Completed
12:11 Memory Management in Custom Metrics
Class Central Classrooms beta
YouTube videos curated by Class Central.
Classroom Contents
Fine-Tuning with Custom Compute Metrics
Automatically move to the next video in the Classroom when playback concludes
- 1 00:00 Introduction to Tracking Loss in Model Training
- 2 00:50 Understanding Custom Metrics
- 3 01:22 Teacher Forced vs. Auto Regressive Decoding
- 4 01:32 Training vs. Inference Libraries
- 5 01:43 Options for Computing Custom Metrics
- 6 02:00 Challenges with Custom Metrics
- 7 03:44 Teacher Forced Decoding Explained
- 8 05:55 Auto Regressive Decoding Explained
- 9 08:19 Combining Training and Inference
- 10 09:22 Implementing Custom Metrics in Training
- 11 09:31 GRPO and Custom Metrics
- 12 12:11 Memory Management in Custom Metrics
- 13 15:44 Practical Demonstration: Setting Up the Environment
- 14 18:05 Loading and Training the Model
- 15 26:14 Setting Up the Training and Evaluation Dataset
- 16 26:31 Formatting the Data for Training
- 17 27:35 Custom Optimizer Setup
- 18 28:03 Loading and Printing Trainable Parameters
- 19 29:30 Fine-Tuning the Model
- 20 30:03 Custom Prediction Step and Metrics
- 21 31:42 Auto Aggressive Decoding and Evaluation
- 22 46:40 Handling Memory and GPU Utilization
- 23 49:55 Conclusion and Resources
