110 lines
3.8 KiB
JSON
110 lines
3.8 KiB
JSON
{
|
|
"_comment": "Example JSON file for AI Experiment Logger plugin. Represents a collection of experiments.",
|
|
"experiments": [
|
|
{
|
|
"_comment": "First experiment - a simple sentiment analysis task",
|
|
"experiment_id": "sentiment-analysis-v1",
|
|
"experiment_name": "Sentiment Analysis with Claude",
|
|
"description": "Experiment to evaluate Claude's sentiment analysis accuracy on customer reviews.",
|
|
"model": "Claude-v2",
|
|
"dataset": "Customer Reviews - Product A",
|
|
"date_created": "2024-01-15",
|
|
"date_completed": "2024-01-16",
|
|
"status": "completed",
|
|
"parameters": {
|
|
"temperature": 0.2,
|
|
"max_tokens": 200,
|
|
"prompt_template": "Analyze the sentiment of the following review: {review}"
|
|
},
|
|
"metrics": {
|
|
"accuracy": 0.92,
|
|
"precision": 0.90,
|
|
"recall": 0.94,
|
|
"f1_score": 0.92
|
|
},
|
|
"artifacts": {
|
|
"results_csv": "sentiment_analysis_results.csv",
|
|
"model_config": "claude_config.json"
|
|
},
|
|
"notes": "High accuracy achieved. Consider testing with more diverse datasets."
|
|
},
|
|
{
|
|
"_comment": "Second experiment - a more complex text summarization task",
|
|
"experiment_id": "text-summarization-v2",
|
|
"experiment_name": "Text Summarization Experiment with Gemini",
|
|
"description": "Experiment to compare different summarization techniques using Gemini.",
|
|
"model": "Gemini-Pro",
|
|
"dataset": "News Articles - Technology",
|
|
"date_created": "2024-01-20",
|
|
"date_completed": "2024-01-22",
|
|
"status": "completed",
|
|
"parameters": {
|
|
"temperature": 0.5,
|
|
"max_tokens": 500,
|
|
"summarization_type": "extractive",
|
|
"prompt_template": "Summarize the following article: {article}"
|
|
},
|
|
"metrics": {
|
|
"rouge_1": 0.45,
|
|
"rouge_2": 0.22,
|
|
"rouge_l": 0.38,
|
|
"bleu_score": 0.28
|
|
},
|
|
"artifacts": {
|
|
"summaries_json": "summaries.json",
|
|
"hyperparameter_search_log": "hyperparameter_search.log"
|
|
},
|
|
"notes": "Extractive summarization performed better than abstractive. Needs further tuning."
|
|
},
|
|
{
|
|
"_comment": "Third experiment - still in progress",
|
|
"experiment_id": "image-classification-v1",
|
|
"experiment_name": "Image Classification with a Custom Model",
|
|
"description": "Training and evaluating a custom image classification model.",
|
|
"model": "Custom CNN",
|
|
"dataset": "Custom Image Dataset - Flowers",
|
|
"date_created": "2024-01-25",
|
|
"date_completed": null,
|
|
"status": "running",
|
|
"parameters": {
|
|
"learning_rate": 0.001,
|
|
"batch_size": 32,
|
|
"epochs": 10
|
|
},
|
|
"metrics": {
|
|
"training_loss": 0.25,
|
|
"validation_accuracy": 0.85
|
|
},
|
|
"artifacts": {
|
|
"model_checkpoint": "model_checkpoint_epoch_5.pth",
|
|
"training_logs": "training.log"
|
|
},
|
|
"notes": "Training in progress. Validation accuracy is improving."
|
|
},
|
|
{
|
|
"_comment": "Fourth experiment - trying to compare the models",
|
|
"experiment_id": "model-comparison-v1",
|
|
"experiment_name": "Model Comparison for Question Answering",
|
|
"description": "Comparing Claude and Gemini for question answering accuracy",
|
|
"model": "Claude-v2, Gemini-Pro",
|
|
"dataset": "SQuAD",
|
|
"date_created": "2024-02-01",
|
|
"date_completed": "2024-02-03",
|
|
"status": "completed",
|
|
"parameters": {
|
|
"question_length": "short",
|
|
"context_length": "medium"
|
|
},
|
|
"metrics": {
|
|
"claude_exact_match": 0.75,
|
|
"claude_f1": 0.82,
|
|
"gemini_exact_match": 0.78,
|
|
"gemini_f1": 0.85
|
|
},
|
|
"artifacts": {
|
|
"results_table": "qa_results.csv"
|
|
},
|
|
"notes": "Gemini slightly outperforms Claude on F1 score. Further investigation needed with longer contexts."
|
|
}
|
|
]
|
|
} |