Iván Sánchez

SSE4 and AVX instruction sets are available on Intel and AMD processors, not ARM processors. Android devices are generally built with ARM processors, that's why you can't compile for, let's...

You have to change the call `tf.contrib.estimator.stop_if_no_increase_hook` to `tf.estimator.experimental.stop_if_no_increase_hook`. For example, in /models/lstm_crf/main.py, im the line 171, change: ``` hook = tf.contrib.estimator.stop_if_no_increase_hook( estimator, 'f1', 500, min_steps=8000, run_every_secs=120) ``` to: ```...

Probably because paper is using Senna embeddings while this repository is using GloVe embeddings

Thank you @Hironsan ! The new features look great. Can't wait to try them! Just to close this properly, I'd like to mention (for future reference) that the problem I...

> Yeah it would be a great help. Kind of tiresome that none of the document transformers (TILT, StructuralLM, layoutlm,StrucTexT) have release their pretraining code, although it is the most...

`prepare_model_for_kbit_training` [freezes all the layers of the model](https://github.com/huggingface/peft/blob/main/src/peft/utils/other.py#L97-L99) so it makes sense they're not trainable after the line: ```python model = prepare_model_for_kbit_training(model) ```

Hi @danielhanchen, I'm not sure this is resolved I was also getting gibberish results when finetuning Llama 3.1 8B instruct on long context samples using unsloth, I was using `neftune_noise_alpha=5`...

> I didn't try to reload the model and use it without neftune after having trained it with neftune to see if this is the problem Update on this. After...

This contribution looks great! Is there any news on this? Why wasn't it merged?

Having this problem as well. Interestingly enough, it happens arbitrarily with some sample once in evaluation edit: Seems to happen with gemma 9b as well