Upload add_sub_sorl_v1_abs10_K1_100K_1L3H510d

add_sub_sorl_v1_abs10_K1_100K_1L3H510d/model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:05915d1d3abf503303872b835b197cd7e1c25afe84a5656f0a395715ce75f417
+oid sha256:eaca29567d0d2a71d057686b57d7e7de3903ae2e8b321def01cd4368a59d7937
 size 634679036

add_sub_sorl_v1_abs10_K1_100K_1L3H510d/train_config.json

@@ -17,10 +17,10 @@
   "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": 468,
+  "warmup_steps": 937,
   "cooldown_frac": 0.4,
   "max_grad_norm": 1.0,
   "vq_abs_pretrain_steps": 0,
@@ -30,7 +30,7 @@
   "vq_abs_pretrain_target_vectors": 20000,
   "batch_size": 64,
   "gradient_accumulation_steps": 1,
-  "num_epochs": 10,
+  "num_epochs": 20,
   "emb_warmup_steps": 0,
   "log_every": 50,
   "eval_every": 1562,
@@ -69,16 +69,16 @@
   "no_wandb": false,
   "n_params": 158593426,
   "run_name": "add_sub_sorl_v1_abs10_K1_100K_1L3H510d",
-  "git_commit": "f447da529caceac8c7d256cbb2cd185cbc50feac",
-  "timestamp": "2026-04-12T09:52:22.660543+00:00",
+  "git_commit": "8d5ee5420119746ef4e2c87570eb250c9718f643",
+  "timestamp": "2026-04-13T00:14:20.891354+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": "v1",
-  "wandb_run_id": "u4e5s4m7",
-  "wandb_url": "https://wandb.ai/nlp_and_interpretability/sorl-arithmetic/runs/u4e5s4m7",
-  "final_accuracy": 0.70625,
-  "sft_accuracy": 0.5979166666666667,
+  "wandb_run_id": "mzxinm9k",
+  "wandb_url": "https://wandb.ai/nlp_and_interpretability/sorl-arithmetic/runs/mzxinm9k",
+  "final_accuracy": 0.68875,
+  "sft_accuracy": 0.62,
   "eval_method": "ArithmeticEvaluator"
 }