ValueError: Unrecognized model in ./MRPC/. Should have a `model_type` key in its config.json, or contain one of the following strings in its name

Question

目标：修改此 Notebook 以使用 Albert 和 Distilbert 模型

内核：conda_pytorch_p36。我做了 Restart & 运行 All，并刷新了工作目录中的文件视图。

第 1.2 节 出现错误，仅针对这 2 个新模型。

对于文件名等，我创建了一个随处使用的变量：

MODEL_NAME = 'albert-base-v2'  # 'distilbert-base-uncased', 'bert-base-uncased'

我将导入替换为：

from transformers import (AutoConfig, AutoModel, AutoTokenizer)
#from transformers import (BertConfig, BertForSequenceClassification, BertTokenizer,)

如 Transformers Documentation - Auto Classes 中所建议。

Instantiating one of AutoConfig, AutoModel, and AutoTokenizer will directly create a class of the relevant architecture.

第 1.2 节：

# load model
model = AutoModel.from_pretrained(configs.output_dir)  # BertForSequenceClassification
model.to(configs.device)


# quantize model
quantized_model = torch.quantization.quantize_dynamic(
    model, {torch.nn.Linear}, dtype=torch.qint8
)

#print(quantized_model)

def print_size_of_model(model):
    torch.save(model.state_dict(), "temp.p")
    print('Size (MB):', os.path.getsize("temp.p")/(1024*1024))
    os.remove('temp.p')

print_size_of_model(model)
print_size_of_model(quantized_model)

回溯：

ValueError: Unrecognized model in ./MRPC/. Should have a `model_type` key in its config.json, or contain one of the following strings in its name: imagegpt, qdqbert, vision-encoder-decoder, trocr, fnet, segformer, vision-text-dual-encoder, perceiver, gptj, layoutlmv2, beit, rembert, visual_bert, canine, roformer, clip, bigbird_pegasus, deit, luke, detr, gpt_neo, big_bird, speech_to_text_2, speech_to_text, vit, wav2vec2, m2m_100, convbert, led, blenderbot-small, retribert, ibert, mt5, t5, mobilebert, distilbert, albert, bert-generation, camembert, xlm-roberta, pegasus, marian, mbart, megatron-bert, mpnet, bart, blenderbot, reformer, longformer, roberta, deberta-v2, deberta, flaubert, fsmt, squeezebert, hubert, bert, openai-gpt, gpt2, transfo-xl, xlnet, xlm-prophetnet, prophetnet, xlm, ctrl, electra, speech-encoder-decoder, encoder-decoder, funnel, lxmert, dpr, layoutlm, rag, tapas, splinter, sew-d, sew, unispeech-sat, unispeech, wavlm

如果我还有什么要补充的，请告诉我 post。

Answer 1

解释：

实例化 AutoModel 时，必须在 ./MRPC/config.json 文件（在笔记本运行时下载）中指定 model_type 参数。

可以找到 model_types 的列表 here。

解决方案：

将 model_type 附加到 config.json 的代码，格式相同：

import json

json_filename = './MRPC/config.json'

with open(json_filename) as json_file:
    json_decoded = json.load(json_file)

json_decoded['model_type'] = # !!

with open(json_filename, 'w') as json_file:
    json.dump(json_decoded, json_file, indent=2, separators=(',', ': '))

config.json:

{
  "attention_probs_dropout_prob": 0.1,
  "finetuning_task": "mrpc",
  "hidden_act": "gelu",
  "hidden_dropout_prob": 0.1,
  "hidden_size": 768,
  "initializer_range": 0.02,
  "intermediate_size": 3072,
  "layer_norm_eps": 1e-12,
  "max_position_embeddings": 512,
  "num_attention_heads": 12,
  "num_hidden_layers": 12,
  "num_labels": 2,
  "output_attentions": false,
  "output_hidden_states": false,
  "pruned_heads": {},
  "torchscript": false,
  "type_vocab_size": 2,
  "vocab_size": 30522,
  "model_type": "albert"
}

ValueError: Unrecognized model in ./MRPC/. Should have a `model_type` key in its config.json, or contain one of the following strings in its name

ValueError: Unrecognized model in ./MRPC/. Should have a `model_type` key in its config.json, or contain one of the following strings in its name

python

tensorflow

huggingface-transformers

bert-language-model

onnx

解释：

解决方案：