hpcaitech · ver217 · Jul 1, 2024 · Jun 24, 2024 · Jun 24, 2024
@@ -1,34 +1,34 @@
 repos:
 
   - repo: https://github.com/PyCQA/autoflake
-    rev: v2.2.1
+    rev: v2.3.1
     hooks:
       - id: autoflake
         name: autoflake (python)
         args: ['--in-place', '--remove-unused-variables', '--remove-all-unused-imports', '--ignore-init-module-imports']
 
   - repo: https://github.com/pycqa/isort
-    rev: 5.12.0
+    rev: 5.13.2
     hooks:
       - id: isort
         name: sort all imports (python)
 
   - repo: https://github.com/psf/black-pre-commit-mirror
-    rev: 23.9.1
+    rev: 24.4.2
     hooks:
     - id: black
       name: black formatter
       args: ['--line-length=120', '--target-version=py37', '--target-version=py38', '--target-version=py39','--target-version=py310']
 
   - repo: https://github.com/pre-commit/mirrors-clang-format
-    rev: v13.0.1
+    rev: v18.1.7
     hooks:
     - id: clang-format
       name: clang formatter
       types_or: [c++, c]
 
   - repo: https://github.com/pre-commit/pre-commit-hooks
-    rev: v4.3.0
+    rev: v4.6.0
     hooks:
       - id: check-yaml
       - id: check-merge-conflict

@@ -83,15 +83,19 @@ def __call__(self, instances: Sequence[Dict[str, List[int]]]) -> Dict[str, torch
 
         # `List[torch.Tensor]`
         batch_input_ids = [
-            torch.LongTensor(instance["input_ids"][: self.max_length])
-            if len(instance["input_ids"]) > self.max_length
-            else torch.LongTensor(instance["input_ids"])
+            (
+                torch.LongTensor(instance["input_ids"][: self.max_length])
+                if len(instance["input_ids"]) > self.max_length
+                else torch.LongTensor(instance["input_ids"])
+            )
             for instance in instances
         ]
         batch_labels = [
-            torch.LongTensor(instance["labels"][: self.max_length])
-            if len(instance["labels"]) > self.max_length
-            else torch.LongTensor(instance["labels"])
+            (
+                torch.LongTensor(instance["labels"][: self.max_length])
+                if len(instance["labels"]) > self.max_length
+                else torch.LongTensor(instance["labels"])
+            )
             for instance in instances
         ]
         if self.tokenizer.padding_side == "right":

@@ -1,6 +1,7 @@
 """
 loss functions
 """
+
 from typing import Optional, Tuple
 
 import torch

@@ -1,6 +1,7 @@
 """
 reward model
 """
+
 from typing import Optional
 
 import torch

@@ -1,6 +1,7 @@
 """
 Training utilities for Coati.
 """
+
 from typing import Any
 
 import torch

@@ -78,7 +78,9 @@ def get_prompt(line: Dict, dataset_name: str, logger: DistributedLogger) -> Dict
             option_string = "ABCDEFG"
             count = len(line["options"])
 
-            input = "问题：" + line["question"] + " " + "从以下选项中选择：" + " ".join(line["options"]) + "\n" + "答案："
+            input = (
+                "问题：" + line["question"] + " " + "从以下选项中选择：" + " ".join(line["options"]) + "\n" + "答案："
+            )
 
             all_classes = list(option_string[0:count])
 
@@ -150,7 +152,15 @@ def combine_prompt(prompt_path, dataset_name, load_explanation=True, chat_mode=F
             )
         elif dataset_name in chinese_qa_datasets:
             question_input = (
-                "问题：" + passage + " " + question + "\n" + "从以下选项中选择：" + " ".join(options) + "\n" + "答案：{}".format(label)
+                "问题："
+                + passage
+                + " "
+                + question
+                + "\n"
+                + "从以下选项中选择："
+                + " ".join(options)
+                + "\n"
+                + "答案：{}".format(label)
             )
         elif dataset_name in english_cloze_datasets:
             question_input = "Question: ".format(idx + 1) + question + "\n" + "Answer: {}".format(answer)

@@ -57,7 +57,11 @@
     "urban_and_rural_planner": ["Urban and Rural Planner", "注册城乡规划师", "Other"],
     "accountant": ["Accountant", "注册会计师", "Other"],
     "fire_engineer": ["Fire Engineer", "注册消防工程师", "Other"],
-    "environmental_impact_assessment_engineer": ["Environmental Impact Assessment Engineer", "环境影响评价工程师", "Other"],
+    "environmental_impact_assessment_engineer": [
+        "Environmental Impact Assessment Engineer",
+        "环境影响评价工程师",
+        "Other",
+    ],
     "tax_accountant": ["Tax Accountant", "税务师", "Other"],
     "physician": ["Physician", "医师资格", "Other"],
 }

@@ -56,9 +56,11 @@ def load(path: str, logger: DistributedLogger, few_shot: bool) -> List[Dict]:
                     "instruction": question["turns"],
                     "input": "",
                     "output": [],
-                    "target": [""] * turn_number
-                    if question["question_id"] not in reference
-                    else reference[question["question_id"]],
+                    "target": (
+                        [""] * turn_number
+                        if question["question_id"] not in reference
+                        else reference[question["question_id"]]
+                    ),
                 }
 
                 if category in dataset["test"]:

@@ -77,7 +77,9 @@ def _get_choices_indices(self, language: str):
             self.indices_for_choices[0].append(
                 self.tokenizer(f"Answer: {choice}", add_special_tokens=False).input_ids[-1]
             )
-            self.indices_for_choices[1].append(self.tokenizer(f"答案：{choice}", add_special_tokens=False).input_ids[-1])
+            self.indices_for_choices[1].append(
+                self.tokenizer(f"答案：{choice}", add_special_tokens=False).input_ids[-1]
+            )
 
     def _load_tokenizer(self, path: str, tokenizer_path: Optional[str], tokenizer_kwargs: dict):
         """

@@ -7,6 +7,7 @@
 https://github.com/langchain-ai/langchain
 The original code is licensed under the MIT license.
 """
+
 from __future__ import annotations
 
 import copy

@@ -8,6 +8,7 @@
 https://github.com/langchain-ai/langchain
 The original code is licensed under the MIT license.
 """
+
 import copy
 from typing import Any, Mapping, Optional, Protocol
 

@@ -7,6 +7,7 @@
 https://github.com/langchain-ai/langchain
 The original code is licensed under the MIT license.
 """
+
 import copy
 from typing import Any, List
 

@@ -2,7 +2,6 @@
 Class for loading table type data. please refer to Pandas-Input/Output for file format details.
 """
 
-
 import glob
 import os
 

@@ -20,6 +20,7 @@
 print(resp)  # super-heavyweight awesome-natured yawning Australian creature!
 
 """
+
 import json
 from typing import Any, Mapping
 

@@ -12,6 +12,7 @@
 logger.info(llm(TEST_PROMPT_CHATGLM, max_new_tokens=100), verbose=True)
 
 """
+
 from typing import Any, List, Mapping, Optional
 
 import torch

@@ -1,6 +1,7 @@
 """
 Generation utilities
 """
+
 import json
 from typing import List
 

@@ -2,6 +2,7 @@
 Implement a memory class for storing conversation history
 Support long term and short term memory
 """
+
 from typing import Any, Dict, List
 
 from colossalqa.chain.memory.summary import ConversationSummaryMemory

@@ -1,6 +1,7 @@
 """
 Class for logging with extra control for debugging
 """
+
 import logging
 
 

@@ -1,6 +1,7 @@
 """
 Script for Chinese retrieval based conversation system backed by ChatGLM
 """
+
 from typing import Tuple
 
 from colossalqa.chain.retrieval_qa.base import RetrievalQA

@@ -1,6 +1,7 @@
 """
 Multilingual retrieval based conversation system
 """
+
 from typing import List
 
 from colossalqa.data_loader.document_loader import DocumentLoader

@@ -1,6 +1,7 @@
 """
 Script for Chinese retrieval based conversation system backed by ChatGLM
 """
+
 from typing import Tuple
 
 from colossalqa.chain.retrieval_qa.base import RetrievalQA

@@ -1,6 +1,7 @@
 """
 Code for custom retriver with incremental update
 """
+
 import copy
 import hashlib
 import os

@@ -1,6 +1,7 @@
 """
 Code for Chinese text splitter
 """
+
 from typing import Any, List, Optional
 
 from colossalqa.text_splitter.utils import get_cleaned_paragraph

@@ -1,6 +1,7 @@
 """
 Script for English retrieval based conversation system backed by LLaMa2
 """
+
 import argparse
 import os
 

@@ -1,6 +1,7 @@
 """
 Script for English retrieval based conversation system backed by LLaMa2
 """
+
 import argparse
 import json
 import os

@@ -1,6 +1,7 @@
 """
 Script for Chinese retrieval based conversation system backed by ChatGLM
 """
+
 import argparse
 import os
 

@@ -1,6 +1,7 @@
 """
 Script for English retrieval based conversation system backed by LLaMa2
 """
+
 import argparse
 import os
 

@@ -107,20 +107,22 @@ def convnd_meta_info(*args, **kwargs) -> Tuple[TrainCycleItem, TrainCycleItem, L
     # NOTE: currently in SPMD solver we always believe that there will be a new tensor created in forward
     fwd_memory_cost = MemoryCost(
         activation=compute_size_in_bytes([input_tensor, output_tensor]),
-        parameter=compute_size_in_bytes([weight_tensor, bias_tensor])
-        if has_bias
-        else compute_size_in_bytes(weight_tensor),
+        parameter=(
+            compute_size_in_bytes([weight_tensor, bias_tensor]) if has_bias else compute_size_in_bytes(weight_tensor)
+        ),
         temp=0,
         buffer=0,
     )
 
     bwd_memory_cost = MemoryCost(
-        activation=compute_size_in_bytes([input_tensor, weight_tensor, bias_tensor])
-        if has_bias
-        else compute_size_in_bytes([input_tensor, weight_tensor]),
-        parameter=compute_size_in_bytes([weight_tensor, bias_tensor])
-        if has_bias
-        else compute_size_in_bytes(weight_tensor),
+        activation=(
+            compute_size_in_bytes([input_tensor, weight_tensor, bias_tensor])
+            if has_bias
+            else compute_size_in_bytes([input_tensor, weight_tensor])
+        ),
+        parameter=(
+            compute_size_in_bytes([weight_tensor, bias_tensor]) if has_bias else compute_size_in_bytes(weight_tensor)
+        ),
         temp=0,
         buffer=0,
     )

@@ -247,16 +247,16 @@ def add_seqs(
                 self._sequences_dict[seq.request_id] = seq
                 self._sequences_indexes[seq.request_id] = self._current_batch_size + i
             # TODO external (rename): modify Sequence.sentence_len to seq_len
-            self._sequence_lengths[
-                self._current_batch_size : self._current_batch_size + num_seqs_to_add
-            ] = torch.tensor([seq.sentence_len for seq in seqs[:num_seqs_to_add]], dtype=torch.int32)
+            self._sequence_lengths[self._current_batch_size : self._current_batch_size + num_seqs_to_add] = (
+                torch.tensor([seq.sentence_len for seq in seqs[:num_seqs_to_add]], dtype=torch.int32)
+            )
             # NOTE block tables to be updated by kvcache manager
             block_tables = self._block_tables[self._current_batch_size : self._current_batch_size + num_seqs_to_add]
             if alloc_block_tables is not None:
                 # copy block ids from provided block tables
-                self._block_tables[
-                    self._current_batch_size : self._current_batch_size + num_seqs_to_add
-                ] = alloc_block_tables
+                self._block_tables[self._current_batch_size : self._current_batch_size + num_seqs_to_add] = (
+                    alloc_block_tables
+                )
             elif alloc_block_tables_fn:
                 alloc_block_tables_fn(
                     block_tables,
-Original file line number
+Diff line change
@@ Expand Up / @@ -2,7 +2,6 @@ @@
     Class for loading table type data. please refer to Pandas-Input/Output for file format details.
     """
     import glob
     import os
@@ Expand Down @@