EasyNLP发布融合语言学和事实知识的中文预训练模型CKBERT

def ltp_process(ltp: LTP, 
                data: List[Dict[str, Union[str, List[Union[int, str]]]]]):
    """use ltp to process the data
    Args:
        Dict ([str, str]): data
        example:
            {'text':['我叫汤姆去拿伞。'],...}
    Returns:
        Dict[str, str]: result
    """
    new_data = list(map(lambda x:x['text'][0].replace(" ", ""), data))
    seg, hiddens = ltp.seg(new_data)
    result = {}
    result['seg'] = seg
    result['ner'] = ltp.ner(hiddens)
    result['dep'] = ltp.dep(hiddens)
    result['sdp'] = ltp.sdp(hiddens)
    for index in range(len(data)):
        data[index]['text'][0] = data[index]['text'][0].replace(" ", "")
        data[index]['seg'] = result['seg'][index]
        data[index]['ner'] = result['ner'][index]
        data[index]['dep'] = result['dep'][index]
        data[index]['sdp'] = result['sdp'][index]

def dep_sdp_mask(left_numbers: List[int], 
                     data_new: List[List[Union[int, str]]], 
                     markers_: List[List[int]], 
                     selected_numbers_: set, 
                     number_: int,
                     marker_attrs: Dict[str, List[int]]) -> int:
        """ mask the `mask_labels` for sdp and dep and record the maskers for each mask item
        Args:
            left_numbers (List[int]): the options that have not been used
            data_new (List[List[Union[int, str]]]): preprocessed data for original dep and sdp
            markers_ (List[List[int]]): a list that is uesd to save the maskers for each mask item
            selected_numbers_ (set): a set that is used to save the selected options
            number_ (int): the number of mask labels
            marker_attrs Dict[str, List[int]]: marker attributes
        Returns:
            int: 0 mean no mask, the others mean the number of masked ids
        """
        np.random.shuffle(left_numbers)
        for item_ in left_numbers:
            target_item = data_new[item_]
            seg_ids = np.array(target_item[:2]) - 1
            delete_ids = np.where(seg_ids < 1)[0]
            seg_ids = np.delete(seg_ids, delete_ids)
            temp_ids = seg2id(seg_ids)
            ids = []
            for item in temp_ids:
                ids += item.copy()
            if check_ids(ids):
                length_ = len(ids)
                if number_ > length_:
                    for id_ in ids:
                        mask_labels[id_] = 1
                    if target_item[2] in marker_attrs:
                        detail_info.append([
                            target_item,
                            [seg_data[target_item[0] - 1],seg_data[target_item[1] - 1]],
                        ])
                        if len(temp_ids) == 1:
                            markers_.append([temp_ids[0][0], temp_ids[0][-1]])
                        elif len(temp_ids) == 2:
                            for i in marker_attrs[target_item[2]]:
                                markers_.append([temp_ids[i][0], temp_ids[i][-1]])
                    selected_numbers_.add(item_)
                    return length_
                else:
                    return 0
        return 0

def get_positive_and_negative_examples(
    self,
    ner_data: str,
    negative_level: int = 3) -> Union[bool, Dict[str, List[str]]]:
    """get the positive examples and negative examples for the ner data
    Args:
        ner_data (str): the ner entity
        negative_level (int, optional): the deepth of the relationship. Defaults to 3.
    Returns:
        Union[bool, Dict[str, List[str]]]: if the `ner_data` not in `konwledge`, return False, otherwise, return the positive and negative examples
    """
    knowledge: Dict[str, Dict[str, str]] = self.Knowledge_G
    common_used = set()
    def get_data(key: str, 
                data: Dict[str, str], 
                results: List[str], 
                deep: int, 
                insert_flag: bool = False):
        """get the negative examples recursively
        Args:
            key (str): the ner
            data (Dict[str, str]): the related data about `key`
            results (List[str]): a list used to save the negative examples
            deep (int): the recursive number
            insert_flag (bool, optional): whether insert data to `results`. Defaults to False.
        """
        nonlocal knowledge
        common_used.add(key)
        if deep == 0:
            return
        else:
            for key_item in data:
                if data[key_item] not in common_used and insert_flag == True:
                    results.append(data[key_item])
                if data[key_item] in knowledge and data[key_item] not in common_used:
                    get_data(data[key_item], knowledge[data[key_item]], results, deep - 1, True)
    all_examples = {
        'ner': ner_data,
        'positive_examples': [],
        'negative_examples': []
    }
    if ner_data in knowledge:
        tp_data = knowledge[ner_data]
        negative_examples = []
        if '描述' in tp_data:
            positive_example = tp_data['描述']
        else:
            keys = list(tp_data.keys())
            choice = np.random.choice([_ for _ in range(len(keys))], 1)[0]
            positive_example = tp_data[keys[choice]]
        # # the description usually contains the ner entity, if not, concate the `ner_data` and the positive example
        if ner_data in positive_example:
            all_examples['positive_examples'].append(positive_example)
        else:
            all_examples['positive_examples'].append(ner_data + positive_example)
        get_data(ner_data, tp_data, negative_examples, negative_level)
        # concate the ner entity and each negative example
        negative_examples = list(map(lambda x: ner_data + x if ner_data not in x else x, negative_examples))
        all_examples['negative_examples'] = negative_examples
        return all_examples
    return False

model.backbone.resize_token_embeddings(len(train_dataset.tokenizer))
model.config.vocab_size = len(train_dataset.tokenizer)

def compute_simcse(self, original_outputs: torch.Tensor, 
                   forward_outputs: torch.Tensor) -> float:
        original_hidden_states = original_outputs['hidden_states'].unsqueeze(-2)
        loss = nn.CrossEntropyLoss()
        forward_outputs = torch.mean(forward_outputs, dim=-2)
        cos_result = self.CosSim(original_hidden_states, forward_outputs)
        cos_result_size = cos_result.size()
        cos_result = cos_result.view(-1, cos_result_size[-1])
        labels = torch.zeros(cos_result.size(0), device=original_outputs['hidden_states'].device).long()
        loss_ = loss(cos_result, labels)
        return loss_

{'text': ['我想,如果我没有去做大学生村官,恐怕我这个在昆明长大的孩子永远都不能切身感受到云南这次60年一遇的特大旱情的严重性,恐怕我只是每天看着新闻上那些缺水的镜头,嘴上说要节水,但事实行动保持不了三天。 我任职的地方在昆明市禄劝县的一个村委会,说实话这里距离禄劝县城不远,自然环境不算很差。目前,只有一个自然村保证不了饮用水。一个自然村基本能保证有饮用水到5月。这里所说的饮用水,是指从山肚子里出来的水,积在小水坝或是水塘里又通过管道输送到村子里的水,和我们城市里真正意义上消过毒的、能安全饮用的饮用水不同。在整个输送的过程中,可能已经产生了有害物质。我觉得是。 没有饮用水的那个自然村叫大海子村,50户,近200多人。地处山头,交通很不便利,走路大概要1个半小时到两个小时,而且坡度比较大,是一个苗族村寨。地理条件限制,基本没有什么经济作物,算是靠天吃饭的那种。今年遇到60年一遇的干旱,村里的两个水窖都基本干了,之前几天,他们村长来反映,几个老人已经是抬个小板凳坐到窖底用碗舀水了。 面对这么严峻的旱情,村委会的领导和各小组长都在想办法。但是上山的路路面情况差,大车重车上不去；周边水源地少。最可行的办法就是从武定那边绕路上去。但每天运水上去也不是办法,长远来看还是要修建一个小水坝。村委会的领导主动捐款,村民也自行筹资,开始自救。 最近每个周末都回家,添置防晒品,因为基本每天都上山下村去了解情况,必须掌握辖区内13个村小组水资源的情况。我每次回家见到朋友们,第一句就是,要节约用水啊~~ 朋友们,你们现在看到的只是简单理解的"缺水"。你们所不知道的是,没水小春作物面临绝收、4月份插秧没有水泡秧田、5月份种烤烟也没有水。。。那么对农民就意味着今年一年就没有了收入。我们现在能努力做好的,只是保证人的饮用水。 上周就在想能不能通过什么渠道帮村民们做点事。叔叔叫我弄个抗旱的基金,他发动周围的朋友来捐赠,希望能口口相传带动多一点朋友。我正在筹备阶段,看怎样才能做到最大的公开透明,让捐赠的人完全放心。 周一接到一个朋友的电话,说他们公司想为旱灾献点爱心,想买点水送去我们那儿。昨天见了负责人,很谦和的一个姐姐,她说大家都想做点什么,觉得捐钱没什么意义,想亲自送水到最需要的地方。 其实人家只是家私营的小公司,但我真的很感谢他们。姐姐还特别交代我,我们只需要找拖拉机下来帮忙把水运上山去,其他的什么都不用管,他们会安排好的。这个周末,将有 400件矿泉水会送到村民家里。我想,应该可以暂时缓解旱情。再次代村民感谢他们！ 下半年,旱情给农民的生产、生活带来的问题还很多。但是我个人的力量很有限,希望能够看到这些帖子的朋友们,如果有能力,有这份心意的话,请给予旱灾地区的农民更多的帮助。 我想大家都知道,昆明80%以上的用水都来自禄劝的云龙水库,云龙的同事们也在"抗旱",他们的工作任务是要保证严格的节约用水,要寻求其他水源用水,从而保证昆明的用水。所以,请每一个昆明人都节水吧,禄劝的很多地方都在缺水,我们那里不算严重的,请珍惜你们现在在用的每一滴水~ 也许,要经历过这样一次触目惊心的大旱才真正知道水的珍贵。希望我们都行动起来,不要再让这样的旱灾侵袭我们的家乡。'], 'title': '旱情记要-----昆明人,请珍惜你们现在在用的每一滴水~'}

红色食品  标签  生活

{"text": ["我想,如果我没有去做大学生村官,恐怕我这个在昆明长大的孩子永远都不能切身感受到云南这次60年一遇的特大旱情的严重性,恐怕我只是每天看着新闻上那些缺水的镜头,嘴上说要节水,但事实行动保持不了三天。我任职的地方在昆明市禄劝县的一个村委会,说实话这里距离禄劝县城不远,自然环境不算很差。目前,只有一个自然村保证不了饮用水。一个自然村基本能保证有饮用水到5月。这里所说的饮用水,是指从山肚子里出来的水,积在小水坝或是水塘里又通过管道输送到村子里的水,和我们城市里真正意义上消过毒的、能安全饮用的饮用水不同。在整个输送的过程中,可能已经产生了有害物质。我觉得是。没有饮用水的那个自然村叫大海子村,50户,近200多人。地处山头,交通很不便利,走路大概要1个半小时到两个小时,而且坡度比较大,是一个苗族村寨。地理条件限制,基本没有什么经济作物,算是靠天吃饭的那种。今年遇到60年一遇的干旱,村里的两个水窖都基本干了,之前几天,他们村长来反映,几个老人已经是抬个小板凳坐到窖底用碗舀水了。面对这么严峻的旱情,村委会的领导和各小组长都在想办法。但是上山的路路面情况差,大车重车上不去；周边水源地少。最可行的办法就是从武定那边绕路上去。但每天运水上去也不是办法,长远来看还是要修建一个小水坝。村委会的领导主动捐款,村民也自行筹资,开始自救。最近每个周末都回家,添置防晒品,因为基本每天都上山下村去了解情况,必须掌握辖区内13个村小组水资源的情况。我每次回家见到朋友们,第一句就是,要节约用水啊~~朋友们,你们现在看到的只是简单理解的\"缺水\"。你们所不知道的是,没水小春作物面临绝收、4月份插秧没有水泡秧田、5月份种烤烟也没有水。。。那么对农民就意味着今年一年就没有了收入。我们现在能努力做好的,只是保证人的饮用水。上周就在想能不能通过什么渠道帮村民们做点事。叔叔叫我弄个抗旱的基金,他发动周围的朋友来捐赠,希望能口口相传带动多一点朋友。我正在筹备阶段,看怎样才能做到最大的公开透明,让捐赠的人完全放心。周一接到一个朋友的电话,说他们公司想为旱灾献点爱心,想买点水送去我们那儿。昨天见了负责人,很谦和的一个姐姐,她说大家都想做点什么,觉得捐钱没什么意义,想亲自送水到最需要的地方。其实人家只是家私营的小公司,但我真的很感谢他们。姐姐还特别交代我,我们只需要找拖拉机下来帮忙把水运上山去,其他的什么都不用管,他们会安排好的。这个周末,将有400件矿泉水会送到村民家里。我想,应该可以暂时缓解旱情。再次代村民感谢他们！下半年,旱情给农民的生产、生活带来的问题还很多。但是我个人的力量很有限,希望能够看到这些帖子的朋友们,如果有能力,有这份心意的话,请给予旱灾地区的农民更多的帮助。我想大家都知道,昆明80%以上的用水都来自禄劝的云龙水库,云龙的同事们也在\"抗旱\",他们的工作任务是要保证严格的节约用水,要寻求其他水源用水,从而保证昆明的用水。所以,请每一个昆明人都节水吧,禄劝的很多地方都在缺水,我们那里不算严重的,请珍惜你们现在在用的每一滴水~也许,要经历过这样一次触目惊心的大旱才真正知道水的珍贵。希望我们都行动起来,不要再让这样的旱灾侵袭我们的家乡。"], "title": "旱情记要-----昆明人,请珍惜你们现在在用的每一滴水~", "seg": ["我", "想", ",", "如果", "我", "没有", "去", "做", "大学生村官", ",", "恐怕", "我", "这个", "在", "昆明长大", "的", "孩子", "永远", "都", "不能切身感受到", "云南", "这次", "60年", "一遇的", "特大旱情", "的", "严重性", ",", "恐怕", "我", "只是", "每天", "看", "着", "新闻上", "那些", "缺水", "的", "镜头", ",", "嘴上说要节水", ",", "但事实行动", "保持不了", "三天", "。", "我", "任职", "的", "地方", "在", "昆明市禄劝县", "的", "一个", "村委会", ",", "说实话", "这里", "距离禄", "劝县城不远", ",", "自然环境", "不算", "很差", "。", "目前", ",", "只有", "一个", "自然村", "保证不了", "饮用水", "。", "一个", "自然村", "基本", "能", "保证", "有", "饮用水", "到", "5月", "。", "这里所说", "的", "饮用水", ",", "是指", "从", "山肚子里", "出来", "的", "水", ",积在", "小水坝", "或是", "水塘里", "又", "通过", "管道", "输送到", "村子里", "的水", ",", "和", "我们", "城市里", "真正意义上", "消过毒的", "、能安全", "饮用", "的", "饮用水不同", "。", "在", "整个", "输送", "的", "过程", "中", ",", "可能", "已经", "产生", "了", "有害物质", "。", "我", "觉得", "是", "。", "没有", "饮用水", "的", "那", "个", "自然村叫大海子村", ",", "50户", ",近", "200多人", "。地处山头", ",", "交通", "很不便利", ",", "走路", "大概要", "1个半小时到", "两个", "小时", ",", "而且坡度", "比较大", ",", "是", "一个", "苗族村寨", "。地理条件", "限制", ",", "基本", "没有什么经济作物", ",", "算是", "靠天", "吃饭", "的", "那种", "。", "今年", "遇到", "60年一遇", "的", "干旱", ",村里", "的", "两个水窖", "都", "基本", "干", "了", ",之前几天", ",", "他们", "村长来反映", ",", "几个老人", "已经", "是", "抬个小板凳坐到窖底用碗舀水", "了", "。面对", "这么", "严峻的", "旱情", ",", "村委会", "的领导", "和", "各小组长", "都", "在", "想", "办法", "。", "但是上山的", "路路面情况", "差", ",", "大车重车上不去", "；", "周边水源地少", "。", "最", "可行", "的", "办法", "就是", "从武定", "那边", "绕路上去", "。", "但", "每天运水上"], "ner": [["Ns", 14, 14], ["Ns", 20, 20], ["Ns", 51, 51]], "dep": [[1, 2, "SBV"], [2, 0, "HED"], [3, 2, "WP"], [4, 6, "ADV"], [5, 6, "SBV"], [6, 2, "VOB"], [7, 6, "COO"], [8, 7, "COO"], [9, 8, "VOB"], [10, 8, "WP"], [11, 7, "COO"], [12, 172, "SBV"], [13, 172, "ADV"], [14, 172, "ADV"], [15, 14, "POB"], [16, 172, "RAD"], [17, 172, "SBV"], [18, 172, "ADV"], [19, 172, "ADV"], [20, 172, "ADV"], [21, 172, "SBV"], [22, 172, "ADV"], [23, 172, "ADV"], [24, 172, "ADV"], [25, 172, "ADV"], [26, 172, "RAD"], [27, 172, "VOB"], [28, 156, "WP"], [29, 33, "ADV"], [30, 33, "SBV"], [31, 33, "ADV"], [32, 33, "ADV"], [33, 156, "COO"], [34, 33, "RAD"], [35, 39, "ATT"], [36, 39, "ATT"], [37, 39, "ATT"], [38, 37, "RAD"], [39, 33, "VOB"], [40, 33, "WP"], [41, 44, "ADV"], [42, 44, "WP"], [43, 44, "ADV"], [44, 33, "COO"], [45, 44, "CMP"], [46, 44, "WP"], [47, 48, "SBV"], [48, 55, "ATT"], [49, 48, "RAD"], [50, 51, "POB"], [51, 48, "ADV"], [52, 51, "POB"], [53, 48, "RAD"], [54, 55, "ATT"], [55, 44, "SBV"], [56, 44, "WP"], [57, 44, "ADV"], [58, 59, "SBV"], [59, 44, "ADV"], [60, 44, "COO"], [61, 44, "WP"], [62, 63, "SBV"], [63, 71, "CMP"], [64, 71, "CMP"], [65, 71, "WP"], [66, 71, "ADV"], [67, 71, "WP"], [68, 71, "ADV"], [69, 70, "ATT"], [70, 71, "SBV"], [71, 44, "COO"], [72, 71, "COO"], [73, 71, "WP"], [74, 75, "ATT"], [75, 71, "SBV"], [76, 78, "ADV"], [77, 78, "ADV"], [78, 44, "COO"], [79, 78, "VOB"], [80, 78, "VOB"], [81, 80, "CMP"], [82, 81, "POB"], [83, 78, "WP"], [84, 86, "ATT"], [85, 84, "RAD"], [86, 78, "COO"], [87, 78, "WP"], [88, 44, "COO"], [89, 91, "ADV"], [90, 89, "POB"], [91, 93, "ATT"], [92, 91, "RAD"], [93, 101, "VOB"], [94, 101, "WP"], [95, 101, "CMP"], [96, 97, "LAD"], [97, 101, "VOB"], [98, 101, "ADV"], [99, 101, "ADV"], [100, 99, "POB"], [101, 156, "COO"], [102, 101, "SBV"], [103, 101, "RAD"], [104, 101, "WP"], [105, 109, "ADV"], [106, 107, "ATT"], [107, 105, "POB"], [108, 109, "ADV"], [109, 156, "COO"], [110, 111, "WP"], [111, 109, "COO"], [112, 109, "RAD"], [113, 109, "COO"], [114, 129, "WP"], [115, 129, "ADV"], [116, 119, "ATT"], [117, 119, "ATT"], [118, 117, "RAD"], [119, 120, "ATT"], [120, 115, "POB"], [121, 129, "WP"], [122, 124, "ADV"], [123, 124, "ADV"], [124, 129, "COO"], [125, 124, "RAD"], [126, 129, "COO"], [127, 129, "WP"], [128, 129, "SBV"], [129, 109, "COO"], [130, 129, "VOB"], [131, 129, "WP"], [132, 133, "COO"], [133, 109, "COO"], [134, 109, "RAD"], [135, 109, "ADV"], [136, 137, "ATT"], [137, 109, "SBV"], [138, 109, "WP"], [139, 109, "ADV"], [140, 109, "WP"], [141, 109, "ADV"], [142, 109, "WP"], [143, 109, "WP"], [144, 109, "COO"], [145, 109, "ADV"], [146, 156, "WP"], [147, 156, "SBV"], [148, 156, "ADV"], [149, 151, "ATT"], [150, 151, "ATT"], [151, 156, "VOB"], [152, 156, "WP"], [153, 156, "ADV"], [154, 156, "ADV"], [155, 156, "WP"], [156, 167, "COO"], [157, 158, "ATT"], [158, 167, "VOB"], [159, 160, "WP"], [160, 167, "COO"], [161, 160, "WP"], [162, 163, "ADV"], [163, 167, "COO"], [164, 165, "WP"], [165, 163, "COO"], [166, 163, "ADV"], [167, 27, "ATT"], [168, 167, "RAD"], [169, 172, "ADV"], [170, 172, "WP"], [171, 172, "ADV"], [172, 7, "COO"], [173, 175, "ATT"], [174, 175, "RAD"], [175, 172, "VOB"], [176, 175, "WP"], [177, 175, "RAD"], [178, 175, "ATT"], [179, 181, "ADV"], [180, 181, "ADV"], [181, 6, "COO"], [182, 181, "RAD"], [183, 181, "WP"], [184, 181, "WP"], [185, 190, "SBV"], [186, 190, "SBV"], [187, 190, "WP"], [188, 190, "SBV"], [189, 190, "ADV"], [190, 181, "COO"], [191, 190, "VOB"], [192, 191, "RAD"], [193, 191, "WP"], [194, 195, "ADV"], [195, 204, "CMP"], [196, 204, "VOB"], [197, 204, "WP"], [198, 204, "SBV"], [199, 198, "RAD"], [200, 204, "LAD"], [201, 204, "SBV"], [202, 204, "ADV"], [203, 204, "ADV"], [204, 191, "COO"], [205, 204, "VOB"], [206, 204, "WP"], [207, 204, "ADV"], [208, 209, "SBV"], [209, 191, "ADV"], [210, 209, "WP"], [211, 209, "SBV"], [212, 209, "WP"], [213, 209, "SBV"], [214, 209, "WP"], [215, 216, "ADV"], [216, 218, "ATT"], [217, 216, "RAD"], [218, 209, "SBV"], [219, 209, "ADV"], [220, 191, "ADV"], [221, 191, "ADV"], [222, 181, "COO"], [223, 181, "WP"], [224, 181, "ADV"], [225, 181, "ADV"]], "sdp": [[1, 2, "AGT"], [1, 129, "AGT"], [2, 0, "Root"], [3, 2, "mPUNC"], [4, 7, "mRELA"], [5, 7, "AGT"], [5, 8, "AGT"], [6, 7, "mNEG"], [6, 8, "mNEG"], [7, 2, "dCONT"], [8, 7, "eSUCC"], [9, 8, "LINK"], [10, 8, "mPUNC"], [11, 8, "eSUCC"], [12, 172, "EXP"], [13, 172, "SCO"], [14, 15, "mRELA"], [15, 167, "LOC"], [15, 172, "LOC"], [16, 172, "mDEPD"], [17, 172, "EXP"], [18, 172, "mDEPD"], [19, 172, "mDEPD"], [20, 172, "mNEG"], [21, 172, "AGT"], [22, 172, "SCO"], [23, 172, "EXP"], [24, 172, "EXP"], [25, 172, "MANN"], [26, 172, "mDEPD"], [27, 172, "CONT"], [28, 172, "mPUNC"], [29, 33, "mDEPD"], [30, 33, "AGT"], [31, 33, "mDEPD"], [32, 33, "mDEPD"], [33, 172, "eSUCC"], [34, 33, "mDEPD"], [35, 39, "FEAT"], [36, 39, "SCO"], [37, 39, "rEXP"], [38, 37, "mDEPD"], [39, 33, "CONT"], [40, 33, "mPUNC"], [41, 44, "LOC"], [42, 44, "mPUNC"], [43, 44, "mRELA"], [44, 33, "eSUCC"], [45, 44, "TIME"], [46, 44, "mPUNC"], [47, 48, "AGT"], [47, 60, "PAT"], [47, 204, "AGT"], [48, 55, "rDATV"], [49, 48, "mDEPD"], [50, 48, "LOC"], [51, 50, "mRELA"], [52, 50, "FEAT"], [53, 48, "mDEPD"], [54, 55, "MEAS"], [55, 44, "EXP"], [56, 44, "mPUNC"], [57, 44, "eCOO"], [58, 59, "EXP"], [59, 44, "eSUCC"], [60, 44, "eSUCC"], [61, 60, "mPUNC"], [62, 60, "PAT"], [63, 60, "eSUCC"], [64, 63, "mDEPD"], [65, 71, "mPUNC"], [66, 71, "TIME"], [67, 66, "mPUNC"], [68, 71, "mDEPD"], [69, 70, "MEAS"], [70, 71, "AGT"], [71, 60, "eCOO"], [72, 71, "dCONT"], [73, 72, "mPUNC"], [74, 75, "MEAS"], [75, 72, "AGT"], [76, 78, "mDEPD"], [77, 78, "mDEPD"], [78, 44, "eSUCC"], [79, 78, "dCONT"], [80, 79, "LINK"], [81, 79, "eCOO"], [82, 81, "TIME"], [83, 91, "mPUNC"], [84, 91, "LOC"], [85, 91, "mDEPD"], [86, 88, "EXP"], [87, 88, "mPUNC"], [88, 91, "mDEPD"], [89, 90, "mRELA"], [90, 91, "LOC"], [91, 79, "eSUCC"], [92, 91, "mDEPD"], [93, 79, "EXP"], [94, 93, "mPUNC"], [95, 93, "FEAT"], [96, 97, "mRELA"], [97, 93, "eCOO"], [98, 78, "mDEPD"], [99, 100, "mRELA"], [100, 101, "MANN"], [101, 78, "dCONT"], [102, 107, "FEAT"], [102, 109, "SCO"], [103, 109, "mDEPD"], [104, 109, "mPUNC"], [105, 107, "mRELA"], [106, 107, "FEAT"], [107, 109, "SCO"], [108, 109, "mDEPD"], [109, 101, "ePREC"], [110, 109, "mPUNC"], [111, 109, "eSUCC"], [112, 109, "mDEPD"], [113, 109, "eSUCC"], [114, 129, "mPUNC"], [115, 119, "mRELA"], [116, 119, "SCO"], [117, 119, "FEAT"], [118, 117, "mDEPD"], [119, 124, "STAT"], [120, 119, "mDEPD"], [121, 119, "mPUNC"], [122, 124, "mDEPD"], [123, 124, "mDEPD"], [124, 129, "dCONT"], [125, 124, "mDEPD"], [126, 129, "dCONT"], [127, 129, "mPUNC"], [128, 129, "AGT"], [129, 109, "eSUCC"], [130, 129, "dCONT"], [131, 129, "mPUNC"], [132, 109, "ePREC"], [133, 109, "eSUCC"], [134, 109, "mDEPD"], [135, 109, "SCO"], [136, 137, "MEAS"], [137, 109, "FEAT"], [138, 109, "mPUNC"], [139, 109, "FEAT"], [140, 109, "mPUNC"], [141, 109, "FEAT"], [142, 109, "mPUNC"], [143, 8, "mPUNC"], [143, 109, "mPUNC"], [144, 109, "eSUCC"], [145, 160, "mDEPD"], [146, 160, "mPUNC"], [147, 160, "eSUCC"], [148, 147, "mDEPD"], [149, 147, "MEAS"], [150, 151, "MEAS"], [151, 147, "TIME"], [152, 147, "mPUNC"], [153, 156, "mRELA"], [154, 156, "mRELA"], [155, 156, "mPUNC"], [156, 160, "mDEPD"], [157, 158, "MEAS"], [158, 156, "LINK"], [159, 156, "mPUNC"], [160, 109, "eSUCC"], [161, 160, "mPUNC"], [162, 163, "mDEPD"], [163, 160, "dEXP"], [164, 165, "mPUNC"], [165, 160, "eCOO"], [166, 167, "mRELA"], [167, 165, "dEXP"], [168, 167, "mDEPD"], [169, 167, "SCO"], [170, 160, "mPUNC"], [171, 172, "TIME"], [172, 11, "dCONT"], [173, 8, "MEAS"], [174, 175, "mDEPD"], [175, 8, "eSUCC"], [176, 175, "mPUNC"], [177, 181, "mDEPD"], [178, 181, "EXP"], [179, 181, "mDEPD"], [180, 181, "mDEPD"], [181, 2, "dCONT"], [182, 44, "mDEPD"], [182, 209, "mDEPD"], [183, 209, "mPUNC"], [184, 209, "mPUNC"], [185, 44, "AGT"], [185, 209, "EXP"], [186, 185, "eCOO"], [187, 209, "mPUNC"], [188, 191, "MEAS"], [189, 191, "mDEPD"], [190, 191, "eSUCC"], [191, 209, "dEXP"], [192, 204, "mDEPD"], [193, 204, "mPUNC"], [194, 195, "SCO"], [195, 204, "FEAT"], [196, 204, "CONT"], [197, 204, "mPUNC"], [198, 204, "AGT"], [199, 48, "mDEPD"], [199, 198, "mDEPD"], [200, 204, "mRELA"], [201, 204, "AGT"], [202, 204, "mDEPD"], [203, 204, "eCOO"], [204, 209, "ePREC"], [205, 204, "CONT"], [206, 204, "mPUNC"], [207, 204, "mDEPD"], [208, 204, "LOC"], [209, 181, "eSUCC"], [210, 209, "mPUNC"], [211, 209, "EXP"], [212, 209, "mPUNC"], [213, 209, "EXP"], [214, 209, "mPUNC"], [215, 216, "mDEPD"], [216, 218, "FEAT"], [217, 37, "mDEPD"], [217, 216, "mDEPD"], [218, 209, "EXP"], [219, 209, "mDEPD"], [220, 221, "mRELA"], [221, 209, "LOC"], [222, 181, "eSUCC"], [223, 181, "mPUNC"], [224, 181, "mRELA"], [225, 181, "SCO"]]}

[['[CLS]', '我', '想', ',', '如', '果', '我', '没', '有', '去', '做', '大', '学', '生', '村', '官', ',', '恐', '怕', '我', '这', '个', '在', '昆', '明', '长', '大', '的', '孩', '子', '永', '远', '都', '不', '能', '切', '身', '感', '受', '到', '云', '南', '这', '次', '6', '0', '年', '一', '遇', '的', '特', '大', '旱', '情', '的', '严', '重', '性', ',', '恐', '怕', '[sdp]', '我', '[sdp]', '只', '是', '每', '天', '[sdp]', '看', '[sdp]', '着', '新', '闻', '上', '那', '些', '缺', '水', '的', '镜', '头', ',', '嘴', '上', '说', '要', '节', '水', ',', '但', '事', '实', '行', '动', '保', '持', '不', '了', '三', '天', '。', '我', '任', '职', '的', '地', '方', '在', '昆', '明', '市', '禄', '劝', '县', '的', '一', '个', '村', '委', '会', ',', '说', '实', '话', '这', '里', '[SEP]'], [0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 0, 0, 0, 0, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], ['昆明长大', '云南', '昆明市禄劝县']]

gpu_number=1
negative_e_number=4
negative_e_length=16
base_dir=$PWD
checkpoint_dir=$base_dir/checkpoints
resources=$base_dir/resources
local_kg=$resources/ownthink_triples_small.txt
local_train_file=$resources/train_small.txt
remote_kg=https://atp-modelzoo-sh.oss-cn-shanghai.aliyuncs.com/release/tutorials/ckbert/ownthink_triples_small.txt
remote_train_file=https://atp-modelzoo-sh.oss-cn-shanghai.aliyuncs.com/release/tutorials/ckbert/train_small.txt
if [ ! -d $checkpoint_dir ];then
mkdir $checkpoint_dir
fi
if [ ! -d $resources ];then
mkdir $resources
fi
if [ ! -f $local_kg ];then
wget -P $resources $remote_kg
fi
if [ ! -f $local_train_file ];then
wget -P $resources $remote_train_file
fi
python -m torch.distributed.launch --nproc_per_node=$gpu_number \
--master_port=52349 \
$base_dir/main.py \
--mode=train \
--worker_gpu=$gpu_number \
--tables=$local_train_file, \
--learning_rate=5e-5  \
--epoch_num=5  \
--logging_steps=10 \
--save_checkpoint_steps=2150 \
--sequence_length=256 \
--train_batch_size=20 \
--checkpoint_dir=$checkpoint_dir \
--app_name=language_modeling \
--use_amp \
--save_all_checkpoints \
--user_defined_parameters="pretrain_model_name_or_path=hfl/macbert-base-zh external_mask_flag=True contrast_learning_flag=True negative_e_number=${negative_e_number} negative_e_length=${negative_e_length} kg_path=${local_kg}"

$ easynlp \
   --mode=train \
   --worker_gpu=1 \
   --tables=train.tsv,dev.tsv \
   --input_schema=label:str:1,sid1:str:1,sid2:str:1,sent1:str:1,sent2:str:1 \
   --first_sequence=sent1 \
   --label_name=label \
   --label_enumerate_values=0,1 \
   --checkpoint_dir=./classification_model \
   --epoch_num=1  \
   --sequence_length=128 \
   --app_name=text_classify \
   --user_defined_parameters='pretrain_model_name_or_path=alibaba-pai/pai-ckbert-base-zh'

$ easynlp \
  --mode=predict \
  --tables=dev.tsv \
  --outputs=dev.pred.tsv \
  --input_schema=label:str:1,sid1:str:1,sid2:str:1,sent1:str:1,sent2:str:1 \
  --output_schema=predictions,probabilities,logits,output \
  --append_cols=label \
  --first_sequence=sent1 \
  --checkpoint_path=./classification_model \
  --app_name=text_classify

from transformers import AutoTokenizer, AutoModelForMaskedLM, FillMaskPipeline
tokenizer = AutoTokenizer.from_pretrained("alibaba-pai/pai-ckbert-base-zh", use_auth_token=True)
model = AutoModelForMaskedLM.from_pretrained("alibaba-pai/pai-ckbert-base-zh", use_auth_token=True)
unmasker = FillMaskPipeline(model, tokenizer)   
unmasker("巴黎是[MASK]国的首都。",top_k=5)
[
    {'score': 0.8580496311187744, 
     'token': 3791, 
     'token_str': '法', 
     'sequence': '巴 黎 是 法 国 的 首 都 。'}, 
    {'score': 0.08550138026475906, 
     'token': 2548, 
     'token_str': '德', 
     'sequence': '巴 黎 是 德 国 的 首 都 。'}, 
    {'score': 0.023137662559747696, 
     'token': 5401, 
     'token_str': '美', 
     'sequence': '巴 黎 是 美 国 的 首 都 。'}, 
    {'score': 0.012281022034585476, 
     'token': 5739, 'token_str': '英', 
     'sequence': '巴 黎 是 英 国 的 首 都 。'}, 
    {'score': 0.005729076452553272, 
     'token': 704, 'token_str': '中', 
     'sequence': '巴 黎 是 中 国 的 首 都 。'}
]

from transformers import AutoTokenizer, AutoModelForMaskedLM
tokenizer = AutoTokenizer.from_pretrained("alibaba-pai/pai-ckbert-base-zh", use_auth_token=True)
model = AutoModelForMaskedLM.from_pretrained("alibaba-pai/pai-ckbert-base-zh", use_auth_token=True)
text = "巴黎是[MASK]国的首都。"
encoded_input = tokenizer(text, return_tensors='pt')
output = model(**encoded_input)