update files

all_results.json

@@ -0,0 +1,7 @@
+{
+    "epoch": 5.0,
+    "train_loss": 0.2349783019721508,
+    "train_runtime": 19900.0608,
+    "train_samples_per_second": 12.859,
+    "train_steps_per_second": 0.201
+}

config.json

@@ -0,0 +1,28 @@
+{
+  "_name_or_path": "./train_1709305042/",
+  "architectures": [
+    "GemmaForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "bos_token_id": 2,
+  "eos_token_id": 1,
+  "head_dim": 256,
+  "hidden_act": "gelu",
+  "hidden_size": 2048,
+  "initializer_range": 0.02,
+  "intermediate_size": 16384,
+  "max_position_embeddings": 8192,
+  "model_type": "gemma",
+  "num_attention_heads": 8,
+  "num_hidden_layers": 18,
+  "num_key_value_heads": 1,
+  "pad_token_id": 0,
+  "rms_norm_eps": 1e-06,
+  "rope_scaling": null,
+  "rope_theta": 10000.0,
+  "torch_dtype": "float16",
+  "transformers_version": "4.38.0",
+  "use_cache": false,
+  "vocab_size": 261568
+}

generation_config.json

@@ -0,0 +1,7 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 2,
+  "eos_token_id": 1,
+  "pad_token_id": 0,
+  "transformers_version": "4.38.0"
+}

model.safetensors.index.json

@@ -0,0 +1,171 @@
+{
+  "metadata": {
+    "total_size": 5035151360
+  },
+  "weight_map": {
+    "model.embed_tokens.weight": "model-00001-of-00002.safetensors",
+    "model.layers.0.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.0.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.0.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.0.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.0.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.0.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.0.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.0.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.0.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.1.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.1.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.1.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.1.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.1.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.1.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.1.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.1.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.1.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.10.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.10.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.10.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.10.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.10.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.10.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.10.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.10.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.10.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.11.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.11.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.11.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.11.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.11.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.11.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.11.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.11.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.11.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.12.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.12.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.12.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.12.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.12.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.12.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.12.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.12.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.12.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.13.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.13.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.13.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.13.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.13.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.13.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.13.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.13.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.13.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.14.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.14.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.14.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.14.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.14.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.14.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.14.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.14.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.14.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.16.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.16.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.16.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.16.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.16.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.16.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.16.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.16.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.16.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.17.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.17.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.17.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.17.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.17.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.17.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.17.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.17.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.17.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.2.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.2.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.2.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.2.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.2.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.2.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.2.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.2.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.2.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.3.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.3.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.3.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.3.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.3.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.3.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.3.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.3.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.3.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.4.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.4.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.4.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.4.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.4.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.4.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.4.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.4.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.4.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.5.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.5.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.5.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.5.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.5.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.5.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.5.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.5.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.5.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.6.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.6.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.6.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.6.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.6.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.6.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.6.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.6.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.6.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.7.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.7.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.7.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.7.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.7.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.7.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.7.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.7.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.7.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.8.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.8.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.8.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.8.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.8.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.8.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.8.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.8.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.8.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.9.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.9.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.9.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.9.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.9.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.9.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.9.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.9.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.9.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.norm.weight": "model-00002-of-00002.safetensors"
+  }
+}

special_tokens_map.json

@@ -0,0 +1,30 @@
+{
+  "bos_token": {
+    "content": "<bos>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "<eos>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<pad>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "unk_token": {
+    "content": "<unk>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

tokenizer_config.json

@@ -0,0 +1,50 @@
+{
+  "add_bos_token": true,
+  "add_eos_token": false,
+  "added_tokens_decoder": {
+    "0": {
+      "content": "<pad>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "1": {
+      "content": "<eos>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "2": {
+      "content": "<bos>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "3": {
+      "content": "<unk>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "bos_token": "<bos>",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<eos>",
+  "model_max_length": 1000000000000000019884624838656,
+  "pad_token": "<pad>",
+  "padding_side": "right",
+  "sp_model_kwargs": {},
+  "spaces_between_special_tokens": false,
+  "split_special_tokens": false,
+  "tokenizer_class": "GemmaTokenizer",
+  "unk_token": "<unk>",
+  "use_default_system_prompt": false
+}

trainer_state.json

@@ -0,0 +1,2830 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 5.0,
+  "eval_steps": 500,
+  "global_step": 4000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.01,
+      "grad_norm": 18.572529250843502,
+      "learning_rate": 4.998749374687344e-05,
+      "loss": 0.6938,
+      "step": 10
+    },
+    {
+      "epoch": 0.03,
+      "grad_norm": 4.9721361054552675,
+      "learning_rate": 4.9874937468734366e-05,
+      "loss": 0.4097,
+      "step": 20
+    },
+    {
+      "epoch": 0.04,
+      "grad_norm": 4.416736002443796,
+      "learning_rate": 4.9749874937468736e-05,
+      "loss": 0.3894,
+      "step": 30
+    },
+    {
+      "epoch": 0.05,
+      "grad_norm": 5.588328371365274,
+      "learning_rate": 4.9624812406203106e-05,
+      "loss": 0.4047,
+      "step": 40
+    },
+    {
+      "epoch": 0.06,
+      "grad_norm": 4.53230725812265,
+      "learning_rate": 4.9499749874937476e-05,
+      "loss": 0.4542,
+      "step": 50
+    },
+    {
+      "epoch": 0.07,
+      "grad_norm": 4.826575762136805,
+      "learning_rate": 4.937468734367184e-05,
+      "loss": 0.4522,
+      "step": 60
+    },
+    {
+      "epoch": 0.09,
+      "grad_norm": 4.525790589117328,
+      "learning_rate": 4.92496248124062e-05,
+      "loss": 0.4822,
+      "step": 70
+    },
+    {
+      "epoch": 0.1,
+      "grad_norm": 5.296156572891334,
+      "learning_rate": 4.912456228114057e-05,
+      "loss": 0.5055,
+      "step": 80
+    },
+    {
+      "epoch": 0.11,
+      "grad_norm": 5.819088718868824,
+      "learning_rate": 4.8999499749874936e-05,
+      "loss": 0.473,
+      "step": 90
+    },
+    {
+      "epoch": 0.12,
+      "grad_norm": 5.433765575947364,
+      "learning_rate": 4.887443721860931e-05,
+      "loss": 0.5087,
+      "step": 100
+    },
+    {
+      "epoch": 0.14,
+      "grad_norm": 5.0965740589803925,
+      "learning_rate": 4.8749374687343676e-05,
+      "loss": 0.5234,
+      "step": 110
+    },
+    {
+      "epoch": 0.15,
+      "grad_norm": 5.850601504449537,
+      "learning_rate": 4.862431215607804e-05,
+      "loss": 0.5141,
+      "step": 120
+    },
+    {
+      "epoch": 0.16,
+      "grad_norm": 4.719305144742639,
+      "learning_rate": 4.849924962481241e-05,
+      "loss": 0.5434,
+      "step": 130
+    },
+    {
+      "epoch": 0.17,
+      "grad_norm": 5.005327628861193,
+      "learning_rate": 4.837418709354677e-05,
+      "loss": 0.5293,
+      "step": 140
+    },
+    {
+      "epoch": 0.19,
+      "grad_norm": 5.146773444206217,
+      "learning_rate": 4.824912456228114e-05,
+      "loss": 0.511,
+      "step": 150
+    },
+    {
+      "epoch": 0.2,
+      "grad_norm": 4.612446424385032,
+      "learning_rate": 4.812406203101551e-05,
+      "loss": 0.5401,
+      "step": 160
+    },
+    {
+      "epoch": 0.21,
+      "grad_norm": 5.293068567439793,
+      "learning_rate": 4.7998999499749876e-05,
+      "loss": 0.6019,
+      "step": 170
+    },
+    {
+      "epoch": 0.23,
+      "grad_norm": 5.1015740901237026,
+      "learning_rate": 4.7873936968484246e-05,
+      "loss": 0.5633,
+      "step": 180
+    },
+    {
+      "epoch": 0.24,
+      "grad_norm": 4.824851299623291,
+      "learning_rate": 4.774887443721861e-05,
+      "loss": 0.583,
+      "step": 190
+    },
+    {
+      "epoch": 0.25,
+      "grad_norm": 5.062655599581182,
+      "learning_rate": 4.762381190595298e-05,
+      "loss": 0.6201,
+      "step": 200
+    },
+    {
+      "epoch": 0.26,
+      "grad_norm": 5.326127827444376,
+      "learning_rate": 4.749874937468735e-05,
+      "loss": 0.6125,
+      "step": 210
+    },
+    {
+      "epoch": 0.28,
+      "grad_norm": 4.9596551153810475,
+      "learning_rate": 4.737368684342171e-05,
+      "loss": 0.5686,
+      "step": 220
+    },
+    {
+      "epoch": 0.29,
+      "grad_norm": 4.658617889684648,
+      "learning_rate": 4.724862431215608e-05,
+      "loss": 0.6173,
+      "step": 230
+    },
+    {
+      "epoch": 0.3,
+      "grad_norm": 4.7094963235295015,
+      "learning_rate": 4.7123561780890446e-05,
+      "loss": 0.6373,
+      "step": 240
+    },
+    {
+      "epoch": 0.31,
+      "grad_norm": 5.322370626831193,
+      "learning_rate": 4.6998499249624816e-05,
+      "loss": 0.5606,
+      "step": 250
+    },
+    {
+      "epoch": 0.33,
+      "grad_norm": 4.959609351002434,
+      "learning_rate": 4.687343671835918e-05,
+      "loss": 0.5903,
+      "step": 260
+    },
+    {
+      "epoch": 0.34,
+      "grad_norm": 5.7940212741786485,
+      "learning_rate": 4.674837418709355e-05,
+      "loss": 0.6201,
+      "step": 270
+    },
+    {
+      "epoch": 0.35,
+      "grad_norm": 6.241362745205275,
+      "learning_rate": 4.662331165582792e-05,
+      "loss": 0.6295,
+      "step": 280
+    },
+    {
+      "epoch": 0.36,
+      "grad_norm": 4.421199895323022,
+      "learning_rate": 4.649824912456228e-05,
+      "loss": 0.6242,
+      "step": 290
+    },
+    {
+      "epoch": 0.38,
+      "grad_norm": 5.693381150794548,
+      "learning_rate": 4.637318659329665e-05,
+      "loss": 0.6343,
+      "step": 300
+    },
+    {
+      "epoch": 0.39,
+      "grad_norm": 5.84570997692131,
+      "learning_rate": 4.6248124062031015e-05,
+      "loss": 0.6293,
+      "step": 310
+    },
+    {
+      "epoch": 0.4,
+      "grad_norm": 4.891858807526382,
+      "learning_rate": 4.6123061530765386e-05,
+      "loss": 0.6699,
+      "step": 320
+    },
+    {
+      "epoch": 0.41,
+      "grad_norm": 5.476686832687266,
+      "learning_rate": 4.5997998999499756e-05,
+      "loss": 0.6448,
+      "step": 330
+    },
+    {
+      "epoch": 0.42,
+      "grad_norm": 5.629121076096647,
+      "learning_rate": 4.587293646823412e-05,
+      "loss": 0.6776,
+      "step": 340
+    },
+    {
+      "epoch": 0.44,
+      "grad_norm": 5.6676894368204485,
+      "learning_rate": 4.574787393696849e-05,
+      "loss": 0.6558,
+      "step": 350
+    },
+    {
+      "epoch": 0.45,
+      "grad_norm": 5.28952155100302,
+      "learning_rate": 4.562281140570285e-05,
+      "loss": 0.705,
+      "step": 360
+    },
+    {
+      "epoch": 0.46,
+      "grad_norm": 4.89682598466814,
+      "learning_rate": 4.5497748874437215e-05,
+      "loss": 0.6564,
+      "step": 370
+    },
+    {
+      "epoch": 0.47,
+      "grad_norm": 4.959965455950147,
+      "learning_rate": 4.5372686343171585e-05,
+      "loss": 0.6734,
+      "step": 380
+    },
+    {
+      "epoch": 0.49,
+      "grad_norm": 5.938681073424107,
+      "learning_rate": 4.5247623811905955e-05,
+      "loss": 0.7523,
+      "step": 390
+    },
+    {
+      "epoch": 0.5,
+      "grad_norm": 5.35665922615818,
+      "learning_rate": 4.5122561280640325e-05,
+      "loss": 0.7124,
+      "step": 400
+    },
+    {
+      "epoch": 0.51,
+      "grad_norm": 4.8123046042668625,
+      "learning_rate": 4.499749874937469e-05,
+      "loss": 0.702,
+      "step": 410
+    },
+    {
+      "epoch": 0.53,
+      "grad_norm": 5.411392852808034,
+      "learning_rate": 4.487243621810905e-05,
+      "loss": 0.6917,
+      "step": 420
+    },
+    {
+      "epoch": 0.54,
+      "grad_norm": 5.009053235776962,
+      "learning_rate": 4.474737368684342e-05,
+      "loss": 0.675,
+      "step": 430
+    },
+    {
+      "epoch": 0.55,
+      "grad_norm": 5.399248296904937,
+      "learning_rate": 4.462231115557779e-05,
+      "loss": 0.6916,
+      "step": 440
+    },
+    {
+      "epoch": 0.56,
+      "grad_norm": 5.195245498569367,
+      "learning_rate": 4.449724862431216e-05,
+      "loss": 0.6664,
+      "step": 450
+    },
+    {
+      "epoch": 0.57,
+      "grad_norm": 6.064114070634714,
+      "learning_rate": 4.4372186093046525e-05,
+      "loss": 0.7281,
+      "step": 460
+    },
+    {
+      "epoch": 0.59,
+      "grad_norm": 4.9950124660638044,
+      "learning_rate": 4.424712356178089e-05,
+      "loss": 0.6826,
+      "step": 470
+    },
+    {
+      "epoch": 0.6,
+      "grad_norm": 4.4353585247328,
+      "learning_rate": 4.412206103051526e-05,
+      "loss": 0.6919,
+      "step": 480
+    },
+    {
+      "epoch": 0.61,
+      "grad_norm": 5.289699318891135,
+      "learning_rate": 4.399699849924963e-05,
+      "loss": 0.689,
+      "step": 490
+    },
+    {
+      "epoch": 0.62,
+      "grad_norm": 6.126514792457231,
+      "learning_rate": 4.3871935967984e-05,
+      "loss": 0.7207,
+      "step": 500
+    },
+    {
+      "epoch": 0.64,
+      "grad_norm": 5.181000464013246,
+      "learning_rate": 4.374687343671836e-05,
+      "loss": 0.7047,
+      "step": 510
+    },
+    {
+      "epoch": 0.65,
+      "grad_norm": 6.228936674875526,
+      "learning_rate": 4.3621810905452725e-05,
+      "loss": 0.6799,
+      "step": 520
+    },
+    {
+      "epoch": 0.66,
+      "grad_norm": 4.499920526438516,
+      "learning_rate": 4.3496748374187095e-05,
+      "loss": 0.7035,
+      "step": 530
+    },
+    {
+      "epoch": 0.68,
+      "grad_norm": 5.258199737028568,
+      "learning_rate": 4.337168584292146e-05,
+      "loss": 0.7166,
+      "step": 540
+    },
+    {
+      "epoch": 0.69,
+      "grad_norm": 5.078292703060163,
+      "learning_rate": 4.324662331165583e-05,
+      "loss": 0.7128,
+      "step": 550
+    },
+    {
+      "epoch": 0.7,
+      "grad_norm": 5.150500612093525,
+      "learning_rate": 4.31215607803902e-05,
+      "loss": 0.7314,
+      "step": 560
+    },
+    {
+      "epoch": 0.71,
+      "grad_norm": 4.789457967587879,
+      "learning_rate": 4.299649824912456e-05,
+      "loss": 0.7351,
+      "step": 570
+    },
+    {
+      "epoch": 0.72,
+      "grad_norm": 5.462318518655166,
+      "learning_rate": 4.287143571785893e-05,
+      "loss": 0.7117,
+      "step": 580
+    },
+    {
+      "epoch": 0.74,
+      "grad_norm": 6.448282045561541,
+      "learning_rate": 4.2746373186593295e-05,
+      "loss": 0.7075,
+      "step": 590
+    },
+    {
+      "epoch": 0.75,
+      "grad_norm": 6.337874374016145,
+      "learning_rate": 4.2621310655327665e-05,
+      "loss": 0.7332,
+      "step": 600
+    },
+    {
+      "epoch": 0.76,
+      "grad_norm": 4.904355017265367,
+      "learning_rate": 4.2496248124062035e-05,
+      "loss": 0.6525,
+      "step": 610
+    },
+    {
+      "epoch": 0.78,
+      "grad_norm": 5.428105097653496,
+      "learning_rate": 4.2371185592796405e-05,
+      "loss": 0.734,
+      "step": 620
+    },
+    {
+      "epoch": 0.79,
+      "grad_norm": 5.489902764520662,
+      "learning_rate": 4.224612306153077e-05,
+      "loss": 0.665,
+      "step": 630
+    },
+    {
+      "epoch": 0.8,
+      "grad_norm": 4.81269221727027,
+      "learning_rate": 4.212106053026513e-05,
+      "loss": 0.7485,
+      "step": 640
+    },
+    {
+      "epoch": 0.81,
+      "grad_norm": 5.90221965868893,
+      "learning_rate": 4.19959979989995e-05,
+      "loss": 0.7342,
+      "step": 650
+    },
+    {
+      "epoch": 0.82,
+      "grad_norm": 5.7730450116082315,
+      "learning_rate": 4.1870935467733865e-05,
+      "loss": 0.7082,
+      "step": 660
+    },
+    {
+      "epoch": 0.84,
+      "grad_norm": 4.274199534648249,
+      "learning_rate": 4.174587293646824e-05,
+      "loss": 0.7262,
+      "step": 670
+    },
+    {
+      "epoch": 0.85,
+      "grad_norm": 5.348282169245115,
+      "learning_rate": 4.1620810405202605e-05,
+      "loss": 0.6719,
+      "step": 680
+    },
+    {
+      "epoch": 0.86,
+      "grad_norm": 5.920417706986164,
+      "learning_rate": 4.149574787393697e-05,
+      "loss": 0.7369,
+      "step": 690
+    },
+    {
+      "epoch": 0.88,
+      "grad_norm": 5.466281181017966,
+      "learning_rate": 4.137068534267134e-05,
+      "loss": 0.7526,
+      "step": 700
+    },
+    {
+      "epoch": 0.89,
+      "grad_norm": 5.421656535471705,
+      "learning_rate": 4.12456228114057e-05,
+      "loss": 0.6593,
+      "step": 710
+    },
+    {
+      "epoch": 0.9,
+      "grad_norm": 4.971718914811608,
+      "learning_rate": 4.112056028014007e-05,
+      "loss": 0.7003,
+      "step": 720
+    },
+    {
+      "epoch": 0.91,
+      "grad_norm": 5.770908323092229,
+      "learning_rate": 4.099549774887444e-05,
+      "loss": 0.7414,
+      "step": 730
+    },
+    {
+      "epoch": 0.93,
+      "grad_norm": 6.190279249754522,
+      "learning_rate": 4.0870435217608805e-05,
+      "loss": 0.7034,
+      "step": 740
+    },
+    {
+      "epoch": 0.94,
+      "grad_norm": 5.246490395350578,
+      "learning_rate": 4.0745372686343175e-05,
+      "loss": 0.6962,
+      "step": 750
+    },
+    {
+      "epoch": 0.95,
+      "grad_norm": 5.581112870841728,
+      "learning_rate": 4.062031015507754e-05,
+      "loss": 0.7137,
+      "step": 760
+    },
+    {
+      "epoch": 0.96,
+      "grad_norm": 5.4016319352096325,
+      "learning_rate": 4.049524762381191e-05,
+      "loss": 0.7035,
+      "step": 770
+    },
+    {
+      "epoch": 0.97,
+      "grad_norm": 4.461968885599671,
+      "learning_rate": 4.037018509254628e-05,
+      "loss": 0.7133,
+      "step": 780
+    },
+    {
+      "epoch": 0.99,
+      "grad_norm": 4.644183814012947,
+      "learning_rate": 4.024512256128064e-05,
+      "loss": 0.7161,
+      "step": 790
+    },
+    {
+      "epoch": 1.0,
+      "grad_norm": 4.826017148028126,
+      "learning_rate": 4.012006003001501e-05,
+      "loss": 0.722,
+      "step": 800
+    },
+    {
+      "epoch": 1.01,
+      "grad_norm": 4.000342354428771,
+      "learning_rate": 3.9994997498749375e-05,
+      "loss": 0.3054,
+      "step": 810
+    },
+    {
+      "epoch": 1.02,
+      "grad_norm": 3.8374628270240616,
+      "learning_rate": 3.9869934967483745e-05,
+      "loss": 0.3016,
+      "step": 820
+    },
+    {
+      "epoch": 1.04,
+      "grad_norm": 5.107383783318757,
+      "learning_rate": 3.974487243621811e-05,
+      "loss": 0.3077,
+      "step": 830
+    },
+    {
+      "epoch": 1.05,
+      "grad_norm": 3.4427943293478123,
+      "learning_rate": 3.961980990495248e-05,
+      "loss": 0.2966,
+      "step": 840
+    },
+    {
+      "epoch": 1.06,
+      "grad_norm": 3.8325782253297174,
+      "learning_rate": 3.949474737368685e-05,
+      "loss": 0.3203,
+      "step": 850
+    },
+    {
+      "epoch": 1.07,
+      "grad_norm": 4.236035011038753,
+      "learning_rate": 3.936968484242121e-05,
+      "loss": 0.3019,
+      "step": 860
+    },
+    {
+      "epoch": 1.09,
+      "grad_norm": 4.491576469992581,
+      "learning_rate": 3.924462231115558e-05,
+      "loss": 0.325,
+      "step": 870
+    },
+    {
+      "epoch": 1.1,
+      "grad_norm": 3.961749774177233,
+      "learning_rate": 3.9119559779889945e-05,
+      "loss": 0.3371,
+      "step": 880
+    },
+    {
+      "epoch": 1.11,
+      "grad_norm": 4.672476739310163,
+      "learning_rate": 3.8994497248624315e-05,
+      "loss": 0.3306,
+      "step": 890
+    },
+    {
+      "epoch": 1.12,
+      "grad_norm": 3.810555384248572,
+      "learning_rate": 3.8869434717358685e-05,
+      "loss": 0.2961,
+      "step": 900
+    },
+    {
+      "epoch": 1.14,
+      "grad_norm": 3.662542292119634,
+      "learning_rate": 3.874437218609305e-05,
+      "loss": 0.3159,
+      "step": 910
+    },
+    {
+      "epoch": 1.15,
+      "grad_norm": 4.615080015205904,
+      "learning_rate": 3.861930965482742e-05,
+      "loss": 0.3321,
+      "step": 920
+    },
+    {
+      "epoch": 1.16,
+      "grad_norm": 3.5514929197773317,
+      "learning_rate": 3.849424712356178e-05,
+      "loss": 0.316,
+      "step": 930
+    },
+    {
+      "epoch": 1.18,
+      "grad_norm": 3.925218593074527,
+      "learning_rate": 3.8369184592296145e-05,
+      "loss": 0.3539,
+      "step": 940
+    },
+    {
+      "epoch": 1.19,
+      "grad_norm": 3.9002231338443205,
+      "learning_rate": 3.824412206103052e-05,
+      "loss": 0.3293,
+      "step": 950
+    },
+    {
+      "epoch": 1.2,
+      "grad_norm": 3.8671256012249478,
+      "learning_rate": 3.8119059529764885e-05,
+      "loss": 0.3169,
+      "step": 960
+    },
+    {
+      "epoch": 1.21,
+      "grad_norm": 3.61716640787587,
+      "learning_rate": 3.7993996998499255e-05,
+      "loss": 0.3321,
+      "step": 970
+    },
+    {
+      "epoch": 1.23,
+      "grad_norm": 4.137379652042079,
+      "learning_rate": 3.786893446723362e-05,
+      "loss": 0.3418,
+      "step": 980
+    },
+    {
+      "epoch": 1.24,
+      "grad_norm": 4.191073131566738,
+      "learning_rate": 3.774387193596798e-05,
+      "loss": 0.3277,
+      "step": 990
+    },
+    {
+      "epoch": 1.25,
+      "grad_norm": 4.654388465252297,
+      "learning_rate": 3.761880940470235e-05,
+      "loss": 0.3371,
+      "step": 1000
+    },
+    {
+      "epoch": 1.26,
+      "grad_norm": 4.900788395574059,
+      "learning_rate": 3.749374687343672e-05,
+      "loss": 0.347,
+      "step": 1010
+    },
+    {
+      "epoch": 1.27,
+      "grad_norm": 3.990193385932706,
+      "learning_rate": 3.736868434217109e-05,
+      "loss": 0.3358,
+      "step": 1020
+    },
+    {
+      "epoch": 1.29,
+      "grad_norm": 4.270444278519215,
+      "learning_rate": 3.7243621810905455e-05,
+      "loss": 0.3461,
+      "step": 1030
+    },
+    {
+      "epoch": 1.3,
+      "grad_norm": 4.286422166946011,
+      "learning_rate": 3.711855927963982e-05,
+      "loss": 0.3362,
+      "step": 1040
+    },
+    {
+      "epoch": 1.31,
+      "grad_norm": 4.78341924787361,
+      "learning_rate": 3.699349674837419e-05,
+      "loss": 0.3418,
+      "step": 1050
+    },
+    {
+      "epoch": 1.32,
+      "grad_norm": 4.193755303051207,
+      "learning_rate": 3.686843421710856e-05,
+      "loss": 0.3372,
+      "step": 1060
+    },
+    {
+      "epoch": 1.34,
+      "grad_norm": 3.989814545760004,
+      "learning_rate": 3.674337168584293e-05,
+      "loss": 0.3254,
+      "step": 1070
+    },
+    {
+      "epoch": 1.35,
+      "grad_norm": 3.823337597588243,
+      "learning_rate": 3.661830915457729e-05,
+      "loss": 0.3471,
+      "step": 1080
+    },
+    {
+      "epoch": 1.36,
+      "grad_norm": 3.793526762626974,
+      "learning_rate": 3.6493246623311655e-05,
+      "loss": 0.3432,
+      "step": 1090
+    },
+    {
+      "epoch": 1.38,
+      "grad_norm": 3.872109227027026,
+      "learning_rate": 3.6368184092046025e-05,
+      "loss": 0.3533,
+      "step": 1100
+    },
+    {
+      "epoch": 1.39,
+      "grad_norm": 4.437307219280221,
+      "learning_rate": 3.624312156078039e-05,
+      "loss": 0.3551,
+      "step": 1110
+    },
+    {
+      "epoch": 1.4,
+      "grad_norm": 4.298491572681636,
+      "learning_rate": 3.611805902951476e-05,
+      "loss": 0.3321,
+      "step": 1120
+    },
+    {
+      "epoch": 1.41,
+      "grad_norm": 3.9890804256002084,
+      "learning_rate": 3.599299649824913e-05,
+      "loss": 0.3339,
+      "step": 1130
+    },
+    {
+      "epoch": 1.43,
+      "grad_norm": 4.437500429825024,
+      "learning_rate": 3.586793396698349e-05,
+      "loss": 0.3272,
+      "step": 1140
+    },
+    {
+      "epoch": 1.44,
+      "grad_norm": 4.218831379423431,
+      "learning_rate": 3.574287143571786e-05,
+      "loss": 0.3527,
+      "step": 1150
+    },
+    {
+      "epoch": 1.45,
+      "grad_norm": 3.747375332703581,
+      "learning_rate": 3.5617808904452225e-05,
+      "loss": 0.3513,
+      "step": 1160
+    },
+    {
+      "epoch": 1.46,
+      "grad_norm": 4.0875968991614355,
+      "learning_rate": 3.5492746373186595e-05,
+      "loss": 0.339,
+      "step": 1170
+    },
+    {
+      "epoch": 1.48,
+      "grad_norm": 4.039233442042736,
+      "learning_rate": 3.5367683841920965e-05,
+      "loss": 0.3511,
+      "step": 1180
+    },
+    {
+      "epoch": 1.49,
+      "grad_norm": 3.5825847169026135,
+      "learning_rate": 3.524262131065533e-05,
+      "loss": 0.3349,
+      "step": 1190
+    },
+    {
+      "epoch": 1.5,
+      "grad_norm": 4.175820759942294,
+      "learning_rate": 3.51175587793897e-05,
+      "loss": 0.3619,
+      "step": 1200
+    },
+    {
+      "epoch": 1.51,
+      "grad_norm": 4.714008298365686,
+      "learning_rate": 3.499249624812406e-05,
+      "loss": 0.3487,
+      "step": 1210
+    },
+    {
+      "epoch": 1.52,
+      "grad_norm": 3.9156939062827925,
+      "learning_rate": 3.486743371685843e-05,
+      "loss": 0.3293,
+      "step": 1220
+    },
+    {
+      "epoch": 1.54,
+      "grad_norm": 3.922432544485633,
+      "learning_rate": 3.4742371185592795e-05,
+      "loss": 0.3424,
+      "step": 1230
+    },
+    {
+      "epoch": 1.55,
+      "grad_norm": 3.871337267070979,
+      "learning_rate": 3.4617308654327165e-05,
+      "loss": 0.3453,
+      "step": 1240
+    },
+    {
+      "epoch": 1.56,
+      "grad_norm": 4.563498191855707,
+      "learning_rate": 3.4492246123061535e-05,
+      "loss": 0.3576,
+      "step": 1250
+    },
+    {
+      "epoch": 1.57,
+      "grad_norm": 4.11645130618609,
+      "learning_rate": 3.43671835917959e-05,
+      "loss": 0.3597,
+      "step": 1260
+    },
+    {
+      "epoch": 1.59,
+      "grad_norm": 4.38717548441809,
+      "learning_rate": 3.424212106053027e-05,
+      "loss": 0.3274,
+      "step": 1270
+    },
+    {
+      "epoch": 1.6,
+      "grad_norm": 4.400823403388009,
+      "learning_rate": 3.411705852926463e-05,
+      "loss": 0.345,
+      "step": 1280
+    },
+    {
+      "epoch": 1.61,
+      "grad_norm": 4.767739149075579,
+      "learning_rate": 3.3991995997999e-05,
+      "loss": 0.3432,
+      "step": 1290
+    },
+    {
+      "epoch": 1.62,
+      "grad_norm": 3.879637250755376,
+      "learning_rate": 3.386693346673337e-05,
+      "loss": 0.3643,
+      "step": 1300
+    },
+    {
+      "epoch": 1.64,
+      "grad_norm": 3.7442488756684345,
+      "learning_rate": 3.3741870935467735e-05,
+      "loss": 0.3488,
+      "step": 1310
+    },
+    {
+      "epoch": 1.65,
+      "grad_norm": 3.8345905122140884,
+      "learning_rate": 3.3616808404202105e-05,
+      "loss": 0.3505,
+      "step": 1320
+    },
+    {
+      "epoch": 1.66,
+      "grad_norm": 4.9315147338510945,
+      "learning_rate": 3.349174587293647e-05,
+      "loss": 0.3577,
+      "step": 1330
+    },
+    {
+      "epoch": 1.68,
+      "grad_norm": 4.151768383452871,
+      "learning_rate": 3.336668334167083e-05,
+      "loss": 0.3471,
+      "step": 1340
+    },
+    {
+      "epoch": 1.69,
+      "grad_norm": 3.751703511185265,
+      "learning_rate": 3.324162081040521e-05,
+      "loss": 0.3489,
+      "step": 1350
+    },
+    {
+      "epoch": 1.7,
+      "grad_norm": 4.52533794005262,
+      "learning_rate": 3.311655827913957e-05,
+      "loss": 0.3362,
+      "step": 1360
+    },
+    {
+      "epoch": 1.71,
+      "grad_norm": 3.938063323883952,
+      "learning_rate": 3.299149574787394e-05,
+      "loss": 0.3496,
+      "step": 1370
+    },
+    {
+      "epoch": 1.73,
+      "grad_norm": 4.321107980845758,
+      "learning_rate": 3.2866433216608305e-05,
+      "loss": 0.3565,
+      "step": 1380
+    },
+    {
+      "epoch": 1.74,
+      "grad_norm": 4.180361684455809,
+      "learning_rate": 3.274137068534267e-05,
+      "loss": 0.3419,
+      "step": 1390
+    },
+    {
+      "epoch": 1.75,
+      "grad_norm": 4.073762043329108,
+      "learning_rate": 3.261630815407704e-05,
+      "loss": 0.3368,
+      "step": 1400
+    },
+    {
+      "epoch": 1.76,
+      "grad_norm": 3.7744796293221623,
+      "learning_rate": 3.249124562281141e-05,
+      "loss": 0.34,
+      "step": 1410
+    },
+    {
+      "epoch": 1.77,
+      "grad_norm": 4.272009909229001,
+      "learning_rate": 3.236618309154578e-05,
+      "loss": 0.346,
+      "step": 1420
+    },
+    {
+      "epoch": 1.79,
+      "grad_norm": 4.400007005165766,
+      "learning_rate": 3.224112056028014e-05,
+      "loss": 0.3422,
+      "step": 1430
+    },
+    {
+      "epoch": 1.8,
+      "grad_norm": 4.245867346769938,
+      "learning_rate": 3.2116058029014505e-05,
+      "loss": 0.3457,
+      "step": 1440
+    },
+    {
+      "epoch": 1.81,
+      "grad_norm": 4.076869025792562,
+      "learning_rate": 3.1990995497748875e-05,
+      "loss": 0.3428,
+      "step": 1450
+    },
+    {
+      "epoch": 1.82,
+      "grad_norm": 3.992331306271212,
+      "learning_rate": 3.1865932966483245e-05,
+      "loss": 0.3423,
+      "step": 1460
+    },
+    {
+      "epoch": 1.84,
+      "grad_norm": 4.329708099040903,
+      "learning_rate": 3.1740870435217615e-05,
+      "loss": 0.3353,
+      "step": 1470
+    },
+    {
+      "epoch": 1.85,
+      "grad_norm": 4.228767990087569,
+      "learning_rate": 3.161580790395198e-05,
+      "loss": 0.3313,
+      "step": 1480
+    },
+    {
+      "epoch": 1.86,
+      "grad_norm": 4.100143216120651,
+      "learning_rate": 3.149074537268634e-05,
+      "loss": 0.3448,
+      "step": 1490
+    },
+    {
+      "epoch": 1.88,
+      "grad_norm": 4.846420241173142,
+      "learning_rate": 3.136568284142071e-05,
+      "loss": 0.3515,
+      "step": 1500
+    },
+    {
+      "epoch": 1.89,
+      "grad_norm": 4.287502873811787,
+      "learning_rate": 3.1240620310155075e-05,
+      "loss": 0.3462,
+      "step": 1510
+    },
+    {
+      "epoch": 1.9,
+      "grad_norm": 3.769572994020805,
+      "learning_rate": 3.111555777888945e-05,
+      "loss": 0.3593,
+      "step": 1520
+    },
+    {
+      "epoch": 1.91,
+      "grad_norm": 3.7731122178800645,
+      "learning_rate": 3.0990495247623815e-05,
+      "loss": 0.3467,
+      "step": 1530
+    },
+    {
+      "epoch": 1.93,
+      "grad_norm": 4.257071726342832,
+      "learning_rate": 3.086543271635818e-05,
+      "loss": 0.3447,
+      "step": 1540
+    },
+    {
+      "epoch": 1.94,
+      "grad_norm": 4.192530556773847,
+      "learning_rate": 3.074037018509255e-05,
+      "loss": 0.3575,
+      "step": 1550
+    },
+    {
+      "epoch": 1.95,
+      "grad_norm": 4.488222817355165,
+      "learning_rate": 3.061530765382691e-05,
+      "loss": 0.3392,
+      "step": 1560
+    },
+    {
+      "epoch": 1.96,
+      "grad_norm": 4.166714909592134,
+      "learning_rate": 3.0490245122561285e-05,
+      "loss": 0.341,
+      "step": 1570
+    },
+    {
+      "epoch": 1.98,
+      "grad_norm": 4.2906704582812445,
+      "learning_rate": 3.0365182591295648e-05,
+      "loss": 0.3337,
+      "step": 1580
+    },
+    {
+      "epoch": 1.99,
+      "grad_norm": 4.631668644360512,
+      "learning_rate": 3.0240120060030014e-05,
+      "loss": 0.3352,
+      "step": 1590
+    },
+    {
+      "epoch": 2.0,
+      "grad_norm": 3.6817020158177343,
+      "learning_rate": 3.0115057528764385e-05,
+      "loss": 0.338,
+      "step": 1600
+    },
+    {
+      "epoch": 2.01,
+      "grad_norm": 2.4770673844893736,
+      "learning_rate": 2.9989994997498748e-05,
+      "loss": 0.1203,
+      "step": 1610
+    },
+    {
+      "epoch": 2.02,
+      "grad_norm": 2.8486557450546752,
+      "learning_rate": 2.986493246623312e-05,
+      "loss": 0.1226,
+      "step": 1620
+    },
+    {
+      "epoch": 2.04,
+      "grad_norm": 2.094604545543018,
+      "learning_rate": 2.9739869934967484e-05,
+      "loss": 0.1236,
+      "step": 1630
+    },
+    {
+      "epoch": 2.05,
+      "grad_norm": 2.892529659135666,
+      "learning_rate": 2.961480740370185e-05,
+      "loss": 0.1242,
+      "step": 1640
+    },
+    {
+      "epoch": 2.06,
+      "grad_norm": 3.0607648332468993,
+      "learning_rate": 2.948974487243622e-05,
+      "loss": 0.1243,
+      "step": 1650
+    },
+    {
+      "epoch": 2.08,
+      "grad_norm": 3.31204465219556,
+      "learning_rate": 2.9364682341170584e-05,
+      "loss": 0.1235,
+      "step": 1660
+    },
+    {
+      "epoch": 2.09,
+      "grad_norm": 2.4729852201434306,
+      "learning_rate": 2.9239619809904954e-05,
+      "loss": 0.1226,
+      "step": 1670
+    },
+    {
+      "epoch": 2.1,
+      "grad_norm": 2.5257406678082583,
+      "learning_rate": 2.911455727863932e-05,
+      "loss": 0.1207,
+      "step": 1680
+    },
+    {
+      "epoch": 2.11,
+      "grad_norm": 3.103672799030984,
+      "learning_rate": 2.8989494747373684e-05,
+      "loss": 0.1324,
+      "step": 1690
+    },
+    {
+      "epoch": 2.12,
+      "grad_norm": 2.3782595301910043,
+      "learning_rate": 2.8864432216108054e-05,
+      "loss": 0.132,
+      "step": 1700
+    },
+    {
+      "epoch": 2.14,
+      "grad_norm": 2.6059740025591505,
+      "learning_rate": 2.873936968484242e-05,
+      "loss": 0.1264,
+      "step": 1710
+    },
+    {
+      "epoch": 2.15,
+      "grad_norm": 2.538398820262338,
+      "learning_rate": 2.861430715357679e-05,
+      "loss": 0.1366,
+      "step": 1720
+    },
+    {
+      "epoch": 2.16,
+      "grad_norm": 2.499812214474419,
+      "learning_rate": 2.8489244622311158e-05,
+      "loss": 0.1377,
+      "step": 1730
+    },
+    {
+      "epoch": 2.17,
+      "grad_norm": 2.3941826372137305,
+      "learning_rate": 2.836418209104552e-05,
+      "loss": 0.1225,
+      "step": 1740
+    },
+    {
+      "epoch": 2.19,
+      "grad_norm": 2.6244609824311014,
+      "learning_rate": 2.823911955977989e-05,
+      "loss": 0.1466,
+      "step": 1750
+    },
+    {
+      "epoch": 2.2,
+      "grad_norm": 2.5276124979370915,
+      "learning_rate": 2.8114057028514258e-05,
+      "loss": 0.1415,
+      "step": 1760
+    },
+    {
+      "epoch": 2.21,
+      "grad_norm": 2.916618637416117,
+      "learning_rate": 2.7988994497248628e-05,
+      "loss": 0.1309,
+      "step": 1770
+    },
+    {
+      "epoch": 2.23,
+      "grad_norm": 2.753008410683784,
+      "learning_rate": 2.786393196598299e-05,
+      "loss": 0.1432,
+      "step": 1780
+    },
+    {
+      "epoch": 2.24,
+      "grad_norm": 3.1016622762701935,
+      "learning_rate": 2.7738869434717364e-05,
+      "loss": 0.1366,
+      "step": 1790
+    },
+    {
+      "epoch": 2.25,
+      "grad_norm": 2.7966281206508032,
+      "learning_rate": 2.7613806903451728e-05,
+      "loss": 0.1408,
+      "step": 1800
+    },
+    {
+      "epoch": 2.26,
+      "grad_norm": 2.439577928684241,
+      "learning_rate": 2.7488744372186094e-05,
+      "loss": 0.1474,
+      "step": 1810
+    },
+    {
+      "epoch": 2.27,
+      "grad_norm": 2.5528279574109884,
+      "learning_rate": 2.7363681840920464e-05,
+      "loss": 0.1393,
+      "step": 1820
+    },
+    {
+      "epoch": 2.29,
+      "grad_norm": 2.882844754170427,
+      "learning_rate": 2.7238619309654828e-05,
+      "loss": 0.1271,
+      "step": 1830
+    },
+    {
+      "epoch": 2.3,
+      "grad_norm": 3.1442381159468424,
+      "learning_rate": 2.7113556778389198e-05,
+      "loss": 0.1401,
+      "step": 1840
+    },
+    {
+      "epoch": 2.31,
+      "grad_norm": 2.3171759061722663,
+      "learning_rate": 2.6988494247123564e-05,
+      "loss": 0.1363,
+      "step": 1850
+    },
+    {
+      "epoch": 2.33,
+      "grad_norm": 2.5545785863215533,
+      "learning_rate": 2.6863431715857928e-05,
+      "loss": 0.1346,
+      "step": 1860
+    },
+    {
+      "epoch": 2.34,
+      "grad_norm": 2.6824086777926412,
+      "learning_rate": 2.6738369184592298e-05,
+      "loss": 0.1341,
+      "step": 1870
+    },
+    {
+      "epoch": 2.35,
+      "grad_norm": 2.935390607845893,
+      "learning_rate": 2.6613306653326664e-05,
+      "loss": 0.1365,
+      "step": 1880
+    },
+    {
+      "epoch": 2.36,
+      "grad_norm": 2.6179278337223364,
+      "learning_rate": 2.6488244122061034e-05,
+      "loss": 0.1302,
+      "step": 1890
+    },
+    {
+      "epoch": 2.38,
+      "grad_norm": 2.5835120129063265,
+      "learning_rate": 2.63631815907954e-05,
+      "loss": 0.1437,
+      "step": 1900
+    },
+    {
+      "epoch": 2.39,
+      "grad_norm": 2.406526227676954,
+      "learning_rate": 2.6238119059529764e-05,
+      "loss": 0.1329,
+      "step": 1910
+    },
+    {
+      "epoch": 2.4,
+      "grad_norm": 2.516912379701269,
+      "learning_rate": 2.6113056528264134e-05,
+      "loss": 0.1307,
+      "step": 1920
+    },
+    {
+      "epoch": 2.41,
+      "grad_norm": 2.4927236524388294,
+      "learning_rate": 2.59879939969985e-05,
+      "loss": 0.1345,
+      "step": 1930
+    },
+    {
+      "epoch": 2.42,
+      "grad_norm": 2.6724439762477457,
+      "learning_rate": 2.586293146573287e-05,
+      "loss": 0.134,
+      "step": 1940
+    },
+    {
+      "epoch": 2.44,
+      "grad_norm": 2.637033763011149,
+      "learning_rate": 2.5737868934467234e-05,
+      "loss": 0.1287,
+      "step": 1950
+    },
+    {
+      "epoch": 2.45,
+      "grad_norm": 2.7214532326123377,
+      "learning_rate": 2.56128064032016e-05,
+      "loss": 0.157,
+      "step": 1960
+    },
+    {
+      "epoch": 2.46,
+      "grad_norm": 2.3869393054803205,
+      "learning_rate": 2.548774387193597e-05,
+      "loss": 0.124,
+      "step": 1970
+    },
+    {
+      "epoch": 2.48,
+      "grad_norm": 3.097006730897294,
+      "learning_rate": 2.5362681340670334e-05,
+      "loss": 0.1403,
+      "step": 1980
+    },
+    {
+      "epoch": 2.49,
+      "grad_norm": 2.703830428494632,
+      "learning_rate": 2.5237618809404708e-05,
+      "loss": 0.1422,
+      "step": 1990
+    },
+    {
+      "epoch": 2.5,
+      "grad_norm": 2.6503565314509916,
+      "learning_rate": 2.511255627813907e-05,
+      "loss": 0.1326,
+      "step": 2000
+    },
+    {
+      "epoch": 2.51,
+      "grad_norm": 2.8501605206174148,
+      "learning_rate": 2.4987493746873438e-05,
+      "loss": 0.1295,
+      "step": 2010
+    },
+    {
+      "epoch": 2.52,
+      "grad_norm": 2.7904621961035954,
+      "learning_rate": 2.4862431215607804e-05,
+      "loss": 0.1314,
+      "step": 2020
+    },
+    {
+      "epoch": 2.54,
+      "grad_norm": 2.555989811347152,
+      "learning_rate": 2.473736868434217e-05,
+      "loss": 0.1346,
+      "step": 2030
+    },
+    {
+      "epoch": 2.55,
+      "grad_norm": 2.5326240483562694,
+      "learning_rate": 2.461230615307654e-05,
+      "loss": 0.134,
+      "step": 2040
+    },
+    {
+      "epoch": 2.56,
+      "grad_norm": 2.878178705134795,
+      "learning_rate": 2.4487243621810908e-05,
+      "loss": 0.1237,
+      "step": 2050
+    },
+    {
+      "epoch": 2.58,
+      "grad_norm": 2.4896292156848974,
+      "learning_rate": 2.4362181090545274e-05,
+      "loss": 0.1374,
+      "step": 2060
+    },
+    {
+      "epoch": 2.59,
+      "grad_norm": 2.4175247992710087,
+      "learning_rate": 2.423711855927964e-05,
+      "loss": 0.1375,
+      "step": 2070
+    },
+    {
+      "epoch": 2.6,
+      "grad_norm": 3.1027213365169675,
+      "learning_rate": 2.4112056028014007e-05,
+      "loss": 0.1392,
+      "step": 2080
+    },
+    {
+      "epoch": 2.61,
+      "grad_norm": 2.6677840295965782,
+      "learning_rate": 2.3986993496748374e-05,
+      "loss": 0.1218,
+      "step": 2090
+    },
+    {
+      "epoch": 2.62,
+      "grad_norm": 2.679408467615217,
+      "learning_rate": 2.3861930965482744e-05,
+      "loss": 0.1304,
+      "step": 2100
+    },
+    {
+      "epoch": 2.64,
+      "grad_norm": 2.7859729569896468,
+      "learning_rate": 2.373686843421711e-05,
+      "loss": 0.1278,
+      "step": 2110
+    },
+    {
+      "epoch": 2.65,
+      "grad_norm": 2.8594418710695764,
+      "learning_rate": 2.3611805902951477e-05,
+      "loss": 0.1297,
+      "step": 2120
+    },
+    {
+      "epoch": 2.66,
+      "grad_norm": 2.684541004807738,
+      "learning_rate": 2.3486743371685844e-05,
+      "loss": 0.1203,
+      "step": 2130
+    },
+    {
+      "epoch": 2.67,
+      "grad_norm": 2.751766677724115,
+      "learning_rate": 2.336168084042021e-05,
+      "loss": 0.118,
+      "step": 2140
+    },
+    {
+      "epoch": 2.69,
+      "grad_norm": 3.110848849081005,
+      "learning_rate": 2.3236618309154577e-05,
+      "loss": 0.132,
+      "step": 2150
+    },
+    {
+      "epoch": 2.7,
+      "grad_norm": 2.6525580385637593,
+      "learning_rate": 2.3111555777888947e-05,
+      "loss": 0.137,
+      "step": 2160
+    },
+    {
+      "epoch": 2.71,
+      "grad_norm": 2.593636751575111,
+      "learning_rate": 2.2986493246623314e-05,
+      "loss": 0.1359,
+      "step": 2170
+    },
+    {
+      "epoch": 2.73,
+      "grad_norm": 3.0950612073524977,
+      "learning_rate": 2.2861430715357677e-05,
+      "loss": 0.1264,
+      "step": 2180
+    },
+    {
+      "epoch": 2.74,
+      "grad_norm": 2.661775116005948,
+      "learning_rate": 2.2736368184092047e-05,
+      "loss": 0.1251,
+      "step": 2190
+    },
+    {
+      "epoch": 2.75,
+      "grad_norm": 3.230211019325226,
+      "learning_rate": 2.2611305652826414e-05,
+      "loss": 0.1296,
+      "step": 2200
+    },
+    {
+      "epoch": 2.76,
+      "grad_norm": 3.081544477462097,
+      "learning_rate": 2.248624312156078e-05,
+      "loss": 0.1247,
+      "step": 2210
+    },
+    {
+      "epoch": 2.77,
+      "grad_norm": 3.1293638659096885,
+      "learning_rate": 2.236118059029515e-05,
+      "loss": 0.1303,
+      "step": 2220
+    },
+    {
+      "epoch": 2.79,
+      "grad_norm": 2.5946152863853404,
+      "learning_rate": 2.2236118059029514e-05,
+      "loss": 0.131,
+      "step": 2230
+    },
+    {
+      "epoch": 2.8,
+      "grad_norm": 2.6311170965776722,
+      "learning_rate": 2.2111055527763884e-05,
+      "loss": 0.1427,
+      "step": 2240
+    },
+    {
+      "epoch": 2.81,
+      "grad_norm": 2.6125582163771894,
+      "learning_rate": 2.198599299649825e-05,
+      "loss": 0.1313,
+      "step": 2250
+    },
+    {
+      "epoch": 2.83,
+      "grad_norm": 2.79415838341794,
+      "learning_rate": 2.1860930465232617e-05,
+      "loss": 0.1306,
+      "step": 2260
+    },
+    {
+      "epoch": 2.84,
+      "grad_norm": 2.4025875805185057,
+      "learning_rate": 2.1735867933966987e-05,
+      "loss": 0.127,
+      "step": 2270
+    },
+    {
+      "epoch": 2.85,
+      "grad_norm": 2.6788154272935376,
+      "learning_rate": 2.161080540270135e-05,
+      "loss": 0.1191,
+      "step": 2280
+    },
+    {
+      "epoch": 2.86,
+      "grad_norm": 2.6992815510003854,
+      "learning_rate": 2.1485742871435717e-05,
+      "loss": 0.1256,
+      "step": 2290
+    },
+    {
+      "epoch": 2.88,
+      "grad_norm": 2.6704557341346087,
+      "learning_rate": 2.1360680340170087e-05,
+      "loss": 0.1238,
+      "step": 2300
+    },
+    {
+      "epoch": 2.89,
+      "grad_norm": 2.7264637751590683,
+      "learning_rate": 2.1235617808904454e-05,
+      "loss": 0.1271,
+      "step": 2310
+    },
+    {
+      "epoch": 2.9,
+      "grad_norm": 2.952720170302931,
+      "learning_rate": 2.111055527763882e-05,
+      "loss": 0.1255,
+      "step": 2320
+    },
+    {
+      "epoch": 2.91,
+      "grad_norm": 3.0158536359854273,
+      "learning_rate": 2.0985492746373187e-05,
+      "loss": 0.1291,
+      "step": 2330
+    },
+    {
+      "epoch": 2.92,
+      "grad_norm": 2.3829743877512035,
+      "learning_rate": 2.0860430215107554e-05,
+      "loss": 0.137,
+      "step": 2340
+    },
+    {
+      "epoch": 2.94,
+      "grad_norm": 2.2748284034151083,
+      "learning_rate": 2.073536768384192e-05,
+      "loss": 0.1274,
+      "step": 2350
+    },
+    {
+      "epoch": 2.95,
+      "grad_norm": 2.526854855449805,
+      "learning_rate": 2.061030515257629e-05,
+      "loss": 0.121,
+      "step": 2360
+    },
+    {
+      "epoch": 2.96,
+      "grad_norm": 2.61083769775896,
+      "learning_rate": 2.0485242621310657e-05,
+      "loss": 0.1221,
+      "step": 2370
+    },
+    {
+      "epoch": 2.98,
+      "grad_norm": 2.732571077670084,
+      "learning_rate": 2.0360180090045024e-05,
+      "loss": 0.1274,
+      "step": 2380
+    },
+    {
+      "epoch": 2.99,
+      "grad_norm": 2.438170511993909,
+      "learning_rate": 2.023511755877939e-05,
+      "loss": 0.1161,
+      "step": 2390
+    },
+    {
+      "epoch": 3.0,
+      "grad_norm": 2.150128950643317,
+      "learning_rate": 2.0110055027513757e-05,
+      "loss": 0.1261,
+      "step": 2400
+    },
+    {
+      "epoch": 3.01,
+      "grad_norm": 1.143682327769968,
+      "learning_rate": 1.9984992496248124e-05,
+      "loss": 0.0474,
+      "step": 2410
+    },
+    {
+      "epoch": 3.02,
+      "grad_norm": 1.6652489672492834,
+      "learning_rate": 1.9859929964982494e-05,
+      "loss": 0.0443,
+      "step": 2420
+    },
+    {
+      "epoch": 3.04,
+      "grad_norm": 1.572882072960937,
+      "learning_rate": 1.9734867433716857e-05,
+      "loss": 0.0424,
+      "step": 2430
+    },
+    {
+      "epoch": 3.05,
+      "grad_norm": 1.3741877497591568,
+      "learning_rate": 1.9609804902451227e-05,
+      "loss": 0.0545,
+      "step": 2440
+    },
+    {
+      "epoch": 3.06,
+      "grad_norm": 1.7064284737834505,
+      "learning_rate": 1.9484742371185594e-05,
+      "loss": 0.0539,
+      "step": 2450
+    },
+    {
+      "epoch": 3.08,
+      "grad_norm": 1.6462372473336482,
+      "learning_rate": 1.935967983991996e-05,
+      "loss": 0.0516,
+      "step": 2460
+    },
+    {
+      "epoch": 3.09,
+      "grad_norm": 1.560256568155276,
+      "learning_rate": 1.923461730865433e-05,
+      "loss": 0.0496,
+      "step": 2470
+    },
+    {
+      "epoch": 3.1,
+      "grad_norm": 1.381325865513328,
+      "learning_rate": 1.9109554777388694e-05,
+      "loss": 0.048,
+      "step": 2480
+    },
+    {
+      "epoch": 3.11,
+      "grad_norm": 1.8124841492058452,
+      "learning_rate": 1.898449224612306e-05,
+      "loss": 0.054,
+      "step": 2490
+    },
+    {
+      "epoch": 3.12,
+      "grad_norm": 1.6535986065065578,
+      "learning_rate": 1.885942971485743e-05,
+      "loss": 0.06,
+      "step": 2500
+    },
+    {
+      "epoch": 3.14,
+      "grad_norm": 1.9819166319488526,
+      "learning_rate": 1.8734367183591797e-05,
+      "loss": 0.056,
+      "step": 2510
+    },
+    {
+      "epoch": 3.15,
+      "grad_norm": 2.087812060027925,
+      "learning_rate": 1.8609304652326164e-05,
+      "loss": 0.0558,
+      "step": 2520
+    },
+    {
+      "epoch": 3.16,
+      "grad_norm": 1.4283212272341697,
+      "learning_rate": 1.848424212106053e-05,
+      "loss": 0.0541,
+      "step": 2530
+    },
+    {
+      "epoch": 3.17,
+      "grad_norm": 2.1967893363996667,
+      "learning_rate": 1.8359179589794897e-05,
+      "loss": 0.0586,
+      "step": 2540
+    },
+    {
+      "epoch": 3.19,
+      "grad_norm": 1.6441632581708676,
+      "learning_rate": 1.8234117058529264e-05,
+      "loss": 0.05,
+      "step": 2550
+    },
+    {
+      "epoch": 3.2,
+      "grad_norm": 1.7647257852834424,
+      "learning_rate": 1.8109054527263634e-05,
+      "loss": 0.0578,
+      "step": 2560
+    },
+    {
+      "epoch": 3.21,
+      "grad_norm": 1.9882813699118946,
+      "learning_rate": 1.7983991995998e-05,
+      "loss": 0.0524,
+      "step": 2570
+    },
+    {
+      "epoch": 3.23,
+      "grad_norm": 1.6533759029481423,
+      "learning_rate": 1.7858929464732367e-05,
+      "loss": 0.0481,
+      "step": 2580
+    },
+    {
+      "epoch": 3.24,
+      "grad_norm": 1.2640217174463038,
+      "learning_rate": 1.7733866933466734e-05,
+      "loss": 0.0475,
+      "step": 2590
+    },
+    {
+      "epoch": 3.25,
+      "grad_norm": 1.8551424723405705,
+      "learning_rate": 1.76088044022011e-05,
+      "loss": 0.0522,
+      "step": 2600
+    },
+    {
+      "epoch": 3.26,
+      "grad_norm": 1.7963597761468264,
+      "learning_rate": 1.748374187093547e-05,
+      "loss": 0.0541,
+      "step": 2610
+    },
+    {
+      "epoch": 3.27,
+      "grad_norm": 1.7321785430153591,
+      "learning_rate": 1.7358679339669837e-05,
+      "loss": 0.0507,
+      "step": 2620
+    },
+    {
+      "epoch": 3.29,
+      "grad_norm": 1.955876601783095,
+      "learning_rate": 1.72336168084042e-05,
+      "loss": 0.0485,
+      "step": 2630
+    },
+    {
+      "epoch": 3.3,
+      "grad_norm": 1.8485106736675452,
+      "learning_rate": 1.710855427713857e-05,
+      "loss": 0.0532,
+      "step": 2640
+    },
+    {
+      "epoch": 3.31,
+      "grad_norm": 1.5996584855623148,
+      "learning_rate": 1.6983491745872937e-05,
+      "loss": 0.0577,
+      "step": 2650
+    },
+    {
+      "epoch": 3.33,
+      "grad_norm": 1.5254709610927397,
+      "learning_rate": 1.6858429214607304e-05,
+      "loss": 0.0456,
+      "step": 2660
+    },
+    {
+      "epoch": 3.34,
+      "grad_norm": 1.4432293700437517,
+      "learning_rate": 1.6733366683341674e-05,
+      "loss": 0.0546,
+      "step": 2670
+    },
+    {
+      "epoch": 3.35,
+      "grad_norm": 1.6784150027520048,
+      "learning_rate": 1.660830415207604e-05,
+      "loss": 0.0535,
+      "step": 2680
+    },
+    {
+      "epoch": 3.36,
+      "grad_norm": 1.7809353600733642,
+      "learning_rate": 1.6483241620810404e-05,
+      "loss": 0.0451,
+      "step": 2690
+    },
+    {
+      "epoch": 3.38,
+      "grad_norm": 2.581167749013489,
+      "learning_rate": 1.6358179089544774e-05,
+      "loss": 0.0547,
+      "step": 2700
+    },
+    {
+      "epoch": 3.39,
+      "grad_norm": 2.0047117997419015,
+      "learning_rate": 1.623311655827914e-05,
+      "loss": 0.0545,
+      "step": 2710
+    },
+    {
+      "epoch": 3.4,
+      "grad_norm": 1.4654369729791492,
+      "learning_rate": 1.6108054027013507e-05,
+      "loss": 0.048,
+      "step": 2720
+    },
+    {
+      "epoch": 3.41,
+      "grad_norm": 1.878234552879185,
+      "learning_rate": 1.5982991495747877e-05,
+      "loss": 0.0508,
+      "step": 2730
+    },
+    {
+      "epoch": 3.42,
+      "grad_norm": 1.4766487217509554,
+      "learning_rate": 1.585792896448224e-05,
+      "loss": 0.048,
+      "step": 2740
+    },
+    {
+      "epoch": 3.44,
+      "grad_norm": 2.015925067182647,
+      "learning_rate": 1.573286643321661e-05,
+      "loss": 0.0535,
+      "step": 2750
+    },
+    {
+      "epoch": 3.45,
+      "grad_norm": 1.2839739616582218,
+      "learning_rate": 1.5607803901950977e-05,
+      "loss": 0.0511,
+      "step": 2760
+    },
+    {
+      "epoch": 3.46,
+      "grad_norm": 1.2605091355298021,
+      "learning_rate": 1.5482741370685344e-05,
+      "loss": 0.0492,
+      "step": 2770
+    },
+    {
+      "epoch": 3.48,
+      "grad_norm": 1.778839488026904,
+      "learning_rate": 1.535767883941971e-05,
+      "loss": 0.0508,
+      "step": 2780
+    },
+    {
+      "epoch": 3.49,
+      "grad_norm": 1.873029054431649,
+      "learning_rate": 1.5232616308154077e-05,
+      "loss": 0.0515,
+      "step": 2790
+    },
+    {
+      "epoch": 3.5,
+      "grad_norm": 1.489240202252438,
+      "learning_rate": 1.5107553776888445e-05,
+      "loss": 0.0469,
+      "step": 2800
+    },
+    {
+      "epoch": 3.51,
+      "grad_norm": 1.1951564612172105,
+      "learning_rate": 1.4982491245622812e-05,
+      "loss": 0.0424,
+      "step": 2810
+    },
+    {
+      "epoch": 3.52,
+      "grad_norm": 1.427494221957884,
+      "learning_rate": 1.485742871435718e-05,
+      "loss": 0.0452,
+      "step": 2820
+    },
+    {
+      "epoch": 3.54,
+      "grad_norm": 1.767643003168799,
+      "learning_rate": 1.4732366183091547e-05,
+      "loss": 0.054,
+      "step": 2830
+    },
+    {
+      "epoch": 3.55,
+      "grad_norm": 1.8461197963044351,
+      "learning_rate": 1.4607303651825912e-05,
+      "loss": 0.05,
+      "step": 2840
+    },
+    {
+      "epoch": 3.56,
+      "grad_norm": 1.6976823547575175,
+      "learning_rate": 1.448224112056028e-05,
+      "loss": 0.0484,
+      "step": 2850
+    },
+    {
+      "epoch": 3.58,
+      "grad_norm": 2.1165349313587103,
+      "learning_rate": 1.4357178589294649e-05,
+      "loss": 0.0495,
+      "step": 2860
+    },
+    {
+      "epoch": 3.59,
+      "grad_norm": 1.4367595921945118,
+      "learning_rate": 1.4232116058029015e-05,
+      "loss": 0.0449,
+      "step": 2870
+    },
+    {
+      "epoch": 3.6,
+      "grad_norm": 1.5701725835763933,
+      "learning_rate": 1.4107053526763384e-05,
+      "loss": 0.0535,
+      "step": 2880
+    },
+    {
+      "epoch": 3.61,
+      "grad_norm": 1.3890521981336823,
+      "learning_rate": 1.3981990995497749e-05,
+      "loss": 0.0497,
+      "step": 2890
+    },
+    {
+      "epoch": 3.62,
+      "grad_norm": 2.3482814470474187,
+      "learning_rate": 1.3856928464232117e-05,
+      "loss": 0.0479,
+      "step": 2900
+    },
+    {
+      "epoch": 3.64,
+      "grad_norm": 1.818720567602742,
+      "learning_rate": 1.3731865932966484e-05,
+      "loss": 0.0456,
+      "step": 2910
+    },
+    {
+      "epoch": 3.65,
+      "grad_norm": 1.580084990497098,
+      "learning_rate": 1.3606803401700852e-05,
+      "loss": 0.0581,
+      "step": 2920
+    },
+    {
+      "epoch": 3.66,
+      "grad_norm": 1.570405568263889,
+      "learning_rate": 1.348174087043522e-05,
+      "loss": 0.0467,
+      "step": 2930
+    },
+    {
+      "epoch": 3.67,
+      "grad_norm": 1.3862471087385182,
+      "learning_rate": 1.3356678339169585e-05,
+      "loss": 0.0427,
+      "step": 2940
+    },
+    {
+      "epoch": 3.69,
+      "grad_norm": 1.905843535066407,
+      "learning_rate": 1.3231615807903952e-05,
+      "loss": 0.0451,
+      "step": 2950
+    },
+    {
+      "epoch": 3.7,
+      "grad_norm": 2.0096560078524535,
+      "learning_rate": 1.310655327663832e-05,
+      "loss": 0.0478,
+      "step": 2960
+    },
+    {
+      "epoch": 3.71,
+      "grad_norm": 1.8387148364015753,
+      "learning_rate": 1.2981490745372687e-05,
+      "loss": 0.0468,
+      "step": 2970
+    },
+    {
+      "epoch": 3.73,
+      "grad_norm": 1.3511210696441773,
+      "learning_rate": 1.2856428214107055e-05,
+      "loss": 0.0433,
+      "step": 2980
+    },
+    {
+      "epoch": 3.74,
+      "grad_norm": 1.343344871307373,
+      "learning_rate": 1.273136568284142e-05,
+      "loss": 0.0491,
+      "step": 2990
+    },
+    {
+      "epoch": 3.75,
+      "grad_norm": 1.6097943398521046,
+      "learning_rate": 1.2606303151575788e-05,
+      "loss": 0.0441,
+      "step": 3000
+    },
+    {
+      "epoch": 3.76,
+      "grad_norm": 1.430264215935809,
+      "learning_rate": 1.2481240620310155e-05,
+      "loss": 0.0442,
+      "step": 3010
+    },
+    {
+      "epoch": 3.77,
+      "grad_norm": 1.986829606239943,
+      "learning_rate": 1.2356178089044523e-05,
+      "loss": 0.0438,
+      "step": 3020
+    },
+    {
+      "epoch": 3.79,
+      "grad_norm": 1.541613560698392,
+      "learning_rate": 1.223111555777889e-05,
+      "loss": 0.05,
+      "step": 3030
+    },
+    {
+      "epoch": 3.8,
+      "grad_norm": 1.4820071156597356,
+      "learning_rate": 1.2106053026513257e-05,
+      "loss": 0.0442,
+      "step": 3040
+    },
+    {
+      "epoch": 3.81,
+      "grad_norm": 1.9180888276280081,
+      "learning_rate": 1.1980990495247623e-05,
+      "loss": 0.0453,
+      "step": 3050
+    },
+    {
+      "epoch": 3.83,
+      "grad_norm": 1.5819512241044464,
+      "learning_rate": 1.1855927963981992e-05,
+      "loss": 0.0444,
+      "step": 3060
+    },
+    {
+      "epoch": 3.84,
+      "grad_norm": 1.502604448747085,
+      "learning_rate": 1.1730865432716358e-05,
+      "loss": 0.0408,
+      "step": 3070
+    },
+    {
+      "epoch": 3.85,
+      "grad_norm": 1.7278176832888912,
+      "learning_rate": 1.1605802901450725e-05,
+      "loss": 0.0392,
+      "step": 3080
+    },
+    {
+      "epoch": 3.86,
+      "grad_norm": 1.3651777045585691,
+      "learning_rate": 1.1480740370185093e-05,
+      "loss": 0.0388,
+      "step": 3090
+    },
+    {
+      "epoch": 3.88,
+      "grad_norm": 1.7185388608772323,
+      "learning_rate": 1.1355677838919462e-05,
+      "loss": 0.0432,
+      "step": 3100
+    },
+    {
+      "epoch": 3.89,
+      "grad_norm": 1.5103350474787707,
+      "learning_rate": 1.1230615307653827e-05,
+      "loss": 0.0441,
+      "step": 3110
+    },
+    {
+      "epoch": 3.9,
+      "grad_norm": 1.4597562206231804,
+      "learning_rate": 1.1105552776388195e-05,
+      "loss": 0.0475,
+      "step": 3120
+    },
+    {
+      "epoch": 3.91,
+      "grad_norm": 1.3774614843117499,
+      "learning_rate": 1.0980490245122562e-05,
+      "loss": 0.0398,
+      "step": 3130
+    },
+    {
+      "epoch": 3.92,
+      "grad_norm": 1.606767412261371,
+      "learning_rate": 1.0855427713856928e-05,
+      "loss": 0.0471,
+      "step": 3140
+    },
+    {
+      "epoch": 3.94,
+      "grad_norm": 1.3861408157622845,
+      "learning_rate": 1.0730365182591297e-05,
+      "loss": 0.0443,
+      "step": 3150
+    },
+    {
+      "epoch": 3.95,
+      "grad_norm": 1.5538838672128699,
+      "learning_rate": 1.0605302651325663e-05,
+      "loss": 0.0472,
+      "step": 3160
+    },
+    {
+      "epoch": 3.96,
+      "grad_norm": 1.5946894102329936,
+      "learning_rate": 1.0480240120060032e-05,
+      "loss": 0.0433,
+      "step": 3170
+    },
+    {
+      "epoch": 3.98,
+      "grad_norm": 1.2468458913211395,
+      "learning_rate": 1.0355177588794397e-05,
+      "loss": 0.0383,
+      "step": 3180
+    },
+    {
+      "epoch": 3.99,
+      "grad_norm": 1.5664978761582478,
+      "learning_rate": 1.0230115057528765e-05,
+      "loss": 0.0459,
+      "step": 3190
+    },
+    {
+      "epoch": 4.0,
+      "grad_norm": 1.809011654121981,
+      "learning_rate": 1.0105052526263133e-05,
+      "loss": 0.0419,
+      "step": 3200
+    },
+    {
+      "epoch": 4.01,
+      "grad_norm": 1.0160616596239802,
+      "learning_rate": 9.979989994997498e-06,
+      "loss": 0.0166,
+      "step": 3210
+    },
+    {
+      "epoch": 4.03,
+      "grad_norm": 0.9631860678272716,
+      "learning_rate": 9.854927463731867e-06,
+      "loss": 0.0147,
+      "step": 3220
+    },
+    {
+      "epoch": 4.04,
+      "grad_norm": 1.265768914223235,
+      "learning_rate": 9.729864932466233e-06,
+      "loss": 0.0157,
+      "step": 3230
+    },
+    {
+      "epoch": 4.05,
+      "grad_norm": 1.110439058880837,
+      "learning_rate": 9.604802401200602e-06,
+      "loss": 0.0188,
+      "step": 3240
+    },
+    {
+      "epoch": 4.06,
+      "grad_norm": 0.7057898144627774,
+      "learning_rate": 9.479739869934968e-06,
+      "loss": 0.0133,
+      "step": 3250
+    },
+    {
+      "epoch": 4.08,
+      "grad_norm": 0.8140341140571917,
+      "learning_rate": 9.354677338669335e-06,
+      "loss": 0.0139,
+      "step": 3260
+    },
+    {
+      "epoch": 4.09,
+      "grad_norm": 0.9005386330187762,
+      "learning_rate": 9.229614807403703e-06,
+      "loss": 0.0178,
+      "step": 3270
+    },
+    {
+      "epoch": 4.1,
+      "grad_norm": 0.6990527589994958,
+      "learning_rate": 9.104552276138068e-06,
+      "loss": 0.0166,
+      "step": 3280
+    },
+    {
+      "epoch": 4.11,
+      "grad_norm": 0.8374763300383368,
+      "learning_rate": 8.979489744872437e-06,
+      "loss": 0.0126,
+      "step": 3290
+    },
+    {
+      "epoch": 4.12,
+      "grad_norm": 0.6026036059465715,
+      "learning_rate": 8.854427213606805e-06,
+      "loss": 0.0194,
+      "step": 3300
+    },
+    {
+      "epoch": 4.14,
+      "grad_norm": 0.6384568858739467,
+      "learning_rate": 8.72936468234117e-06,
+      "loss": 0.0185,
+      "step": 3310
+    },
+    {
+      "epoch": 4.15,
+      "grad_norm": 0.7979584602452928,
+      "learning_rate": 8.604302151075538e-06,
+      "loss": 0.0127,
+      "step": 3320
+    },
+    {
+      "epoch": 4.16,
+      "grad_norm": 1.0334510433669584,
+      "learning_rate": 8.479239619809905e-06,
+      "loss": 0.0153,
+      "step": 3330
+    },
+    {
+      "epoch": 4.17,
+      "grad_norm": 0.788681126956416,
+      "learning_rate": 8.354177088544273e-06,
+      "loss": 0.0196,
+      "step": 3340
+    },
+    {
+      "epoch": 4.19,
+      "grad_norm": 0.6685423145269257,
+      "learning_rate": 8.22911455727864e-06,
+      "loss": 0.0157,
+      "step": 3350
+    },
+    {
+      "epoch": 4.2,
+      "grad_norm": 0.7287135532877159,
+      "learning_rate": 8.104052026013006e-06,
+      "loss": 0.0161,
+      "step": 3360
+    },
+    {
+      "epoch": 4.21,
+      "grad_norm": 0.7064113955783903,
+      "learning_rate": 7.978989494747375e-06,
+      "loss": 0.0147,
+      "step": 3370
+    },
+    {
+      "epoch": 4.22,
+      "grad_norm": 0.6607611553955022,
+      "learning_rate": 7.85392696348174e-06,
+      "loss": 0.0133,
+      "step": 3380
+    },
+    {
+      "epoch": 4.24,
+      "grad_norm": 1.067040055954057,
+      "learning_rate": 7.728864432216108e-06,
+      "loss": 0.0148,
+      "step": 3390
+    },
+    {
+      "epoch": 4.25,
+      "grad_norm": 1.0324328603459556,
+      "learning_rate": 7.603801900950476e-06,
+      "loss": 0.0142,
+      "step": 3400
+    },
+    {
+      "epoch": 4.26,
+      "grad_norm": 0.9592195055570732,
+      "learning_rate": 7.478739369684842e-06,
+      "loss": 0.016,
+      "step": 3410
+    },
+    {
+      "epoch": 4.28,
+      "grad_norm": 0.9230560560572981,
+      "learning_rate": 7.35367683841921e-06,
+      "loss": 0.0144,
+      "step": 3420
+    },
+    {
+      "epoch": 4.29,
+      "grad_norm": 0.6457139232953377,
+      "learning_rate": 7.228614307153578e-06,
+      "loss": 0.0145,
+      "step": 3430
+    },
+    {
+      "epoch": 4.3,
+      "grad_norm": 0.3527501285452713,
+      "learning_rate": 7.103551775887944e-06,
+      "loss": 0.0144,
+      "step": 3440
+    },
+    {
+      "epoch": 4.31,
+      "grad_norm": 0.6072232084037488,
+      "learning_rate": 6.978489244622312e-06,
+      "loss": 0.0134,
+      "step": 3450
+    },
+    {
+      "epoch": 4.33,
+      "grad_norm": 0.6554608368424355,
+      "learning_rate": 6.853426713356678e-06,
+      "loss": 0.0136,
+      "step": 3460
+    },
+    {
+      "epoch": 4.34,
+      "grad_norm": 0.7345293166981702,
+      "learning_rate": 6.7283641820910456e-06,
+      "loss": 0.017,
+      "step": 3470
+    },
+    {
+      "epoch": 4.35,
+      "grad_norm": 0.6630110101404149,
+      "learning_rate": 6.603301650825414e-06,
+      "loss": 0.0138,
+      "step": 3480
+    },
+    {
+      "epoch": 4.36,
+      "grad_norm": 0.6014101157456893,
+      "learning_rate": 6.47823911955978e-06,
+      "loss": 0.0131,
+      "step": 3490
+    },
+    {
+      "epoch": 4.38,
+      "grad_norm": 0.78056530512705,
+      "learning_rate": 6.353176588294148e-06,
+      "loss": 0.0152,
+      "step": 3500
+    },
+    {
+      "epoch": 4.39,
+      "grad_norm": 0.625830313370077,
+      "learning_rate": 6.228114057028515e-06,
+      "loss": 0.0136,
+      "step": 3510
+    },
+    {
+      "epoch": 4.4,
+      "grad_norm": 1.04947606820658,
+      "learning_rate": 6.103051525762881e-06,
+      "loss": 0.0158,
+      "step": 3520
+    },
+    {
+      "epoch": 4.41,
+      "grad_norm": 0.6295492545834148,
+      "learning_rate": 5.977988994497249e-06,
+      "loss": 0.0143,
+      "step": 3530
+    },
+    {
+      "epoch": 4.42,
+      "grad_norm": 0.6227005381610481,
+      "learning_rate": 5.8529264632316155e-06,
+      "loss": 0.0133,
+      "step": 3540
+    },
+    {
+      "epoch": 4.44,
+      "grad_norm": 0.8799433375127763,
+      "learning_rate": 5.727863931965984e-06,
+      "loss": 0.0154,
+      "step": 3550
+    },
+    {
+      "epoch": 4.45,
+      "grad_norm": 0.5503927714060292,
+      "learning_rate": 5.6028014007003505e-06,
+      "loss": 0.0114,
+      "step": 3560
+    },
+    {
+      "epoch": 4.46,
+      "grad_norm": 1.165400459405073,
+      "learning_rate": 5.477738869434718e-06,
+      "loss": 0.0135,
+      "step": 3570
+    },
+    {
+      "epoch": 4.47,
+      "grad_norm": 0.5467153588664332,
+      "learning_rate": 5.352676338169085e-06,
+      "loss": 0.0131,
+      "step": 3580
+    },
+    {
+      "epoch": 4.49,
+      "grad_norm": 0.9263285375289982,
+      "learning_rate": 5.227613806903451e-06,
+      "loss": 0.0128,
+      "step": 3590
+    },
+    {
+      "epoch": 4.5,
+      "grad_norm": 0.8415201179153123,
+      "learning_rate": 5.10255127563782e-06,
+      "loss": 0.0112,
+      "step": 3600
+    },
+    {
+      "epoch": 4.51,
+      "grad_norm": 0.6593144935941556,
+      "learning_rate": 4.977488744372186e-06,
+      "loss": 0.0154,
+      "step": 3610
+    },
+    {
+      "epoch": 4.53,
+      "grad_norm": 0.914570463491157,
+      "learning_rate": 4.852426213106554e-06,
+      "loss": 0.0131,
+      "step": 3620
+    },
+    {
+      "epoch": 4.54,
+      "grad_norm": 0.5307450980608305,
+      "learning_rate": 4.7273636818409205e-06,
+      "loss": 0.0152,
+      "step": 3630
+    },
+    {
+      "epoch": 4.55,
+      "grad_norm": 0.5720744936742393,
+      "learning_rate": 4.602301150575288e-06,
+      "loss": 0.01,
+      "step": 3640
+    },
+    {
+      "epoch": 4.56,
+      "grad_norm": 0.6906646371919622,
+      "learning_rate": 4.4772386193096554e-06,
+      "loss": 0.0158,
+      "step": 3650
+    },
+    {
+      "epoch": 4.58,
+      "grad_norm": 0.750217922657539,
+      "learning_rate": 4.352176088044022e-06,
+      "loss": 0.0137,
+      "step": 3660
+    },
+    {
+      "epoch": 4.59,
+      "grad_norm": 0.6507958418548583,
+      "learning_rate": 4.22711355677839e-06,
+      "loss": 0.0108,
+      "step": 3670
+    },
+    {
+      "epoch": 4.6,
+      "grad_norm": 0.4417488321613792,
+      "learning_rate": 4.102051025512756e-06,
+      "loss": 0.0125,
+      "step": 3680
+    },
+    {
+      "epoch": 4.61,
+      "grad_norm": 0.8625362167494579,
+      "learning_rate": 3.976988494247124e-06,
+      "loss": 0.0157,
+      "step": 3690
+    },
+    {
+      "epoch": 4.62,
+      "grad_norm": 0.7856141458518725,
+      "learning_rate": 3.851925962981491e-06,
+      "loss": 0.0109,
+      "step": 3700
+    },
+    {
+      "epoch": 4.64,
+      "grad_norm": 0.6760897290661452,
+      "learning_rate": 3.726863431715858e-06,
+      "loss": 0.012,
+      "step": 3710
+    },
+    {
+      "epoch": 4.65,
+      "grad_norm": 0.8928787426379452,
+      "learning_rate": 3.601800900450225e-06,
+      "loss": 0.0118,
+      "step": 3720
+    },
+    {
+      "epoch": 4.66,
+      "grad_norm": 0.9518079569519863,
+      "learning_rate": 3.476738369184592e-06,
+      "loss": 0.0136,
+      "step": 3730
+    },
+    {
+      "epoch": 4.67,
+      "grad_norm": 1.1730903426890404,
+      "learning_rate": 3.35167583791896e-06,
+      "loss": 0.0162,
+      "step": 3740
+    },
+    {
+      "epoch": 4.69,
+      "grad_norm": 0.7279487019873496,
+      "learning_rate": 3.226613306653327e-06,
+      "loss": 0.0115,
+      "step": 3750
+    },
+    {
+      "epoch": 4.7,
+      "grad_norm": 0.9401650695234949,
+      "learning_rate": 3.101550775387694e-06,
+      "loss": 0.0096,
+      "step": 3760
+    },
+    {
+      "epoch": 4.71,
+      "grad_norm": 0.6111755424427511,
+      "learning_rate": 2.976488244122061e-06,
+      "loss": 0.0172,
+      "step": 3770
+    },
+    {
+      "epoch": 4.72,
+      "grad_norm": 0.3495626578205311,
+      "learning_rate": 2.8514257128564283e-06,
+      "loss": 0.0123,
+      "step": 3780
+    },
+    {
+      "epoch": 4.74,
+      "grad_norm": 0.4299090247891363,
+      "learning_rate": 2.7263631815907953e-06,
+      "loss": 0.0111,
+      "step": 3790
+    },
+    {
+      "epoch": 4.75,
+      "grad_norm": 0.6609027411517305,
+      "learning_rate": 2.601300650325163e-06,
+      "loss": 0.0143,
+      "step": 3800
+    },
+    {
+      "epoch": 4.76,
+      "grad_norm": 0.6019036947918726,
+      "learning_rate": 2.47623811905953e-06,
+      "loss": 0.0106,
+      "step": 3810
+    },
+    {
+      "epoch": 4.78,
+      "grad_norm": 0.8931005105357686,
+      "learning_rate": 2.351175587793897e-06,
+      "loss": 0.0103,
+      "step": 3820
+    },
+    {
+      "epoch": 4.79,
+      "grad_norm": 0.7662844250959648,
+      "learning_rate": 2.226113056528264e-06,
+      "loss": 0.0112,
+      "step": 3830
+    },
+    {
+      "epoch": 4.8,
+      "grad_norm": 1.0339128043635872,
+      "learning_rate": 2.1010505252626315e-06,
+      "loss": 0.0136,
+      "step": 3840
+    },
+    {
+      "epoch": 4.81,
+      "grad_norm": 0.6241448751379605,
+      "learning_rate": 1.9759879939969986e-06,
+      "loss": 0.0143,
+      "step": 3850
+    },
+    {
+      "epoch": 4.83,
+      "grad_norm": 0.6831824891172988,
+      "learning_rate": 1.8509254627313657e-06,
+      "loss": 0.0121,
+      "step": 3860
+    },
+    {
+      "epoch": 4.84,
+      "grad_norm": 0.7889281432328193,
+      "learning_rate": 1.725862931465733e-06,
+      "loss": 0.0134,
+      "step": 3870
+    },
+    {
+      "epoch": 4.85,
+      "grad_norm": 1.013124998305625,
+      "learning_rate": 1.6008004002001e-06,
+      "loss": 0.0117,
+      "step": 3880
+    },
+    {
+      "epoch": 4.86,
+      "grad_norm": 0.7583032179361607,
+      "learning_rate": 1.4757378689344673e-06,
+      "loss": 0.0117,
+      "step": 3890
+    },
+    {
+      "epoch": 4.88,
+      "grad_norm": 0.7353255835981027,
+      "learning_rate": 1.3506753376688344e-06,
+      "loss": 0.0104,
+      "step": 3900
+    },
+    {
+      "epoch": 4.89,
+      "grad_norm": 0.8789979335860438,
+      "learning_rate": 1.2256128064032017e-06,
+      "loss": 0.0127,
+      "step": 3910
+    },
+    {
+      "epoch": 4.9,
+      "grad_norm": 0.7867461850365051,
+      "learning_rate": 1.1005502751375688e-06,
+      "loss": 0.0108,
+      "step": 3920
+    },
+    {
+      "epoch": 4.91,
+      "grad_norm": 0.6118577850304773,
+      "learning_rate": 9.75487743871936e-07,
+      "loss": 0.0114,
+      "step": 3930
+    },
+    {
+      "epoch": 4.92,
+      "grad_norm": 0.45045792869989826,
+      "learning_rate": 8.504252126063032e-07,
+      "loss": 0.0103,
+      "step": 3940
+    },
+    {
+      "epoch": 4.94,
+      "grad_norm": 0.5684631882075091,
+      "learning_rate": 7.253626813406703e-07,
+      "loss": 0.0129,
+      "step": 3950
+    },
+    {
+      "epoch": 4.95,
+      "grad_norm": 0.7805656487508835,
+      "learning_rate": 6.003001500750376e-07,
+      "loss": 0.0147,
+      "step": 3960
+    },
+    {
+      "epoch": 4.96,
+      "grad_norm": 0.41471654590454743,
+      "learning_rate": 4.7523761880940473e-07,
+      "loss": 0.0138,
+      "step": 3970
+    },
+    {
+      "epoch": 4.97,
+      "grad_norm": 0.5741494584607629,
+      "learning_rate": 3.501750875437719e-07,
+      "loss": 0.0121,
+      "step": 3980
+    },
+    {
+      "epoch": 4.99,
+      "grad_norm": 1.0545689268922978,
+      "learning_rate": 2.2511255627813906e-07,
+      "loss": 0.0102,
+      "step": 3990
+    },
+    {
+      "epoch": 5.0,
+      "grad_norm": 0.5955044024086137,
+      "learning_rate": 1.0005002501250625e-07,
+      "loss": 0.0094,
+      "step": 4000
+    },
+    {
+      "epoch": 5.0,
+      "step": 4000,
+      "total_flos": 18796948488192.0,
+      "train_loss": 0.2349783019721508,
+      "train_runtime": 19900.0608,
+      "train_samples_per_second": 12.859,
+      "train_steps_per_second": 0.201
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 4000,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 5,
+  "save_steps": 50000,
+  "total_flos": 18796948488192.0,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}