Upload add_sub_baseline_100K_1L3H510d

add_sub_baseline_100K_1L3H510d/model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3ba90e0fbbdac3c126de2b39f542a7f66ad9d062ae36bc9296b4dd0cfa2ede84
+oid sha256:0b85ef5533b228f5e46463fbed81a2b2470589d326d11ab23624303456dfb058
 size 634642298

add_sub_baseline_100K_1L3H510d/train_config.json

@@ -17,7 +17,7 @@
   "target_vocab_util": 0.8,
   "min_abs_ppl": 0.0,
   "zipf_alpha": 1.0,
-  "lr": 4e-05,
+  "lr": 8e-05,
   "emb_lr_mult": 1.0,
   "weight_decay": 0.01,
   "warmup_steps": 937,
@@ -69,16 +69,16 @@
   "no_wandb": false,
   "n_params": 158584246,
   "run_name": "add_sub_baseline_100K_1L3H510d",
-  "git_commit": "f447da529caceac8c7d256cbb2cd185cbc50feac",
-  "timestamp": "2026-04-12T18:32:15.366122+00:00",
+  "git_commit": "8d5ee5420119746ef4e2c87570eb250c9718f643",
+  "timestamp": "2026-04-12T23:07:20.411593+00:00",
   "tokenizer": "Qwen/Qwen3-0.6B",
   "dataset_repo": "thoughtworks/arithmetic-sorl-data",
   "dataset_config": "add_sub_6digit",
   "model_repo": "thoughtworks/arithmetic-sorl",
   "trainer_version": "sft",
-  "wandb_run_id": "a545bryb",
-  "wandb_url": "https://wandb.ai/nlp_and_interpretability/sorl-arithmetic/runs/a545bryb",
-  "final_accuracy": 0.6416666666666667,
-  "sft_accuracy": 0.6416666666666667,
+  "wandb_run_id": "8b7rrjnk",
+  "wandb_url": "https://wandb.ai/nlp_and_interpretability/sorl-arithmetic/runs/8b7rrjnk",
+  "final_accuracy": 0.6070833333333333,
+  "sft_accuracy": 0.6070833333333333,
   "eval_method": "ArithmeticEvaluator"
 }