Upload folder using huggingface_hub

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f7ffb7bae90b8b5c78a9a13cdfe0a421560d34bb5871890f58ee7261c464bc64
+oid sha256:06673c96202830b503b6e83acb14987c3d249b2811f2f9192c8a5c414315461d
 size 121915776

optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:fdafc2afc39a8132321f56d6e60f80865fa74f0aa4159e89c08146a2111c68ee
+oid sha256:f436c2cd11fc1a4166ddc4846d77e8224fffe93f49b63d05ae8889c5885050ba
 size 243883194

rng_state.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5b2c699dc57ecf42633a0bfe4f07a4b91bc05a4b7e5079fe8a31469f27538ed5
+oid sha256:a6a82aa8dc18e09c52a17abf1ac8d6c9de5fe230e09a3faf36beb1eeb47a03bd
 size 14244

scheduler.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5ccb54d0139ddf32a93ee725362dd8028f6a8c59e1e2faa2e66ab1b6583eb8b9
+oid sha256:0fc8f1838e22027d1c8795abfafd0327a77e02ab1b7dd271ee107ff8dcf6d8d1
 size 1064

trainer_state.json

@@ -1,218 +1,164 @@
 {
   "best_metric": null,
   "best_model_checkpoint": null,
-  "epoch": 0.9990680335507922,
+  "epoch": 0.9846153846153847,
   "eval_steps": 500,
-  "global_step": 67,
+  "global_step": 48,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
-      "epoch": 0.03,
-      "learning_rate": 1.3333333333333333e-05,
-      "loss": 2.5624,
+      "epoch": 0.04,
+      "learning_rate": 2e-05,
+      "loss": 2.12,
       "step": 2
     },
     {
-      "epoch": 0.06,
-      "learning_rate": 1.9987954562051724e-05,
-      "loss": 1.5913,
+      "epoch": 0.08,
+      "learning_rate": 1.9906859460363307e-05,
+      "loss": 2.5155,
       "step": 4
     },
     {
-      "epoch": 0.09,
-      "learning_rate": 1.989176509964781e-05,
-      "loss": 1.6086,
+      "epoch": 0.12,
+      "learning_rate": 1.9629172873477995e-05,
+      "loss": 2.3475,
       "step": 6
     },
     {
-      "epoch": 0.12,
-      "learning_rate": 1.9700312531945444e-05,
-      "loss": 1.5026,
+      "epoch": 0.16,
+      "learning_rate": 1.917211301505453e-05,
+      "loss": 2.4742,
       "step": 8
     },
     {
-      "epoch": 0.15,
-      "learning_rate": 1.941544065183021e-05,
-      "loss": 1.2094,
+      "epoch": 0.21,
+      "learning_rate": 1.8544194045464888e-05,
+      "loss": 2.4053,
       "step": 10
     },
     {
-      "epoch": 0.18,
-      "learning_rate": 1.9039892931234434e-05,
-      "loss": 1.4642,
+      "epoch": 0.25,
+      "learning_rate": 1.77571129070442e-05,
+      "loss": 2.9106,
       "step": 12
     },
     {
-      "epoch": 0.21,
-      "learning_rate": 1.8577286100002723e-05,
-      "loss": 1.7891,
+      "epoch": 0.29,
+      "learning_rate": 1.6825531432186545e-05,
+      "loss": 1.9816,
       "step": 14
     },
     {
-      "epoch": 0.24,
-      "learning_rate": 1.803207531480645e-05,
-      "loss": 1.9994,
+      "epoch": 0.33,
+      "learning_rate": 1.5766803221148676e-05,
+      "loss": 2.2862,
       "step": 16
     },
     {
-      "epoch": 0.27,
-      "learning_rate": 1.7409511253549592e-05,
-      "loss": 1.999,
+      "epoch": 0.37,
+      "learning_rate": 1.4600650377311523e-05,
+      "loss": 2.3261,
       "step": 18
     },
     {
-      "epoch": 0.3,
-      "learning_rate": 1.6715589548470187e-05,
-      "loss": 1.4883,
+      "epoch": 0.41,
+      "learning_rate": 1.3348796121709862e-05,
+      "loss": 2.4281,
       "step": 20
     },
     {
-      "epoch": 0.33,
-      "learning_rate": 1.5956993044924334e-05,
-      "loss": 1.4515,
+      "epoch": 0.45,
+      "learning_rate": 1.2034560130526341e-05,
+      "loss": 2.3129,
       "step": 22
     },
     {
-      "epoch": 0.36,
-      "learning_rate": 1.5141027441932217e-05,
-      "loss": 1.4227,
+      "epoch": 0.49,
+      "learning_rate": 1.0682424133646712e-05,
+      "loss": 2.4867,
       "step": 24
     },
     {
-      "epoch": 0.39,
-      "learning_rate": 1.4275550934302822e-05,
-      "loss": 1.3913,
+      "epoch": 0.53,
+      "learning_rate": 9.317575866353293e-06,
+      "loss": 1.9649,
       "step": 26
     },
     {
-      "epoch": 0.42,
-      "learning_rate": 1.3368898533922202e-05,
-      "loss": 1.5564,
+      "epoch": 0.57,
+      "learning_rate": 7.965439869473664e-06,
+      "loss": 2.2729,
       "step": 28
     },
     {
-      "epoch": 0.45,
-      "learning_rate": 1.242980179903264e-05,
-      "loss": 1.837,
+      "epoch": 0.62,
+      "learning_rate": 6.651203878290139e-06,
+      "loss": 2.2966,
       "step": 30
     },
     {
-      "epoch": 0.48,
-      "learning_rate": 1.1467304744553618e-05,
-      "loss": 1.8164,
+      "epoch": 0.66,
+      "learning_rate": 5.399349622688479e-06,
+      "loss": 2.3374,
       "step": 32
     },
     {
-      "epoch": 0.51,
-      "learning_rate": 1.0490676743274181e-05,
-      "loss": 2.4233,
+      "epoch": 0.7,
+      "learning_rate": 4.2331967788513295e-06,
+      "loss": 2.396,
       "step": 34
     },
     {
-      "epoch": 0.54,
-      "learning_rate": 9.50932325672582e-06,
-      "loss": 1.8559,
+      "epoch": 0.74,
+      "learning_rate": 3.174468567813461e-06,
+      "loss": 2.5016,
       "step": 36
     },
     {
-      "epoch": 0.57,
-      "learning_rate": 8.532695255446384e-06,
-      "loss": 1.5067,
+      "epoch": 0.78,
+      "learning_rate": 2.2428870929558012e-06,
+      "loss": 2.156,
       "step": 38
     },
     {
-      "epoch": 0.6,
-      "learning_rate": 7.570198200967363e-06,
-      "loss": 1.3854,
+      "epoch": 0.82,
+      "learning_rate": 1.4558059545351144e-06,
+      "loss": 2.3286,
       "step": 40
     },
     {
-      "epoch": 0.63,
-      "learning_rate": 6.631101466077801e-06,
-      "loss": 1.2381,
+      "epoch": 0.86,
+      "learning_rate": 8.278869849454718e-07,
+      "loss": 2.4654,
       "step": 42
     },
     {
-      "epoch": 0.66,
-      "learning_rate": 5.724449065697182e-06,
-      "loss": 1.4576,
+      "epoch": 0.9,
+      "learning_rate": 3.708271265220087e-07,
+      "loss": 2.394,
       "step": 44
     },
     {
-      "epoch": 0.69,
-      "learning_rate": 4.858972558067784e-06,
-      "loss": 1.7201,
+      "epoch": 0.94,
+      "learning_rate": 9.314053963669245e-08,
+      "loss": 2.4198,
       "step": 46
     },
-    {
-      "epoch": 0.72,
-      "learning_rate": 4.043006955075667e-06,
-      "loss": 1.6459,
-      "step": 48
-    },
-    {
-      "epoch": 0.75,
-      "learning_rate": 3.284410451529816e-06,
-      "loss": 2.7212,
-      "step": 50
-    },
-    {
-      "epoch": 0.78,
-      "learning_rate": 2.5904887464504115e-06,
-      "loss": 1.5635,
-      "step": 52
-    },
-    {
-      "epoch": 0.81,
-      "learning_rate": 1.967924685193552e-06,
-      "loss": 1.3143,
-      "step": 54
-    },
-    {
-      "epoch": 0.84,
-      "learning_rate": 1.4227138999972801e-06,
-      "loss": 1.5857,
-      "step": 56
-    },
-    {
-      "epoch": 0.86,
-      "learning_rate": 9.601070687655667e-07,
-      "loss": 1.3948,
-      "step": 58
-    },
-    {
-      "epoch": 0.89,
-      "learning_rate": 5.845593481697931e-07,
-      "loss": 1.2734,
-      "step": 60
-    },
-    {
-      "epoch": 0.92,
-      "learning_rate": 2.996874680545603e-07,
-      "loss": 1.4232,
-      "step": 62
-    },
-    {
-      "epoch": 0.95,
-      "learning_rate": 1.0823490035218986e-07,
-      "loss": 1.7512,
-      "step": 64
-    },
     {
       "epoch": 0.98,
-      "learning_rate": 1.2045437948275952e-08,
-      "loss": 1.3898,
-      "step": 66
+      "learning_rate": 0.0,
+      "loss": 2.7688,
+      "step": 48
     }
   ],
   "logging_steps": 2,
-  "max_steps": 67,
+  "max_steps": 48,
   "num_input_tokens_seen": 0,
   "num_train_epochs": 1,
   "save_steps": 500,
-  "total_flos": 4856523023781888.0,
+  "total_flos": 3312099646525440.0,
   "train_batch_size": 1,
   "trial_name": null,
   "trial_params": null

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ba6a6332806d9e827a1ab53dc34d98e7840e0b213435f5b8fb4fc8d68f2dfb74
+oid sha256:0ff763329515fd373f72fad313525c90bcaf4f217f616d926b8ca8e7604c24d0
 size 4728