AWS bedrock documentation change
Summary
Removed 'Automatic stopping criteria' section and simplified job status tracking steps. Updated real-time monitoring description with clearer formatting.
Security assessment
Changes involve removing documentation about training optimization features (automatic stopping criteria) and simplifying workflow descriptions. No security vulnerabilities, mitigations, or security features are mentioned in the changes. The modifications appear focused on streamlining documentation rather than addressing security concerns.
Diff
diff --git a/bedrock/latest/userguide/rft-monitor-job.md b/bedrock/latest/userguide/rft-monitor-job.md index 206725245..0d0d8d7b8 100644 --- a//bedrock/latest/userguide/rft-monitor-job.md +++ b//bedrock/latest/userguide/rft-monitor-job.md @@ -5 +5 @@ -Real-time training metricsAutomatic stopping criteriaJob status tracking +Real-time training metricsJob status tracking @@ -15,2 +14,0 @@ During reinforcement fine-tuning, you can monitor training progress in real-time - * Automatic stopping criteria - @@ -41 +39 @@ Amazon Bedrock provides real-time monitoring during RFT training with visual gra -The console displays interactive graphs that update in real-time as your RFT job progresses. These visualizations can help you. +The console displays interactive graphs that update in real-time as your RFT job progresses. These visualizations can help you: @@ -54,17 +51,0 @@ The console displays interactive graphs that update in real-time as your RFT job -## Automatic stopping criteria - -Amazon Bedrock implements intelligent stopping criteria to prevent overtraining and optimize resource usage. - -### Convergence detection - -Training automatically stops when multiple key performance indicators show consistent stagnation. - - * **Training scores** \- Monitor immediate actor model performance improvements - - * **Average validation scores** \- Assess model generalization to unseen data - - * **Scalarized policy gradient values** \- Evaluate optimization trajectory stability - - - - @@ -77,5 +58 @@ Monitor your RFT job status through the Amazon Bedrock console. - 1. **Initializing** \- Setting up training environment and loading data - - 2. **Training** \- Active model training with reward computation - - 3. **Validating** \- Evaluating model performance on validation data + 1. Validation @@ -83 +60 @@ Monitor your RFT job status through the Amazon Bedrock console. - 4. **Completing** \- Finalizing model artifacts and cleanup + 2. Training