[ACL 2023] One Embedder, Any Task: Instruction-Finetuned Text Embeddings
Apache License 2.0
1.85k
stars
134
forks
source link
Some weights of the model checkpoint at /home/rnd/wmj/instructor-large/instructor-embedding/output/checkpoint-6500/ were not used when initializing T5EncoderModel: ['2.linear.weight'] - This IS expected if you are initializing T5EncoderModel from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model). - This IS NOT expected if you are initializing T5EncoderModel from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model). max_seq_length 512 #117
Hello, first of all, thank you very much for your excellent open-source code. However, I encountered some problems while using it. I used the provided Instructor-large model weight finetune to access our data, but in the output checkpoint, I first found that its checkpoint folder is not a loadable model file (it is missing multiple files, such as the 1_Pooling, 2_Dense folder, config. json). So, I copied these files from the downloaded Instructor-large folder, but I still got the following error. How should I use the finetune checkpoint？
Some weights of the model checkpoint at /home/rnd/wmj/instructor-large/instructor-embedding/output/checkpoint-6500/ were not used when initializing T5EncoderModel: ['2.linear.weight']
This IS expected if you are initializing T5EncoderModel from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).
This IS NOT expected if you are initializing T5EncoderModel from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).
max_seq_length 512
Hello, first of all, thank you very much for your excellent open-source code. However, I encountered some problems while using it. I used the provided Instructor-large model weight finetune to access our data, but in the output checkpoint, I first found that its checkpoint folder is not a loadable model file (it is missing multiple files, such as the 1_Pooling, 2_Dense folder, config. json). So, I copied these files from the downloaded Instructor-large folder, but I still got the following error. How should I use the finetune checkpoint？
Some weights of the model checkpoint at /home/rnd/wmj/instructor-large/instructor-embedding/output/checkpoint-6500/ were not used when initializing T5EncoderModel: ['2.linear.weight']