latest = step 4000 (nDCG@10=0.4708)

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:fee63fc43fad49735a01f83f3930e8e185482e88e728de25a8f1adde93016768
+oid sha256:1fc86a9adb931d8156ecfd4ed2b56add71480fff05643cc048b7e04d7aeb1cbd
 size 280019040

trainer_state.json

@@ -1,10 +1,10 @@
 {
-  "best_global_step": 3500,
-  "best_metric": 0.4608934483931739,
-  "best_model_checkpoint": "checkpoints/med-ruri-v3-70m-v2-from-ruri/checkpoint-3500",
-  "epoch": 0.5682740704659848,
+  "best_global_step": 4000,
+  "best_metric": 0.4708195055654097,
+  "best_model_checkpoint": "checkpoints/med-ruri-v3-70m-v2-from-ruri/checkpoint-4000",
+  "epoch": 0.649456080532554,
   "eval_steps": 500,
-  "global_step": 3500,
+  "global_step": 4000,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
@@ -631,6 +631,95 @@
       "eval_samples_per_second": 0.0,
       "eval_steps_per_second": 0.0,
       "step": 3500
+    },
+    {
+      "epoch": 0.5763922714726417,
+      "grad_norm": 22.564878463745117,
+      "learning_rate": 4.460775935737481e-06,
+      "loss": 0.5085312652587891,
+      "step": 3550
+    },
+    {
+      "epoch": 0.5845104724792985,
+      "grad_norm": 25.2183780670166,
+      "learning_rate": 4.375320458041361e-06,
+      "loss": 0.4952116012573242,
+      "step": 3600
+    },
+    {
+      "epoch": 0.5926286734859555,
+      "grad_norm": 27.938398361206055,
+      "learning_rate": 4.28986498034524e-06,
+      "loss": 0.5000743103027344,
+      "step": 3650
+    },
+    {
+      "epoch": 0.6007468744926124,
+      "grad_norm": 25.634239196777344,
+      "learning_rate": 4.20440950264912e-06,
+      "loss": 0.5024850082397461,
+      "step": 3700
+    },
+    {
+      "epoch": 0.6088650754992694,
+      "grad_norm": 24.961740493774414,
+      "learning_rate": 4.118954024952999e-06,
+      "loss": 0.5098614120483398,
+      "step": 3750
+    },
+    {
+      "epoch": 0.6169832765059263,
+      "grad_norm": 20.451208114624023,
+      "learning_rate": 4.033498547256879e-06,
+      "loss": 0.47686138153076174,
+      "step": 3800
+    },
+    {
+      "epoch": 0.6251014775125832,
+      "grad_norm": 26.467348098754883,
+      "learning_rate": 3.948043069560759e-06,
+      "loss": 0.47324317932128906,
+      "step": 3850
+    },
+    {
+      "epoch": 0.6332196785192401,
+      "grad_norm": 24.046497344970703,
+      "learning_rate": 3.862587591864639e-06,
+      "loss": 0.48282413482666015,
+      "step": 3900
+    },
+    {
+      "epoch": 0.641337879525897,
+      "grad_norm": 27.17749786376953,
+      "learning_rate": 3.7771321141685185e-06,
+      "loss": 0.48658458709716795,
+      "step": 3950
+    },
+    {
+      "epoch": 0.649456080532554,
+      "grad_norm": 35.05656051635742,
+      "learning_rate": 3.6916766364723984e-06,
+      "loss": 0.4936316680908203,
+      "step": 4000
+    },
+    {
+      "epoch": 0.649456080532554,
+      "eval_med-ir_cosine_accuracy@1": 0.259,
+      "eval_med-ir_cosine_accuracy@10": 0.71,
+      "eval_med-ir_cosine_accuracy@5": 0.59,
+      "eval_med-ir_cosine_map@10": 0.3956890873015873,
+      "eval_med-ir_cosine_mrr@10": 0.3956890873015883,
+      "eval_med-ir_cosine_ndcg@10": 0.4708195055654097,
+      "eval_med-ir_cosine_precision@1": 0.259,
+      "eval_med-ir_cosine_precision@10": 0.071,
+      "eval_med-ir_cosine_precision@5": 0.118,
+      "eval_med-ir_cosine_recall@1": 0.259,
+      "eval_med-ir_cosine_recall@10": 0.71,
+      "eval_med-ir_cosine_recall@5": 0.59,
+      "eval_runtime": 53.0112,
+      "eval_samples_per_second": 0.0,
+      "eval_steps_per_second": 0.0,
+      "step": 4000
     }
   ],
   "logging_steps": 50,