I am training LlamaForSequenceClassification . It seems that it doesnt include the linear mapping in the model nn.Linear(config.hidden_size, self.num_labels, bias=False) . Namely it doesn’t upload this matrix with the model. (I can see that during the inference I get different results even if I fix the seed and the temperature). Is this true?
Related topics
| Topic | Replies | Views | Activity | |
|---|---|---|---|---|
| Replacing the LlamaDecoderLayer Class hugging Face With New LongNet | 0 | 741 | March 30, 2024 | |
| LlamaForSequenceClassification class and its results? | 1 | 235 | October 15, 2024 | |
| Llama-2 Sequence Classification: Much lower accuracy on inference from checkpoint compared to model | 5 | 5878 | February 20, 2024 | |
| How was LlamaForSequenceClassification Pretrained | 0 | 291 | July 15, 2023 | |
| Unable to Fine Tuning LLama2 for sequence classification | 0 | 219 | January 31, 2024 |