(question, original_question, state, stopping_strings=None, is_chat=False)
| 310 | |
| 311 | |
| 312 | def generate_reply_HF(question, original_question, state, stopping_strings=None, is_chat=False): |
| 313 | import torch |
| 314 | import transformers |
| 315 | from transformers import LogitsProcessorList |
| 316 | |
| 317 | from modules.grammar.grammar_utils import initialize_grammar |
| 318 | from modules.grammar.logits_process import ( |
| 319 | GrammarConstrainedLogitsProcessor |
| 320 | ) |
| 321 | from modules.torch_utils import clear_torch_cache, get_device |
| 322 | from modules.transformers_loader import ( |
| 323 | Stream, |
| 324 | _StopEverythingStoppingCriteria |
| 325 | ) |
| 326 | |
| 327 | if shared.args.loader == 'Transformers': |
| 328 | clear_torch_cache() |
| 329 | |
| 330 | seed = set_manual_seed(state['seed']) |
| 331 | |
| 332 | generate_params = {} |
| 333 | for k in [ |
| 334 | 'temperature', |
| 335 | 'dynatemp_low', |
| 336 | 'dynatemp_high', |
| 337 | 'dynatemp_exponent', |
| 338 | 'smoothing_factor', |
| 339 | 'smoothing_curve', |
| 340 | 'min_p', |
| 341 | 'top_p', |
| 342 | 'top_k', |
| 343 | 'typical_p', |
| 344 | 'xtc_threshold', |
| 345 | 'xtc_probability', |
| 346 | 'tfs', |
| 347 | 'top_a', |
| 348 | 'top_n_sigma', |
| 349 | 'adaptive_target', |
| 350 | 'adaptive_decay', |
| 351 | 'dry_multiplier', |
| 352 | 'dry_allowed_length', |
| 353 | 'dry_base', |
| 354 | 'repetition_penalty', |
| 355 | 'frequency_penalty', |
| 356 | 'presence_penalty', |
| 357 | 'encoder_repetition_penalty', |
| 358 | 'no_repeat_ngram_size', |
| 359 | 'repetition_penalty_range', |
| 360 | 'penalty_alpha', |
| 361 | 'guidance_scale', |
| 362 | 'mirostat_mode', |
| 363 | 'mirostat_tau', |
| 364 | 'mirostat_eta', |
| 365 | 'max_new_tokens', |
| 366 | 'do_sample', |
| 367 | 'dynamic_temperature', |
| 368 | 'temperature_last', |
| 369 | 'dry_sequence_breakers', |
nothing calls this directly
no test coverage detected