Initial upload of checkpoint-4382

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+tokenizer.json filter=lfs diff=lfs merge=lfs -text

config.json

@@ -0,0 +1,38 @@
+{
+  "architectures": [
+    "LlamaForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "bos_token_id": 128000,
+  "dtype": "bfloat16",
+  "eos_token_id": 128009,
+  "head_dim": 128,
+  "hidden_act": "silu",
+  "hidden_size": 3072,
+  "initializer_range": 0.02,
+  "intermediate_size": 8192,
+  "max_position_embeddings": 131072,
+  "mlp_bias": false,
+  "model_type": "llama",
+  "num_attention_heads": 24,
+  "num_hidden_layers": 28,
+  "num_key_value_heads": 8,
+  "pad_token_id": 128004,
+  "pretraining_tp": 1,
+  "rms_norm_eps": 1e-05,
+  "rope_scaling": {
+    "factor": 32.0,
+    "high_freq_factor": 4.0,
+    "low_freq_factor": 1.0,
+    "original_max_position_embeddings": 8192,
+    "rope_type": "llama3"
+  },
+  "rope_theta": 500000.0,
+  "tie_word_embeddings": true,
+  "transformers_version": "4.57.2",
+  "unsloth_fixed": true,
+  "unsloth_version": "2025.11.6",
+  "use_cache": true,
+  "vocab_size": 156940
+}

generation_config.json

@@ -0,0 +1,11 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 128000,
+  "do_sample": true,
+  "eos_token_id": 128009,
+  "max_length": 131072,
+  "pad_token_id": 128004,
+  "temperature": 0.6,
+  "top_p": 0.9,
+  "transformers_version": "4.57.2"
+}

model-00001-of-00002.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e99351714441c0edfa195b6102f1893b9c636dc957097e4f0ba391329c2a4d20
+size 4991037968

model-00002-of-00002.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:87ae423c3725a5eacac72b1c186f28ed7e950a55b6271af29c960dba35ac4fc5
+size 1610725592

model.safetensors.index.json

@@ -0,0 +1,262 @@
+{
+  "metadata": {
+    "total_parameters": 3300867072,
+    "total_size": 6601734144
+  },
+  "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-00001-of-00002.safetensors",
+    "model.layers.17.mlp.down_proj.weight": "model-00001-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-00001-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.18.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.18.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.18.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.18.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.18.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.18.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.18.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.18.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.18.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.19.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.19.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.19.mlp.gate_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.19.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.19.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.19.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.19.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.19.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.19.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.20.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.20.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.20.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.20.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.20.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.20.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.20.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.20.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.20.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.21.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.21.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.21.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.21.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.21.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.21.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.21.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.21.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.21.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.22.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.22.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.22.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.22.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.22.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.22.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.22.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.22.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.22.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.23.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.23.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.23.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.23.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.23.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.23.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.23.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.23.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.23.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.24.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.24.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.24.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.24.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.24.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.24.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.24.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.24.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.24.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.25.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.25.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.25.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.25.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.25.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.25.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.25.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.25.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.25.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.26.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.26.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.26.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.26.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.26.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.26.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.26.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.26.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.26.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.27.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.27.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.27.mlp.gate_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.27.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.27.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.27.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.27.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.27.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.27.self_attn.v_proj.weight": "model-00002-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"
+  }
+}

optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:89a07ea6cbfbafc31f60c871625ae3fa7911fa336e59a361f3518e741182728d
+size 13203690391

rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e25b745caac240acf85f4d993ccdf1355dc409b319795bdaf284ce913f84fcba
+size 16389

rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:55c34b50cbc9ef0ee16f78568ab8cb9f39fda2528fb528565e4c4e6bbef4b5d1
+size 16389

rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e6fd420f2a650afc8853dcf4be7b848033a4553571318be351980db83cb07753
+size 16389

rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f24da66a8bbec7799b070f8ddce5f2e44fd2bdcde052433a0474aadf2b3e3028
+size 16389

rng_state_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ce4ed0ec5b6b99af767428bd1c70e550c3a87cfcf6b471b0391c3c278894a784
+size 16389

rng_state_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:72eb96462bd64bca2c87e7ff1077864a9d49b25c89f2addef49d882400d6839b
+size 16389

rng_state_6.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cc83999425f72eeaf8c12a8957680422762a46701bcd05a47249a6fa69666239
+size 16389

rng_state_7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:77250f2f6a628fe10f48d8e92f5eac168e6eceb0a77a739ad018bcfa675df1a0
+size 16389

scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:369e40521235ea74dbfb0c4aba515ec8d7ece6b6b57c25be5b00416758fa43dd
+size 1465

special_tokens_map.json

@@ -0,0 +1,3 @@
+{
+  "pad_token": "<|finetune_right_pad_id|>"
+}

tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fc3fecb199b4170636dbfab986d25f628157268d37b861f9cadaca60b1353bce
+size 22849547

trainer_state.json

@@ -0,0 +1,3132 @@
+{
+  "best_global_step": 626,
+  "best_metric": 4.402504920959473,
+  "best_model_checkpoint": "/home/deployer/laion/Orpheus-3B-Continued-2E-V4-WithGen/checkpoint-626",
+  "epoch": 0.12624531117654175,
+  "eval_steps": 313,
+  "global_step": 4382,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.00020166982616060984,
+      "grad_norm": 0.2099609375,
+      "learning_rate": 1.8e-05,
+      "loss": 4.9769,
+      "step": 10
+    },
+    {
+      "epoch": 0.0004033396523212197,
+      "grad_norm": 0.1552734375,
+      "learning_rate": 3.8e-05,
+      "loss": 4.8382,
+      "step": 20
+    },
+    {
+      "epoch": 0.0006050094784818296,
+      "grad_norm": 0.07666015625,
+      "learning_rate": 5.8e-05,
+      "loss": 4.6796,
+      "step": 30
+    },
+    {
+      "epoch": 0.0008066793046424393,
+      "grad_norm": 0.06787109375,
+      "learning_rate": 7.800000000000001e-05,
+      "loss": 4.6557,
+      "step": 40
+    },
+    {
+      "epoch": 0.0010083491308030493,
+      "grad_norm": 0.058837890625,
+      "learning_rate": 9.8e-05,
+      "loss": 4.6592,
+      "step": 50
+    },
+    {
+      "epoch": 0.0012100189569636591,
+      "grad_norm": 0.06982421875,
+      "learning_rate": 0.000118,
+      "loss": 4.612,
+      "step": 60
+    },
+    {
+      "epoch": 0.001411688783124269,
+      "grad_norm": 0.06396484375,
+      "learning_rate": 0.000138,
+      "loss": 4.5519,
+      "step": 70
+    },
+    {
+      "epoch": 0.0016133586092848787,
+      "grad_norm": 0.06787109375,
+      "learning_rate": 0.00015800000000000002,
+      "loss": 4.5649,
+      "step": 80
+    },
+    {
+      "epoch": 0.0018150284354454887,
+      "grad_norm": 0.07177734375,
+      "learning_rate": 0.00017800000000000002,
+      "loss": 4.5665,
+      "step": 90
+    },
+    {
+      "epoch": 0.0020166982616060987,
+      "grad_norm": 0.0615234375,
+      "learning_rate": 0.00019800000000000002,
+      "loss": 4.5854,
+      "step": 100
+    },
+    {
+      "epoch": 0.0022183680877667085,
+      "grad_norm": 0.06298828125,
+      "learning_rate": 0.00019999998367737306,
+      "loss": 4.6197,
+      "step": 110
+    },
+    {
+      "epoch": 0.0024200379139273183,
+      "grad_norm": 0.0693359375,
+      "learning_rate": 0.00019999992725348425,
+      "loss": 4.5047,
+      "step": 120
+    },
+    {
+      "epoch": 0.002621707740087928,
+      "grad_norm": 0.0732421875,
+      "learning_rate": 0.00019999983052684242,
+      "loss": 4.5378,
+      "step": 130
+    },
+    {
+      "epoch": 0.002823377566248538,
+      "grad_norm": 0.07763671875,
+      "learning_rate": 0.0001999996934974865,
+      "loss": 4.5702,
+      "step": 140
+    },
+    {
+      "epoch": 0.0030250473924091476,
+      "grad_norm": 0.08154296875,
+      "learning_rate": 0.00019999951616547182,
+      "loss": 4.5161,
+      "step": 150
+    },
+    {
+      "epoch": 0.0032267172185697574,
+      "grad_norm": 0.07177734375,
+      "learning_rate": 0.00019999929853086975,
+      "loss": 4.5427,
+      "step": 160
+    },
+    {
+      "epoch": 0.0034283870447303676,
+      "grad_norm": 0.064453125,
+      "learning_rate": 0.00019999904059376803,
+      "loss": 4.5205,
+      "step": 170
+    },
+    {
+      "epoch": 0.0036300568708909774,
+      "grad_norm": 0.06298828125,
+      "learning_rate": 0.00019999874235427067,
+      "loss": 4.5382,
+      "step": 180
+    },
+    {
+      "epoch": 0.003831726697051587,
+      "grad_norm": 0.0693359375,
+      "learning_rate": 0.0001999984038124978,
+      "loss": 4.5279,
+      "step": 190
+    },
+    {
+      "epoch": 0.004033396523212197,
+      "grad_norm": 0.06640625,
+      "learning_rate": 0.0001999980249685859,
+      "loss": 4.5047,
+      "step": 200
+    },
+    {
+      "epoch": 0.004235066349372807,
+      "grad_norm": 0.06494140625,
+      "learning_rate": 0.00019999760582268763,
+      "loss": 4.5041,
+      "step": 210
+    },
+    {
+      "epoch": 0.004436736175533417,
+      "grad_norm": 0.06689453125,
+      "learning_rate": 0.00019999714637497192,
+      "loss": 4.5513,
+      "step": 220
+    },
+    {
+      "epoch": 0.004638406001694026,
+      "grad_norm": 0.0654296875,
+      "learning_rate": 0.00019999664662562398,
+      "loss": 4.5115,
+      "step": 230
+    },
+    {
+      "epoch": 0.0048400758278546365,
+      "grad_norm": 0.06396484375,
+      "learning_rate": 0.0001999961065748452,
+      "loss": 4.5027,
+      "step": 240
+    },
+    {
+      "epoch": 0.005041745654015246,
+      "grad_norm": 0.07373046875,
+      "learning_rate": 0.00019999552622285317,
+      "loss": 4.448,
+      "step": 250
+    },
+    {
+      "epoch": 0.005243415480175856,
+      "grad_norm": 0.06689453125,
+      "learning_rate": 0.0001999949055698819,
+      "loss": 4.5337,
+      "step": 260
+    },
+    {
+      "epoch": 0.005445085306336466,
+      "grad_norm": 0.0654296875,
+      "learning_rate": 0.00019999424461618145,
+      "loss": 4.5358,
+      "step": 270
+    },
+    {
+      "epoch": 0.005646755132497076,
+      "grad_norm": 0.06103515625,
+      "learning_rate": 0.00019999354336201828,
+      "loss": 4.5168,
+      "step": 280
+    },
+    {
+      "epoch": 0.005848424958657686,
+      "grad_norm": 0.0634765625,
+      "learning_rate": 0.0001999928018076749,
+      "loss": 4.4455,
+      "step": 290
+    },
+    {
+      "epoch": 0.006050094784818295,
+      "grad_norm": 0.06591796875,
+      "learning_rate": 0.00019999201995345026,
+      "loss": 4.4747,
+      "step": 300
+    },
+    {
+      "epoch": 0.006251764610978905,
+      "grad_norm": 0.064453125,
+      "learning_rate": 0.00019999119779965947,
+      "loss": 4.4948,
+      "step": 310
+    },
+    {
+      "epoch": 0.006312265558827088,
+      "eval_loss": 4.493932723999023,
+      "eval_runtime": 8.8208,
+      "eval_samples_per_second": 22.674,
+      "eval_steps_per_second": 1.474,
+      "step": 313
+    },
+    {
+      "epoch": 0.00028233775662485377,
+      "grad_norm": 0.046875,
+      "learning_rate": 0.00019996118655688004,
+      "loss": 4.4607,
+      "step": 320
+    },
+    {
+      "epoch": 0.0006856774089460735,
+      "grad_norm": 0.04541015625,
+      "learning_rate": 0.00019995756127956854,
+      "loss": 4.4336,
+      "step": 330
+    },
+    {
+      "epoch": 0.0010890170612672932,
+      "grad_norm": 0.046142578125,
+      "learning_rate": 0.00019995377420631467,
+      "loss": 4.4016,
+      "step": 340
+    },
+    {
+      "epoch": 0.001492356713588513,
+      "grad_norm": 0.04541015625,
+      "learning_rate": 0.00019994982534324835,
+      "loss": 4.2976,
+      "step": 350
+    },
+    {
+      "epoch": 0.0018956963659097325,
+      "grad_norm": 0.05224609375,
+      "learning_rate": 0.00019994571469676142,
+      "loss": 4.2416,
+      "step": 360
+    },
+    {
+      "epoch": 0.0022990360182309523,
+      "grad_norm": 0.052734375,
+      "learning_rate": 0.00019994144227350756,
+      "loss": 4.1895,
+      "step": 370
+    },
+    {
+      "epoch": 0.002702375670552172,
+      "grad_norm": 0.058349609375,
+      "learning_rate": 0.00019993700808040233,
+      "loss": 4.1082,
+      "step": 380
+    },
+    {
+      "epoch": 0.003105715322873392,
+      "grad_norm": 0.057861328125,
+      "learning_rate": 0.0001999324121246231,
+      "loss": 4.148,
+      "step": 390
+    },
+    {
+      "epoch": 0.0035090549751946114,
+      "grad_norm": 0.06494140625,
+      "learning_rate": 0.00019992765441360905,
+      "loss": 4.1184,
+      "step": 400
+    },
+    {
+      "epoch": 0.003912394627515831,
+      "grad_norm": 0.068359375,
+      "learning_rate": 0.00019992273495506133,
+      "loss": 4.1018,
+      "step": 410
+    },
+    {
+      "epoch": 0.004315734279837051,
+      "grad_norm": 0.06005859375,
+      "learning_rate": 0.00019991765375694276,
+      "loss": 4.0995,
+      "step": 420
+    },
+    {
+      "epoch": 0.0047190739321582706,
+      "grad_norm": 0.0634765625,
+      "learning_rate": 0.00019991241082747795,
+      "loss": 4.1194,
+      "step": 430
+    },
+    {
+      "epoch": 0.00512241358447949,
+      "grad_norm": 0.0634765625,
+      "learning_rate": 0.00019990700617515344,
+      "loss": 4.0612,
+      "step": 440
+    },
+    {
+      "epoch": 0.00552575323680071,
+      "grad_norm": 0.06298828125,
+      "learning_rate": 0.00019990143980871738,
+      "loss": 4.1001,
+      "step": 450
+    },
+    {
+      "epoch": 0.005929092889121929,
+      "grad_norm": 0.0634765625,
+      "learning_rate": 0.00019989571173717975,
+      "loss": 4.075,
+      "step": 460
+    },
+    {
+      "epoch": 0.00633243254144315,
+      "grad_norm": 0.07275390625,
+      "learning_rate": 0.00019988982196981233,
+      "loss": 4.1117,
+      "step": 470
+    },
+    {
+      "epoch": 0.006735772193764369,
+      "grad_norm": 0.0517578125,
+      "learning_rate": 0.00019988377051614854,
+      "loss": 4.5104,
+      "step": 480
+    },
+    {
+      "epoch": 0.007139111846085589,
+      "grad_norm": 0.047607421875,
+      "learning_rate": 0.00019987755738598356,
+      "loss": 4.5086,
+      "step": 490
+    },
+    {
+      "epoch": 0.007542451498406808,
+      "grad_norm": 0.047607421875,
+      "learning_rate": 0.00019987118258937416,
+      "loss": 4.4517,
+      "step": 500
+    },
+    {
+      "epoch": 0.007945791150728028,
+      "grad_norm": 0.044677734375,
+      "learning_rate": 0.000199864646136639,
+      "loss": 4.419,
+      "step": 510
+    },
+    {
+      "epoch": 0.008349130803049248,
+      "grad_norm": 0.046142578125,
+      "learning_rate": 0.00019985794803835825,
+      "loss": 4.4749,
+      "step": 520
+    },
+    {
+      "epoch": 0.008752470455370467,
+      "grad_norm": 0.0478515625,
+      "learning_rate": 0.00019985108830537372,
+      "loss": 4.4646,
+      "step": 530
+    },
+    {
+      "epoch": 0.009155810107691688,
+      "grad_norm": 0.04345703125,
+      "learning_rate": 0.00019984406694878895,
+      "loss": 4.4207,
+      "step": 540
+    },
+    {
+      "epoch": 0.009559149760012906,
+      "grad_norm": 0.044677734375,
+      "learning_rate": 0.00019983688397996898,
+      "loss": 4.4308,
+      "step": 550
+    },
+    {
+      "epoch": 0.009962489412334127,
+      "grad_norm": 0.04345703125,
+      "learning_rate": 0.00019982953941054054,
+      "loss": 4.4311,
+      "step": 560
+    },
+    {
+      "epoch": 0.010365829064655347,
+      "grad_norm": 0.04541015625,
+      "learning_rate": 0.00019982203325239186,
+      "loss": 4.4623,
+      "step": 570
+    },
+    {
+      "epoch": 0.010769168716976566,
+      "grad_norm": 0.04833984375,
+      "learning_rate": 0.00019981436551767275,
+      "loss": 4.4461,
+      "step": 580
+    },
+    {
+      "epoch": 0.011172508369297786,
+      "grad_norm": 0.046630859375,
+      "learning_rate": 0.00019980653621879462,
+      "loss": 4.4197,
+      "step": 590
+    },
+    {
+      "epoch": 0.011575848021619005,
+      "grad_norm": 0.047119140625,
+      "learning_rate": 0.00019979854536843027,
+      "loss": 4.4135,
+      "step": 600
+    },
+    {
+      "epoch": 0.011979187673940225,
+      "grad_norm": 0.0498046875,
+      "learning_rate": 0.0001997903929795141,
+      "loss": 4.4105,
+      "step": 610
+    },
+    {
+      "epoch": 0.012382527326261444,
+      "grad_norm": 0.04638671875,
+      "learning_rate": 0.00019978207906524192,
+      "loss": 4.4107,
+      "step": 620
+    },
+    {
+      "epoch": 0.012624531117654176,
+      "eval_loss": 4.402504920959473,
+      "eval_runtime": 3.1085,
+      "eval_samples_per_second": 64.339,
+      "eval_steps_per_second": 8.042,
+      "step": 626
+    },
+    {
+      "epoch": 0.00016133586092848787,
+      "grad_norm": 0.047119140625,
+      "learning_rate": 0.000199773603639071,
+      "loss": 4.2779,
+      "step": 630
+    },
+    {
+      "epoch": 0.0005646755132497075,
+      "grad_norm": 0.0517578125,
+      "learning_rate": 0.0001997649667147201,
+      "loss": 4.2592,
+      "step": 640
+    },
+    {
+      "epoch": 0.0009680151655709273,
+      "grad_norm": 0.053466796875,
+      "learning_rate": 0.00019975616830616937,
+      "loss": 4.2228,
+      "step": 650
+    },
+    {
+      "epoch": 0.001371354817892147,
+      "grad_norm": 0.055908203125,
+      "learning_rate": 0.00019974720842766023,
+      "loss": 4.122,
+      "step": 660
+    },
+    {
+      "epoch": 0.0017746944702133668,
+      "grad_norm": 0.057373046875,
+      "learning_rate": 0.00019973808709369565,
+      "loss": 4.0099,
+      "step": 670
+    },
+    {
+      "epoch": 0.0021780341225345863,
+      "grad_norm": 0.0703125,
+      "learning_rate": 0.00019972880431903977,
+      "loss": 3.9065,
+      "step": 680
+    },
+    {
+      "epoch": 0.002581373774855806,
+      "grad_norm": 0.06982421875,
+      "learning_rate": 0.00019971936011871816,
+      "loss": 3.7953,
+      "step": 690
+    },
+    {
+      "epoch": 0.002984713427177026,
+      "grad_norm": 0.08349609375,
+      "learning_rate": 0.00019970975450801762,
+      "loss": 3.7788,
+      "step": 700
+    },
+    {
+      "epoch": 0.0033880530794982455,
+      "grad_norm": 0.08544921875,
+      "learning_rate": 0.00019969998750248626,
+      "loss": 3.7368,
+      "step": 710
+    },
+    {
+      "epoch": 0.003791392731819465,
+      "grad_norm": 0.09375,
+      "learning_rate": 0.0001996900591179334,
+      "loss": 3.7229,
+      "step": 720
+    },
+    {
+      "epoch": 0.004194732384140685,
+      "grad_norm": 0.095703125,
+      "learning_rate": 0.0001996799693704296,
+      "loss": 3.7192,
+      "step": 730
+    },
+    {
+      "epoch": 0.004598072036461905,
+      "grad_norm": 0.09228515625,
+      "learning_rate": 0.00019966971827630654,
+      "loss": 3.7457,
+      "step": 740
+    },
+    {
+      "epoch": 0.005001411688783124,
+      "grad_norm": 0.0986328125,
+      "learning_rate": 0.00019965930585215714,
+      "loss": 3.7031,
+      "step": 750
+    },
+    {
+      "epoch": 0.005404751341104344,
+      "grad_norm": 0.10595703125,
+      "learning_rate": 0.00019964873211483547,
+      "loss": 3.7335,
+      "step": 760
+    },
+    {
+      "epoch": 0.005808090993425563,
+      "grad_norm": 0.10986328125,
+      "learning_rate": 0.00019963799708145664,
+      "loss": 3.6902,
+      "step": 770
+    },
+    {
+      "epoch": 0.006211430645746784,
+      "grad_norm": 0.10400390625,
+      "learning_rate": 0.00019962710076939686,
+      "loss": 3.7408,
+      "step": 780
+    },
+    {
+      "epoch": 0.006614770298068003,
+      "grad_norm": 0.062255859375,
+      "learning_rate": 0.00019961604319629342,
+      "loss": 4.204,
+      "step": 790
+    },
+    {
+      "epoch": 0.007018109950389223,
+      "grad_norm": 0.053955078125,
+      "learning_rate": 0.00019960482438004462,
+      "loss": 4.364,
+      "step": 800
+    },
+    {
+      "epoch": 0.007421449602710442,
+      "grad_norm": 0.053955078125,
+      "learning_rate": 0.00019959344433880978,
+      "loss": 4.3305,
+      "step": 810
+    },
+    {
+      "epoch": 0.007824789255031663,
+      "grad_norm": 0.04833984375,
+      "learning_rate": 0.0001995819030910091,
+      "loss": 4.2992,
+      "step": 820
+    },
+    {
+      "epoch": 0.008228128907352882,
+      "grad_norm": 0.049072265625,
+      "learning_rate": 0.00019957020065532386,
+      "loss": 4.3347,
+      "step": 830
+    },
+    {
+      "epoch": 0.008631468559674102,
+      "grad_norm": 0.04931640625,
+      "learning_rate": 0.0001995583370506961,
+      "loss": 4.321,
+      "step": 840
+    },
+    {
+      "epoch": 0.00903480821199532,
+      "grad_norm": 0.0517578125,
+      "learning_rate": 0.00019954631229632884,
+      "loss": 4.3122,
+      "step": 850
+    },
+    {
+      "epoch": 0.009438147864316541,
+      "grad_norm": 0.05322265625,
+      "learning_rate": 0.00019953412641168588,
+      "loss": 4.2918,
+      "step": 860
+    },
+    {
+      "epoch": 0.00984148751663776,
+      "grad_norm": 0.052490234375,
+      "learning_rate": 0.00019952177941649185,
+      "loss": 4.2793,
+      "step": 870
+    },
+    {
+      "epoch": 0.01024482716895898,
+      "grad_norm": 0.048828125,
+      "learning_rate": 0.00019950927133073222,
+      "loss": 4.3363,
+      "step": 880
+    },
+    {
+      "epoch": 0.0106481668212802,
+      "grad_norm": 0.05126953125,
+      "learning_rate": 0.00019949660217465307,
+      "loss": 4.311,
+      "step": 890
+    },
+    {
+      "epoch": 0.01105150647360142,
+      "grad_norm": 0.05126953125,
+      "learning_rate": 0.00019948377196876138,
+      "loss": 4.2843,
+      "step": 900
+    },
+    {
+      "epoch": 0.01145484612592264,
+      "grad_norm": 0.053466796875,
+      "learning_rate": 0.00019947078073382466,
+      "loss": 4.2645,
+      "step": 910
+    },
+    {
+      "epoch": 0.011858185778243858,
+      "grad_norm": 0.05029296875,
+      "learning_rate": 0.00019945762849087113,
+      "loss": 4.2695,
+      "step": 920
+    },
+    {
+      "epoch": 0.012261525430565079,
+      "grad_norm": 0.054931640625,
+      "learning_rate": 0.00019944431526118964,
+      "loss": 4.279,
+      "step": 930
+    },
+    {
+      "epoch": 0.012624531117654176,
+      "eval_loss": 4.440088272094727,
+      "eval_runtime": 3.1398,
+      "eval_samples_per_second": 63.698,
+      "eval_steps_per_second": 7.962,
+      "step": 939
+    },
+    {
+      "epoch": 4.033396523212197e-05,
+      "grad_norm": 0.058837890625,
+      "learning_rate": 0.0,
+      "loss": 4.0975,
+      "step": 940
+    },
+    {
+      "epoch": 0.0004436736175533417,
+      "grad_norm": 0.07421875,
+      "learning_rate": 2e-05,
+      "loss": 4.1075,
+      "step": 950
+    },
+    {
+      "epoch": 0.0008470132698745614,
+      "grad_norm": 0.0703125,
+      "learning_rate": 4e-05,
+      "loss": 4.015,
+      "step": 960
+    },
+    {
+      "epoch": 0.001250352922195781,
+      "grad_norm": 0.07373046875,
+      "learning_rate": 6e-05,
+      "loss": 3.8564,
+      "step": 970
+    },
+    {
+      "epoch": 0.0016536925745170008,
+      "grad_norm": 0.1279296875,
+      "learning_rate": 8e-05,
+      "loss": 3.5457,
+      "step": 980
+    },
+    {
+      "epoch": 0.0020570322268382204,
+      "grad_norm": 0.1982421875,
+      "learning_rate": 0.0001,
+      "loss": 3.2341,
+      "step": 990
+    },
+    {
+      "epoch": 0.00246037187915944,
+      "grad_norm": 0.234375,
+      "learning_rate": 0.00012,
+      "loss": 3.0956,
+      "step": 1000
+    },
+    {
+      "epoch": 0.00286371153148066,
+      "grad_norm": 0.27734375,
+      "learning_rate": 0.00014,
+      "loss": 2.9331,
+      "step": 1010
+    },
+    {
+      "epoch": 0.0032670511838018795,
+      "grad_norm": 0.337890625,
+      "learning_rate": 0.00016,
+      "loss": 2.8938,
+      "step": 1020
+    },
+    {
+      "epoch": 0.003670390836123099,
+      "grad_norm": 0.328125,
+      "learning_rate": 0.00018,
+      "loss": 2.8145,
+      "step": 1030
+    },
+    {
+      "epoch": 0.004073730488444319,
+      "grad_norm": 0.296875,
+      "learning_rate": 0.0002,
+      "loss": 2.9022,
+      "step": 1040
+    },
+    {
+      "epoch": 0.004477070140765539,
+      "grad_norm": 0.279296875,
+      "learning_rate": 0.0001999999190676822,
+      "loss": 2.9572,
+      "step": 1050
+    },
+    {
+      "epoch": 0.004880409793086758,
+      "grad_norm": 0.33203125,
+      "learning_rate": 0.00019999967627085973,
+      "loss": 2.9381,
+      "step": 1060
+    },
+    {
+      "epoch": 0.005283749445407978,
+      "grad_norm": 0.2470703125,
+      "learning_rate": 0.00019999927160992563,
+      "loss": 2.9392,
+      "step": 1070
+    },
+    {
+      "epoch": 0.005687089097729197,
+      "grad_norm": 0.267578125,
+      "learning_rate": 0.00019999870508553488,
+      "loss": 2.8675,
+      "step": 1080
+    },
+    {
+      "epoch": 0.006090428750050418,
+      "grad_norm": 0.279296875,
+      "learning_rate": 0.00019999797669860455,
+      "loss": 2.9042,
+      "step": 1090
+    },
+    {
+      "epoch": 0.006493768402371637,
+      "grad_norm": 0.162109375,
+      "learning_rate": 0.00019999708645031353,
+      "loss": 3.4063,
+      "step": 1100
+    },
+    {
+      "epoch": 0.006897108054692857,
+      "grad_norm": 0.08349609375,
+      "learning_rate": 0.00019999603434210292,
+      "loss": 4.137,
+      "step": 1110
+    },
+    {
+      "epoch": 0.0073004477070140765,
+      "grad_norm": 0.0732421875,
+      "learning_rate": 0.00019999482037567565,
+      "loss": 4.1305,
+      "step": 1120
+    },
+    {
+      "epoch": 0.007703787359335296,
+      "grad_norm": 0.06591796875,
+      "learning_rate": 0.00019999344455299674,
+      "loss": 4.0303,
+      "step": 1130
+    },
+    {
+      "epoch": 0.008107127011656516,
+      "grad_norm": 0.0634765625,
+      "learning_rate": 0.0001999919068762931,
+      "loss": 4.0717,
+      "step": 1140
+    },
+    {
+      "epoch": 0.008510466663977735,
+      "grad_norm": 0.059326171875,
+      "learning_rate": 0.00019999020734805373,
+      "loss": 4.0664,
+      "step": 1150
+    },
+    {
+      "epoch": 0.008913806316298956,
+      "grad_norm": 0.060302734375,
+      "learning_rate": 0.0001999883459710296,
+      "loss": 4.0298,
+      "step": 1160
+    },
+    {
+      "epoch": 0.009317145968620174,
+      "grad_norm": 0.057861328125,
+      "learning_rate": 0.00019998632274823358,
+      "loss": 4.0348,
+      "step": 1170
+    },
+    {
+      "epoch": 0.009720485620941395,
+      "grad_norm": 0.06201171875,
+      "learning_rate": 0.00019998413768294052,
+      "loss": 4.0192,
+      "step": 1180
+    },
+    {
+      "epoch": 0.010123825273262615,
+      "grad_norm": 0.064453125,
+      "learning_rate": 0.0001999817907786873,
+      "loss": 4.033,
+      "step": 1190
+    },
+    {
+      "epoch": 0.010527164925583834,
+      "grad_norm": 0.0654296875,
+      "learning_rate": 0.00019997928203927275,
+      "loss": 4.0413,
+      "step": 1200
+    },
+    {
+      "epoch": 0.010930504577905054,
+      "grad_norm": 0.06884765625,
+      "learning_rate": 0.00019997661146875758,
+      "loss": 4.0011,
+      "step": 1210
+    },
+    {
+      "epoch": 0.011333844230226273,
+      "grad_norm": 0.068359375,
+      "learning_rate": 0.00019997377907146459,
+      "loss": 3.9817,
+      "step": 1220
+    },
+    {
+      "epoch": 0.011737183882547493,
+      "grad_norm": 0.06884765625,
+      "learning_rate": 0.0001999707848519783,
+      "loss": 4.0007,
+      "step": 1230
+    },
+    {
+      "epoch": 0.012140523534868712,
+      "grad_norm": 0.0654296875,
+      "learning_rate": 0.0001999676288151454,
+      "loss": 4.0265,
+      "step": 1240
+    },
+    {
+      "epoch": 0.012543863187189933,
+      "grad_norm": 0.060302734375,
+      "learning_rate": 0.00019996431096607438,
+      "loss": 4.172,
+      "step": 1250
+    },
+    {
+      "epoch": 0.012624531117654176,
+      "eval_loss": 4.49122428894043,
+      "eval_runtime": 3.1912,
+      "eval_samples_per_second": 62.672,
+      "eval_steps_per_second": 7.834,
+      "step": 1252
+    },
+    {
+      "epoch": 0.00032267172185697574,
+      "grad_norm": 0.08203125,
+      "learning_rate": 1.4000000000000001e-06,
+      "loss": 4.1169,
+      "step": 1260
+    },
+    {
+      "epoch": 0.0007260113741781954,
+      "grad_norm": 0.1220703125,
+      "learning_rate": 3.4000000000000005e-06,
+      "loss": 4.0418,
+      "step": 1270
+    },
+    {
+      "epoch": 0.001129351026499415,
+      "grad_norm": 0.1943359375,
+      "learning_rate": 5.400000000000001e-06,
+      "loss": 3.9211,
+      "step": 1280
+    },
+    {
+      "epoch": 0.0015326906788206349,
+      "grad_norm": 0.328125,
+      "learning_rate": 7.4e-06,
+      "loss": 3.6888,
+      "step": 1290
+    },
+    {
+      "epoch": 0.0019360303311418546,
+      "grad_norm": 0.375,
+      "learning_rate": 9.4e-06,
+      "loss": 3.4768,
+      "step": 1300
+    },
+    {
+      "epoch": 0.0023393699834630744,
+      "grad_norm": 0.408203125,
+      "learning_rate": 1.14e-05,
+      "loss": 3.242,
+      "step": 1310
+    },
+    {
+      "epoch": 0.002742709635784294,
+      "grad_norm": 0.3671875,
+      "learning_rate": 1.3400000000000002e-05,
+      "loss": 2.9401,
+      "step": 1320
+    },
+    {
+      "epoch": 0.0031460492881055136,
+      "grad_norm": 0.265625,
+      "learning_rate": 1.54e-05,
+      "loss": 2.7321,
+      "step": 1330
+    },
+    {
+      "epoch": 0.0035493889404267336,
+      "grad_norm": 0.23828125,
+      "learning_rate": 1.7400000000000003e-05,
+      "loss": 2.413,
+      "step": 1340
+    },
+    {
+      "epoch": 0.003952728592747953,
+      "grad_norm": 0.2353515625,
+      "learning_rate": 1.94e-05,
+      "loss": 2.2395,
+      "step": 1350
+    },
+    {
+      "epoch": 0.004356068245069173,
+      "grad_norm": 0.2353515625,
+      "learning_rate": 1.9999999012581816e-05,
+      "loss": 2.2062,
+      "step": 1360
+    },
+    {
+      "epoch": 0.004759407897390392,
+      "grad_norm": 0.232421875,
+      "learning_rate": 1.999999417624832e-05,
+      "loss": 2.1174,
+      "step": 1370
+    },
+    {
+      "epoch": 0.005162747549711612,
+      "grad_norm": 0.2265625,
+      "learning_rate": 1.999998530963894e-05,
+      "loss": 1.9498,
+      "step": 1380
+    },
+    {
+      "epoch": 0.005566087202032832,
+      "grad_norm": 0.23046875,
+      "learning_rate": 1.999997241275724e-05,
+      "loss": 1.8648,
+      "step": 1390
+    },
+    {
+      "epoch": 0.005969426854354052,
+      "grad_norm": 0.2392578125,
+      "learning_rate": 1.9999955485608426e-05,
+      "loss": 1.8192,
+      "step": 1400
+    },
+    {
+      "epoch": 0.006372766506675271,
+      "grad_norm": 0.296875,
+      "learning_rate": 1.999993452819932e-05,
+      "loss": 2.0548,
+      "step": 1410
+    },
+    {
+      "epoch": 0.006776106158996491,
+      "grad_norm": 0.34765625,
+      "learning_rate": 1.999990954053836e-05,
+      "loss": 3.5804,
+      "step": 1420
+    },
+    {
+      "epoch": 0.0071794458113177105,
+      "grad_norm": 0.21484375,
+      "learning_rate": 1.9999880522635625e-05,
+      "loss": 3.8265,
+      "step": 1430
+    },
+    {
+      "epoch": 0.00758278546363893,
+      "grad_norm": 0.130859375,
+      "learning_rate": 1.999984747450281e-05,
+      "loss": 3.7124,
+      "step": 1440
+    },
+    {
+      "epoch": 0.00798612511596015,
+      "grad_norm": 0.1142578125,
+      "learning_rate": 1.9999810396153232e-05,
+      "loss": 3.7061,
+      "step": 1450
+    },
+    {
+      "epoch": 0.00838946476828137,
+      "grad_norm": 0.103515625,
+      "learning_rate": 1.9999769287601834e-05,
+      "loss": 3.68,
+      "step": 1460
+    },
+    {
+      "epoch": 0.008792804420602589,
+      "grad_norm": 0.09814453125,
+      "learning_rate": 1.9999724148865183e-05,
+      "loss": 3.6099,
+      "step": 1470
+    },
+    {
+      "epoch": 0.00919614407292381,
+      "grad_norm": 0.09423828125,
+      "learning_rate": 1.9999674979961473e-05,
+      "loss": 3.613,
+      "step": 1480
+    },
+    {
+      "epoch": 0.00959948372524503,
+      "grad_norm": 0.0966796875,
+      "learning_rate": 1.999962178091052e-05,
+      "loss": 3.5696,
+      "step": 1490
+    },
+    {
+      "epoch": 0.010002823377566248,
+      "grad_norm": 0.10986328125,
+      "learning_rate": 1.9999564551733764e-05,
+      "loss": 3.5333,
+      "step": 1500
+    },
+    {
+      "epoch": 0.010406163029887469,
+      "grad_norm": 0.09375,
+      "learning_rate": 1.9999503292454275e-05,
+      "loss": 3.5119,
+      "step": 1510
+    },
+    {
+      "epoch": 0.010809502682208687,
+      "grad_norm": 0.103515625,
+      "learning_rate": 1.9999438003096733e-05,
+      "loss": 3.403,
+      "step": 1520
+    },
+    {
+      "epoch": 0.011212842334529908,
+      "grad_norm": 0.10546875,
+      "learning_rate": 1.9999368683687457e-05,
+      "loss": 3.3882,
+      "step": 1530
+    },
+    {
+      "epoch": 0.011616181986851127,
+      "grad_norm": 0.0986328125,
+      "learning_rate": 1.999929533425439e-05,
+      "loss": 3.3576,
+      "step": 1540
+    },
+    {
+      "epoch": 0.012019521639172347,
+      "grad_norm": 0.107421875,
+      "learning_rate": 1.999921795482708e-05,
+      "loss": 3.4566,
+      "step": 1550
+    },
+    {
+      "epoch": 0.012422861291493567,
+      "grad_norm": 0.158203125,
+      "learning_rate": 1.9999136545436727e-05,
+      "loss": 3.7767,
+      "step": 1560
+    },
+    {
+      "epoch": 0.012826200943814786,
+      "grad_norm": 0.2177734375,
+      "learning_rate": 1.999905110611613e-05,
+      "loss": 4.5098,
+      "step": 1570
+    },
+    {
+      "epoch": 0.013229540596136007,
+      "grad_norm": 0.1552734375,
+      "learning_rate": 1.9998961636899736e-05,
+      "loss": 4.6336,
+      "step": 1580
+    },
+    {
+      "epoch": 0.013632880248457225,
+      "grad_norm": 0.12255859375,
+      "learning_rate": 1.999886813782359e-05,
+      "loss": 4.5697,
+      "step": 1590
+    },
+    {
+      "epoch": 0.014036219900778446,
+      "grad_norm": 0.10498046875,
+      "learning_rate": 1.999877060892538e-05,
+      "loss": 4.5598,
+      "step": 1600
+    },
+    {
+      "epoch": 0.014439559553099664,
+      "grad_norm": 0.08984375,
+      "learning_rate": 1.9998669050244416e-05,
+      "loss": 4.5326,
+      "step": 1610
+    },
+    {
+      "epoch": 0.014842899205420885,
+      "grad_norm": 0.07958984375,
+      "learning_rate": 1.999856346182163e-05,
+      "loss": 4.5219,
+      "step": 1620
+    },
+    {
+      "epoch": 0.015246238857742105,
+      "grad_norm": 0.07177734375,
+      "learning_rate": 1.999845384369957e-05,
+      "loss": 4.5013,
+      "step": 1630
+    },
+    {
+      "epoch": 0.015649578510063326,
+      "grad_norm": 0.07080078125,
+      "learning_rate": 1.9998340195922418e-05,
+      "loss": 4.4859,
+      "step": 1640
+    },
+    {
+      "epoch": 0.016052918162384543,
+      "grad_norm": 0.061767578125,
+      "learning_rate": 1.999822251853598e-05,
+      "loss": 4.5328,
+      "step": 1650
+    },
+    {
+      "epoch": 0.016456257814705763,
+      "grad_norm": 0.05859375,
+      "learning_rate": 1.9998100811587686e-05,
+      "loss": 4.5024,
+      "step": 1660
+    },
+    {
+      "epoch": 0.016859597467026984,
+      "grad_norm": 0.06201171875,
+      "learning_rate": 1.9997975075126573e-05,
+      "loss": 4.5178,
+      "step": 1670
+    },
+    {
+      "epoch": 0.017262937119348204,
+      "grad_norm": 0.05908203125,
+      "learning_rate": 1.9997845309203333e-05,
+      "loss": 4.4892,
+      "step": 1680
+    },
+    {
+      "epoch": 0.017666276771669424,
+      "grad_norm": 0.055908203125,
+      "learning_rate": 1.9997711513870257e-05,
+      "loss": 4.4645,
+      "step": 1690
+    },
+    {
+      "epoch": 0.01806961642399064,
+      "grad_norm": 0.0546875,
+      "learning_rate": 1.9997573689181272e-05,
+      "loss": 4.4891,
+      "step": 1700
+    },
+    {
+      "epoch": 0.018472956076311862,
+      "grad_norm": 0.054443359375,
+      "learning_rate": 1.999743183519192e-05,
+      "loss": 4.4604,
+      "step": 1710
+    },
+    {
+      "epoch": 0.018876295728633082,
+      "grad_norm": 0.05224609375,
+      "learning_rate": 1.9997285951959372e-05,
+      "loss": 4.4935,
+      "step": 1720
+    },
+    {
+      "epoch": 0.019279635380954303,
+      "grad_norm": 0.052978515625,
+      "learning_rate": 1.999713603954243e-05,
+      "loss": 4.4723,
+      "step": 1730
+    },
+    {
+      "epoch": 0.01968297503327552,
+      "grad_norm": 0.052734375,
+      "learning_rate": 1.9996982098001508e-05,
+      "loss": 4.4923,
+      "step": 1740
+    },
+    {
+      "epoch": 0.02008631468559674,
+      "grad_norm": 0.050537109375,
+      "learning_rate": 1.9996824127398648e-05,
+      "loss": 4.4402,
+      "step": 1750
+    },
+    {
+      "epoch": 0.02048965433791796,
+      "grad_norm": 0.05078125,
+      "learning_rate": 1.999666212779752e-05,
+      "loss": 4.4716,
+      "step": 1760
+    },
+    {
+      "epoch": 0.02089299399023918,
+      "grad_norm": 0.0478515625,
+      "learning_rate": 1.999649609926341e-05,
+      "loss": 4.4952,
+      "step": 1770
+    },
+    {
+      "epoch": 0.0212963336425604,
+      "grad_norm": 0.052490234375,
+      "learning_rate": 1.9996326041863236e-05,
+      "loss": 4.4755,
+      "step": 1780
+    },
+    {
+      "epoch": 0.02169967329488162,
+      "grad_norm": 0.04931640625,
+      "learning_rate": 1.9996151955665535e-05,
+      "loss": 4.4645,
+      "step": 1790
+    },
+    {
+      "epoch": 0.02210301294720284,
+      "grad_norm": 0.05029296875,
+      "learning_rate": 1.9995973840740467e-05,
+      "loss": 4.474,
+      "step": 1800
+    },
+    {
+      "epoch": 0.02250635259952406,
+      "grad_norm": 0.05029296875,
+      "learning_rate": 1.999579169715982e-05,
+      "loss": 4.4614,
+      "step": 1810
+    },
+    {
+      "epoch": 0.02290969225184528,
+      "grad_norm": 0.049560546875,
+      "learning_rate": 1.9995605524996996e-05,
+      "loss": 4.4622,
+      "step": 1820
+    },
+    {
+      "epoch": 0.0233130319041665,
+      "grad_norm": 0.052490234375,
+      "learning_rate": 1.9995415324327038e-05,
+      "loss": 4.4785,
+      "step": 1830
+    },
+    {
+      "epoch": 0.023716371556487717,
+      "grad_norm": 0.048828125,
+      "learning_rate": 1.9995221095226597e-05,
+      "loss": 4.5104,
+      "step": 1840
+    },
+    {
+      "epoch": 0.024119711208808937,
+      "grad_norm": 0.048095703125,
+      "learning_rate": 1.999502283777395e-05,
+      "loss": 4.4759,
+      "step": 1850
+    },
+    {
+      "epoch": 0.024523050861130158,
+      "grad_norm": 0.0458984375,
+      "learning_rate": 1.9994820552049002e-05,
+      "loss": 4.4359,
+      "step": 1860
+    },
+    {
+      "epoch": 0.02492639051345138,
+      "grad_norm": 0.045166015625,
+      "learning_rate": 1.9994614238133282e-05,
+      "loss": 4.4895,
+      "step": 1870
+    },
+    {
+      "epoch": 0.0253297301657726,
+      "grad_norm": 0.04541015625,
+      "learning_rate": 1.9994403896109942e-05,
+      "loss": 4.4755,
+      "step": 1880
+    },
+    {
+      "epoch": 0.025733069818093816,
+      "grad_norm": 0.045654296875,
+      "learning_rate": 1.9994189526063746e-05,
+      "loss": 4.4661,
+      "step": 1890
+    },
+    {
+      "epoch": 0.026136409470415036,
+      "grad_norm": 0.046630859375,
+      "learning_rate": 1.99939711280811e-05,
+      "loss": 4.4775,
+      "step": 1900
+    },
+    {
+      "epoch": 0.026539749122736257,
+      "grad_norm": 0.04931640625,
+      "learning_rate": 1.999374870225003e-05,
+      "loss": 4.4954,
+      "step": 1910
+    },
+    {
+      "epoch": 0.026943088775057477,
+      "grad_norm": 0.046142578125,
+      "learning_rate": 1.9993522248660163e-05,
+      "loss": 4.4905,
+      "step": 1920
+    },
+    {
+      "epoch": 0.027346428427378694,
+      "grad_norm": 0.049072265625,
+      "learning_rate": 1.9993291767402776e-05,
+      "loss": 4.4545,
+      "step": 1930
+    },
+    {
+      "epoch": 0.027749768079699914,
+      "grad_norm": 0.045166015625,
+      "learning_rate": 1.9993057258570762e-05,
+      "loss": 4.4737,
+      "step": 1940
+    },
+    {
+      "epoch": 0.028153107732021135,
+      "grad_norm": 0.046630859375,
+      "learning_rate": 1.9992818722258626e-05,
+      "loss": 4.4908,
+      "step": 1950
+    },
+    {
+      "epoch": 0.028556447384342355,
+      "grad_norm": 0.047607421875,
+      "learning_rate": 1.9992576158562515e-05,
+      "loss": 4.4893,
+      "step": 1960
+    },
+    {
+      "epoch": 0.028959787036663576,
+      "grad_norm": 0.04736328125,
+      "learning_rate": 1.999232956758018e-05,
+      "loss": 4.4452,
+      "step": 1970
+    },
+    {
+      "epoch": 0.029363126688984793,
+      "grad_norm": 0.044677734375,
+      "learning_rate": 1.999207894941101e-05,
+      "loss": 4.4901,
+      "step": 1980
+    },
+    {
+      "epoch": 0.029766466341306013,
+      "grad_norm": 0.0478515625,
+      "learning_rate": 1.9991824304156006e-05,
+      "loss": 4.419,
+      "step": 1990
+    },
+    {
+      "epoch": 0.030169805993627234,
+      "grad_norm": 0.048828125,
+      "learning_rate": 1.99915656319178e-05,
+      "loss": 4.4521,
+      "step": 2000
+    },
+    {
+      "epoch": 0.030573145645948454,
+      "grad_norm": 0.046875,
+      "learning_rate": 1.999130293280065e-05,
+      "loss": 4.4561,
+      "step": 2010
+    },
+    {
+      "epoch": 0.030976485298269674,
+      "grad_norm": 0.0595703125,
+      "learning_rate": 1.9991036206910417e-05,
+      "loss": 4.4544,
+      "step": 2020
+    },
+    {
+      "epoch": 0.03137982495059089,
+      "grad_norm": 0.048095703125,
+      "learning_rate": 1.999076545435461e-05,
+      "loss": 4.4713,
+      "step": 2030
+    },
+    {
+      "epoch": 0.03178316460291211,
+      "grad_norm": 0.047119140625,
+      "learning_rate": 1.999049067524235e-05,
+      "loss": 4.4408,
+      "step": 2040
+    },
+    {
+      "epoch": 0.03218650425523333,
+      "grad_norm": 0.04541015625,
+      "learning_rate": 1.9990211869684374e-05,
+      "loss": 4.4686,
+      "step": 2050
+    },
+    {
+      "epoch": 0.03258984390755455,
+      "grad_norm": 0.047119140625,
+      "learning_rate": 1.998992903779305e-05,
+      "loss": 4.4673,
+      "step": 2060
+    },
+    {
+      "epoch": 0.03299318355987577,
+      "grad_norm": 0.046630859375,
+      "learning_rate": 1.9989642179682374e-05,
+      "loss": 4.4302,
+      "step": 2070
+    },
+    {
+      "epoch": 0.033396523212196993,
+      "grad_norm": 0.046875,
+      "learning_rate": 1.998935129546795e-05,
+      "loss": 4.4844,
+      "step": 2080
+    },
+    {
+      "epoch": 0.033799862864518214,
+      "grad_norm": 0.0439453125,
+      "learning_rate": 1.9989056385267015e-05,
+      "loss": 4.4739,
+      "step": 2090
+    },
+    {
+      "epoch": 0.03420320251683943,
+      "grad_norm": 0.047119140625,
+      "learning_rate": 1.9988757449198428e-05,
+      "loss": 4.499,
+      "step": 2100
+    },
+    {
+      "epoch": 0.03460654216916065,
+      "grad_norm": 0.046875,
+      "learning_rate": 1.9988454487382667e-05,
+      "loss": 4.4517,
+      "step": 2110
+    },
+    {
+      "epoch": 0.03500988182148187,
+      "grad_norm": 0.044677734375,
+      "learning_rate": 1.9988147499941832e-05,
+      "loss": 4.4608,
+      "step": 2120
+    },
+    {
+      "epoch": 0.03541322147380309,
+      "grad_norm": 0.0458984375,
+      "learning_rate": 1.998783648699965e-05,
+      "loss": 4.4515,
+      "step": 2130
+    },
+    {
+      "epoch": 0.03581656112612431,
+      "grad_norm": 0.04443359375,
+      "learning_rate": 1.9987521448681465e-05,
+      "loss": 4.4646,
+      "step": 2140
+    },
+    {
+      "epoch": 0.03621990077844553,
+      "grad_norm": 0.05908203125,
+      "learning_rate": 1.9987202385114252e-05,
+      "loss": 4.4586,
+      "step": 2150
+    },
+    {
+      "epoch": 0.03662324043076675,
+      "grad_norm": 0.04736328125,
+      "learning_rate": 1.99868792964266e-05,
+      "loss": 4.4699,
+      "step": 2160
+    },
+    {
+      "epoch": 0.03702658008308797,
+      "grad_norm": 0.044189453125,
+      "learning_rate": 1.9986552182748715e-05,
+      "loss": 4.4678,
+      "step": 2170
+    },
+    {
+      "epoch": 0.03742991973540919,
+      "grad_norm": 0.04541015625,
+      "learning_rate": 1.9986221044212442e-05,
+      "loss": 4.4581,
+      "step": 2180
+    },
+    {
+      "epoch": 0.037833259387730404,
+      "grad_norm": 0.045654296875,
+      "learning_rate": 1.998588588095124e-05,
+      "loss": 4.4273,
+      "step": 2190
+    },
+    {
+      "epoch": 0.038236599040051625,
+      "grad_norm": 0.044677734375,
+      "learning_rate": 1.9985546693100186e-05,
+      "loss": 4.4602,
+      "step": 2200
+    },
+    {
+      "epoch": 0.038639938692372845,
+      "grad_norm": 0.04931640625,
+      "learning_rate": 1.9985203480795977e-05,
+      "loss": 4.4766,
+      "step": 2210
+    },
+    {
+      "epoch": 0.039043278344694066,
+      "grad_norm": 0.045166015625,
+      "learning_rate": 1.9984856244176948e-05,
+      "loss": 4.4579,
+      "step": 2220
+    },
+    {
+      "epoch": 0.039446617997015286,
+      "grad_norm": 0.04638671875,
+      "learning_rate": 1.998450498338303e-05,
+      "loss": 4.4661,
+      "step": 2230
+    },
+    {
+      "epoch": 0.039849957649336507,
+      "grad_norm": 0.043701171875,
+      "learning_rate": 1.9984149698555808e-05,
+      "loss": 4.4415,
+      "step": 2240
+    },
+    {
+      "epoch": 0.04025329730165773,
+      "grad_norm": 0.046875,
+      "learning_rate": 1.998379038983846e-05,
+      "loss": 4.4461,
+      "step": 2250
+    },
+    {
+      "epoch": 0.04065663695397895,
+      "grad_norm": 0.04443359375,
+      "learning_rate": 1.9983427057375802e-05,
+      "loss": 4.4627,
+      "step": 2260
+    },
+    {
+      "epoch": 0.04105997660630017,
+      "grad_norm": 0.0458984375,
+      "learning_rate": 1.9983059701314267e-05,
+      "loss": 4.4399,
+      "step": 2270
+    },
+    {
+      "epoch": 0.04146331625862139,
+      "grad_norm": 0.045654296875,
+      "learning_rate": 1.9982688321801906e-05,
+      "loss": 4.4673,
+      "step": 2280
+    },
+    {
+      "epoch": 0.0418666559109426,
+      "grad_norm": 0.046142578125,
+      "learning_rate": 1.99823129189884e-05,
+      "loss": 4.478,
+      "step": 2290
+    },
+    {
+      "epoch": 0.04226999556326382,
+      "grad_norm": 0.045166015625,
+      "learning_rate": 1.9981933493025044e-05,
+      "loss": 4.4433,
+      "step": 2300
+    },
+    {
+      "epoch": 0.04267333521558504,
+      "grad_norm": 0.045166015625,
+      "learning_rate": 1.9981550044064756e-05,
+      "loss": 4.4535,
+      "step": 2310
+    },
+    {
+      "epoch": 0.04307667486790626,
+      "grad_norm": 0.046142578125,
+      "learning_rate": 1.998116257226208e-05,
+      "loss": 4.4665,
+      "step": 2320
+    },
+    {
+      "epoch": 0.043480014520227483,
+      "grad_norm": 0.045166015625,
+      "learning_rate": 1.9980771077773177e-05,
+      "loss": 4.4419,
+      "step": 2330
+    },
+    {
+      "epoch": 0.043883354172548704,
+      "grad_norm": 0.046875,
+      "learning_rate": 1.9980375560755833e-05,
+      "loss": 4.4724,
+      "step": 2340
+    },
+    {
+      "epoch": 0.044286693824869924,
+      "grad_norm": 0.044921875,
+      "learning_rate": 1.997997602136944e-05,
+      "loss": 4.4485,
+      "step": 2350
+    },
+    {
+      "epoch": 0.044690033477191145,
+      "grad_norm": 0.044921875,
+      "learning_rate": 1.997957245977504e-05,
+      "loss": 4.4484,
+      "step": 2360
+    },
+    {
+      "epoch": 0.045093373129512365,
+      "grad_norm": 0.044921875,
+      "learning_rate": 1.997916487613527e-05,
+      "loss": 4.4594,
+      "step": 2370
+    },
+    {
+      "epoch": 0.04549671278183358,
+      "grad_norm": 0.04638671875,
+      "learning_rate": 1.9978753270614403e-05,
+      "loss": 4.4836,
+      "step": 2380
+    },
+    {
+      "epoch": 0.0459000524341548,
+      "grad_norm": 0.044921875,
+      "learning_rate": 1.997833764337832e-05,
+      "loss": 4.4767,
+      "step": 2390
+    },
+    {
+      "epoch": 0.04630339208647602,
+      "grad_norm": 0.04736328125,
+      "learning_rate": 1.9977917994594537e-05,
+      "loss": 4.4358,
+      "step": 2400
+    },
+    {
+      "epoch": 0.04670673173879724,
+      "grad_norm": 0.04541015625,
+      "learning_rate": 1.997749432443218e-05,
+      "loss": 4.4552,
+      "step": 2410
+    },
+    {
+      "epoch": 0.04711007139111846,
+      "grad_norm": 0.044677734375,
+      "learning_rate": 1.9977066633062002e-05,
+      "loss": 4.4383,
+      "step": 2420
+    },
+    {
+      "epoch": 0.04751341104343968,
+      "grad_norm": 0.045654296875,
+      "learning_rate": 1.9976634920656374e-05,
+      "loss": 4.4605,
+      "step": 2430
+    },
+    {
+      "epoch": 0.0479167506957609,
+      "grad_norm": 0.042724609375,
+      "learning_rate": 1.9976199187389286e-05,
+      "loss": 4.4496,
+      "step": 2440
+    },
+    {
+      "epoch": 0.04832009034808212,
+      "grad_norm": 0.04638671875,
+      "learning_rate": 1.997575943343635e-05,
+      "loss": 4.454,
+      "step": 2450
+    },
+    {
+      "epoch": 0.04872343000040334,
+      "grad_norm": 0.044189453125,
+      "learning_rate": 1.997531565897481e-05,
+      "loss": 4.4212,
+      "step": 2460
+    },
+    {
+      "epoch": 0.04912676965272456,
+      "grad_norm": 0.043212890625,
+      "learning_rate": 1.9974867864183508e-05,
+      "loss": 4.4441,
+      "step": 2470
+    },
+    {
+      "epoch": 0.049530109305045776,
+      "grad_norm": 0.04541015625,
+      "learning_rate": 1.997441604924292e-05,
+      "loss": 4.4085,
+      "step": 2480
+    },
+    {
+      "epoch": 0.049933448957366997,
+      "grad_norm": 0.044189453125,
+      "learning_rate": 1.997396021433514e-05,
+      "loss": 4.4616,
+      "step": 2490
+    },
+    {
+      "epoch": 0.05033678860968822,
+      "grad_norm": 0.04833984375,
+      "learning_rate": 1.9973500359643885e-05,
+      "loss": 4.4544,
+      "step": 2500
+    },
+    {
+      "epoch": 0.05074012826200944,
+      "grad_norm": 0.044677734375,
+      "learning_rate": 1.9973036485354485e-05,
+      "loss": 4.453,
+      "step": 2510
+    },
+    {
+      "epoch": 0.05114346791433066,
+      "grad_norm": 0.04638671875,
+      "learning_rate": 1.99725685916539e-05,
+      "loss": 4.4722,
+      "step": 2520
+    },
+    {
+      "epoch": 0.05154680756665188,
+      "grad_norm": 0.04296875,
+      "learning_rate": 1.99720966787307e-05,
+      "loss": 4.4332,
+      "step": 2530
+    },
+    {
+      "epoch": 0.0519501472189731,
+      "grad_norm": 0.04638671875,
+      "learning_rate": 1.9971620746775077e-05,
+      "loss": 4.4757,
+      "step": 2540
+    },
+    {
+      "epoch": 0.05235348687129432,
+      "grad_norm": 0.04638671875,
+      "learning_rate": 1.997114079597885e-05,
+      "loss": 4.4264,
+      "step": 2550
+    },
+    {
+      "epoch": 0.05275682652361554,
+      "grad_norm": 0.04638671875,
+      "learning_rate": 1.997065682653545e-05,
+      "loss": 4.4372,
+      "step": 2560
+    },
+    {
+      "epoch": 0.05316016617593675,
+      "grad_norm": 0.045166015625,
+      "learning_rate": 1.997016883863993e-05,
+      "loss": 4.4566,
+      "step": 2570
+    },
+    {
+      "epoch": 0.05356350582825797,
+      "grad_norm": 0.04345703125,
+      "learning_rate": 1.9969676832488965e-05,
+      "loss": 4.4309,
+      "step": 2580
+    },
+    {
+      "epoch": 0.053966845480579194,
+      "grad_norm": 0.046142578125,
+      "learning_rate": 1.9969180808280845e-05,
+      "loss": 4.4621,
+      "step": 2590
+    },
+    {
+      "epoch": 0.054370185132900414,
+      "grad_norm": 0.04443359375,
+      "learning_rate": 1.9968680766215477e-05,
+      "loss": 4.4465,
+      "step": 2600
+    },
+    {
+      "epoch": 0.054773524785221635,
+      "grad_norm": 0.04541015625,
+      "learning_rate": 1.9968176706494403e-05,
+      "loss": 4.4239,
+      "step": 2610
+    },
+    {
+      "epoch": 0.055176864437542855,
+      "grad_norm": 0.042724609375,
+      "learning_rate": 1.996766862932076e-05,
+      "loss": 4.4567,
+      "step": 2620
+    },
+    {
+      "epoch": 0.055580204089864076,
+      "grad_norm": 0.046142578125,
+      "learning_rate": 1.996715653489933e-05,
+      "loss": 4.423,
+      "step": 2630
+    },
+    {
+      "epoch": 0.055983543742185296,
+      "grad_norm": 0.04248046875,
+      "learning_rate": 1.9966640423436492e-05,
+      "loss": 4.3849,
+      "step": 2640
+    },
+    {
+      "epoch": 0.056386883394506516,
+      "grad_norm": 0.0458984375,
+      "learning_rate": 1.9966120295140258e-05,
+      "loss": 4.4788,
+      "step": 2650
+    },
+    {
+      "epoch": 0.05679022304682774,
+      "grad_norm": 0.045166015625,
+      "learning_rate": 1.996559615022025e-05,
+      "loss": 4.4313,
+      "step": 2660
+    },
+    {
+      "epoch": 0.05719356269914895,
+      "grad_norm": 0.044921875,
+      "learning_rate": 1.996506798888772e-05,
+      "loss": 4.4493,
+      "step": 2670
+    },
+    {
+      "epoch": 0.05759690235147017,
+      "grad_norm": 0.047607421875,
+      "learning_rate": 1.9964535811355524e-05,
+      "loss": 4.4438,
+      "step": 2680
+    },
+    {
+      "epoch": 0.05800024200379139,
+      "grad_norm": 0.04296875,
+      "learning_rate": 1.996399961783815e-05,
+      "loss": 4.413,
+      "step": 2690
+    },
+    {
+      "epoch": 0.05840358165611261,
+      "grad_norm": 0.043701171875,
+      "learning_rate": 1.9963459408551693e-05,
+      "loss": 4.4538,
+      "step": 2700
+    },
+    {
+      "epoch": 0.05880692130843383,
+      "grad_norm": 0.0439453125,
+      "learning_rate": 1.996291518371388e-05,
+      "loss": 4.466,
+      "step": 2710
+    },
+    {
+      "epoch": 0.05921026096075505,
+      "grad_norm": 0.045654296875,
+      "learning_rate": 1.9962366943544045e-05,
+      "loss": 4.4963,
+      "step": 2720
+    },
+    {
+      "epoch": 0.05961360061307627,
+      "grad_norm": 0.04541015625,
+      "learning_rate": 1.9961814688263138e-05,
+      "loss": 4.46,
+      "step": 2730
+    },
+    {
+      "epoch": 0.06001694026539749,
+      "grad_norm": 0.04345703125,
+      "learning_rate": 1.9961258418093745e-05,
+      "loss": 4.4481,
+      "step": 2740
+    },
+    {
+      "epoch": 0.060420279917718714,
+      "grad_norm": 0.044677734375,
+      "learning_rate": 1.9960698133260053e-05,
+      "loss": 4.442,
+      "step": 2750
+    },
+    {
+      "epoch": 0.06082361957003993,
+      "grad_norm": 0.052490234375,
+      "learning_rate": 1.9960133833987866e-05,
+      "loss": 4.4473,
+      "step": 2760
+    },
+    {
+      "epoch": 0.06122695922236115,
+      "grad_norm": 0.07958984375,
+      "learning_rate": 1.9959565520504625e-05,
+      "loss": 4.4286,
+      "step": 2770
+    },
+    {
+      "epoch": 0.06163029887468237,
+      "grad_norm": 0.042724609375,
+      "learning_rate": 1.9958993193039365e-05,
+      "loss": 4.4616,
+      "step": 2780
+    },
+    {
+      "epoch": 0.06203363852700359,
+      "grad_norm": 0.04296875,
+      "learning_rate": 1.9958416851822755e-05,
+      "loss": 4.4409,
+      "step": 2790
+    },
+    {
+      "epoch": 0.06243697817932481,
+      "grad_norm": 0.042236328125,
+      "learning_rate": 1.9957836497087074e-05,
+      "loss": 4.4416,
+      "step": 2800
+    },
+    {
+      "epoch": 0.06284031783164602,
+      "grad_norm": 0.04541015625,
+      "learning_rate": 1.9957252129066227e-05,
+      "loss": 4.4552,
+      "step": 2810
+    },
+    {
+      "epoch": 0.06324365748396725,
+      "grad_norm": 0.046875,
+      "learning_rate": 1.9956663747995724e-05,
+      "loss": 4.4536,
+      "step": 2820
+    },
+    {
+      "epoch": 0.06364699713628846,
+      "grad_norm": 0.045654296875,
+      "learning_rate": 1.99560713541127e-05,
+      "loss": 4.4544,
+      "step": 2830
+    },
+    {
+      "epoch": 0.06405033678860969,
+      "grad_norm": 0.0419921875,
+      "learning_rate": 1.9955474947655912e-05,
+      "loss": 4.4288,
+      "step": 2840
+    },
+    {
+      "epoch": 0.0644536764409309,
+      "grad_norm": 0.041748046875,
+      "learning_rate": 1.995487452886572e-05,
+      "loss": 4.4397,
+      "step": 2850
+    },
+    {
+      "epoch": 0.06485701609325213,
+      "grad_norm": 0.044677734375,
+      "learning_rate": 1.995427009798411e-05,
+      "loss": 4.4909,
+      "step": 2860
+    },
+    {
+      "epoch": 0.06526035574557335,
+      "grad_norm": 0.043701171875,
+      "learning_rate": 1.9953661655254695e-05,
+      "loss": 4.4528,
+      "step": 2870
+    },
+    {
+      "epoch": 0.06566369539789457,
+      "grad_norm": 0.045654296875,
+      "learning_rate": 1.9953049200922684e-05,
+      "loss": 4.4308,
+      "step": 2880
+    },
+    {
+      "epoch": 0.06606703505021579,
+      "grad_norm": 0.04541015625,
+      "learning_rate": 1.9952432735234918e-05,
+      "loss": 4.4585,
+      "step": 2890
+    },
+    {
+      "epoch": 0.066470374702537,
+      "grad_norm": 0.046630859375,
+      "learning_rate": 1.9951812258439846e-05,
+      "loss": 4.4663,
+      "step": 2900
+    },
+    {
+      "epoch": 0.06687371435485823,
+      "grad_norm": 0.0439453125,
+      "learning_rate": 1.995118777078754e-05,
+      "loss": 4.4549,
+      "step": 2910
+    },
+    {
+      "epoch": 0.06727705400717944,
+      "grad_norm": 0.04931640625,
+      "learning_rate": 1.9950559272529686e-05,
+      "loss": 4.4434,
+      "step": 2920
+    },
+    {
+      "epoch": 0.06768039365950067,
+      "grad_norm": 0.04443359375,
+      "learning_rate": 1.9949926763919586e-05,
+      "loss": 4.4086,
+      "step": 2930
+    },
+    {
+      "epoch": 0.06808373331182188,
+      "grad_norm": 0.045166015625,
+      "learning_rate": 1.9949290245212157e-05,
+      "loss": 4.4456,
+      "step": 2940
+    },
+    {
+      "epoch": 0.06848707296414311,
+      "grad_norm": 0.042724609375,
+      "learning_rate": 1.9948649716663936e-05,
+      "loss": 4.4395,
+      "step": 2950
+    },
+    {
+      "epoch": 0.06889041261646432,
+      "grad_norm": 0.048095703125,
+      "learning_rate": 1.994800517853307e-05,
+      "loss": 4.4357,
+      "step": 2960
+    },
+    {
+      "epoch": 0.06929375226878555,
+      "grad_norm": 0.041015625,
+      "learning_rate": 1.9947356631079337e-05,
+      "loss": 4.4129,
+      "step": 2970
+    },
+    {
+      "epoch": 0.06969709192110676,
+      "grad_norm": 0.042724609375,
+      "learning_rate": 1.9946704074564105e-05,
+      "loss": 4.4587,
+      "step": 2980
+    },
+    {
+      "epoch": 0.07010043157342799,
+      "grad_norm": 0.04248046875,
+      "learning_rate": 1.994604750925038e-05,
+      "loss": 4.4844,
+      "step": 2990
+    },
+    {
+      "epoch": 0.0705037712257492,
+      "grad_norm": 0.044921875,
+      "learning_rate": 1.9945386935402775e-05,
+      "loss": 4.4578,
+      "step": 3000
+    },
+    {
+      "epoch": 0.07090711087807042,
+      "grad_norm": 0.04443359375,
+      "learning_rate": 1.9944722353287518e-05,
+      "loss": 4.4306,
+      "step": 3010
+    },
+    {
+      "epoch": 0.07131045053039164,
+      "grad_norm": 0.04248046875,
+      "learning_rate": 1.994405376317246e-05,
+      "loss": 4.4542,
+      "step": 3020
+    },
+    {
+      "epoch": 0.07171379018271286,
+      "grad_norm": 0.044921875,
+      "learning_rate": 1.9943381165327053e-05,
+      "loss": 4.4166,
+      "step": 3030
+    },
+    {
+      "epoch": 0.07211712983503409,
+      "grad_norm": 0.045166015625,
+      "learning_rate": 1.9942704560022378e-05,
+      "loss": 4.4745,
+      "step": 3040
+    },
+    {
+      "epoch": 0.0725204694873553,
+      "grad_norm": 0.045166015625,
+      "learning_rate": 1.9942023947531122e-05,
+      "loss": 4.4555,
+      "step": 3050
+    },
+    {
+      "epoch": 0.07292380913967653,
+      "grad_norm": 0.047119140625,
+      "learning_rate": 1.99413393281276e-05,
+      "loss": 4.4133,
+      "step": 3060
+    },
+    {
+      "epoch": 0.07332714879199774,
+      "grad_norm": 0.04638671875,
+      "learning_rate": 1.9940650702087718e-05,
+      "loss": 4.4555,
+      "step": 3070
+    },
+    {
+      "epoch": 0.07373048844431897,
+      "grad_norm": 0.044677734375,
+      "learning_rate": 1.9939958069689026e-05,
+      "loss": 4.4011,
+      "step": 3080
+    },
+    {
+      "epoch": 0.07413382809664018,
+      "grad_norm": 0.04541015625,
+      "learning_rate": 1.9939261431210664e-05,
+      "loss": 4.4595,
+      "step": 3090
+    },
+    {
+      "epoch": 0.0745371677489614,
+      "grad_norm": 0.04296875,
+      "learning_rate": 1.9938560786933398e-05,
+      "loss": 4.452,
+      "step": 3100
+    },
+    {
+      "epoch": 0.07494050740128262,
+      "grad_norm": 0.04345703125,
+      "learning_rate": 1.9937856137139612e-05,
+      "loss": 4.4497,
+      "step": 3110
+    },
+    {
+      "epoch": 0.07534384705360384,
+      "grad_norm": 0.044677734375,
+      "learning_rate": 1.9937147482113296e-05,
+      "loss": 4.4514,
+      "step": 3120
+    },
+    {
+      "epoch": 0.07574718670592506,
+      "grad_norm": 0.04541015625,
+      "learning_rate": 1.993643482214006e-05,
+      "loss": 4.4674,
+      "step": 3130
+    },
+    {
+      "epoch": 0.07615052635824628,
+      "grad_norm": 0.0419921875,
+      "learning_rate": 1.9935718157507124e-05,
+      "loss": 4.4503,
+      "step": 3140
+    },
+    {
+      "epoch": 0.0765538660105675,
+      "grad_norm": 0.044677734375,
+      "learning_rate": 1.9934997488503325e-05,
+      "loss": 4.4512,
+      "step": 3150
+    },
+    {
+      "epoch": 0.07695720566288872,
+      "grad_norm": 0.044677734375,
+      "learning_rate": 1.993427281541911e-05,
+      "loss": 4.4441,
+      "step": 3160
+    },
+    {
+      "epoch": 0.07736054531520994,
+      "grad_norm": 0.044189453125,
+      "learning_rate": 1.9933544138546542e-05,
+      "loss": 4.4542,
+      "step": 3170
+    },
+    {
+      "epoch": 0.07776388496753116,
+      "grad_norm": 0.0439453125,
+      "learning_rate": 1.9932811458179305e-05,
+      "loss": 4.4436,
+      "step": 3180
+    },
+    {
+      "epoch": 0.07816722461985237,
+      "grad_norm": 0.047607421875,
+      "learning_rate": 1.993207477461268e-05,
+      "loss": 4.4158,
+      "step": 3190
+    },
+    {
+      "epoch": 0.0785705642721736,
+      "grad_norm": 0.04443359375,
+      "learning_rate": 1.993133408814358e-05,
+      "loss": 4.4518,
+      "step": 3200
+    },
+    {
+      "epoch": 0.07897390392449481,
+      "grad_norm": 0.044921875,
+      "learning_rate": 1.9930589399070515e-05,
+      "loss": 4.4289,
+      "step": 3210
+    },
+    {
+      "epoch": 0.07937724357681604,
+      "grad_norm": 0.045654296875,
+      "learning_rate": 1.9929840707693618e-05,
+      "loss": 4.4452,
+      "step": 3220
+    },
+    {
+      "epoch": 0.07978058322913725,
+      "grad_norm": 0.042724609375,
+      "learning_rate": 1.9929088014314636e-05,
+      "loss": 4.462,
+      "step": 3230
+    },
+    {
+      "epoch": 0.08018392288145848,
+      "grad_norm": 0.04345703125,
+      "learning_rate": 1.992833131923692e-05,
+      "loss": 4.4282,
+      "step": 3240
+    },
+    {
+      "epoch": 0.0805872625337797,
+      "grad_norm": 0.0439453125,
+      "learning_rate": 1.9927570622765443e-05,
+      "loss": 4.4584,
+      "step": 3250
+    },
+    {
+      "epoch": 0.08099060218610092,
+      "grad_norm": 0.045166015625,
+      "learning_rate": 1.9926805925206784e-05,
+      "loss": 4.455,
+      "step": 3260
+    },
+    {
+      "epoch": 0.08139394183842213,
+      "grad_norm": 0.043212890625,
+      "learning_rate": 1.992603722686914e-05,
+      "loss": 4.4233,
+      "step": 3270
+    },
+    {
+      "epoch": 0.08179728149074335,
+      "grad_norm": 0.043701171875,
+      "learning_rate": 1.9925264528062317e-05,
+      "loss": 4.4435,
+      "step": 3280
+    },
+    {
+      "epoch": 0.08220062114306458,
+      "grad_norm": 0.04541015625,
+      "learning_rate": 1.9924487829097733e-05,
+      "loss": 4.4385,
+      "step": 3290
+    },
+    {
+      "epoch": 0.08260396079538579,
+      "grad_norm": 0.04248046875,
+      "learning_rate": 1.9923707130288415e-05,
+      "loss": 4.4204,
+      "step": 3300
+    },
+    {
+      "epoch": 0.08300730044770702,
+      "grad_norm": 0.043701171875,
+      "learning_rate": 1.9922922431949017e-05,
+      "loss": 4.4202,
+      "step": 3310
+    },
+    {
+      "epoch": 0.08341064010002823,
+      "grad_norm": 0.04296875,
+      "learning_rate": 1.9922133734395787e-05,
+      "loss": 4.394,
+      "step": 3320
+    },
+    {
+      "epoch": 0.08381397975234946,
+      "grad_norm": 0.04443359375,
+      "learning_rate": 1.9921341037946592e-05,
+      "loss": 4.4216,
+      "step": 3330
+    },
+    {
+      "epoch": 0.08421731940467067,
+      "grad_norm": 0.044677734375,
+      "learning_rate": 1.9920544342920913e-05,
+      "loss": 4.4231,
+      "step": 3340
+    },
+    {
+      "epoch": 0.0846206590569919,
+      "grad_norm": 0.0458984375,
+      "learning_rate": 1.991974364963984e-05,
+      "loss": 4.4847,
+      "step": 3350
+    },
+    {
+      "epoch": 0.08502399870931311,
+      "grad_norm": 0.04248046875,
+      "learning_rate": 1.9918938958426075e-05,
+      "loss": 4.4063,
+      "step": 3360
+    },
+    {
+      "epoch": 0.08542733836163434,
+      "grad_norm": 0.0458984375,
+      "learning_rate": 1.9918130269603926e-05,
+      "loss": 4.4668,
+      "step": 3370
+    },
+    {
+      "epoch": 0.08583067801395555,
+      "grad_norm": 0.0458984375,
+      "learning_rate": 1.991731758349933e-05,
+      "loss": 4.4454,
+      "step": 3380
+    },
+    {
+      "epoch": 0.08623401766627677,
+      "grad_norm": 0.04541015625,
+      "learning_rate": 1.9916500900439806e-05,
+      "loss": 4.4527,
+      "step": 3390
+    },
+    {
+      "epoch": 0.086637357318598,
+      "grad_norm": 0.042724609375,
+      "learning_rate": 1.991568022075451e-05,
+      "loss": 4.4057,
+      "step": 3400
+    },
+    {
+      "epoch": 0.0870406969709192,
+      "grad_norm": 0.044189453125,
+      "learning_rate": 1.9914855544774195e-05,
+      "loss": 4.4503,
+      "step": 3410
+    },
+    {
+      "epoch": 0.08744403662324043,
+      "grad_norm": 0.043212890625,
+      "learning_rate": 1.991402687283123e-05,
+      "loss": 4.4968,
+      "step": 3420
+    },
+    {
+      "epoch": 0.08784737627556165,
+      "grad_norm": 0.045166015625,
+      "learning_rate": 1.9913194205259595e-05,
+      "loss": 4.4642,
+      "step": 3430
+    },
+    {
+      "epoch": 0.08825071592788288,
+      "grad_norm": 0.0439453125,
+      "learning_rate": 1.9912357542394873e-05,
+      "loss": 4.4283,
+      "step": 3440
+    },
+    {
+      "epoch": 0.08865405558020409,
+      "grad_norm": 0.0458984375,
+      "learning_rate": 1.9911516884574262e-05,
+      "loss": 4.4776,
+      "step": 3450
+    },
+    {
+      "epoch": 0.08905739523252532,
+      "grad_norm": 0.045166015625,
+      "learning_rate": 1.9910672232136578e-05,
+      "loss": 4.4578,
+      "step": 3460
+    },
+    {
+      "epoch": 0.08946073488484653,
+      "grad_norm": 0.04443359375,
+      "learning_rate": 1.990982358542223e-05,
+      "loss": 4.4718,
+      "step": 3470
+    },
+    {
+      "epoch": 0.08986407453716774,
+      "grad_norm": 0.043212890625,
+      "learning_rate": 1.9908970944773255e-05,
+      "loss": 4.4575,
+      "step": 3480
+    },
+    {
+      "epoch": 0.09026741418948897,
+      "grad_norm": 0.04296875,
+      "learning_rate": 1.9908114310533285e-05,
+      "loss": 4.4147,
+      "step": 3490
+    },
+    {
+      "epoch": 0.09067075384181018,
+      "grad_norm": 0.048095703125,
+      "learning_rate": 1.990725368304757e-05,
+      "loss": 4.4363,
+      "step": 3500
+    },
+    {
+      "epoch": 0.09107409349413141,
+      "grad_norm": 0.0439453125,
+      "learning_rate": 1.990638906266297e-05,
+      "loss": 4.4575,
+      "step": 3510
+    },
+    {
+      "epoch": 0.09147743314645262,
+      "grad_norm": 0.043212890625,
+      "learning_rate": 1.990552044972794e-05,
+      "loss": 4.4545,
+      "step": 3520
+    },
+    {
+      "epoch": 0.09188077279877385,
+      "grad_norm": 0.044189453125,
+      "learning_rate": 1.9904647844592572e-05,
+      "loss": 4.4439,
+      "step": 3530
+    },
+    {
+      "epoch": 0.09228411245109507,
+      "grad_norm": 0.043701171875,
+      "learning_rate": 1.9903771247608535e-05,
+      "loss": 4.4192,
+      "step": 3540
+    },
+    {
+      "epoch": 0.09268745210341629,
+      "grad_norm": 0.044677734375,
+      "learning_rate": 1.9902890659129125e-05,
+      "loss": 4.4251,
+      "step": 3550
+    },
+    {
+      "epoch": 0.0930907917557375,
+      "grad_norm": 0.041259765625,
+      "learning_rate": 1.990200607950925e-05,
+      "loss": 4.4339,
+      "step": 3560
+    },
+    {
+      "epoch": 0.09349413140805872,
+      "grad_norm": 0.04541015625,
+      "learning_rate": 1.9901117509105417e-05,
+      "loss": 4.4635,
+      "step": 3570
+    },
+    {
+      "epoch": 0.09389747106037995,
+      "grad_norm": 0.044921875,
+      "learning_rate": 1.990022494827574e-05,
+      "loss": 4.4255,
+      "step": 3580
+    },
+    {
+      "epoch": 0.09430081071270116,
+      "grad_norm": 0.044677734375,
+      "learning_rate": 1.9899328397379955e-05,
+      "loss": 4.4412,
+      "step": 3590
+    },
+    {
+      "epoch": 0.09470415036502239,
+      "grad_norm": 0.042724609375,
+      "learning_rate": 1.989842785677939e-05,
+      "loss": 4.437,
+      "step": 3600
+    },
+    {
+      "epoch": 0.0951074900173436,
+      "grad_norm": 0.04541015625,
+      "learning_rate": 1.9897523326836987e-05,
+      "loss": 4.4136,
+      "step": 3610
+    },
+    {
+      "epoch": 0.09551082966966483,
+      "grad_norm": 0.047607421875,
+      "learning_rate": 1.98966148079173e-05,
+      "loss": 4.4353,
+      "step": 3620
+    },
+    {
+      "epoch": 0.09591416932198604,
+      "grad_norm": 0.044677734375,
+      "learning_rate": 1.989570230038649e-05,
+      "loss": 4.4331,
+      "step": 3630
+    },
+    {
+      "epoch": 0.09631750897430727,
+      "grad_norm": 0.043212890625,
+      "learning_rate": 1.989478580461232e-05,
+      "loss": 4.4311,
+      "step": 3640
+    },
+    {
+      "epoch": 0.09672084862662848,
+      "grad_norm": 0.04443359375,
+      "learning_rate": 1.9893865320964162e-05,
+      "loss": 4.3961,
+      "step": 3650
+    },
+    {
+      "epoch": 0.0971241882789497,
+      "grad_norm": 0.044921875,
+      "learning_rate": 1.9892940849812997e-05,
+      "loss": 4.4645,
+      "step": 3660
+    },
+    {
+      "epoch": 0.09752752793127092,
+      "grad_norm": 0.046142578125,
+      "learning_rate": 1.9892012391531413e-05,
+      "loss": 4.495,
+      "step": 3670
+    },
+    {
+      "epoch": 0.09793086758359214,
+      "grad_norm": 0.044677734375,
+      "learning_rate": 1.989107994649361e-05,
+      "loss": 4.4074,
+      "step": 3680
+    },
+    {
+      "epoch": 0.09833420723591337,
+      "grad_norm": 0.04638671875,
+      "learning_rate": 1.989014351507538e-05,
+      "loss": 4.441,
+      "step": 3690
+    },
+    {
+      "epoch": 0.09873754688823458,
+      "grad_norm": 0.045654296875,
+      "learning_rate": 1.988920309765413e-05,
+      "loss": 4.446,
+      "step": 3700
+    },
+    {
+      "epoch": 0.0991408865405558,
+      "grad_norm": 0.045166015625,
+      "learning_rate": 1.9888258694608886e-05,
+      "loss": 4.4378,
+      "step": 3710
+    },
+    {
+      "epoch": 0.09954422619287702,
+      "grad_norm": 0.044189453125,
+      "learning_rate": 1.988731030632026e-05,
+      "loss": 4.4068,
+      "step": 3720
+    },
+    {
+      "epoch": 0.09994756584519825,
+      "grad_norm": 0.0458984375,
+      "learning_rate": 1.988635793317048e-05,
+      "loss": 4.4587,
+      "step": 3730
+    },
+    {
+      "epoch": 0.10035090549751946,
+      "grad_norm": 0.042236328125,
+      "learning_rate": 1.9885401575543384e-05,
+      "loss": 4.4586,
+      "step": 3740
+    },
+    {
+      "epoch": 0.10075424514984069,
+      "grad_norm": 0.044677734375,
+      "learning_rate": 1.98844412338244e-05,
+      "loss": 4.4342,
+      "step": 3750
+    },
+    {
+      "epoch": 0.1011575848021619,
+      "grad_norm": 0.04638671875,
+      "learning_rate": 1.9883476908400587e-05,
+      "loss": 4.4647,
+      "step": 3760
+    },
+    {
+      "epoch": 0.10156092445448311,
+      "grad_norm": 0.04345703125,
+      "learning_rate": 1.9882508599660583e-05,
+      "loss": 4.4362,
+      "step": 3770
+    },
+    {
+      "epoch": 0.10196426410680434,
+      "grad_norm": 0.04248046875,
+      "learning_rate": 1.9881536307994645e-05,
+      "loss": 4.4402,
+      "step": 3780
+    },
+    {
+      "epoch": 0.10236760375912556,
+      "grad_norm": 0.0439453125,
+      "learning_rate": 1.9880560033794637e-05,
+      "loss": 4.4238,
+      "step": 3790
+    },
+    {
+      "epoch": 0.10277094341144678,
+      "grad_norm": 0.04443359375,
+      "learning_rate": 1.9879579777454027e-05,
+      "loss": 4.4556,
+      "step": 3800
+    },
+    {
+      "epoch": 0.103174283063768,
+      "grad_norm": 0.049560546875,
+      "learning_rate": 1.987859553936788e-05,
+      "loss": 4.4287,
+      "step": 3810
+    },
+    {
+      "epoch": 0.10357762271608922,
+      "grad_norm": 0.045654296875,
+      "learning_rate": 1.9877607319932872e-05,
+      "loss": 4.421,
+      "step": 3820
+    },
+    {
+      "epoch": 0.10398096236841044,
+      "grad_norm": 0.04638671875,
+      "learning_rate": 1.9876615119547286e-05,
+      "loss": 4.4607,
+      "step": 3830
+    },
+    {
+      "epoch": 0.10438430202073166,
+      "grad_norm": 0.04638671875,
+      "learning_rate": 1.9875618938611008e-05,
+      "loss": 4.4433,
+      "step": 3840
+    },
+    {
+      "epoch": 0.10478764167305288,
+      "grad_norm": 0.04443359375,
+      "learning_rate": 1.987461877752552e-05,
+      "loss": 4.4713,
+      "step": 3850
+    },
+    {
+      "epoch": 0.10519098132537409,
+      "grad_norm": 0.043701171875,
+      "learning_rate": 1.9873614636693918e-05,
+      "loss": 4.4631,
+      "step": 3860
+    },
+    {
+      "epoch": 0.10559432097769532,
+      "grad_norm": 0.045166015625,
+      "learning_rate": 1.9872606516520898e-05,
+      "loss": 4.3911,
+      "step": 3870
+    },
+    {
+      "epoch": 0.10599766063001653,
+      "grad_norm": 0.0458984375,
+      "learning_rate": 1.9871594417412763e-05,
+      "loss": 4.451,
+      "step": 3880
+    },
+    {
+      "epoch": 0.10640100028233776,
+      "grad_norm": 0.04541015625,
+      "learning_rate": 1.9870578339777416e-05,
+      "loss": 4.4256,
+      "step": 3890
+    },
+    {
+      "epoch": 0.10680433993465897,
+      "grad_norm": 0.04541015625,
+      "learning_rate": 1.9869558284024363e-05,
+      "loss": 4.4336,
+      "step": 3900
+    },
+    {
+      "epoch": 0.1072076795869802,
+      "grad_norm": 0.044921875,
+      "learning_rate": 1.9868534250564713e-05,
+      "loss": 4.464,
+      "step": 3910
+    },
+    {
+      "epoch": 0.10761101923930141,
+      "grad_norm": 0.042236328125,
+      "learning_rate": 1.9867506239811188e-05,
+      "loss": 4.4258,
+      "step": 3920
+    },
+    {
+      "epoch": 0.10801435889162264,
+      "grad_norm": 0.04736328125,
+      "learning_rate": 1.9866474252178096e-05,
+      "loss": 4.4037,
+      "step": 3930
+    },
+    {
+      "epoch": 0.10841769854394386,
+      "grad_norm": 0.04345703125,
+      "learning_rate": 1.9865438288081366e-05,
+      "loss": 4.383,
+      "step": 3940
+    },
+    {
+      "epoch": 0.10882103819626507,
+      "grad_norm": 0.0439453125,
+      "learning_rate": 1.986439834793851e-05,
+      "loss": 4.4667,
+      "step": 3950
+    },
+    {
+      "epoch": 0.1092243778485863,
+      "grad_norm": 0.044677734375,
+      "learning_rate": 1.986335443216866e-05,
+      "loss": 4.4176,
+      "step": 3960
+    },
+    {
+      "epoch": 0.10962771750090751,
+      "grad_norm": 0.044189453125,
+      "learning_rate": 1.9862306541192536e-05,
+      "loss": 4.4293,
+      "step": 3970
+    },
+    {
+      "epoch": 0.11003105715322874,
+      "grad_norm": 0.047119140625,
+      "learning_rate": 1.9861254675432478e-05,
+      "loss": 4.4302,
+      "step": 3980
+    },
+    {
+      "epoch": 0.11043439680554995,
+      "grad_norm": 0.04541015625,
+      "learning_rate": 1.9860198835312408e-05,
+      "loss": 4.4271,
+      "step": 3990
+    },
+    {
+      "epoch": 0.11083773645787118,
+      "grad_norm": 0.044677734375,
+      "learning_rate": 1.985913902125786e-05,
+      "loss": 4.4409,
+      "step": 4000
+    },
+    {
+      "epoch": 0.11124107611019239,
+      "grad_norm": 0.04541015625,
+      "learning_rate": 1.9858075233695974e-05,
+      "loss": 4.4272,
+      "step": 4010
+    },
+    {
+      "epoch": 0.11164441576251362,
+      "grad_norm": 0.0439453125,
+      "learning_rate": 1.9857007473055482e-05,
+      "loss": 4.426,
+      "step": 4020
+    },
+    {
+      "epoch": 0.11204775541483483,
+      "grad_norm": 0.0439453125,
+      "learning_rate": 1.9855935739766724e-05,
+      "loss": 4.4782,
+      "step": 4030
+    },
+    {
+      "epoch": 0.11245109506715605,
+      "grad_norm": 0.047119140625,
+      "learning_rate": 1.9854860034261635e-05,
+      "loss": 4.4529,
+      "step": 4040
+    },
+    {
+      "epoch": 0.11285443471947727,
+      "grad_norm": 0.044677734375,
+      "learning_rate": 1.9853780356973757e-05,
+      "loss": 4.413,
+      "step": 4050
+    },
+    {
+      "epoch": 0.11325777437179849,
+      "grad_norm": 0.045166015625,
+      "learning_rate": 1.9852696708338224e-05,
+      "loss": 4.4179,
+      "step": 4060
+    },
+    {
+      "epoch": 0.11366111402411971,
+      "grad_norm": 0.04443359375,
+      "learning_rate": 1.9851609088791783e-05,
+      "loss": 4.4679,
+      "step": 4070
+    },
+    {
+      "epoch": 0.11406445367644093,
+      "grad_norm": 0.04345703125,
+      "learning_rate": 1.9850517498772775e-05,
+      "loss": 4.4556,
+      "step": 4080
+    },
+    {
+      "epoch": 0.11446779332876215,
+      "grad_norm": 0.043701171875,
+      "learning_rate": 1.9849421938721137e-05,
+      "loss": 4.4282,
+      "step": 4090
+    },
+    {
+      "epoch": 0.11487113298108337,
+      "grad_norm": 0.0439453125,
+      "learning_rate": 1.9848322409078412e-05,
+      "loss": 4.4264,
+      "step": 4100
+    },
+    {
+      "epoch": 0.1152744726334046,
+      "grad_norm": 0.045166015625,
+      "learning_rate": 1.9847218910287743e-05,
+      "loss": 4.4565,
+      "step": 4110
+    },
+    {
+      "epoch": 0.11567781228572581,
+      "grad_norm": 0.04736328125,
+      "learning_rate": 1.9846111442793866e-05,
+      "loss": 4.4284,
+      "step": 4120
+    },
+    {
+      "epoch": 0.11608115193804702,
+      "grad_norm": 0.045166015625,
+      "learning_rate": 1.984500000704313e-05,
+      "loss": 4.4254,
+      "step": 4130
+    },
+    {
+      "epoch": 0.11648449159036825,
+      "grad_norm": 0.0458984375,
+      "learning_rate": 1.9843884603483464e-05,
+      "loss": 4.4437,
+      "step": 4140
+    },
+    {
+      "epoch": 0.11688783124268946,
+      "grad_norm": 0.0439453125,
+      "learning_rate": 1.9842765232564415e-05,
+      "loss": 4.4201,
+      "step": 4150
+    },
+    {
+      "epoch": 0.11729117089501069,
+      "grad_norm": 0.044921875,
+      "learning_rate": 1.9841641894737113e-05,
+      "loss": 4.442,
+      "step": 4160
+    },
+    {
+      "epoch": 0.1176945105473319,
+      "grad_norm": 0.044189453125,
+      "learning_rate": 1.98405145904543e-05,
+      "loss": 4.4023,
+      "step": 4170
+    },
+    {
+      "epoch": 0.11809785019965313,
+      "grad_norm": 0.044921875,
+      "learning_rate": 1.9839383320170308e-05,
+      "loss": 4.4158,
+      "step": 4180
+    },
+    {
+      "epoch": 0.11850118985197435,
+      "grad_norm": 0.0439453125,
+      "learning_rate": 1.9838248084341077e-05,
+      "loss": 4.4407,
+      "step": 4190
+    },
+    {
+      "epoch": 0.11890452950429557,
+      "grad_norm": 0.044189453125,
+      "learning_rate": 1.9837108883424128e-05,
+      "loss": 4.4053,
+      "step": 4200
+    },
+    {
+      "epoch": 0.11930786915661679,
+      "grad_norm": 0.044189453125,
+      "learning_rate": 1.98359657178786e-05,
+      "loss": 4.453,
+      "step": 4210
+    },
+    {
+      "epoch": 0.11971120880893801,
+      "grad_norm": 0.0439453125,
+      "learning_rate": 1.9834818588165216e-05,
+      "loss": 4.428,
+      "step": 4220
+    },
+    {
+      "epoch": 0.12011454846125923,
+      "grad_norm": 0.046875,
+      "learning_rate": 1.98336674947463e-05,
+      "loss": 4.4208,
+      "step": 4230
+    },
+    {
+      "epoch": 0.12051788811358044,
+      "grad_norm": 0.0439453125,
+      "learning_rate": 1.9832512438085776e-05,
+      "loss": 4.4668,
+      "step": 4240
+    },
+    {
+      "epoch": 0.12092122776590167,
+      "grad_norm": 0.04345703125,
+      "learning_rate": 1.9831353418649168e-05,
+      "loss": 4.4151,
+      "step": 4250
+    },
+    {
+      "epoch": 0.12132456741822288,
+      "grad_norm": 0.0478515625,
+      "learning_rate": 1.9830190436903587e-05,
+      "loss": 4.4326,
+      "step": 4260
+    },
+    {
+      "epoch": 0.12172790707054411,
+      "grad_norm": 0.044677734375,
+      "learning_rate": 1.982902349331775e-05,
+      "loss": 4.4254,
+      "step": 4270
+    },
+    {
+      "epoch": 0.12213124672286532,
+      "grad_norm": 0.043212890625,
+      "learning_rate": 1.9827852588361966e-05,
+      "loss": 4.46,
+      "step": 4280
+    },
+    {
+      "epoch": 0.12253458637518655,
+      "grad_norm": 0.04541015625,
+      "learning_rate": 1.982667772250815e-05,
+      "loss": 4.4317,
+      "step": 4290
+    },
+    {
+      "epoch": 0.12293792602750776,
+      "grad_norm": 0.044677734375,
+      "learning_rate": 1.9825498896229793e-05,
+      "loss": 4.4348,
+      "step": 4300
+    },
+    {
+      "epoch": 0.12334126567982899,
+      "grad_norm": 0.043701171875,
+      "learning_rate": 1.9824316110002e-05,
+      "loss": 4.4246,
+      "step": 4310
+    },
+    {
+      "epoch": 0.1237446053321502,
+      "grad_norm": 0.0439453125,
+      "learning_rate": 1.9823129364301474e-05,
+      "loss": 4.4576,
+      "step": 4320
+    },
+    {
+      "epoch": 0.12414794498447142,
+      "grad_norm": 0.04541015625,
+      "learning_rate": 1.9821938659606496e-05,
+      "loss": 4.4288,
+      "step": 4330
+    },
+    {
+      "epoch": 0.12455128463679264,
+      "grad_norm": 0.0439453125,
+      "learning_rate": 1.9820743996396957e-05,
+      "loss": 4.4153,
+      "step": 4340
+    },
+    {
+      "epoch": 0.12495462428911386,
+      "grad_norm": 0.046142578125,
+      "learning_rate": 1.981954537515434e-05,
+      "loss": 4.4412,
+      "step": 4350
+    },
+    {
+      "epoch": 0.12535796394143509,
+      "grad_norm": 0.0458984375,
+      "learning_rate": 1.9818342796361723e-05,
+      "loss": 4.4165,
+      "step": 4360
+    },
+    {
+      "epoch": 0.1257613035937563,
+      "grad_norm": 0.046630859375,
+      "learning_rate": 1.981713626050378e-05,
+      "loss": 4.4294,
+      "step": 4370
+    },
+    {
+      "epoch": 0.1261646432460775,
+      "grad_norm": 0.045166015625,
+      "learning_rate": 1.9815925768066776e-05,
+      "loss": 4.4297,
+      "step": 4380
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 49586,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 2,
+  "save_steps": 313,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.014857922095612e+19,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:851261daa18b1a1011c324f3d85b62e997899ab3bdb437a88e53633344259ebc
+size 5841