David Thrower
David Thrower
## TLDR: - Deletion of the duplicative `import tensorflow as tf` after we train the baseline GPT model, before we train the Cerebros model, appears to lower val_binary_accuracy on the...
It appears a phishing_email_detection_gpt2.py line 407 needs the zero mask set to False. This is an AI - introduced error. ``` embedded = tf.keras.layers.Embedding( input_dim=VOCABULARY_SIZE, output_dim=EMBEDDING_DIM, input_length=max_seq_length, mask_zero=True)(tokens) ```
Duplicative import: https://github.com/david-thrower/cerebros-core-algorithm-alpha/blob/main/phishing_email_detection_gpt2.py line 190, 191 ` from transformers import AutoTokenizer import tensorflow as tf `
Add the table in the PR and release for #172 to readme.md and a good explanation of how we scale in O(n) timing with sequence length to practically unlimited length.
# Complete this ... ## The code thus far ```python3 # Hard set these 2 params and do separate studies on each MINIMUM_LEVELS = 2 MAXIMUM_LEVELS = 2 NAMESPACE =...
Working from #156 ... ## It appears the positional embedding addition to the embedding model was a winner. GRUs alone deterriorated performance. LayerNorm dis as well. We have shown that...
### What happened? # Issue described: ## Environment details: - Python 3.11.9 (Canonical managed Kubeflow Jupyter environment) - Kubeflow SDK versions - kubeflow: 0.0.1rc0 - kubeflow-katib: 0.18.0 - kubeflow-training: 1.9.0...
## Issue described The documentation found in [1] depicts the code block below, which mentions the class method KatibClient.tune(), which the page for KatibClient [2] does not list. The source...
1. Add a demo of our generative LLM. 2. Clean up the directory / folder structure.