Functions — github.com/AndrewWTY/UNIT

hub / github.com/AndrewWTY/UNIT / functions

Functions113 in github.com/AndrewWTY/UNIT

⨍ Functions 113◇ Types & classes 12

↓ 6 callersMethodclose

Close the connection to the database.

src/factscore_retrieval.py:43

↓ 5 callersFunctionis_info

(d)

src/make_data_ref.py:127

↓ 4 callersFunctionformat_prompt

(tokenizer, system, input, no_system=False)

src/vllm_inference_for_ccp.py:95

↓ 3 callersFunctionbatchify

Split the list `lst` into sublists of size `batch_size`.

src/compute_ccp.py:28

↓ 3 callersMethodget_passages

(self, topic, question, k)

src/wild_retrieval.py:156

↓ 2 callersFunctionask_open_ai

(args, prompts, sub_cache_name, batch_result, system_msg="You are a helpful AI assistant.")

src/extract_sentence_facts.py:35

↓ 2 callersFunctionbatchify

Split the list `lst` into sublists of size `batch_size`.

src/factcheck.py:226

↓ 2 callersFunctioncompute_ranked_error_four_lists

(incorrect_in_reflection, correct_in_reflection, incorrect_in_unreflection, correct_in_unreflection, threshold

src/evaluate_calibration.py:292

↓ 2 callersFunctioncreate_answers_async

(model, messages, cache_path, batch_size=5, seed=42, temperature=0)

src/factcheck.py:344

↓ 2 callersFunctioncreate_answers_async

(model, messages, cache_path, generation_args, batch_size=5)

src/make_data_cut.py:191

↓ 2 callersFunctioncreate_answers_async

Send prompts in parallel batches (with simple on-disk caching).

src/instruction_classification.py:69

↓ 2 callersFunctionextract_sorting_response

(input_text)

src/evaluate_calibration.py:215

↓ 2 callersFunctionfind_quantile

(x, data)

src/evaluate_calibration.py:254

↓ 2 callersMethodget_all_passages

(self, entity)

src/wild_retrieval.py:103

↓ 2 callersFunctionget_input_price

(input_str, model)

src/factcheck.py:63

↓ 2 callersFunctionget_input_price

(model, input_len=None)

src/make_data_cut.py:95

↓ 2 callersFunctionget_output_price

(output_str, model)

src/factcheck.py:70

↓ 2 callersFunctionget_output_price

(model, output_len=None)

src/make_data_cut.py:100

↓ 2 callersFunctionload_jsonl

Loads a JSONL file into a list of dictionaries. Args: file_path: The path to the JSONL file. Returns: A list of dictionaries

src/evaluate_calibration.py:109

↓ 2 callersFunctionranked_error_for_group

(values)

src/evaluate_calibration.py:310

↓ 2 callersFunctionsave_jsonl

(rows, path)

src/make_data_ref.py:63

↓ 2 callersFunctiontoken_price

(txt, model, price_dict)

src/instruction_classification.py:94

↓ 2 callersFunctiontransform_data

(rows, all_info_seeking=False, no_surgery=False)

src/make_data_ref.py:235

↓ 1 callersMethod_async_send_request

(self, client, message)

src/factcheck.py:192

↓ 1 callersMethod_async_send_request

(self, client, message)

src/compute_helpfulness.py:183

↓ 1 callersMethod_format_prompt

(self, system, input_text, no_system=False)

src/inference_pipeline.py:61

↓ 1 callersMethod_initialize_database

(self)

src/wild_retrieval.py:69

↓ 1 callersMethod_send_request

(self, messages)

src/factcheck.py:201

↓ 1 callersMethod_send_request

(self, messages)

src/compute_helpfulness.py:192

↓ 1 callersMethod_send_request

(self, messages)

src/make_data_cut.py:155

↓ 1 callersMethod_tokenize_and_store

(self, data)

src/wild_retrieval.py:83

↓ 1 callersFunctionachat

(client, model, message, seed=42, temperature=0)

src/factcheck.py:334

↓ 1 callersFunctionachat

(model, messages, generation_args)

src/make_data_cut.py:175

↓ 1 callersFunctionachat

(client, model, msg, seed=42, temperature=0)

src/instruction_classification.py:63

↓ 1 callersFunctionadd_certainty_fields

Adds 'info-seeking', 'certain_claims', 'uncertain_claims' columns.

src/make_data_ref.py:169

↓ 1 callersFunctionalign

Identifies token indices in `sent_tokens` that align with matching characters (marked by '^') in `match_str`. All tokens, which textual rep

src/compute_ccp.py:185

↓ 1 callersFunctionalign_facts

Given target output texts, and extracted facts from the output texts. Align the tokens of extracted facts to tokens in output texts.

src/compute_ccp.py:232

↓ 1 callersFunctionapi_call

()

src/factcheck.py:492

↓ 1 callersFunctionappend_exceeding_claims_with_confidence

Attach a 'surgery_response' that includes reflection text.

src/make_data_ref.py:193

↓ 1 callersMethodask

(self, message: str)

src/compute_helpfulness.py:96

↓ 1 callersFunctionask_open_ai

(args, prompts, batch_result, sub_cache_name, system_prompt=SYSTEM_PROMPT_JUDGE, real_time=False)

src/compute_helpfulness.py:222

↓ 1 callersMethodasync_ask

(self, message_list)

src/factcheck.py:138

↓ 1 callersFunctionbatchify

Split the list `lst` into sublists of size `batch_size`.

src/make_data_cut.py:186

↓ 1 callersFunctionbatchify

(lst, n)

src/instruction_classification.py:59

↓ 1 callersFunctionbatchify_claims

(extraction_results, output_texts, batch_size)

src/compute_ccp.py:33

↓ 1 callersMethodbuild_db

(self, db_path, data_path)

src/factscore_retrieval.py:47

↓ 1 callersFunctioncapitalize_first_letter

(input_string)

src/vllm_inference_for_ccp.py:109

↓ 1 callersFunctionextract_and_plot_ccp

(data)

src/make_data_ref.py:81

↓ 1 callersFunctionextract_claims_ccp

(claim_uncertainty_text)

src/evaluate_calibration.py:238

↓ 1 callersFunctionextract_facts

(args, output_texts)

src/extract_sentence_facts.py:199

↓ 1 callersFunctionextract_reflections

Extracts reflections from the response text after the <reflection> token. Strips numbering from each reflection.

src/evaluate_calibration.py:189

↓ 1 callersFunctionfilter_lora

(lora_path)

src/inference_pipeline.py:15

↓ 1 callersMethodgenerate_responses

Generates responses based on provided sampling parameters.

src/inference_pipeline.py:51

↓ 1 callersFunctionget_args

()

src/vllm_inference_for_ccp.py:117

↓ 1 callersMethodget_bm25_passages

(self, topic, query, passages, k)

src/factscore_retrieval.py:171

↓ 1 callersMethodget_bm25_passages

(self, entity, query_sentence, top_k=5)

src/wild_retrieval.py:112

↓ 1 callersMethodget_gtr_passages

(self, topic, retrieval_query, passages, k)

src/factscore_retrieval.py:182

↓ 1 callersMethodget_gtr_passages

(self, entity, query_sentence, top_k=5)

src/wild_retrieval.py:130

↓ 1 callersFunctionget_input_price

(input_str, model)

src/compute_helpfulness.py:54

↓ 1 callersFunctionget_output_price

(output_str, model)

src/compute_helpfulness.py:61

↓ 1 callersMethodget_passages

(self, topic, question, k)

src/factscore_retrieval.py:199

↓ 1 callersFunctionget_ranks

(values, sorted_list)

src/evaluate_calibration.py:307

↓ 1 callersMethodget_text_from_title

Fetch the raw text of the doc for 'doc_id'.

src/factscore_retrieval.py:100

↓ 1 callersFunctionis_info

(d)

src/make_data_cut.py:260

↓ 1 callersFunctionlabel_quantiles

Add booleans (ccp_above_qXX / max_prob_above_qXX) to each uncertainty block. Returns (new_data, ccp_q75).

src/make_data_ref.py:124

↓ 1 callersFunctionload_datasets

(file_paths, question_columns)

src/inference_pipeline.py:74

↓ 1 callersMethodload_encoder

(self)

src/factscore_retrieval.py:132

↓ 1 callersMethodload_encoder

(self)

src/wild_retrieval.py:96

↓ 1 callersFunctionload_jsonl

(path)

src/make_data_ref.py:52

↓ 1 callersFunctionmain

()

src/factcheck.py:391

↓ 1 callersFunctionmain

(file, setting)

src/compute_truthfulness.py:5

↓ 1 callersFunctionmain

()

src/compute_helpfulness.py:324

↓ 1 callersFunctionmain

()

src/evaluate_calibration.py:320

↓ 1 callersFunctionmain

()

src/make_data_cut.py:244

↓ 1 callersFunctionmain

()

src/get_atomic_claims.py:27

↓ 1 callersFunctionmain

(file, setting)

src/compute_honesty_upper_bound.py:6

↓ 1 callersFunctionmain

(config)

src/inference_pipeline.py:130

↓ 1 callersFunctionmain

(args)

src/vllm_inference_for_ccp.py:134

↓ 1 callersFunctionmain

()

src/make_data_ref.py:295

↓ 1 callersFunctionmatch_string

Greedily matching words from `match_words` to `sent`. Parameters: sent (str): sentence string match_words (List[str]): li

src/extract_sentence_facts.py:138

↓ 1 callersFunctionmerge_configs

(config, args, parser)

src/inference_pipeline.py:122

↓ 1 callersFunctionparse_args

()

src/make_data_ref.py:15

↓ 1 callersFunctionprint_claims

(claims, number=True)

src/make_data_cut.py:105

↓ 1 callersFunctionremove_entries_with_none_response

(data)

src/make_data_ref.py:71

Method__call__

(self, system_prompt, user_prompt, max_tokens = 300, temperature = 0.7, response_format = { "type": "text" })

src/evaluate_calibration.py:27

Method__call__

Calculates the statistics of probabilities at each token position in the output text. Parameters: input_texts (List[

src/compute_ccp.py:75

Method__call__

Calculates the statistics of probabilities at each token position in the output text. Parameters: input_texts (List[

src/vllm_inference_for_ccp.py:38

Method__enter__

(self)

src/factscore_retrieval.py:33

Method__exit__

(self, *args)

src/factscore_retrieval.py:36

Method__init__

Parameters ---------- openai_model: str the model to use in OpenAI to chat.

src/factcheck.py:82

Method__init__

Parameters ---------- openai_model: str the model to use in OpenAI to chat.

src/compute_helpfulness.py:73

Method__init__

(self, model)

src/evaluate_calibration.py:23

Method__init__

( self, model_name: str = None, cache_path: str = "litellm_cache",

src/make_data_cut.py:117

Method__init__

(self, n_alternatives: int = 10)

src/compute_ccp.py:72

Method__init__

(self, db_path=None, data_path=None)

src/factscore_retrieval.py:21

Method__init__

(self, db, cache_path, embed_cache_path, retrieval_type="gtr-t5-large", batch_size=None)

src/factscore_retrieval.py:116

Method__init__

(self, model_type, llm, tokenizer)

src/inference_pipeline.py:45

Method__init__

(self, n_alternatives: int = 10)

src/vllm_inference_for_ccp.py:35

Method__init__

(self, cache_path, batch_size, embed_cache_path, db_path='factcheck_cache/wildhallu.db', retrieval_type="gtr-t

src/wild_retrieval.py:15

Methodask

(self, message: str)

src/factcheck.py:105

next →1–100 of 113, ranked by callers