwargs,
)
generation_params = {}
if self.temperature is not None:
generation_params["temperature"] = self.temperature
if self.max_new_tokens is not None:
generation_params["max_new_tokens"] = self.max_new_tokens
generation_params = {**generation_params, **additional_params}
if generation_params:
self._text_generation_params, _ = self._split_generation_params(
generation_params
)
else:
self._text_generation_params = None
self._client = api_client
self._model = ModelInference(
model_id=model_id,
deployment_id=deployment_id,
credentials=(
Credentials.from_dict(
{
key: value.get_secret_value() if value else None
for key, value in self._get_credential_kwargs().items()
},
_verify=self.verify,
)
if creds
else None
),
params=self._text_generation_params,
project_id=self.project_id,
space_id=self.space_id,
api_client=api_client,
validate=validate_model,
)
self._model_info = None
self._deployment_info = None
class Config:
validate_assignment = True
@property
def model_info(self):
if self._model.model_id and self._model_info is None:
self._model_info = self._model.get_details()
return self._model_info
@property
def deployment_info(self):
if self._model.deployment_id and self._deployment_info is None:
self._deployment_info = self._model.get_details()
return self._deployment_info
@classmethod
def class_name(cls) -> str:
"""Get Class Name."""
return "WatsonxLLM"
def _get_credential_kwargs(self) -> Dict[str, SecretStr | None]:
return {
"url": self.url,
"apikey": self.apikey,
"token": self.token,
"password": self.password,
"username": self.username,
"instance_id": self.instance_id,
"version": self.version,
}
@property
def metadata(self) -> LLMMetadata:
if self.model_id:
return LLMMetadata(
context_window=(
self.model_info.get("model_limits", {}).get("max_sequence_length")
),
num_output=(self.max_new_tokens or DEFAULT_MAX_TOKENS),
model_name=self.model_id,
)
else:
model_id = self.deployment_info.get("entity", {}).get("base_model_id")
context_window = (
self._model._client.foundation_models.get_model_specs(model_id=model_id)
.get("model_limits", {})
.get("max_sequence_length")
)
return LLMMetadata(
context_window=context_window
or self._context_window
or DEFAULT_CONTEXT_WINDOW,
num_output=(self.max_new_tokens or DEFAULT_MAX_TOKENS),
model_name=model_id or self._model.deployment_id,
)
@property
def sample_generation_text_params(self) -> Dict[str, Any]:
"""Example of Model generation text kwargs that a user can pass to the model."""
return GenTextParamsMetaNames().get_example_values()
def _split_generation_params(
self, data: Dict[str, Any]
) -> Tuple[Dict[str, Any] | None, Dict[str, Any]]:
params = {}
kwargs = {}
sample_generation_kwargs_keys = set(self.sample_generation_text_params.keys())
sample_generation_kwargs_keys.add("prompt_variables")
for key, value in data.items():
if key in sample_generation_kwargs_keys:
params.update({key: value})
else:
kwargs.update({key: value})
return params if params else None, kwargs
@llm_completion_callback()
def complete(
self, prompt: str, formatted: bool = False, **kwargs: Any
) -> CompletionResponse:
params, generation_kwargs = self._split_generation_params(kwargs)
response = self._model.generate(
prompt=prompt,
params=self._text_generation_params or params,
**generation_kwargs,
)
return CompletionResponse(
text=self._model._return_guardrails_stats(response).get("generated_text"),
raw=response,
)
@llm_completion_callback()
def stream_complete(
self, prompt: str, formatted: bool = False, **kwargs: Any
) -> CompletionResponseGen:
params, generation_kwargs = self._split_generation_params(kwargs)
stream_response = self._model.generate_text_stream(
prompt=prompt,
params=self._text_generation_params or params,
**generation_kwargs,
)
def gen() -> CompletionResponseGen:
content = ""
if kwargs.get("raw_response"):
for stream_delta in stream_response:
stream_delta_text = self._model._return_guardrails_stats(
stream_delta
).get("generated_text", "")
content += stream_delta_text
yield CompletionResponse(
text=content, delta=stream_delta_text, raw=stream_delta
)
else:
for stream_delta in stream_response:
content += stream_delta
yield CompletionResponse(text=content, delta=stream_delta)
return gen()
@llm_chat_callback()
def chat(self, messages: Sequence[ChatMessage], **kwargs: Any) -> ChatResponse:
chat_fn = completion_to_chat_decorator(self.complete)
return chat_fn(messages, **kwargs)
@llm_chat_callback()
def stream_chat(
self, messages: Sequence[ChatMessage], **kwargs: Any
) -> ChatResponseGen:
chat_stream_fn = stream_completion_to_chat_decorator(self.stream_complete)
return chat_stream_fn(messages, **kwargs)
|