Upload add_sub_baseline_10K_2L1H128d
Browse files
add_sub_baseline_10K_2L1H128d/train_config.json
CHANGED
|
@@ -69,15 +69,15 @@
|
|
| 69 |
"no_wandb": false,
|
| 70 |
"n_params": 39346560,
|
| 71 |
"run_name": "add_sub_baseline_10K_2L1H128d",
|
| 72 |
-
"git_commit": "
|
| 73 |
-
"timestamp": "2026-04-
|
| 74 |
"tokenizer": "Qwen/Qwen3-0.6B",
|
| 75 |
"dataset_repo": "thoughtworks/arithmetic-sorl-data",
|
| 76 |
"dataset_config": "add_sub_6digit",
|
| 77 |
"model_repo": "thoughtworks/arithmetic-sorl",
|
| 78 |
"trainer_version": "sft",
|
| 79 |
-
"wandb_run_id": "
|
| 80 |
-
"wandb_url": "https://wandb.ai/nlp_and_interpretability/sorl-arithmetic/runs/
|
| 81 |
"final_accuracy": 0.0,
|
| 82 |
"sft_accuracy": 0.0,
|
| 83 |
"eval_method": "ArithmeticEvaluator"
|
|
|
|
| 69 |
"no_wandb": false,
|
| 70 |
"n_params": 39346560,
|
| 71 |
"run_name": "add_sub_baseline_10K_2L1H128d",
|
| 72 |
+
"git_commit": "17e935f460a7f9595b705c1d614101a6b0e520f7",
|
| 73 |
+
"timestamp": "2026-04-14T06:22:14.025912+00:00",
|
| 74 |
"tokenizer": "Qwen/Qwen3-0.6B",
|
| 75 |
"dataset_repo": "thoughtworks/arithmetic-sorl-data",
|
| 76 |
"dataset_config": "add_sub_6digit",
|
| 77 |
"model_repo": "thoughtworks/arithmetic-sorl",
|
| 78 |
"trainer_version": "sft",
|
| 79 |
+
"wandb_run_id": "lhp7f41c",
|
| 80 |
+
"wandb_url": "https://wandb.ai/nlp_and_interpretability/sorl-arithmetic/runs/lhp7f41c",
|
| 81 |
"final_accuracy": 0.0,
|
| 82 |
"sft_accuracy": 0.0,
|
| 83 |
"eval_method": "ArithmeticEvaluator"
|