Upload folder using huggingface_hub

README.md

@@ -0,0 +1,60 @@
+---
+license: llama2
+library_name: peft
+tags:
+- llama-factory
+- lora
+- generated_from_trainer
+base_model: sudipto-ducs/InLegalLLaMA
+model-index:
+- name: sudipto-ducs/InLegalLLaMA-Instruct
+  results: []
+---
+
+<!-- This model card has been generated automatically according to the information the Trainer had access to. You
+should probably proofread and complete it, then remove this comment. -->
+
+# inlegalllama-instruct-hf
+
+This model is a fine-tuned version of [sudipto-ducs/InLegalLLaMA](https://huggingface.co/sudipto-ducs/InLegalLLaMA) on the legalkg_dataset_prompts, the legal_semantic_segmentation and the lima datasets.
+
+## Model description
+
+More information needed
+
+## Intended uses & limitations
+
+More information needed
+
+## Training and evaluation data
+
+More information needed
+
+## Training procedure
+
+### Training hyperparameters
+
+The following hyperparameters were used during training:
+- learning_rate: 3e-05
+- train_batch_size: 2
+- eval_batch_size: 8
+- seed: 42
+- gradient_accumulation_steps: 8
+- total_train_batch_size: 16
+- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
+- lr_scheduler_type: cosine
+- lr_scheduler_warmup_steps: 1000
+- num_epochs: 3.0
+- mixed_precision_training: Native AMP
+
+### Training results
+
+
+
+### Framework versions
+
+- PEFT 0.10.0
+- Transformers 4.39.0
+- Pytorch 2.2.1+cu121
+- Datasets 2.18.0
+- Tokenizers 0.15.2

all_results.json

@@ -0,0 +1,7 @@
+{
+    "epoch": 3.0,
+    "train_loss": 0.5588170812518586,
+    "train_runtime": 28902.9641,
+    "train_samples_per_second": 1.508,
+    "train_steps_per_second": 0.094
+}

config.json

@@ -0,0 +1,28 @@
+{
+  "_name_or_path": "/home/user/Desktop/Sudipto/major/models/inlegalllama-hf",
+  "architectures": [
+    "LlamaForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "bos_token_id": 1,
+  "eos_token_id": 2,
+  "hidden_act": "silu",
+  "hidden_size": 4096,
+  "initializer_range": 0.02,
+  "intermediate_size": 11008,
+  "max_position_embeddings": 4096,
+  "model_type": "llama",
+  "num_attention_heads": 32,
+  "num_hidden_layers": 32,
+  "num_key_value_heads": 32,
+  "pretraining_tp": 1,
+  "rms_norm_eps": 1e-05,
+  "rope_scaling": null,
+  "rope_theta": 10000.0,
+  "tie_word_embeddings": false,
+  "torch_dtype": "float16",
+  "transformers_version": "4.39.0",
+  "use_cache": true,
+  "vocab_size": 32000
+}

generation_config.json

@@ -0,0 +1,10 @@
+{
+  "bos_token_id": 1,
+  "do_sample": true,
+  "eos_token_id": 2,
+  "max_length": 4096,
+  "pad_token_id": 0,
+  "temperature": 0.6,
+  "top_p": 0.9,
+  "transformers_version": "4.39.0"
+}

model-00001-of-00014.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:692ad66fd25265bfdfab256e2b7b976d66e1929219176d257bb1c7a8c25f963a
+size 981485352

model-00002-of-00014.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7b4d5d03ace17f109340a83aa93b32ec5e0111755eb8ad1c2535b301656984ff
+size 966838840

model-00003-of-00014.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b7ef30e2613be3f3cf9fe3e1a80396a5d5942f0d20cc7010a0f5d6745991f1f4
+size 966822232

model-00004-of-00014.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e70b9909559cdac9f7da5fe6c8c701d6a7853dba996e72efb5c9a1d2bf78bdd2
+size 989907400

model-00005-of-00014.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a560697c29937448d9e21d982a33749ba3b187ffcc117b854317ccbc057b26b5
+size 943753688

model-00006-of-00014.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a76fdee9c74a21dd83b02bb18d468b8046ddffd6a45e2c1c124dfb4d02cffbc5
+size 989890808

model-00007-of-00014.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:adccdc5e3f967042e53bdd6df0bbc25b8c7051170593638c7962b1e4af1e15d7
+size 966838864

model-00008-of-00014.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:375e29d6cec30f7211fb2561bdc477cb35f683856f7adae9c0ea91f7b2dd5bda
+size 966822256

model-00009-of-00014.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3c71390281cba2d6ce8b33b4b16e24c59052271a0ed7b6e5ec000d3dc39c33b7
+size 989907416

model-00010-of-00014.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b9dac30afba12bfe22a77ad990afb66961b941858b14f8cb07dfe0db959f9781
+size 943753696

model-00011-of-00014.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:74aa659ffa036776f6442c231031ea4e399cce50ccc56e9710d7eb3522c214de
+size 989890808

model-00012-of-00014.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c3736e06f715d0bdb20fa6cd686ffba39b48ba106443edabd6345e0e0ba45ea0
+size 966838864

model-00013-of-00014.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fbd3fd3d97645933e24cccab595e88718c7aafd5d1dbbb6b3e74ae25a498fec3
+size 966822256

model-00014-of-00014.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c6d7eb8ea26d35e47fbc70468eb82bbf3a72296827ee79e41bc1aeae437a9b86
+size 847292056

model.safetensors.index.json

@@ -0,0 +1,298 @@
+{
+  "metadata": {
+    "total_size": 13476831232
+  },
+  "weight_map": {
+    "lm_head.weight": "model-00014-of-00014.safetensors",
+    "model.embed_tokens.weight": "model-00001-of-00014.safetensors",
+    "model.layers.0.input_layernorm.weight": "model-00001-of-00014.safetensors",
+    "model.layers.0.mlp.down_proj.weight": "model-00001-of-00014.safetensors",
+    "model.layers.0.mlp.gate_proj.weight": "model-00001-of-00014.safetensors",
+    "model.layers.0.mlp.up_proj.weight": "model-00001-of-00014.safetensors",
+    "model.layers.0.post_attention_layernorm.weight": "model-00001-of-00014.safetensors",
+    "model.layers.0.self_attn.k_proj.weight": "model-00001-of-00014.safetensors",
+    "model.layers.0.self_attn.o_proj.weight": "model-00001-of-00014.safetensors",
+    "model.layers.0.self_attn.q_proj.weight": "model-00001-of-00014.safetensors",
+    "model.layers.0.self_attn.v_proj.weight": "model-00001-of-00014.safetensors",
+    "model.layers.1.input_layernorm.weight": "model-00002-of-00014.safetensors",
+    "model.layers.1.mlp.down_proj.weight": "model-00002-of-00014.safetensors",
+    "model.layers.1.mlp.gate_proj.weight": "model-00001-of-00014.safetensors",
+    "model.layers.1.mlp.up_proj.weight": "model-00001-of-00014.safetensors",
+    "model.layers.1.post_attention_layernorm.weight": "model-00002-of-00014.safetensors",
+    "model.layers.1.self_attn.k_proj.weight": "model-00001-of-00014.safetensors",
+    "model.layers.1.self_attn.o_proj.weight": "model-00001-of-00014.safetensors",
+    "model.layers.1.self_attn.q_proj.weight": "model-00001-of-00014.safetensors",
+    "model.layers.1.self_attn.v_proj.weight": "model-00001-of-00014.safetensors",
+    "model.layers.10.input_layernorm.weight": "model-00005-of-00014.safetensors",
+    "model.layers.10.mlp.down_proj.weight": "model-00005-of-00014.safetensors",
+    "model.layers.10.mlp.gate_proj.weight": "model-00005-of-00014.safetensors",
+    "model.layers.10.mlp.up_proj.weight": "model-00005-of-00014.safetensors",
+    "model.layers.10.post_attention_layernorm.weight": "model-00005-of-00014.safetensors",
+    "model.layers.10.self_attn.k_proj.weight": "model-00005-of-00014.safetensors",
+    "model.layers.10.self_attn.o_proj.weight": "model-00005-of-00014.safetensors",
+    "model.layers.10.self_attn.q_proj.weight": "model-00005-of-00014.safetensors",
+    "model.layers.10.self_attn.v_proj.weight": "model-00005-of-00014.safetensors",
+    "model.layers.11.input_layernorm.weight": "model-00006-of-00014.safetensors",
+    "model.layers.11.mlp.down_proj.weight": "model-00006-of-00014.safetensors",
+    "model.layers.11.mlp.gate_proj.weight": "model-00006-of-00014.safetensors",
+    "model.layers.11.mlp.up_proj.weight": "model-00006-of-00014.safetensors",
+    "model.layers.11.post_attention_layernorm.weight": "model-00006-of-00014.safetensors",
+    "model.layers.11.self_attn.k_proj.weight": "model-00005-of-00014.safetensors",
+    "model.layers.11.self_attn.o_proj.weight": "model-00005-of-00014.safetensors",
+    "model.layers.11.self_attn.q_proj.weight": "model-00005-of-00014.safetensors",
+    "model.layers.11.self_attn.v_proj.weight": "model-00005-of-00014.safetensors",
+    "model.layers.12.input_layernorm.weight": "model-00006-of-00014.safetensors",
+    "model.layers.12.mlp.down_proj.weight": "model-00006-of-00014.safetensors",
+    "model.layers.12.mlp.gate_proj.weight": "model-00006-of-00014.safetensors",
+    "model.layers.12.mlp.up_proj.weight": "model-00006-of-00014.safetensors",
+    "model.layers.12.post_attention_layernorm.weight": "model-00006-of-00014.safetensors",
+    "model.layers.12.self_attn.k_proj.weight": "model-00006-of-00014.safetensors",
+    "model.layers.12.self_attn.o_proj.weight": "model-00006-of-00014.safetensors",
+    "model.layers.12.self_attn.q_proj.weight": "model-00006-of-00014.safetensors",
+    "model.layers.12.self_attn.v_proj.weight": "model-00006-of-00014.safetensors",
+    "model.layers.13.input_layernorm.weight": "model-00007-of-00014.safetensors",
+    "model.layers.13.mlp.down_proj.weight": "model-00007-of-00014.safetensors",
+    "model.layers.13.mlp.gate_proj.weight": "model-00006-of-00014.safetensors",
+    "model.layers.13.mlp.up_proj.weight": "model-00006-of-00014.safetensors",
+    "model.layers.13.post_attention_layernorm.weight": "model-00007-of-00014.safetensors",
+    "model.layers.13.self_attn.k_proj.weight": "model-00006-of-00014.safetensors",
+    "model.layers.13.self_attn.o_proj.weight": "model-00006-of-00014.safetensors",
+    "model.layers.13.self_attn.q_proj.weight": "model-00006-of-00014.safetensors",
+    "model.layers.13.self_attn.v_proj.weight": "model-00006-of-00014.safetensors",
+    "model.layers.14.input_layernorm.weight": "model-00007-of-00014.safetensors",
+    "model.layers.14.mlp.down_proj.weight": "model-00007-of-00014.safetensors",
+    "model.layers.14.mlp.gate_proj.weight": "model-00007-of-00014.safetensors",
+    "model.layers.14.mlp.up_proj.weight": "model-00007-of-00014.safetensors",
+    "model.layers.14.post_attention_layernorm.weight": "model-00007-of-00014.safetensors",
+    "model.layers.14.self_attn.k_proj.weight": "model-00007-of-00014.safetensors",
+    "model.layers.14.self_attn.o_proj.weight": "model-00007-of-00014.safetensors",
+    "model.layers.14.self_attn.q_proj.weight": "model-00007-of-00014.safetensors",
+    "model.layers.14.self_attn.v_proj.weight": "model-00007-of-00014.safetensors",
+    "model.layers.15.input_layernorm.weight": "model-00007-of-00014.safetensors",
+    "model.layers.15.mlp.down_proj.weight": "model-00007-of-00014.safetensors",
+    "model.layers.15.mlp.gate_proj.weight": "model-00007-of-00014.safetensors",
+    "model.layers.15.mlp.up_proj.weight": "model-00007-of-00014.safetensors",
+    "model.layers.15.post_attention_layernorm.weight": "model-00007-of-00014.safetensors",
+    "model.layers.15.self_attn.k_proj.weight": "model-00007-of-00014.safetensors",
+    "model.layers.15.self_attn.o_proj.weight": "model-00007-of-00014.safetensors",
+    "model.layers.15.self_attn.q_proj.weight": "model-00007-of-00014.safetensors",
+    "model.layers.15.self_attn.v_proj.weight": "model-00007-of-00014.safetensors",
+    "model.layers.16.input_layernorm.weight": "model-00008-of-00014.safetensors",
+    "model.layers.16.mlp.down_proj.weight": "model-00008-of-00014.safetensors",
+    "model.layers.16.mlp.gate_proj.weight": "model-00008-of-00014.safetensors",
+    "model.layers.16.mlp.up_proj.weight": "model-00008-of-00014.safetensors",
+    "model.layers.16.post_attention_layernorm.weight": "model-00008-of-00014.safetensors",
+    "model.layers.16.self_attn.k_proj.weight": "model-00007-of-00014.safetensors",
+    "model.layers.16.self_attn.o_proj.weight": "model-00008-of-00014.safetensors",
+    "model.layers.16.self_attn.q_proj.weight": "model-00007-of-00014.safetensors",
+    "model.layers.16.self_attn.v_proj.weight": "model-00008-of-00014.safetensors",
+    "model.layers.17.input_layernorm.weight": "model-00008-of-00014.safetensors",
+    "model.layers.17.mlp.down_proj.weight": "model-00008-of-00014.safetensors",
+    "model.layers.17.mlp.gate_proj.weight": "model-00008-of-00014.safetensors",
+    "model.layers.17.mlp.up_proj.weight": "model-00008-of-00014.safetensors",
+    "model.layers.17.post_attention_layernorm.weight": "model-00008-of-00014.safetensors",
+    "model.layers.17.self_attn.k_proj.weight": "model-00008-of-00014.safetensors",
+    "model.layers.17.self_attn.o_proj.weight": "model-00008-of-00014.safetensors",
+    "model.layers.17.self_attn.q_proj.weight": "model-00008-of-00014.safetensors",
+    "model.layers.17.self_attn.v_proj.weight": "model-00008-of-00014.safetensors",
+    "model.layers.18.input_layernorm.weight": "model-00009-of-00014.safetensors",
+    "model.layers.18.mlp.down_proj.weight": "model-00009-of-00014.safetensors",
+    "model.layers.18.mlp.gate_proj.weight": "model-00008-of-00014.safetensors",
+    "model.layers.18.mlp.up_proj.weight": "model-00009-of-00014.safetensors",
+    "model.layers.18.post_attention_layernorm.weight": "model-00009-of-00014.safetensors",
+    "model.layers.18.self_attn.k_proj.weight": "model-00008-of-00014.safetensors",
+    "model.layers.18.self_attn.o_proj.weight": "model-00008-of-00014.safetensors",
+    "model.layers.18.self_attn.q_proj.weight": "model-00008-of-00014.safetensors",
+    "model.layers.18.self_attn.v_proj.weight": "model-00008-of-00014.safetensors",
+    "model.layers.19.input_layernorm.weight": "model-00009-of-00014.safetensors",
+    "model.layers.19.mlp.down_proj.weight": "model-00009-of-00014.safetensors",
+    "model.layers.19.mlp.gate_proj.weight": "model-00009-of-00014.safetensors",
+    "model.layers.19.mlp.up_proj.weight": "model-00009-of-00014.safetensors",
+    "model.layers.19.post_attention_layernorm.weight": "model-00009-of-00014.safetensors",
+    "model.layers.19.self_attn.k_proj.weight": "model-00009-of-00014.safetensors",
+    "model.layers.19.self_attn.o_proj.weight": "model-00009-of-00014.safetensors",
+    "model.layers.19.self_attn.q_proj.weight": "model-00009-of-00014.safetensors",
+    "model.layers.19.self_attn.v_proj.weight": "model-00009-of-00014.safetensors",
+    "model.layers.2.input_layernorm.weight": "model-00002-of-00014.safetensors",
+    "model.layers.2.mlp.down_proj.weight": "model-00002-of-00014.safetensors",
+    "model.layers.2.mlp.gate_proj.weight": "model-00002-of-00014.safetensors",
+    "model.layers.2.mlp.up_proj.weight": "model-00002-of-00014.safetensors",
+    "model.layers.2.post_attention_layernorm.weight": "model-00002-of-00014.safetensors",
+    "model.layers.2.self_attn.k_proj.weight": "model-00002-of-00014.safetensors",
+    "model.layers.2.self_attn.o_proj.weight": "model-00002-of-00014.safetensors",
+    "model.layers.2.self_attn.q_proj.weight": "model-00002-of-00014.safetensors",
+    "model.layers.2.self_attn.v_proj.weight": "model-00002-of-00014.safetensors",
+    "model.layers.20.input_layernorm.weight": "model-00009-of-00014.safetensors",
+    "model.layers.20.mlp.down_proj.weight": "model-00009-of-00014.safetensors",
+    "model.layers.20.mlp.gate_proj.weight": "model-00009-of-00014.safetensors",
+    "model.layers.20.mlp.up_proj.weight": "model-00009-of-00014.safetensors",
+    "model.layers.20.post_attention_layernorm.weight": "model-00009-of-00014.safetensors",
+    "model.layers.20.self_attn.k_proj.weight": "model-00009-of-00014.safetensors",
+    "model.layers.20.self_attn.o_proj.weight": "model-00009-of-00014.safetensors",
+    "model.layers.20.self_attn.q_proj.weight": "model-00009-of-00014.safetensors",
+    "model.layers.20.self_attn.v_proj.weight": "model-00009-of-00014.safetensors",
+    "model.layers.21.input_layernorm.weight": "model-00010-of-00014.safetensors",
+    "model.layers.21.mlp.down_proj.weight": "model-00010-of-00014.safetensors",
+    "model.layers.21.mlp.gate_proj.weight": "model-00010-of-00014.safetensors",
+    "model.layers.21.mlp.up_proj.weight": "model-00010-of-00014.safetensors",
+    "model.layers.21.post_attention_layernorm.weight": "model-00010-of-00014.safetensors",
+    "model.layers.21.self_attn.k_proj.weight": "model-00010-of-00014.safetensors",
+    "model.layers.21.self_attn.o_proj.weight": "model-00010-of-00014.safetensors",
+    "model.layers.21.self_attn.q_proj.weight": "model-00010-of-00014.safetensors",
+    "model.layers.21.self_attn.v_proj.weight": "model-00010-of-00014.safetensors",
+    "model.layers.22.input_layernorm.weight": "model-00010-of-00014.safetensors",
+    "model.layers.22.mlp.down_proj.weight": "model-00010-of-00014.safetensors",
+    "model.layers.22.mlp.gate_proj.weight": "model-00010-of-00014.safetensors",
+    "model.layers.22.mlp.up_proj.weight": "model-00010-of-00014.safetensors",
+    "model.layers.22.post_attention_layernorm.weight": "model-00010-of-00014.safetensors",
+    "model.layers.22.self_attn.k_proj.weight": "model-00010-of-00014.safetensors",
+    "model.layers.22.self_attn.o_proj.weight": "model-00010-of-00014.safetensors",
+    "model.layers.22.self_attn.q_proj.weight": "model-00010-of-00014.safetensors",
+    "model.layers.22.self_attn.v_proj.weight": "model-00010-of-00014.safetensors",
+    "model.layers.23.input_layernorm.weight": "model-00011-of-00014.safetensors",
+    "model.layers.23.mlp.down_proj.weight": "model-00011-of-00014.safetensors",
+    "model.layers.23.mlp.gate_proj.weight": "model-00011-of-00014.safetensors",
+    "model.layers.23.mlp.up_proj.weight": "model-00011-of-00014.safetensors",
+    "model.layers.23.post_attention_layernorm.weight": "model-00011-of-00014.safetensors",
+    "model.layers.23.self_attn.k_proj.weight": "model-00010-of-00014.safetensors",
+    "model.layers.23.self_attn.o_proj.weight": "model-00010-of-00014.safetensors",
+    "model.layers.23.self_attn.q_proj.weight": "model-00010-of-00014.safetensors",
+    "model.layers.23.self_attn.v_proj.weight": "model-00010-of-00014.safetensors",
+    "model.layers.24.input_layernorm.weight": "model-00011-of-00014.safetensors",
+    "model.layers.24.mlp.down_proj.weight": "model-00011-of-00014.safetensors",
+    "model.layers.24.mlp.gate_proj.weight": "model-00011-of-00014.safetensors",
+    "model.layers.24.mlp.up_proj.weight": "model-00011-of-00014.safetensors",
+    "model.layers.24.post_attention_layernorm.weight": "model-00011-of-00014.safetensors",
+    "model.layers.24.self_attn.k_proj.weight": "model-00011-of-00014.safetensors",
+    "model.layers.24.self_attn.o_proj.weight": "model-00011-of-00014.safetensors",
+    "model.layers.24.self_attn.q_proj.weight": "model-00011-of-00014.safetensors",
+    "model.layers.24.self_attn.v_proj.weight": "model-00011-of-00014.safetensors",
+    "model.layers.25.input_layernorm.weight": "model-00012-of-00014.safetensors",
+    "model.layers.25.mlp.down_proj.weight": "model-00012-of-00014.safetensors",
+    "model.layers.25.mlp.gate_proj.weight": "model-00011-of-00014.safetensors",
+    "model.layers.25.mlp.up_proj.weight": "model-00011-of-00014.safetensors",
+    "model.layers.25.post_attention_layernorm.weight": "model-00012-of-00014.safetensors",
+    "model.layers.25.self_attn.k_proj.weight": "model-00011-of-00014.safetensors",
+    "model.layers.25.self_attn.o_proj.weight": "model-00011-of-00014.safetensors",
+    "model.layers.25.self_attn.q_proj.weight": "model-00011-of-00014.safetensors",
+    "model.layers.25.self_attn.v_proj.weight": "model-00011-of-00014.safetensors",
+    "model.layers.26.input_layernorm.weight": "model-00012-of-00014.safetensors",
+    "model.layers.26.mlp.down_proj.weight": "model-00012-of-00014.safetensors",
+    "model.layers.26.mlp.gate_proj.weight": "model-00012-of-00014.safetensors",
+    "model.layers.26.mlp.up_proj.weight": "model-00012-of-00014.safetensors",
+    "model.layers.26.post_attention_layernorm.weight": "model-00012-of-00014.safetensors",
+    "model.layers.26.self_attn.k_proj.weight": "model-00012-of-00014.safetensors",
+    "model.layers.26.self_attn.o_proj.weight": "model-00012-of-00014.safetensors",
+    "model.layers.26.self_attn.q_proj.weight": "model-00012-of-00014.safetensors",
+    "model.layers.26.self_attn.v_proj.weight": "model-00012-of-00014.safetensors",
+    "model.layers.27.input_layernorm.weight": "model-00012-of-00014.safetensors",
+    "model.layers.27.mlp.down_proj.weight": "model-00012-of-00014.safetensors",
+    "model.layers.27.mlp.gate_proj.weight": "model-00012-of-00014.safetensors",
+    "model.layers.27.mlp.up_proj.weight": "model-00012-of-00014.safetensors",
+    "model.layers.27.post_attention_layernorm.weight": "model-00012-of-00014.safetensors",
+    "model.layers.27.self_attn.k_proj.weight": "model-00012-of-00014.safetensors",
+    "model.layers.27.self_attn.o_proj.weight": "model-00012-of-00014.safetensors",
+    "model.layers.27.self_attn.q_proj.weight": "model-00012-of-00014.safetensors",
+    "model.layers.27.self_attn.v_proj.weight": "model-00012-of-00014.safetensors",
+    "model.layers.28.input_layernorm.weight": "model-00013-of-00014.safetensors",
+    "model.layers.28.mlp.down_proj.weight": "model-00013-of-00014.safetensors",
+    "model.layers.28.mlp.gate_proj.weight": "model-00013-of-00014.safetensors",
+    "model.layers.28.mlp.up_proj.weight": "model-00013-of-00014.safetensors",
+    "model.layers.28.post_attention_layernorm.weight": "model-00013-of-00014.safetensors",
+    "model.layers.28.self_attn.k_proj.weight": "model-00012-of-00014.safetensors",
+    "model.layers.28.self_attn.o_proj.weight": "model-00013-of-00014.safetensors",
+    "model.layers.28.self_attn.q_proj.weight": "model-00012-of-00014.safetensors",
+    "model.layers.28.self_attn.v_proj.weight": "model-00013-of-00014.safetensors",
+    "model.layers.29.input_layernorm.weight": "model-00013-of-00014.safetensors",
+    "model.layers.29.mlp.down_proj.weight": "model-00013-of-00014.safetensors",
+    "model.layers.29.mlp.gate_proj.weight": "model-00013-of-00014.safetensors",
+    "model.layers.29.mlp.up_proj.weight": "model-00013-of-00014.safetensors",
+    "model.layers.29.post_attention_layernorm.weight": "model-00013-of-00014.safetensors",
+    "model.layers.29.self_attn.k_proj.weight": "model-00013-of-00014.safetensors",
+    "model.layers.29.self_attn.o_proj.weight": "model-00013-of-00014.safetensors",
+    "model.layers.29.self_attn.q_proj.weight": "model-00013-of-00014.safetensors",
+    "model.layers.29.self_attn.v_proj.weight": "model-00013-of-00014.safetensors",
+    "model.layers.3.input_layernorm.weight": "model-00002-of-00014.safetensors",
+    "model.layers.3.mlp.down_proj.weight": "model-00002-of-00014.safetensors",
+    "model.layers.3.mlp.gate_proj.weight": "model-00002-of-00014.safetensors",
+    "model.layers.3.mlp.up_proj.weight": "model-00002-of-00014.safetensors",
+    "model.layers.3.post_attention_layernorm.weight": "model-00002-of-00014.safetensors",
+    "model.layers.3.self_attn.k_proj.weight": "model-00002-of-00014.safetensors",
+    "model.layers.3.self_attn.o_proj.weight": "model-00002-of-00014.safetensors",
+    "model.layers.3.self_attn.q_proj.weight": "model-00002-of-00014.safetensors",
+    "model.layers.3.self_attn.v_proj.weight": "model-00002-of-00014.safetensors",
+    "model.layers.30.input_layernorm.weight": "model-00014-of-00014.safetensors",
+    "model.layers.30.mlp.down_proj.weight": "model-00014-of-00014.safetensors",
+    "model.layers.30.mlp.gate_proj.weight": "model-00013-of-00014.safetensors",
+    "model.layers.30.mlp.up_proj.weight": "model-00014-of-00014.safetensors",
+    "model.layers.30.post_attention_layernorm.weight": "model-00014-of-00014.safetensors",
+    "model.layers.30.self_attn.k_proj.weight": "model-00013-of-00014.safetensors",
+    "model.layers.30.self_attn.o_proj.weight": "model-00013-of-00014.safetensors",
+    "model.layers.30.self_attn.q_proj.weight": "model-00013-of-00014.safetensors",
+    "model.layers.30.self_attn.v_proj.weight": "model-00013-of-00014.safetensors",
+    "model.layers.31.input_layernorm.weight": "model-00014-of-00014.safetensors",
+    "model.layers.31.mlp.down_proj.weight": "model-00014-of-00014.safetensors",
+    "model.layers.31.mlp.gate_proj.weight": "model-00014-of-00014.safetensors",
+    "model.layers.31.mlp.up_proj.weight": "model-00014-of-00014.safetensors",
+    "model.layers.31.post_attention_layernorm.weight": "model-00014-of-00014.safetensors",
+    "model.layers.31.self_attn.k_proj.weight": "model-00014-of-00014.safetensors",
+    "model.layers.31.self_attn.o_proj.weight": "model-00014-of-00014.safetensors",
+    "model.layers.31.self_attn.q_proj.weight": "model-00014-of-00014.safetensors",
+    "model.layers.31.self_attn.v_proj.weight": "model-00014-of-00014.safetensors",
+    "model.layers.4.input_layernorm.weight": "model-00003-of-00014.safetensors",
+    "model.layers.4.mlp.down_proj.weight": "model-00003-of-00014.safetensors",
+    "model.layers.4.mlp.gate_proj.weight": "model-00003-of-00014.safetensors",
+    "model.layers.4.mlp.up_proj.weight": "model-00003-of-00014.safetensors",
+    "model.layers.4.post_attention_layernorm.weight": "model-00003-of-00014.safetensors",
+    "model.layers.4.self_attn.k_proj.weight": "model-00002-of-00014.safetensors",
+    "model.layers.4.self_attn.o_proj.weight": "model-00003-of-00014.safetensors",
+    "model.layers.4.self_attn.q_proj.weight": "model-00002-of-00014.safetensors",
+    "model.layers.4.self_attn.v_proj.weight": "model-00003-of-00014.safetensors",
+    "model.layers.5.input_layernorm.weight": "model-00003-of-00014.safetensors",
+    "model.layers.5.mlp.down_proj.weight": "model-00003-of-00014.safetensors",
+    "model.layers.5.mlp.gate_proj.weight": "model-00003-of-00014.safetensors",
+    "model.layers.5.mlp.up_proj.weight": "model-00003-of-00014.safetensors",
+    "model.layers.5.post_attention_layernorm.weight": "model-00003-of-00014.safetensors",
+    "model.layers.5.self_attn.k_proj.weight": "model-00003-of-00014.safetensors",
+    "model.layers.5.self_attn.o_proj.weight": "model-00003-of-00014.safetensors",
+    "model.layers.5.self_attn.q_proj.weight": "model-00003-of-00014.safetensors",
+    "model.layers.5.self_attn.v_proj.weight": "model-00003-of-00014.safetensors",
+    "model.layers.6.input_layernorm.weight": "model-00004-of-00014.safetensors",
+    "model.layers.6.mlp.down_proj.weight": "model-00004-of-00014.safetensors",
+    "model.layers.6.mlp.gate_proj.weight": "model-00003-of-00014.safetensors",
+    "model.layers.6.mlp.up_proj.weight": "model-00004-of-00014.safetensors",
+    "model.layers.6.post_attention_layernorm.weight": "model-00004-of-00014.safetensors",
+    "model.layers.6.self_attn.k_proj.weight": "model-00003-of-00014.safetensors",
+    "model.layers.6.self_attn.o_proj.weight": "model-00003-of-00014.safetensors",
+    "model.layers.6.self_attn.q_proj.weight": "model-00003-of-00014.safetensors",
+    "model.layers.6.self_attn.v_proj.weight": "model-00003-of-00014.safetensors",
+    "model.layers.7.input_layernorm.weight": "model-00004-of-00014.safetensors",
+    "model.layers.7.mlp.down_proj.weight": "model-00004-of-00014.safetensors",
+    "model.layers.7.mlp.gate_proj.weight": "model-00004-of-00014.safetensors",
+    "model.layers.7.mlp.up_proj.weight": "model-00004-of-00014.safetensors",
+    "model.layers.7.post_attention_layernorm.weight": "model-00004-of-00014.safetensors",
+    "model.layers.7.self_attn.k_proj.weight": "model-00004-of-00014.safetensors",
+    "model.layers.7.self_attn.o_proj.weight": "model-00004-of-00014.safetensors",
+    "model.layers.7.self_attn.q_proj.weight": "model-00004-of-00014.safetensors",
+    "model.layers.7.self_attn.v_proj.weight": "model-00004-of-00014.safetensors",
+    "model.layers.8.input_layernorm.weight": "model-00004-of-00014.safetensors",
+    "model.layers.8.mlp.down_proj.weight": "model-00004-of-00014.safetensors",
+    "model.layers.8.mlp.gate_proj.weight": "model-00004-of-00014.safetensors",
+    "model.layers.8.mlp.up_proj.weight": "model-00004-of-00014.safetensors",
+    "model.layers.8.post_attention_layernorm.weight": "model-00004-of-00014.safetensors",
+    "model.layers.8.self_attn.k_proj.weight": "model-00004-of-00014.safetensors",
+    "model.layers.8.self_attn.o_proj.weight": "model-00004-of-00014.safetensors",
+    "model.layers.8.self_attn.q_proj.weight": "model-00004-of-00014.safetensors",
+    "model.layers.8.self_attn.v_proj.weight": "model-00004-of-00014.safetensors",
+    "model.layers.9.input_layernorm.weight": "model-00005-of-00014.safetensors",
+    "model.layers.9.mlp.down_proj.weight": "model-00005-of-00014.safetensors",
+    "model.layers.9.mlp.gate_proj.weight": "model-00005-of-00014.safetensors",
+    "model.layers.9.mlp.up_proj.weight": "model-00005-of-00014.safetensors",
+    "model.layers.9.post_attention_layernorm.weight": "model-00005-of-00014.safetensors",
+    "model.layers.9.self_attn.k_proj.weight": "model-00005-of-00014.safetensors",
+    "model.layers.9.self_attn.o_proj.weight": "model-00005-of-00014.safetensors",
+    "model.layers.9.self_attn.q_proj.weight": "model-00005-of-00014.safetensors",
+    "model.layers.9.self_attn.v_proj.weight": "model-00005-of-00014.safetensors",
+    "model.norm.weight": "model-00014-of-00014.safetensors"
+  }
+}

special_tokens_map.json

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

tokenizer.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9e556afd44213b6bd1be2b850ebbbd98f5481437a8021afaf58ee7fb1818d347
+size 499723

tokenizer_config.json

@@ -0,0 +1,45 @@
+{
+  "add_bos_token": true,
+  "add_eos_token": false,
+  "add_prefix_space": true,
+  "added_tokens_decoder": {
+    "0": {
+      "content": "<unk>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "1": {
+      "content": "<s>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "2": {
+      "content": "</s>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "bos_token": "<s>",
+  "chat_template": "{% set system_message = 'You are a helpful, respectful and honest assistant. Always answer as helpfully as possible, while being safe. Your answers should not include any harmful, unethical, racist, sexist, toxic, dangerous, or illegal content. Please ensure that your responses are socially unbiased and positive in nature.\\n\\nIf a question does not make any sense, or is not factually coherent, explain why instead of answering something not correct. If you don\\'t know the answer to a question, please don\\'t share false information.' %}{% if messages[0]['role'] == 'system' %}{% set system_message = messages[0]['content'] %}{% endif %}{% for message in messages %}{% set content = message['content'] %}{% if loop.index0 == 0 and system_message is defined %}{% set content = '<<SYS>>\\n' + system_message + '\\n<</SYS>>\\n\\n' + message['content'] %}{% endif %}{% if message['role'] == 'user' %}{{ '<s>' + '[INST] ' + content + ' [/INST]' }}{% elif message['role'] == 'assistant' %}{{ content + '</s>' }}{% endif %}{% endfor %}",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "</s>",
+  "legacy": false,
+  "model_max_length": 1000000000000000019884624838656,
+  "pad_token": "</s>",
+  "padding_side": "left",
+  "sp_model_kwargs": {},
+  "spaces_between_special_tokens": false,
+  "split_special_tokens": false,
+  "tokenizer_class": "LlamaTokenizer",
+  "unk_token": "<unk>",
+  "use_default_system_prompt": false
+}

train_results.json

@@ -0,0 +1,7 @@
+{
+    "epoch": 3.0,
+    "train_loss": 0.5588170812518586,
+    "train_runtime": 28902.9641,
+    "train_samples_per_second": 1.508,
+    "train_steps_per_second": 0.094
+}

trainer_state.json

@@ -0,0 +1,3838 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.999174236168456,
+  "eval_steps": 500,
+  "global_step": 2724,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.01,
+      "grad_norm": 4.723361968994141,
+      "learning_rate": 1.5000000000000002e-07,
+      "loss": 3.607,
+      "step": 5
+    },
+    {
+      "epoch": 0.01,
+      "grad_norm": 2.801504611968994,
+      "learning_rate": 3.0000000000000004e-07,
+      "loss": 3.6149,
+      "step": 10
+    },
+    {
+      "epoch": 0.02,
+      "grad_norm": 5.641049861907959,
+      "learning_rate": 4.5e-07,
+      "loss": 3.933,
+      "step": 15
+    },
+    {
+      "epoch": 0.02,
+      "grad_norm": 5.0696587562561035,
+      "learning_rate": 6.000000000000001e-07,
+      "loss": 3.5431,
+      "step": 20
+    },
+    {
+      "epoch": 0.03,
+      "grad_norm": 6.305172920227051,
+      "learning_rate": 7.5e-07,
+      "loss": 3.5036,
+      "step": 25
+    },
+    {
+      "epoch": 0.03,
+      "grad_norm": 5.998531818389893,
+      "learning_rate": 9e-07,
+      "loss": 3.7618,
+      "step": 30
+    },
+    {
+      "epoch": 0.04,
+      "grad_norm": 4.491389751434326,
+      "learning_rate": 1.0500000000000001e-06,
+      "loss": 3.3489,
+      "step": 35
+    },
+    {
+      "epoch": 0.04,
+      "grad_norm": 5.935448169708252,
+      "learning_rate": 1.2000000000000002e-06,
+      "loss": 3.9085,
+      "step": 40
+    },
+    {
+      "epoch": 0.05,
+      "grad_norm": 5.686679363250732,
+      "learning_rate": 1.35e-06,
+      "loss": 3.6536,
+      "step": 45
+    },
+    {
+      "epoch": 0.06,
+      "grad_norm": 4.586507797241211,
+      "learning_rate": 1.5e-06,
+      "loss": 3.4313,
+      "step": 50
+    },
+    {
+      "epoch": 0.06,
+      "grad_norm": 4.8034772872924805,
+      "learning_rate": 1.65e-06,
+      "loss": 3.3082,
+      "step": 55
+    },
+    {
+      "epoch": 0.07,
+      "grad_norm": 5.47461462020874,
+      "learning_rate": 1.8e-06,
+      "loss": 3.5749,
+      "step": 60
+    },
+    {
+      "epoch": 0.07,
+      "grad_norm": 6.304111957550049,
+      "learning_rate": 1.95e-06,
+      "loss": 3.5684,
+      "step": 65
+    },
+    {
+      "epoch": 0.08,
+      "grad_norm": 6.649425983428955,
+      "learning_rate": 2.1000000000000002e-06,
+      "loss": 3.25,
+      "step": 70
+    },
+    {
+      "epoch": 0.08,
+      "grad_norm": 6.285486698150635,
+      "learning_rate": 2.25e-06,
+      "loss": 3.0361,
+      "step": 75
+    },
+    {
+      "epoch": 0.09,
+      "grad_norm": 8.364280700683594,
+      "learning_rate": 2.4000000000000003e-06,
+      "loss": 3.4918,
+      "step": 80
+    },
+    {
+      "epoch": 0.09,
+      "grad_norm": 8.540417671203613,
+      "learning_rate": 2.55e-06,
+      "loss": 3.2424,
+      "step": 85
+    },
+    {
+      "epoch": 0.1,
+      "grad_norm": 3.8771581649780273,
+      "learning_rate": 2.7e-06,
+      "loss": 2.7386,
+      "step": 90
+    },
+    {
+      "epoch": 0.1,
+      "grad_norm": 6.713310241699219,
+      "learning_rate": 2.8500000000000002e-06,
+      "loss": 2.5229,
+      "step": 95
+    },
+    {
+      "epoch": 0.11,
+      "grad_norm": 7.163495063781738,
+      "learning_rate": 3e-06,
+      "loss": 2.3543,
+      "step": 100
+    },
+    {
+      "epoch": 0.12,
+      "grad_norm": 6.106879234313965,
+      "learning_rate": 3.15e-06,
+      "loss": 2.1774,
+      "step": 105
+    },
+    {
+      "epoch": 0.12,
+      "grad_norm": 6.056279182434082,
+      "learning_rate": 3.3e-06,
+      "loss": 2.102,
+      "step": 110
+    },
+    {
+      "epoch": 0.13,
+      "grad_norm": 6.470534324645996,
+      "learning_rate": 3.4500000000000004e-06,
+      "loss": 1.8336,
+      "step": 115
+    },
+    {
+      "epoch": 0.13,
+      "grad_norm": 5.0930376052856445,
+      "learning_rate": 3.6e-06,
+      "loss": 1.4314,
+      "step": 120
+    },
+    {
+      "epoch": 0.14,
+      "grad_norm": 4.7424397468566895,
+      "learning_rate": 3.75e-06,
+      "loss": 1.3507,
+      "step": 125
+    },
+    {
+      "epoch": 0.14,
+      "grad_norm": 4.056792736053467,
+      "learning_rate": 3.9e-06,
+      "loss": 1.2108,
+      "step": 130
+    },
+    {
+      "epoch": 0.15,
+      "grad_norm": 3.122858762741089,
+      "learning_rate": 4.05e-06,
+      "loss": 0.7046,
+      "step": 135
+    },
+    {
+      "epoch": 0.15,
+      "grad_norm": 2.3814079761505127,
+      "learning_rate": 4.2000000000000004e-06,
+      "loss": 0.7918,
+      "step": 140
+    },
+    {
+      "epoch": 0.16,
+      "grad_norm": 3.257763385772705,
+      "learning_rate": 4.35e-06,
+      "loss": 0.8256,
+      "step": 145
+    },
+    {
+      "epoch": 0.17,
+      "grad_norm": 2.871866464614868,
+      "learning_rate": 4.5e-06,
+      "loss": 0.6759,
+      "step": 150
+    },
+    {
+      "epoch": 0.17,
+      "grad_norm": 3.0206196308135986,
+      "learning_rate": 4.65e-06,
+      "loss": 0.7936,
+      "step": 155
+    },
+    {
+      "epoch": 0.18,
+      "grad_norm": 2.52962589263916,
+      "learning_rate": 4.800000000000001e-06,
+      "loss": 0.5798,
+      "step": 160
+    },
+    {
+      "epoch": 0.18,
+      "grad_norm": 1.2966934442520142,
+      "learning_rate": 4.95e-06,
+      "loss": 0.6066,
+      "step": 165
+    },
+    {
+      "epoch": 0.19,
+      "grad_norm": 1.9257962703704834,
+      "learning_rate": 5.1e-06,
+      "loss": 0.6452,
+      "step": 170
+    },
+    {
+      "epoch": 0.19,
+      "grad_norm": 2.903165340423584,
+      "learning_rate": 5.25e-06,
+      "loss": 0.6818,
+      "step": 175
+    },
+    {
+      "epoch": 0.2,
+      "grad_norm": 2.3169002532958984,
+      "learning_rate": 5.4e-06,
+      "loss": 0.7669,
+      "step": 180
+    },
+    {
+      "epoch": 0.2,
+      "grad_norm": 2.465702533721924,
+      "learning_rate": 5.55e-06,
+      "loss": 0.5847,
+      "step": 185
+    },
+    {
+      "epoch": 0.21,
+      "grad_norm": 2.0977537631988525,
+      "learning_rate": 5.7000000000000005e-06,
+      "loss": 0.6548,
+      "step": 190
+    },
+    {
+      "epoch": 0.21,
+      "grad_norm": 2.9947569370269775,
+      "learning_rate": 5.850000000000001e-06,
+      "loss": 0.6992,
+      "step": 195
+    },
+    {
+      "epoch": 0.22,
+      "grad_norm": 2.6077871322631836,
+      "learning_rate": 6e-06,
+      "loss": 0.6392,
+      "step": 200
+    },
+    {
+      "epoch": 0.23,
+      "grad_norm": 2.76350736618042,
+      "learning_rate": 6.1499999999999996e-06,
+      "loss": 0.5201,
+      "step": 205
+    },
+    {
+      "epoch": 0.23,
+      "grad_norm": 1.9421017169952393,
+      "learning_rate": 6.3e-06,
+      "loss": 0.5464,
+      "step": 210
+    },
+    {
+      "epoch": 0.24,
+      "grad_norm": 1.9033386707305908,
+      "learning_rate": 6.45e-06,
+      "loss": 0.567,
+      "step": 215
+    },
+    {
+      "epoch": 0.24,
+      "grad_norm": 2.0922622680664062,
+      "learning_rate": 6.6e-06,
+      "loss": 0.6302,
+      "step": 220
+    },
+    {
+      "epoch": 0.25,
+      "grad_norm": 2.7234456539154053,
+      "learning_rate": 6.750000000000001e-06,
+      "loss": 0.8481,
+      "step": 225
+    },
+    {
+      "epoch": 0.25,
+      "grad_norm": 2.4856104850769043,
+      "learning_rate": 6.900000000000001e-06,
+      "loss": 0.5044,
+      "step": 230
+    },
+    {
+      "epoch": 0.26,
+      "grad_norm": 2.776841163635254,
+      "learning_rate": 7.049999999999999e-06,
+      "loss": 0.4163,
+      "step": 235
+    },
+    {
+      "epoch": 0.26,
+      "grad_norm": 1.360962986946106,
+      "learning_rate": 7.2e-06,
+      "loss": 0.5952,
+      "step": 240
+    },
+    {
+      "epoch": 0.27,
+      "grad_norm": 1.9951967000961304,
+      "learning_rate": 7.35e-06,
+      "loss": 0.4798,
+      "step": 245
+    },
+    {
+      "epoch": 0.28,
+      "grad_norm": 1.2397233247756958,
+      "learning_rate": 7.5e-06,
+      "loss": 0.3873,
+      "step": 250
+    },
+    {
+      "epoch": 0.28,
+      "grad_norm": 1.793684720993042,
+      "learning_rate": 7.65e-06,
+      "loss": 0.5231,
+      "step": 255
+    },
+    {
+      "epoch": 0.29,
+      "grad_norm": 4.247454643249512,
+      "learning_rate": 7.8e-06,
+      "loss": 0.4825,
+      "step": 260
+    },
+    {
+      "epoch": 0.29,
+      "grad_norm": 2.6619770526885986,
+      "learning_rate": 7.95e-06,
+      "loss": 0.4743,
+      "step": 265
+    },
+    {
+      "epoch": 0.3,
+      "grad_norm": 2.1707732677459717,
+      "learning_rate": 8.1e-06,
+      "loss": 0.3325,
+      "step": 270
+    },
+    {
+      "epoch": 0.3,
+      "grad_norm": 1.470760464668274,
+      "learning_rate": 8.25e-06,
+      "loss": 0.6128,
+      "step": 275
+    },
+    {
+      "epoch": 0.31,
+      "grad_norm": 2.647947072982788,
+      "learning_rate": 8.400000000000001e-06,
+      "loss": 0.6124,
+      "step": 280
+    },
+    {
+      "epoch": 0.31,
+      "grad_norm": 5.327182769775391,
+      "learning_rate": 8.55e-06,
+      "loss": 0.5567,
+      "step": 285
+    },
+    {
+      "epoch": 0.32,
+      "grad_norm": 2.314610719680786,
+      "learning_rate": 8.7e-06,
+      "loss": 0.4464,
+      "step": 290
+    },
+    {
+      "epoch": 0.32,
+      "grad_norm": 2.571409225463867,
+      "learning_rate": 8.85e-06,
+      "loss": 0.5127,
+      "step": 295
+    },
+    {
+      "epoch": 0.33,
+      "grad_norm": 2.6803112030029297,
+      "learning_rate": 9e-06,
+      "loss": 0.4206,
+      "step": 300
+    },
+    {
+      "epoch": 0.34,
+      "grad_norm": 1.8423502445220947,
+      "learning_rate": 9.15e-06,
+      "loss": 0.4846,
+      "step": 305
+    },
+    {
+      "epoch": 0.34,
+      "grad_norm": 1.603908658027649,
+      "learning_rate": 9.3e-06,
+      "loss": 0.5434,
+      "step": 310
+    },
+    {
+      "epoch": 0.35,
+      "grad_norm": 2.6442675590515137,
+      "learning_rate": 9.450000000000001e-06,
+      "loss": 0.56,
+      "step": 315
+    },
+    {
+      "epoch": 0.35,
+      "grad_norm": 4.274781703948975,
+      "learning_rate": 9.600000000000001e-06,
+      "loss": 0.4854,
+      "step": 320
+    },
+    {
+      "epoch": 0.36,
+      "grad_norm": 3.596944570541382,
+      "learning_rate": 9.75e-06,
+      "loss": 0.6743,
+      "step": 325
+    },
+    {
+      "epoch": 0.36,
+      "grad_norm": 2.466118574142456,
+      "learning_rate": 9.9e-06,
+      "loss": 0.5149,
+      "step": 330
+    },
+    {
+      "epoch": 0.37,
+      "grad_norm": 1.8980193138122559,
+      "learning_rate": 1.005e-05,
+      "loss": 0.5721,
+      "step": 335
+    },
+    {
+      "epoch": 0.37,
+      "grad_norm": 1.9017683267593384,
+      "learning_rate": 1.02e-05,
+      "loss": 0.4637,
+      "step": 340
+    },
+    {
+      "epoch": 0.38,
+      "grad_norm": 2.220961809158325,
+      "learning_rate": 1.035e-05,
+      "loss": 0.4672,
+      "step": 345
+    },
+    {
+      "epoch": 0.39,
+      "grad_norm": 2.758296489715576,
+      "learning_rate": 1.05e-05,
+      "loss": 0.4269,
+      "step": 350
+    },
+    {
+      "epoch": 0.39,
+      "grad_norm": 1.6211260557174683,
+      "learning_rate": 1.065e-05,
+      "loss": 0.4208,
+      "step": 355
+    },
+    {
+      "epoch": 0.4,
+      "grad_norm": 1.5319430828094482,
+      "learning_rate": 1.08e-05,
+      "loss": 0.5012,
+      "step": 360
+    },
+    {
+      "epoch": 0.4,
+      "grad_norm": 2.5233569145202637,
+      "learning_rate": 1.095e-05,
+      "loss": 0.2889,
+      "step": 365
+    },
+    {
+      "epoch": 0.41,
+      "grad_norm": 1.5497692823410034,
+      "learning_rate": 1.11e-05,
+      "loss": 0.3052,
+      "step": 370
+    },
+    {
+      "epoch": 0.41,
+      "grad_norm": 3.8792238235473633,
+      "learning_rate": 1.125e-05,
+      "loss": 0.6235,
+      "step": 375
+    },
+    {
+      "epoch": 0.42,
+      "grad_norm": 2.3589935302734375,
+      "learning_rate": 1.1400000000000001e-05,
+      "loss": 0.5508,
+      "step": 380
+    },
+    {
+      "epoch": 0.42,
+      "grad_norm": 2.8691861629486084,
+      "learning_rate": 1.1550000000000001e-05,
+      "loss": 0.4495,
+      "step": 385
+    },
+    {
+      "epoch": 0.43,
+      "grad_norm": 2.3704771995544434,
+      "learning_rate": 1.1700000000000001e-05,
+      "loss": 0.4749,
+      "step": 390
+    },
+    {
+      "epoch": 0.43,
+      "grad_norm": 2.102457046508789,
+      "learning_rate": 1.185e-05,
+      "loss": 0.4324,
+      "step": 395
+    },
+    {
+      "epoch": 0.44,
+      "grad_norm": 6.848997116088867,
+      "learning_rate": 1.2e-05,
+      "loss": 0.4796,
+      "step": 400
+    },
+    {
+      "epoch": 0.45,
+      "grad_norm": 1.6692250967025757,
+      "learning_rate": 1.215e-05,
+      "loss": 0.3403,
+      "step": 405
+    },
+    {
+      "epoch": 0.45,
+      "grad_norm": 2.407634973526001,
+      "learning_rate": 1.2299999999999999e-05,
+      "loss": 0.4553,
+      "step": 410
+    },
+    {
+      "epoch": 0.46,
+      "grad_norm": 2.6080591678619385,
+      "learning_rate": 1.245e-05,
+      "loss": 0.3708,
+      "step": 415
+    },
+    {
+      "epoch": 0.46,
+      "grad_norm": 2.4701273441314697,
+      "learning_rate": 1.26e-05,
+      "loss": 0.3112,
+      "step": 420
+    },
+    {
+      "epoch": 0.47,
+      "grad_norm": 2.7998530864715576,
+      "learning_rate": 1.275e-05,
+      "loss": 0.3935,
+      "step": 425
+    },
+    {
+      "epoch": 0.47,
+      "grad_norm": 2.0203258991241455,
+      "learning_rate": 1.29e-05,
+      "loss": 0.3994,
+      "step": 430
+    },
+    {
+      "epoch": 0.48,
+      "grad_norm": 2.6193652153015137,
+      "learning_rate": 1.305e-05,
+      "loss": 0.5095,
+      "step": 435
+    },
+    {
+      "epoch": 0.48,
+      "grad_norm": 1.9212027788162231,
+      "learning_rate": 1.32e-05,
+      "loss": 0.3615,
+      "step": 440
+    },
+    {
+      "epoch": 0.49,
+      "grad_norm": 3.352229595184326,
+      "learning_rate": 1.3350000000000001e-05,
+      "loss": 0.6153,
+      "step": 445
+    },
+    {
+      "epoch": 0.5,
+      "grad_norm": 1.7962478399276733,
+      "learning_rate": 1.3500000000000001e-05,
+      "loss": 0.6112,
+      "step": 450
+    },
+    {
+      "epoch": 0.5,
+      "grad_norm": 4.245258331298828,
+      "learning_rate": 1.3650000000000001e-05,
+      "loss": 0.5036,
+      "step": 455
+    },
+    {
+      "epoch": 0.51,
+      "grad_norm": 1.9633055925369263,
+      "learning_rate": 1.3800000000000002e-05,
+      "loss": 0.5212,
+      "step": 460
+    },
+    {
+      "epoch": 0.51,
+      "grad_norm": 2.65956711769104,
+      "learning_rate": 1.395e-05,
+      "loss": 0.4709,
+      "step": 465
+    },
+    {
+      "epoch": 0.52,
+      "grad_norm": 1.38247811794281,
+      "learning_rate": 1.4099999999999999e-05,
+      "loss": 0.7051,
+      "step": 470
+    },
+    {
+      "epoch": 0.52,
+      "grad_norm": 1.2942965030670166,
+      "learning_rate": 1.4249999999999999e-05,
+      "loss": 0.5219,
+      "step": 475
+    },
+    {
+      "epoch": 0.53,
+      "grad_norm": 1.6196554899215698,
+      "learning_rate": 1.44e-05,
+      "loss": 0.4838,
+      "step": 480
+    },
+    {
+      "epoch": 0.53,
+      "grad_norm": 2.9407804012298584,
+      "learning_rate": 1.455e-05,
+      "loss": 0.4163,
+      "step": 485
+    },
+    {
+      "epoch": 0.54,
+      "grad_norm": 2.036487340927124,
+      "learning_rate": 1.47e-05,
+      "loss": 0.3517,
+      "step": 490
+    },
+    {
+      "epoch": 0.55,
+      "grad_norm": 4.017712116241455,
+      "learning_rate": 1.485e-05,
+      "loss": 0.3407,
+      "step": 495
+    },
+    {
+      "epoch": 0.55,
+      "grad_norm": 1.1127279996871948,
+      "learning_rate": 1.5e-05,
+      "loss": 0.5691,
+      "step": 500
+    },
+    {
+      "epoch": 0.56,
+      "grad_norm": 2.890312910079956,
+      "learning_rate": 1.515e-05,
+      "loss": 0.3602,
+      "step": 505
+    },
+    {
+      "epoch": 0.56,
+      "grad_norm": 2.6266424655914307,
+      "learning_rate": 1.53e-05,
+      "loss": 0.2707,
+      "step": 510
+    },
+    {
+      "epoch": 0.57,
+      "grad_norm": 1.9139853715896606,
+      "learning_rate": 1.545e-05,
+      "loss": 0.4348,
+      "step": 515
+    },
+    {
+      "epoch": 0.57,
+      "grad_norm": 1.4371010065078735,
+      "learning_rate": 1.56e-05,
+      "loss": 0.4869,
+      "step": 520
+    },
+    {
+      "epoch": 0.58,
+      "grad_norm": 2.060549020767212,
+      "learning_rate": 1.575e-05,
+      "loss": 0.446,
+      "step": 525
+    },
+    {
+      "epoch": 0.58,
+      "grad_norm": 2.383223295211792,
+      "learning_rate": 1.59e-05,
+      "loss": 0.4541,
+      "step": 530
+    },
+    {
+      "epoch": 0.59,
+      "grad_norm": 3.066714286804199,
+      "learning_rate": 1.605e-05,
+      "loss": 0.654,
+      "step": 535
+    },
+    {
+      "epoch": 0.59,
+      "grad_norm": 1.502321481704712,
+      "learning_rate": 1.62e-05,
+      "loss": 0.6171,
+      "step": 540
+    },
+    {
+      "epoch": 0.6,
+      "grad_norm": 3.31970477104187,
+      "learning_rate": 1.635e-05,
+      "loss": 0.4521,
+      "step": 545
+    },
+    {
+      "epoch": 0.61,
+      "grad_norm": 2.2695887088775635,
+      "learning_rate": 1.65e-05,
+      "loss": 0.5606,
+      "step": 550
+    },
+    {
+      "epoch": 0.61,
+      "grad_norm": 1.9954079389572144,
+      "learning_rate": 1.665e-05,
+      "loss": 0.4517,
+      "step": 555
+    },
+    {
+      "epoch": 0.62,
+      "grad_norm": 3.29400372505188,
+      "learning_rate": 1.6800000000000002e-05,
+      "loss": 0.6205,
+      "step": 560
+    },
+    {
+      "epoch": 0.62,
+      "grad_norm": 2.401451587677002,
+      "learning_rate": 1.695e-05,
+      "loss": 0.5424,
+      "step": 565
+    },
+    {
+      "epoch": 0.63,
+      "grad_norm": 3.064767837524414,
+      "learning_rate": 1.71e-05,
+      "loss": 0.5388,
+      "step": 570
+    },
+    {
+      "epoch": 0.63,
+      "grad_norm": 2.0694704055786133,
+      "learning_rate": 1.725e-05,
+      "loss": 0.3733,
+      "step": 575
+    },
+    {
+      "epoch": 0.64,
+      "grad_norm": 1.1812100410461426,
+      "learning_rate": 1.74e-05,
+      "loss": 0.4173,
+      "step": 580
+    },
+    {
+      "epoch": 0.64,
+      "grad_norm": 1.7811343669891357,
+      "learning_rate": 1.755e-05,
+      "loss": 0.7022,
+      "step": 585
+    },
+    {
+      "epoch": 0.65,
+      "grad_norm": 1.5441101789474487,
+      "learning_rate": 1.77e-05,
+      "loss": 0.6539,
+      "step": 590
+    },
+    {
+      "epoch": 0.66,
+      "grad_norm": 2.039987087249756,
+      "learning_rate": 1.785e-05,
+      "loss": 0.3081,
+      "step": 595
+    },
+    {
+      "epoch": 0.66,
+      "grad_norm": 2.5088701248168945,
+      "learning_rate": 1.8e-05,
+      "loss": 0.3314,
+      "step": 600
+    },
+    {
+      "epoch": 0.67,
+      "grad_norm": 2.6933419704437256,
+      "learning_rate": 1.815e-05,
+      "loss": 0.5642,
+      "step": 605
+    },
+    {
+      "epoch": 0.67,
+      "grad_norm": 1.9799847602844238,
+      "learning_rate": 1.83e-05,
+      "loss": 0.3405,
+      "step": 610
+    },
+    {
+      "epoch": 0.68,
+      "grad_norm": 2.641113758087158,
+      "learning_rate": 1.845e-05,
+      "loss": 0.4678,
+      "step": 615
+    },
+    {
+      "epoch": 0.68,
+      "grad_norm": 1.5051275491714478,
+      "learning_rate": 1.86e-05,
+      "loss": 0.5578,
+      "step": 620
+    },
+    {
+      "epoch": 0.69,
+      "grad_norm": 1.5316920280456543,
+      "learning_rate": 1.8750000000000002e-05,
+      "loss": 0.5345,
+      "step": 625
+    },
+    {
+      "epoch": 0.69,
+      "grad_norm": 0.8923272490501404,
+      "learning_rate": 1.8900000000000002e-05,
+      "loss": 0.4797,
+      "step": 630
+    },
+    {
+      "epoch": 0.7,
+      "grad_norm": 1.6083885431289673,
+      "learning_rate": 1.9050000000000002e-05,
+      "loss": 0.4117,
+      "step": 635
+    },
+    {
+      "epoch": 0.7,
+      "grad_norm": 1.3130501508712769,
+      "learning_rate": 1.9200000000000003e-05,
+      "loss": 0.4314,
+      "step": 640
+    },
+    {
+      "epoch": 0.71,
+      "grad_norm": 1.2643862962722778,
+      "learning_rate": 1.935e-05,
+      "loss": 0.3503,
+      "step": 645
+    },
+    {
+      "epoch": 0.72,
+      "grad_norm": 2.3432857990264893,
+      "learning_rate": 1.95e-05,
+      "loss": 0.4664,
+      "step": 650
+    },
+    {
+      "epoch": 0.72,
+      "grad_norm": 1.6133382320404053,
+      "learning_rate": 1.965e-05,
+      "loss": 0.5415,
+      "step": 655
+    },
+    {
+      "epoch": 0.73,
+      "grad_norm": 2.1628336906433105,
+      "learning_rate": 1.98e-05,
+      "loss": 0.3583,
+      "step": 660
+    },
+    {
+      "epoch": 0.73,
+      "grad_norm": 1.1896861791610718,
+      "learning_rate": 1.995e-05,
+      "loss": 0.3906,
+      "step": 665
+    },
+    {
+      "epoch": 0.74,
+      "grad_norm": 1.399902582168579,
+      "learning_rate": 2.01e-05,
+      "loss": 0.5711,
+      "step": 670
+    },
+    {
+      "epoch": 0.74,
+      "grad_norm": 1.9324532747268677,
+      "learning_rate": 2.025e-05,
+      "loss": 0.447,
+      "step": 675
+    },
+    {
+      "epoch": 0.75,
+      "grad_norm": 1.4938113689422607,
+      "learning_rate": 2.04e-05,
+      "loss": 0.6049,
+      "step": 680
+    },
+    {
+      "epoch": 0.75,
+      "grad_norm": 0.6185623407363892,
+      "learning_rate": 2.055e-05,
+      "loss": 0.44,
+      "step": 685
+    },
+    {
+      "epoch": 0.76,
+      "grad_norm": 1.0526788234710693,
+      "learning_rate": 2.07e-05,
+      "loss": 0.5259,
+      "step": 690
+    },
+    {
+      "epoch": 0.77,
+      "grad_norm": 1.5097798109054565,
+      "learning_rate": 2.085e-05,
+      "loss": 0.4513,
+      "step": 695
+    },
+    {
+      "epoch": 0.77,
+      "grad_norm": 1.4816968441009521,
+      "learning_rate": 2.1e-05,
+      "loss": 0.6861,
+      "step": 700
+    },
+    {
+      "epoch": 0.78,
+      "grad_norm": 2.5907509326934814,
+      "learning_rate": 2.115e-05,
+      "loss": 0.5984,
+      "step": 705
+    },
+    {
+      "epoch": 0.78,
+      "grad_norm": 1.7099113464355469,
+      "learning_rate": 2.13e-05,
+      "loss": 0.2523,
+      "step": 710
+    },
+    {
+      "epoch": 0.79,
+      "grad_norm": 1.2753609418869019,
+      "learning_rate": 2.145e-05,
+      "loss": 0.5409,
+      "step": 715
+    },
+    {
+      "epoch": 0.79,
+      "grad_norm": 0.9064136147499084,
+      "learning_rate": 2.16e-05,
+      "loss": 0.3004,
+      "step": 720
+    },
+    {
+      "epoch": 0.8,
+      "grad_norm": 1.4393867254257202,
+      "learning_rate": 2.175e-05,
+      "loss": 0.4156,
+      "step": 725
+    },
+    {
+      "epoch": 0.8,
+      "grad_norm": 1.8317877054214478,
+      "learning_rate": 2.19e-05,
+      "loss": 0.5019,
+      "step": 730
+    },
+    {
+      "epoch": 0.81,
+      "grad_norm": 1.2695151567459106,
+      "learning_rate": 2.205e-05,
+      "loss": 0.6344,
+      "step": 735
+    },
+    {
+      "epoch": 0.81,
+      "grad_norm": 0.9795133471488953,
+      "learning_rate": 2.22e-05,
+      "loss": 0.4634,
+      "step": 740
+    },
+    {
+      "epoch": 0.82,
+      "grad_norm": 2.9170186519622803,
+      "learning_rate": 2.235e-05,
+      "loss": 0.5323,
+      "step": 745
+    },
+    {
+      "epoch": 0.83,
+      "grad_norm": 1.4109234809875488,
+      "learning_rate": 2.25e-05,
+      "loss": 0.4622,
+      "step": 750
+    },
+    {
+      "epoch": 0.83,
+      "grad_norm": 1.0275310277938843,
+      "learning_rate": 2.265e-05,
+      "loss": 0.4223,
+      "step": 755
+    },
+    {
+      "epoch": 0.84,
+      "grad_norm": 2.929098129272461,
+      "learning_rate": 2.2800000000000002e-05,
+      "loss": 0.4939,
+      "step": 760
+    },
+    {
+      "epoch": 0.84,
+      "grad_norm": 1.730406641960144,
+      "learning_rate": 2.2950000000000002e-05,
+      "loss": 0.6818,
+      "step": 765
+    },
+    {
+      "epoch": 0.85,
+      "grad_norm": 0.8586449027061462,
+      "learning_rate": 2.3100000000000002e-05,
+      "loss": 0.5406,
+      "step": 770
+    },
+    {
+      "epoch": 0.85,
+      "grad_norm": 1.2498224973678589,
+      "learning_rate": 2.3250000000000003e-05,
+      "loss": 0.3736,
+      "step": 775
+    },
+    {
+      "epoch": 0.86,
+      "grad_norm": 0.6030346751213074,
+      "learning_rate": 2.3400000000000003e-05,
+      "loss": 0.5458,
+      "step": 780
+    },
+    {
+      "epoch": 0.86,
+      "grad_norm": 1.4750096797943115,
+      "learning_rate": 2.3550000000000003e-05,
+      "loss": 0.5234,
+      "step": 785
+    },
+    {
+      "epoch": 0.87,
+      "grad_norm": 1.3055192232131958,
+      "learning_rate": 2.37e-05,
+      "loss": 0.5177,
+      "step": 790
+    },
+    {
+      "epoch": 0.88,
+      "grad_norm": 1.990779995918274,
+      "learning_rate": 2.385e-05,
+      "loss": 0.4851,
+      "step": 795
+    },
+    {
+      "epoch": 0.88,
+      "grad_norm": 2.6766905784606934,
+      "learning_rate": 2.4e-05,
+      "loss": 0.5505,
+      "step": 800
+    },
+    {
+      "epoch": 0.89,
+      "grad_norm": 2.486478090286255,
+      "learning_rate": 2.415e-05,
+      "loss": 0.4049,
+      "step": 805
+    },
+    {
+      "epoch": 0.89,
+      "grad_norm": 1.191136360168457,
+      "learning_rate": 2.43e-05,
+      "loss": 0.3537,
+      "step": 810
+    },
+    {
+      "epoch": 0.9,
+      "grad_norm": 1.3190600872039795,
+      "learning_rate": 2.4449999999999998e-05,
+      "loss": 0.5542,
+      "step": 815
+    },
+    {
+      "epoch": 0.9,
+      "grad_norm": 1.4110169410705566,
+      "learning_rate": 2.4599999999999998e-05,
+      "loss": 0.4228,
+      "step": 820
+    },
+    {
+      "epoch": 0.91,
+      "grad_norm": 2.436192512512207,
+      "learning_rate": 2.475e-05,
+      "loss": 0.5922,
+      "step": 825
+    },
+    {
+      "epoch": 0.91,
+      "grad_norm": 2.8494598865509033,
+      "learning_rate": 2.49e-05,
+      "loss": 0.6034,
+      "step": 830
+    },
+    {
+      "epoch": 0.92,
+      "grad_norm": 1.5710737705230713,
+      "learning_rate": 2.505e-05,
+      "loss": 0.2955,
+      "step": 835
+    },
+    {
+      "epoch": 0.92,
+      "grad_norm": 2.6547062397003174,
+      "learning_rate": 2.52e-05,
+      "loss": 0.4286,
+      "step": 840
+    },
+    {
+      "epoch": 0.93,
+      "grad_norm": 1.1925177574157715,
+      "learning_rate": 2.535e-05,
+      "loss": 0.4727,
+      "step": 845
+    },
+    {
+      "epoch": 0.94,
+      "grad_norm": 1.3642709255218506,
+      "learning_rate": 2.55e-05,
+      "loss": 0.406,
+      "step": 850
+    },
+    {
+      "epoch": 0.94,
+      "grad_norm": 1.1482707262039185,
+      "learning_rate": 2.565e-05,
+      "loss": 0.4667,
+      "step": 855
+    },
+    {
+      "epoch": 0.95,
+      "grad_norm": 2.0437512397766113,
+      "learning_rate": 2.58e-05,
+      "loss": 0.5288,
+      "step": 860
+    },
+    {
+      "epoch": 0.95,
+      "grad_norm": 1.0377088785171509,
+      "learning_rate": 2.595e-05,
+      "loss": 0.3604,
+      "step": 865
+    },
+    {
+      "epoch": 0.96,
+      "grad_norm": 2.0166819095611572,
+      "learning_rate": 2.61e-05,
+      "loss": 0.4019,
+      "step": 870
+    },
+    {
+      "epoch": 0.96,
+      "grad_norm": 1.1202322244644165,
+      "learning_rate": 2.625e-05,
+      "loss": 0.397,
+      "step": 875
+    },
+    {
+      "epoch": 0.97,
+      "grad_norm": 1.1196401119232178,
+      "learning_rate": 2.64e-05,
+      "loss": 0.3056,
+      "step": 880
+    },
+    {
+      "epoch": 0.97,
+      "grad_norm": 1.8753234148025513,
+      "learning_rate": 2.655e-05,
+      "loss": 0.3347,
+      "step": 885
+    },
+    {
+      "epoch": 0.98,
+      "grad_norm": 2.1503918170928955,
+      "learning_rate": 2.6700000000000002e-05,
+      "loss": 0.6701,
+      "step": 890
+    },
+    {
+      "epoch": 0.99,
+      "grad_norm": 1.2472453117370605,
+      "learning_rate": 2.6850000000000002e-05,
+      "loss": 0.513,
+      "step": 895
+    },
+    {
+      "epoch": 0.99,
+      "grad_norm": 1.2424818277359009,
+      "learning_rate": 2.7000000000000002e-05,
+      "loss": 0.3437,
+      "step": 900
+    },
+    {
+      "epoch": 1.0,
+      "grad_norm": 1.36464524269104,
+      "learning_rate": 2.7150000000000003e-05,
+      "loss": 0.5147,
+      "step": 905
+    },
+    {
+      "epoch": 1.0,
+      "grad_norm": 2.0129966735839844,
+      "learning_rate": 2.7300000000000003e-05,
+      "loss": 0.5189,
+      "step": 910
+    },
+    {
+      "epoch": 1.01,
+      "grad_norm": 3.107072114944458,
+      "learning_rate": 2.7450000000000003e-05,
+      "loss": 0.527,
+      "step": 915
+    },
+    {
+      "epoch": 1.01,
+      "grad_norm": 1.9841065406799316,
+      "learning_rate": 2.7600000000000003e-05,
+      "loss": 0.3458,
+      "step": 920
+    },
+    {
+      "epoch": 1.02,
+      "grad_norm": 1.4352362155914307,
+      "learning_rate": 2.7750000000000004e-05,
+      "loss": 0.4081,
+      "step": 925
+    },
+    {
+      "epoch": 1.02,
+      "grad_norm": 1.5732287168502808,
+      "learning_rate": 2.79e-05,
+      "loss": 0.6428,
+      "step": 930
+    },
+    {
+      "epoch": 1.03,
+      "grad_norm": 1.5462892055511475,
+      "learning_rate": 2.805e-05,
+      "loss": 0.5453,
+      "step": 935
+    },
+    {
+      "epoch": 1.03,
+      "grad_norm": 2.118161201477051,
+      "learning_rate": 2.8199999999999998e-05,
+      "loss": 0.5021,
+      "step": 940
+    },
+    {
+      "epoch": 1.04,
+      "grad_norm": 2.953655481338501,
+      "learning_rate": 2.8349999999999998e-05,
+      "loss": 0.4168,
+      "step": 945
+    },
+    {
+      "epoch": 1.05,
+      "grad_norm": 0.539341390132904,
+      "learning_rate": 2.8499999999999998e-05,
+      "loss": 0.5838,
+      "step": 950
+    },
+    {
+      "epoch": 1.05,
+      "grad_norm": 0.8679121732711792,
+      "learning_rate": 2.865e-05,
+      "loss": 0.4803,
+      "step": 955
+    },
+    {
+      "epoch": 1.06,
+      "grad_norm": 0.6618215441703796,
+      "learning_rate": 2.88e-05,
+      "loss": 0.4741,
+      "step": 960
+    },
+    {
+      "epoch": 1.06,
+      "grad_norm": 1.349791407585144,
+      "learning_rate": 2.895e-05,
+      "loss": 0.539,
+      "step": 965
+    },
+    {
+      "epoch": 1.07,
+      "grad_norm": 0.9783216714859009,
+      "learning_rate": 2.91e-05,
+      "loss": 0.4964,
+      "step": 970
+    },
+    {
+      "epoch": 1.07,
+      "grad_norm": 0.7639239430427551,
+      "learning_rate": 2.925e-05,
+      "loss": 0.305,
+      "step": 975
+    },
+    {
+      "epoch": 1.08,
+      "grad_norm": 1.376918911933899,
+      "learning_rate": 2.94e-05,
+      "loss": 0.415,
+      "step": 980
+    },
+    {
+      "epoch": 1.08,
+      "grad_norm": 0.8644680380821228,
+      "learning_rate": 2.955e-05,
+      "loss": 0.4427,
+      "step": 985
+    },
+    {
+      "epoch": 1.09,
+      "grad_norm": 0.9307350516319275,
+      "learning_rate": 2.97e-05,
+      "loss": 0.3651,
+      "step": 990
+    },
+    {
+      "epoch": 1.1,
+      "grad_norm": 1.0594663619995117,
+      "learning_rate": 2.985e-05,
+      "loss": 0.3768,
+      "step": 995
+    },
+    {
+      "epoch": 1.1,
+      "grad_norm": 1.9165922403335571,
+      "learning_rate": 3e-05,
+      "loss": 0.3997,
+      "step": 1000
+    },
+    {
+      "epoch": 1.11,
+      "grad_norm": 1.4573155641555786,
+      "learning_rate": 2.999937737939374e-05,
+      "loss": 0.2581,
+      "step": 1005
+    },
+    {
+      "epoch": 1.11,
+      "grad_norm": 1.2896963357925415,
+      "learning_rate": 2.9997509569262485e-05,
+      "loss": 0.3903,
+      "step": 1010
+    },
+    {
+      "epoch": 1.12,
+      "grad_norm": 1.1831902265548706,
+      "learning_rate": 2.9994396724664514e-05,
+      "loss": 0.49,
+      "step": 1015
+    },
+    {
+      "epoch": 1.12,
+      "grad_norm": 2.8245599269866943,
+      "learning_rate": 2.9990039104015984e-05,
+      "loss": 0.4482,
+      "step": 1020
+    },
+    {
+      "epoch": 1.13,
+      "grad_norm": 1.2168867588043213,
+      "learning_rate": 2.998443706906948e-05,
+      "loss": 0.5133,
+      "step": 1025
+    },
+    {
+      "epoch": 1.13,
+      "grad_norm": 1.0247105360031128,
+      "learning_rate": 2.9977591084883992e-05,
+      "loss": 0.4762,
+      "step": 1030
+    },
+    {
+      "epoch": 1.14,
+      "grad_norm": 0.7047167420387268,
+      "learning_rate": 2.9969501719786296e-05,
+      "loss": 0.6164,
+      "step": 1035
+    },
+    {
+      "epoch": 1.15,
+      "grad_norm": 1.1248735189437866,
+      "learning_rate": 2.9960169645323774e-05,
+      "loss": 0.3757,
+      "step": 1040
+    },
+    {
+      "epoch": 1.15,
+      "grad_norm": 0.8849346041679382,
+      "learning_rate": 2.9949595636208678e-05,
+      "loss": 0.3092,
+      "step": 1045
+    },
+    {
+      "epoch": 1.16,
+      "grad_norm": 1.1959189176559448,
+      "learning_rate": 2.9937780570253807e-05,
+      "loss": 0.3765,
+      "step": 1050
+    },
+    {
+      "epoch": 1.16,
+      "grad_norm": 0.845991313457489,
+      "learning_rate": 2.9924725428299625e-05,
+      "loss": 0.5164,
+      "step": 1055
+    },
+    {
+      "epoch": 1.17,
+      "grad_norm": 0.6847428679466248,
+      "learning_rate": 2.991043129413285e-05,
+      "loss": 0.439,
+      "step": 1060
+    },
+    {
+      "epoch": 1.17,
+      "grad_norm": 0.9967267513275146,
+      "learning_rate": 2.9894899354396496e-05,
+      "loss": 0.4436,
+      "step": 1065
+    },
+    {
+      "epoch": 1.18,
+      "grad_norm": 0.8753231167793274,
+      "learning_rate": 2.9878130898491303e-05,
+      "loss": 0.3638,
+      "step": 1070
+    },
+    {
+      "epoch": 1.18,
+      "grad_norm": 0.7683019042015076,
+      "learning_rate": 2.9860127318468782e-05,
+      "loss": 0.5267,
+      "step": 1075
+    },
+    {
+      "epoch": 1.19,
+      "grad_norm": 1.2402276992797852,
+      "learning_rate": 2.9840890108915572e-05,
+      "loss": 0.66,
+      "step": 1080
+    },
+    {
+      "epoch": 1.19,
+      "grad_norm": 0.9644812345504761,
+      "learning_rate": 2.9820420866829433e-05,
+      "loss": 0.1985,
+      "step": 1085
+    },
+    {
+      "epoch": 1.2,
+      "grad_norm": 1.2949899435043335,
+      "learning_rate": 2.979872129148661e-05,
+      "loss": 0.4225,
+      "step": 1090
+    },
+    {
+      "epoch": 1.21,
+      "grad_norm": 1.2814561128616333,
+      "learning_rate": 2.9775793184300798e-05,
+      "loss": 0.2877,
+      "step": 1095
+    },
+    {
+      "epoch": 1.21,
+      "grad_norm": 2.1358540058135986,
+      "learning_rate": 2.9751638448673612e-05,
+      "loss": 0.609,
+      "step": 1100
+    },
+    {
+      "epoch": 1.22,
+      "grad_norm": NaN,
+      "learning_rate": 2.9731432831792347e-05,
+      "loss": 0.5548,
+      "step": 1105
+    },
+    {
+      "epoch": 1.22,
+      "grad_norm": 1.0352325439453125,
+      "learning_rate": 2.9705075286477642e-05,
+      "loss": 0.5043,
+      "step": 1110
+    },
+    {
+      "epoch": 1.23,
+      "grad_norm": 1.3879212141036987,
+      "learning_rate": 2.9677496983444274e-05,
+      "loss": 0.3721,
+      "step": 1115
+    },
+    {
+      "epoch": 1.23,
+      "grad_norm": 1.3792262077331543,
+      "learning_rate": 2.9648700212134885e-05,
+      "loss": 0.4842,
+      "step": 1120
+    },
+    {
+      "epoch": 1.24,
+      "grad_norm": 1.1040356159210205,
+      "learning_rate": 2.961868736314456e-05,
+      "loss": 0.4203,
+      "step": 1125
+    },
+    {
+      "epoch": 1.24,
+      "grad_norm": 1.0859042406082153,
+      "learning_rate": 2.9587460928022404e-05,
+      "loss": 0.4129,
+      "step": 1130
+    },
+    {
+      "epoch": 1.25,
+      "grad_norm": 0.9380947351455688,
+      "learning_rate": 2.9555023499064677e-05,
+      "loss": 0.3209,
+      "step": 1135
+    },
+    {
+      "epoch": 1.26,
+      "grad_norm": 0.6131200194358826,
+      "learning_rate": 2.9521377769099603e-05,
+      "loss": 0.4211,
+      "step": 1140
+    },
+    {
+      "epoch": 1.26,
+      "grad_norm": 1.2759392261505127,
+      "learning_rate": 2.948652653126382e-05,
+      "loss": 0.7952,
+      "step": 1145
+    },
+    {
+      "epoch": 1.27,
+      "grad_norm": 2.88456130027771,
+      "learning_rate": 2.9450472678770505e-05,
+      "loss": 0.5054,
+      "step": 1150
+    },
+    {
+      "epoch": 1.27,
+      "grad_norm": 0.833156943321228,
+      "learning_rate": 2.9413219204669195e-05,
+      "loss": 0.37,
+      "step": 1155
+    },
+    {
+      "epoch": 1.28,
+      "grad_norm": 1.2013107538223267,
+      "learning_rate": 2.9374769201597305e-05,
+      "loss": 0.4595,
+      "step": 1160
+    },
+    {
+      "epoch": 1.28,
+      "grad_norm": 1.4866896867752075,
+      "learning_rate": 2.9335125861523395e-05,
+      "loss": 0.318,
+      "step": 1165
+    },
+    {
+      "epoch": 1.29,
+      "grad_norm": 3.088609218597412,
+      "learning_rate": 2.9294292475482192e-05,
+      "loss": 0.5222,
+      "step": 1170
+    },
+    {
+      "epoch": 1.29,
+      "grad_norm": 1.1780695915222168,
+      "learning_rate": 2.9252272433301376e-05,
+      "loss": 0.6204,
+      "step": 1175
+    },
+    {
+      "epoch": 1.3,
+      "grad_norm": 0.6830143928527832,
+      "learning_rate": 2.920906922332016e-05,
+      "loss": 0.5025,
+      "step": 1180
+    },
+    {
+      "epoch": 1.3,
+      "grad_norm": 1.2696876525878906,
+      "learning_rate": 2.9164686432099713e-05,
+      "loss": 0.3402,
+      "step": 1185
+    },
+    {
+      "epoch": 1.31,
+      "grad_norm": 0.5633934736251831,
+      "learning_rate": 2.9119127744125428e-05,
+      "loss": 0.4001,
+      "step": 1190
+    },
+    {
+      "epoch": 1.32,
+      "grad_norm": 0.7630770802497864,
+      "learning_rate": 2.9072396941501023e-05,
+      "loss": 0.4415,
+      "step": 1195
+    },
+    {
+      "epoch": 1.32,
+      "grad_norm": 1.4518195390701294,
+      "learning_rate": 2.9024497903634584e-05,
+      "loss": 0.5651,
+      "step": 1200
+    },
+    {
+      "epoch": 1.33,
+      "grad_norm": 1.5180546045303345,
+      "learning_rate": 2.8975434606916515e-05,
+      "loss": 0.4689,
+      "step": 1205
+    },
+    {
+      "epoch": 1.33,
+      "grad_norm": 1.4250961542129517,
+      "learning_rate": 2.8925211124389423e-05,
+      "loss": 0.2893,
+      "step": 1210
+    },
+    {
+      "epoch": 1.34,
+      "grad_norm": 1.938849687576294,
+      "learning_rate": 2.887383162540999e-05,
+      "loss": 0.3407,
+      "step": 1215
+    },
+    {
+      "epoch": 1.34,
+      "grad_norm": 1.0794405937194824,
+      "learning_rate": 2.8821300375302858e-05,
+      "loss": 0.374,
+      "step": 1220
+    },
+    {
+      "epoch": 1.35,
+      "grad_norm": 1.5490267276763916,
+      "learning_rate": 2.876762173500653e-05,
+      "loss": 0.5285,
+      "step": 1225
+    },
+    {
+      "epoch": 1.35,
+      "grad_norm": 1.1713478565216064,
+      "learning_rate": 2.8712800160711353e-05,
+      "loss": 0.5842,
+      "step": 1230
+    },
+    {
+      "epoch": 1.36,
+      "grad_norm": 1.5063586235046387,
+      "learning_rate": 2.8656840203489565e-05,
+      "loss": 0.4172,
+      "step": 1235
+    },
+    {
+      "epoch": 1.37,
+      "grad_norm": 0.477967232465744,
+      "learning_rate": 2.8599746508917496e-05,
+      "loss": 0.4052,
+      "step": 1240
+    },
+    {
+      "epoch": 1.37,
+      "grad_norm": 1.269463062286377,
+      "learning_rate": 2.8541523816689917e-05,
+      "loss": 0.5223,
+      "step": 1245
+    },
+    {
+      "epoch": 1.38,
+      "grad_norm": 0.9727627038955688,
+      "learning_rate": 2.8482176960226555e-05,
+      "loss": 0.3865,
+      "step": 1250
+    },
+    {
+      "epoch": 1.38,
+      "grad_norm": 1.1779907941818237,
+      "learning_rate": 2.842171086627083e-05,
+      "loss": 0.4907,
+      "step": 1255
+    },
+    {
+      "epoch": 1.39,
+      "grad_norm": 1.321874976158142,
+      "learning_rate": 2.8360130554480895e-05,
+      "loss": 0.4799,
+      "step": 1260
+    },
+    {
+      "epoch": 1.39,
+      "grad_norm": 1.128961205482483,
+      "learning_rate": 2.829744113701289e-05,
+      "loss": 0.2856,
+      "step": 1265
+    },
+    {
+      "epoch": 1.4,
+      "grad_norm": 0.7564502358436584,
+      "learning_rate": 2.8233647818096562e-05,
+      "loss": 0.4064,
+      "step": 1270
+    },
+    {
+      "epoch": 1.4,
+      "grad_norm": 1.3942064046859741,
+      "learning_rate": 2.8168755893603233e-05,
+      "loss": 0.4904,
+      "step": 1275
+    },
+    {
+      "epoch": 1.41,
+      "grad_norm": 0.8835464715957642,
+      "learning_rate": 2.8102770750606147e-05,
+      "loss": 0.4211,
+      "step": 1280
+    },
+    {
+      "epoch": 1.41,
+      "grad_norm": 1.4010355472564697,
+      "learning_rate": 2.8035697866933277e-05,
+      "loss": 0.3368,
+      "step": 1285
+    },
+    {
+      "epoch": 1.42,
+      "grad_norm": 0.8351401686668396,
+      "learning_rate": 2.7967542810712548e-05,
+      "loss": 0.4576,
+      "step": 1290
+    },
+    {
+      "epoch": 1.43,
+      "grad_norm": 1.1750423908233643,
+      "learning_rate": 2.789831123990962e-05,
+      "loss": 0.2992,
+      "step": 1295
+    },
+    {
+      "epoch": 1.43,
+      "grad_norm": 0.9141637682914734,
+      "learning_rate": 2.7828008901858175e-05,
+      "loss": 0.5354,
+      "step": 1300
+    },
+    {
+      "epoch": 1.44,
+      "grad_norm": 1.4612001180648804,
+      "learning_rate": 2.775664163278278e-05,
+      "loss": 0.4124,
+      "step": 1305
+    },
+    {
+      "epoch": 1.44,
+      "grad_norm": 0.8740494847297668,
+      "learning_rate": 2.7684215357314428e-05,
+      "loss": 0.4102,
+      "step": 1310
+    },
+    {
+      "epoch": 1.45,
+      "grad_norm": 1.0325684547424316,
+      "learning_rate": 2.7610736087998648e-05,
+      "loss": 0.3746,
+      "step": 1315
+    },
+    {
+      "epoch": 1.45,
+      "grad_norm": 1.2742629051208496,
+      "learning_rate": 2.7536209924796407e-05,
+      "loss": 0.5872,
+      "step": 1320
+    },
+    {
+      "epoch": 1.46,
+      "grad_norm": 0.6541928052902222,
+      "learning_rate": 2.7460643054577684e-05,
+      "loss": 0.4492,
+      "step": 1325
+    },
+    {
+      "epoch": 1.46,
+      "grad_norm": 0.9362463355064392,
+      "learning_rate": 2.7384041750607895e-05,
+      "loss": 0.3903,
+      "step": 1330
+    },
+    {
+      "epoch": 1.47,
+      "grad_norm": 1.3968509435653687,
+      "learning_rate": 2.7306412372027082e-05,
+      "loss": 0.557,
+      "step": 1335
+    },
+    {
+      "epoch": 1.48,
+      "grad_norm": 1.368221640586853,
+      "learning_rate": 2.7227761363322006e-05,
+      "loss": 0.446,
+      "step": 1340
+    },
+    {
+      "epoch": 1.48,
+      "grad_norm": 1.2684072256088257,
+      "learning_rate": 2.7148095253791174e-05,
+      "loss": 0.6294,
+      "step": 1345
+    },
+    {
+      "epoch": 1.49,
+      "grad_norm": 1.466238260269165,
+      "learning_rate": 2.706742065700276e-05,
+      "loss": 0.4555,
+      "step": 1350
+    },
+    {
+      "epoch": 1.49,
+      "grad_norm": 1.2599902153015137,
+      "learning_rate": 2.6985744270245627e-05,
+      "loss": 0.4199,
+      "step": 1355
+    },
+    {
+      "epoch": 1.5,
+      "grad_norm": 0.8082879185676575,
+      "learning_rate": 2.690307287397329e-05,
+      "loss": 0.3501,
+      "step": 1360
+    },
+    {
+      "epoch": 1.5,
+      "grad_norm": 1.3563495874404907,
+      "learning_rate": 2.681941333124107e-05,
+      "loss": 0.4526,
+      "step": 1365
+    },
+    {
+      "epoch": 1.51,
+      "grad_norm": 0.7122817039489746,
+      "learning_rate": 2.6734772587136324e-05,
+      "loss": 0.4239,
+      "step": 1370
+    },
+    {
+      "epoch": 1.51,
+      "grad_norm": 1.2698917388916016,
+      "learning_rate": 2.664915766820191e-05,
+      "loss": 0.3165,
+      "step": 1375
+    },
+    {
+      "epoch": 1.52,
+      "grad_norm": 1.7459849119186401,
+      "learning_rate": 2.656257568185286e-05,
+      "loss": 0.3812,
+      "step": 1380
+    },
+    {
+      "epoch": 1.52,
+      "grad_norm": 0.7778897285461426,
+      "learning_rate": 2.6475033815786353e-05,
+      "loss": 0.4149,
+      "step": 1385
+    },
+    {
+      "epoch": 1.53,
+      "grad_norm": 1.0717759132385254,
+      "learning_rate": 2.6386539337385012e-05,
+      "loss": 0.515,
+      "step": 1390
+    },
+    {
+      "epoch": 1.54,
+      "grad_norm": 0.9191924929618835,
+      "learning_rate": 2.629709959311361e-05,
+      "loss": 0.3977,
+      "step": 1395
+    },
+    {
+      "epoch": 1.54,
+      "grad_norm": 1.0447479486465454,
+      "learning_rate": 2.62067220079092e-05,
+      "loss": 0.335,
+      "step": 1400
+    },
+    {
+      "epoch": 1.55,
+      "grad_norm": 1.2984446287155151,
+      "learning_rate": 2.6115414084564682e-05,
+      "loss": 0.5425,
+      "step": 1405
+    },
+    {
+      "epoch": 1.55,
+      "grad_norm": 0.9010884761810303,
+      "learning_rate": 2.6023183403106014e-05,
+      "loss": 0.4465,
+      "step": 1410
+    },
+    {
+      "epoch": 1.56,
+      "grad_norm": 1.8324975967407227,
+      "learning_rate": 2.59300376201629e-05,
+      "loss": 0.3131,
+      "step": 1415
+    },
+    {
+      "epoch": 1.56,
+      "grad_norm": 1.0951191186904907,
+      "learning_rate": 2.583598446833319e-05,
+      "loss": 0.4113,
+      "step": 1420
+    },
+    {
+      "epoch": 1.57,
+      "grad_norm": 0.9710536003112793,
+      "learning_rate": 2.5741031755540932e-05,
+      "loss": 0.368,
+      "step": 1425
+    },
+    {
+      "epoch": 1.57,
+      "grad_norm": 0.8006061911582947,
+      "learning_rate": 2.564518736438821e-05,
+      "loss": 0.5833,
+      "step": 1430
+    },
+    {
+      "epoch": 1.58,
+      "grad_norm": 0.47641128301620483,
+      "learning_rate": 2.5548459251500747e-05,
+      "loss": 0.5812,
+      "step": 1435
+    },
+    {
+      "epoch": 1.59,
+      "grad_norm": 0.873199999332428,
+      "learning_rate": 2.5450855446867384e-05,
+      "loss": 0.5852,
+      "step": 1440
+    },
+    {
+      "epoch": 1.59,
+      "grad_norm": 1.6408307552337646,
+      "learning_rate": 2.5352384053173453e-05,
+      "loss": 0.4245,
+      "step": 1445
+    },
+    {
+      "epoch": 1.6,
+      "grad_norm": 0.8571744561195374,
+      "learning_rate": 2.5253053245128135e-05,
+      "loss": 0.3346,
+      "step": 1450
+    },
+    {
+      "epoch": 1.6,
+      "grad_norm": 0.6984312534332275,
+      "learning_rate": 2.5152871268785813e-05,
+      "loss": 0.3652,
+      "step": 1455
+    },
+    {
+      "epoch": 1.61,
+      "grad_norm": 0.6111830472946167,
+      "learning_rate": 2.5051846440861545e-05,
+      "loss": 0.2863,
+      "step": 1460
+    },
+    {
+      "epoch": 1.61,
+      "grad_norm": 1.2058312892913818,
+      "learning_rate": 2.4949987148040608e-05,
+      "loss": 0.5868,
+      "step": 1465
+    },
+    {
+      "epoch": 1.62,
+      "grad_norm": 0.940091609954834,
+      "learning_rate": 2.4847301846282277e-05,
+      "loss": 0.4137,
+      "step": 1470
+    },
+    {
+      "epoch": 1.62,
+      "grad_norm": 1.831299901008606,
+      "learning_rate": 2.474379906011788e-05,
+      "loss": 0.4452,
+      "step": 1475
+    },
+    {
+      "epoch": 1.63,
+      "grad_norm": 0.8370305299758911,
+      "learning_rate": 2.4639487381943075e-05,
+      "loss": 0.4705,
+      "step": 1480
+    },
+    {
+      "epoch": 1.64,
+      "grad_norm": 1.1725839376449585,
+      "learning_rate": 2.4534375471304563e-05,
+      "loss": 0.3157,
+      "step": 1485
+    },
+    {
+      "epoch": 1.64,
+      "grad_norm": 1.3950480222702026,
+      "learning_rate": 2.442847205418122e-05,
+      "loss": 0.4179,
+      "step": 1490
+    },
+    {
+      "epoch": 1.65,
+      "grad_norm": 1.3916176557540894,
+      "learning_rate": 2.4321785922259685e-05,
+      "loss": 0.5267,
+      "step": 1495
+    },
+    {
+      "epoch": 1.65,
+      "grad_norm": 1.6205462217330933,
+      "learning_rate": 2.42143259322045e-05,
+      "loss": 0.5793,
+      "step": 1500
+    },
+    {
+      "epoch": 1.66,
+      "grad_norm": 1.0348162651062012,
+      "learning_rate": 2.4106101004922893e-05,
+      "loss": 0.4365,
+      "step": 1505
+    },
+    {
+      "epoch": 1.66,
+      "grad_norm": 0.5200983285903931,
+      "learning_rate": 2.3997120124824178e-05,
+      "loss": 0.3882,
+      "step": 1510
+    },
+    {
+      "epoch": 1.67,
+      "grad_norm": 1.8840643167495728,
+      "learning_rate": 2.3887392339073898e-05,
+      "loss": 0.3899,
+      "step": 1515
+    },
+    {
+      "epoch": 1.67,
+      "grad_norm": 1.1195874214172363,
+      "learning_rate": 2.3776926756842787e-05,
+      "loss": 0.3071,
+      "step": 1520
+    },
+    {
+      "epoch": 1.68,
+      "grad_norm": 0.9049373269081116,
+      "learning_rate": 2.3665732548550558e-05,
+      "loss": 0.3486,
+      "step": 1525
+    },
+    {
+      "epoch": 1.68,
+      "grad_norm": 1.0285577774047852,
+      "learning_rate": 2.3553818945104588e-05,
+      "loss": 0.4388,
+      "step": 1530
+    },
+    {
+      "epoch": 1.69,
+      "grad_norm": 1.3187785148620605,
+      "learning_rate": 2.3441195237133624e-05,
+      "loss": 0.4953,
+      "step": 1535
+    },
+    {
+      "epoch": 1.7,
+      "grad_norm": 1.222777247428894,
+      "learning_rate": 2.3327870774216515e-05,
+      "loss": 0.4594,
+      "step": 1540
+    },
+    {
+      "epoch": 1.7,
+      "grad_norm": 0.6450194120407104,
+      "learning_rate": 2.3213854964106035e-05,
+      "loss": 0.4665,
+      "step": 1545
+    },
+    {
+      "epoch": 1.71,
+      "grad_norm": 1.990315318107605,
+      "learning_rate": 2.309915727194789e-05,
+      "loss": 0.3313,
+      "step": 1550
+    },
+    {
+      "epoch": 1.71,
+      "grad_norm": 1.7571722269058228,
+      "learning_rate": 2.2983787219494964e-05,
+      "loss": 0.3571,
+      "step": 1555
+    },
+    {
+      "epoch": 1.72,
+      "grad_norm": 0.9484684467315674,
+      "learning_rate": 2.2867754384316865e-05,
+      "loss": 0.4045,
+      "step": 1560
+    },
+    {
+      "epoch": 1.72,
+      "grad_norm": 0.9541623592376709,
+      "learning_rate": 2.2751068399004806e-05,
+      "loss": 0.5165,
+      "step": 1565
+    },
+    {
+      "epoch": 1.73,
+      "grad_norm": 1.3149410486221313,
+      "learning_rate": 2.2633738950371984e-05,
+      "loss": 0.3404,
+      "step": 1570
+    },
+    {
+      "epoch": 1.73,
+      "grad_norm": 0.7877027988433838,
+      "learning_rate": 2.2515775778649386e-05,
+      "loss": 0.3857,
+      "step": 1575
+    },
+    {
+      "epoch": 1.74,
+      "grad_norm": 0.8127877116203308,
+      "learning_rate": 2.2397188676677217e-05,
+      "loss": 0.5441,
+      "step": 1580
+    },
+    {
+      "epoch": 1.75,
+      "grad_norm": 1.8573230504989624,
+      "learning_rate": 2.227798748909191e-05,
+      "loss": 0.4012,
+      "step": 1585
+    },
+    {
+      "epoch": 1.75,
+      "grad_norm": 1.7995444536209106,
+      "learning_rate": 2.2158182111508904e-05,
+      "loss": 0.3189,
+      "step": 1590
+    },
+    {
+      "epoch": 1.76,
+      "grad_norm": 0.9786883592605591,
+      "learning_rate": 2.20377824897011e-05,
+      "loss": 0.4802,
+      "step": 1595
+    },
+    {
+      "epoch": 1.76,
+      "grad_norm": 0.9040762186050415,
+      "learning_rate": 2.191679861877323e-05,
+      "loss": 0.4147,
+      "step": 1600
+    },
+    {
+      "epoch": 1.77,
+      "grad_norm": 0.962874174118042,
+      "learning_rate": 2.179524054233211e-05,
+      "loss": 0.3856,
+      "step": 1605
+    },
+    {
+      "epoch": 1.77,
+      "grad_norm": 1.2043737173080444,
+      "learning_rate": 2.1673118351652843e-05,
+      "loss": 0.5292,
+      "step": 1610
+    },
+    {
+      "epoch": 1.78,
+      "grad_norm": 0.5223831534385681,
+      "learning_rate": 2.1550442184841072e-05,
+      "loss": 0.4305,
+      "step": 1615
+    },
+    {
+      "epoch": 1.78,
+      "grad_norm": 0.9191122055053711,
+      "learning_rate": 2.1427222225991383e-05,
+      "loss": 0.472,
+      "step": 1620
+    },
+    {
+      "epoch": 1.79,
+      "grad_norm": 1.261473536491394,
+      "learning_rate": 2.130346870434184e-05,
+      "loss": 0.4266,
+      "step": 1625
+    },
+    {
+      "epoch": 1.79,
+      "grad_norm": 1.202399730682373,
+      "learning_rate": 2.1179191893424797e-05,
+      "loss": 0.3269,
+      "step": 1630
+    },
+    {
+      "epoch": 1.8,
+      "grad_norm": 0.7132131457328796,
+      "learning_rate": 2.105440211021404e-05,
+      "loss": 0.2125,
+      "step": 1635
+    },
+    {
+      "epoch": 1.81,
+      "grad_norm": 1.0556139945983887,
+      "learning_rate": 2.09291097142683e-05,
+      "loss": 0.3205,
+      "step": 1640
+    },
+    {
+      "epoch": 1.81,
+      "grad_norm": 0.7523159980773926,
+      "learning_rate": 2.0803325106871234e-05,
+      "loss": 0.3239,
+      "step": 1645
+    },
+    {
+      "epoch": 1.82,
+      "grad_norm": 1.4122850894927979,
+      "learning_rate": 2.0677058730168e-05,
+      "loss": 0.4792,
+      "step": 1650
+    },
+    {
+      "epoch": 1.82,
+      "grad_norm": 1.4816789627075195,
+      "learning_rate": 2.055032106629831e-05,
+      "loss": 0.6169,
+      "step": 1655
+    },
+    {
+      "epoch": 1.83,
+      "grad_norm": 0.8011027574539185,
+      "learning_rate": 2.0423122636526325e-05,
+      "loss": 0.4804,
+      "step": 1660
+    },
+    {
+      "epoch": 1.83,
+      "grad_norm": 0.6560605764389038,
+      "learning_rate": 2.0295474000367173e-05,
+      "loss": 0.3411,
+      "step": 1665
+    },
+    {
+      "epoch": 1.84,
+      "grad_norm": 1.2358989715576172,
+      "learning_rate": 2.0167385754710346e-05,
+      "loss": 0.3476,
+      "step": 1670
+    },
+    {
+      "epoch": 1.84,
+      "grad_norm": 1.342943787574768,
+      "learning_rate": 2.0038868532940003e-05,
+      "loss": 0.3981,
+      "step": 1675
+    },
+    {
+      "epoch": 1.85,
+      "grad_norm": 1.1446216106414795,
+      "learning_rate": 1.990993300405222e-05,
+      "loss": 0.406,
+      "step": 1680
+    },
+    {
+      "epoch": 1.86,
+      "grad_norm": 1.2379969358444214,
+      "learning_rate": 1.9780589871769272e-05,
+      "loss": 0.4369,
+      "step": 1685
+    },
+    {
+      "epoch": 1.86,
+      "grad_norm": 1.4717028141021729,
+      "learning_rate": 1.9650849873651102e-05,
+      "loss": 0.5346,
+      "step": 1690
+    },
+    {
+      "epoch": 1.87,
+      "grad_norm": 0.824568510055542,
+      "learning_rate": 1.952072378020387e-05,
+      "loss": 0.3544,
+      "step": 1695
+    },
+    {
+      "epoch": 1.87,
+      "grad_norm": 1.0470279455184937,
+      "learning_rate": 1.9390222393985866e-05,
+      "loss": 0.2773,
+      "step": 1700
+    },
+    {
+      "epoch": 1.88,
+      "grad_norm": 0.9954514503479004,
+      "learning_rate": 1.9259356548710722e-05,
+      "loss": 0.3972,
+      "step": 1705
+    },
+    {
+      "epoch": 1.88,
+      "grad_norm": 0.49930986762046814,
+      "learning_rate": 1.912813710834803e-05,
+      "loss": 0.3004,
+      "step": 1710
+    },
+    {
+      "epoch": 1.89,
+      "grad_norm": 0.7324523329734802,
+      "learning_rate": 1.8996574966221453e-05,
+      "loss": 0.7305,
+      "step": 1715
+    },
+    {
+      "epoch": 1.89,
+      "grad_norm": 0.9635807275772095,
+      "learning_rate": 1.886468104410442e-05,
+      "loss": 0.4353,
+      "step": 1720
+    },
+    {
+      "epoch": 1.9,
+      "grad_norm": 1.794856309890747,
+      "learning_rate": 1.873246629131343e-05,
+      "loss": 0.3805,
+      "step": 1725
+    },
+    {
+      "epoch": 1.9,
+      "grad_norm": 1.0289092063903809,
+      "learning_rate": 1.8599941683799087e-05,
+      "loss": 0.4203,
+      "step": 1730
+    },
+    {
+      "epoch": 1.91,
+      "grad_norm": 0.6733255982398987,
+      "learning_rate": 1.846711822323492e-05,
+      "loss": 0.3358,
+      "step": 1735
+    },
+    {
+      "epoch": 1.92,
+      "grad_norm": 0.44114547967910767,
+      "learning_rate": 1.8334006936104077e-05,
+      "loss": 0.4077,
+      "step": 1740
+    },
+    {
+      "epoch": 1.92,
+      "grad_norm": 1.2705578804016113,
+      "learning_rate": 1.8200618872783917e-05,
+      "loss": 0.2637,
+      "step": 1745
+    },
+    {
+      "epoch": 1.93,
+      "grad_norm": 1.1656275987625122,
+      "learning_rate": 1.8066965106628698e-05,
+      "loss": 0.2701,
+      "step": 1750
+    },
+    {
+      "epoch": 1.93,
+      "grad_norm": 1.0239408016204834,
+      "learning_rate": 1.7933056733050267e-05,
+      "loss": 0.3862,
+      "step": 1755
+    },
+    {
+      "epoch": 1.94,
+      "grad_norm": 1.3315787315368652,
+      "learning_rate": 1.7798904868596994e-05,
+      "loss": 0.4242,
+      "step": 1760
+    },
+    {
+      "epoch": 1.94,
+      "grad_norm": 1.190731406211853,
+      "learning_rate": 1.7664520650030903e-05,
+      "loss": 0.1688,
+      "step": 1765
+    },
+    {
+      "epoch": 1.95,
+      "grad_norm": 1.2880607843399048,
+      "learning_rate": 1.7529915233403145e-05,
+      "loss": 0.3489,
+      "step": 1770
+    },
+    {
+      "epoch": 1.95,
+      "grad_norm": 0.499487966299057,
+      "learning_rate": 1.7395099793127865e-05,
+      "loss": 0.4446,
+      "step": 1775
+    },
+    {
+      "epoch": 1.96,
+      "grad_norm": 1.118751049041748,
+      "learning_rate": 1.726008552105455e-05,
+      "loss": 0.4587,
+      "step": 1780
+    },
+    {
+      "epoch": 1.97,
+      "grad_norm": 1.0625907182693481,
+      "learning_rate": 1.712488362553893e-05,
+      "loss": 0.6011,
+      "step": 1785
+    },
+    {
+      "epoch": 1.97,
+      "grad_norm": 0.5367158055305481,
+      "learning_rate": 1.6989505330512484e-05,
+      "loss": 0.292,
+      "step": 1790
+    },
+    {
+      "epoch": 1.98,
+      "grad_norm": 0.6846025586128235,
+      "learning_rate": 1.6853961874550698e-05,
+      "loss": 0.396,
+      "step": 1795
+    },
+    {
+      "epoch": 1.98,
+      "grad_norm": 1.0111944675445557,
+      "learning_rate": 1.6718264509940076e-05,
+      "loss": 0.3749,
+      "step": 1800
+    },
+    {
+      "epoch": 1.99,
+      "grad_norm": 1.2397122383117676,
+      "learning_rate": 1.6582424501743996e-05,
+      "loss": 0.2241,
+      "step": 1805
+    },
+    {
+      "epoch": 1.99,
+      "grad_norm": 1.4244718551635742,
+      "learning_rate": 1.644645312686757e-05,
+      "loss": 0.3892,
+      "step": 1810
+    },
+    {
+      "epoch": 2.0,
+      "grad_norm": 2.0830795764923096,
+      "learning_rate": 1.631036167312144e-05,
+      "loss": 0.4523,
+      "step": 1815
+    },
+    {
+      "epoch": 2.0,
+      "grad_norm": 0.6778771281242371,
+      "learning_rate": 1.617416143828473e-05,
+      "loss": 0.3769,
+      "step": 1820
+    },
+    {
+      "epoch": 2.01,
+      "grad_norm": 0.6887485384941101,
+      "learning_rate": 1.603786372916714e-05,
+      "loss": 0.4108,
+      "step": 1825
+    },
+    {
+      "epoch": 2.01,
+      "grad_norm": 1.05043625831604,
+      "learning_rate": 1.5901479860670323e-05,
+      "loss": 0.4774,
+      "step": 1830
+    },
+    {
+      "epoch": 2.02,
+      "grad_norm": 0.8635022044181824,
+      "learning_rate": 1.576502115484852e-05,
+      "loss": 0.46,
+      "step": 1835
+    },
+    {
+      "epoch": 2.03,
+      "grad_norm": 0.7978819608688354,
+      "learning_rate": 1.5628498939968686e-05,
+      "loss": 0.4967,
+      "step": 1840
+    },
+    {
+      "epoch": 2.03,
+      "grad_norm": 0.591788113117218,
+      "learning_rate": 1.549192454957005e-05,
+      "loss": 0.4264,
+      "step": 1845
+    },
+    {
+      "epoch": 2.04,
+      "grad_norm": 0.7562071084976196,
+      "learning_rate": 1.5355309321523237e-05,
+      "loss": 0.5705,
+      "step": 1850
+    },
+    {
+      "epoch": 2.04,
+      "grad_norm": 0.5904907584190369,
+      "learning_rate": 1.5218664597089071e-05,
+      "loss": 0.6413,
+      "step": 1855
+    },
+    {
+      "epoch": 2.05,
+      "grad_norm": 1.1089377403259277,
+      "learning_rate": 1.508200171997704e-05,
+      "loss": 0.4454,
+      "step": 1860
+    },
+    {
+      "epoch": 2.05,
+      "grad_norm": 0.8817293643951416,
+      "learning_rate": 1.4945332035403587e-05,
+      "loss": 0.4438,
+      "step": 1865
+    },
+    {
+      "epoch": 2.06,
+      "grad_norm": 1.1409955024719238,
+      "learning_rate": 1.4808666889150303e-05,
+      "loss": 0.4391,
+      "step": 1870
+    },
+    {
+      "epoch": 2.06,
+      "grad_norm": 0.8762648701667786,
+      "learning_rate": 1.4672017626622023e-05,
+      "loss": 0.3385,
+      "step": 1875
+    },
+    {
+      "epoch": 2.07,
+      "grad_norm": 0.9974231123924255,
+      "learning_rate": 1.4535395591904958e-05,
+      "loss": 0.3412,
+      "step": 1880
+    },
+    {
+      "epoch": 2.08,
+      "grad_norm": 0.7490105032920837,
+      "learning_rate": 1.439881212682499e-05,
+      "loss": 0.4903,
+      "step": 1885
+    },
+    {
+      "epoch": 2.08,
+      "grad_norm": 1.0734269618988037,
+      "learning_rate": 1.4262278570006103e-05,
+      "loss": 0.5753,
+      "step": 1890
+    },
+    {
+      "epoch": 2.09,
+      "grad_norm": 1.4163916110992432,
+      "learning_rate": 1.4125806255929076e-05,
+      "loss": 0.2856,
+      "step": 1895
+    },
+    {
+      "epoch": 2.09,
+      "grad_norm": 0.6236549019813538,
+      "learning_rate": 1.3989406513990575e-05,
+      "loss": 0.286,
+      "step": 1900
+    },
+    {
+      "epoch": 2.1,
+      "grad_norm": 0.8455921411514282,
+      "learning_rate": 1.38530906675626e-05,
+      "loss": 0.4819,
+      "step": 1905
+    },
+    {
+      "epoch": 2.1,
+      "grad_norm": 0.7641191482543945,
+      "learning_rate": 1.3716870033052476e-05,
+      "loss": 0.497,
+      "step": 1910
+    },
+    {
+      "epoch": 2.11,
+      "grad_norm": 0.9699380993843079,
+      "learning_rate": 1.358075591896341e-05,
+      "loss": 0.1826,
+      "step": 1915
+    },
+    {
+      "epoch": 2.11,
+      "grad_norm": 1.1498082876205444,
+      "learning_rate": 1.34447596249557e-05,
+      "loss": 0.2888,
+      "step": 1920
+    },
+    {
+      "epoch": 2.12,
+      "grad_norm": 1.7836366891860962,
+      "learning_rate": 1.3308892440908678e-05,
+      "loss": 0.5541,
+      "step": 1925
+    },
+    {
+      "epoch": 2.12,
+      "grad_norm": 1.6860030889511108,
+      "learning_rate": 1.3173165645983482e-05,
+      "loss": 0.3234,
+      "step": 1930
+    },
+    {
+      "epoch": 2.13,
+      "grad_norm": 0.7591866850852966,
+      "learning_rate": 1.3037590507686695e-05,
+      "loss": 0.5428,
+      "step": 1935
+    },
+    {
+      "epoch": 2.14,
+      "grad_norm": 1.1677800416946411,
+      "learning_rate": 1.2902178280934948e-05,
+      "loss": 0.3434,
+      "step": 1940
+    },
+    {
+      "epoch": 2.14,
+      "grad_norm": 1.8365012407302856,
+      "learning_rate": 1.2766940207120609e-05,
+      "loss": 0.2709,
+      "step": 1945
+    },
+    {
+      "epoch": 2.15,
+      "grad_norm": 0.8034731149673462,
+      "learning_rate": 1.2631887513178544e-05,
+      "loss": 0.5227,
+      "step": 1950
+    },
+    {
+      "epoch": 2.15,
+      "grad_norm": 0.7503282427787781,
+      "learning_rate": 1.249703141065411e-05,
+      "loss": 0.3255,
+      "step": 1955
+    },
+    {
+      "epoch": 2.16,
+      "grad_norm": 1.1902406215667725,
+      "learning_rate": 1.2362383094772413e-05,
+      "loss": 0.518,
+      "step": 1960
+    },
+    {
+      "epoch": 2.16,
+      "grad_norm": 0.50048828125,
+      "learning_rate": 1.2227953743508935e-05,
+      "loss": 0.412,
+      "step": 1965
+    },
+    {
+      "epoch": 2.17,
+      "grad_norm": 1.088340163230896,
+      "learning_rate": 1.209375451666156e-05,
+      "loss": 0.3134,
+      "step": 1970
+    },
+    {
+      "epoch": 2.17,
+      "grad_norm": 0.9749330878257751,
+      "learning_rate": 1.1959796554924155e-05,
+      "loss": 0.2601,
+      "step": 1975
+    },
+    {
+      "epoch": 2.18,
+      "grad_norm": 1.1269935369491577,
+      "learning_rate": 1.1826090978961706e-05,
+      "loss": 0.4026,
+      "step": 1980
+    },
+    {
+      "epoch": 2.19,
+      "grad_norm": 1.243455171585083,
+      "learning_rate": 1.1692648888487106e-05,
+      "loss": 0.3724,
+      "step": 1985
+    },
+    {
+      "epoch": 2.19,
+      "grad_norm": 1.3622498512268066,
+      "learning_rate": 1.1559481361339723e-05,
+      "loss": 0.3994,
+      "step": 1990
+    },
+    {
+      "epoch": 2.2,
+      "grad_norm": 1.5660467147827148,
+      "learning_rate": 1.142659945256576e-05,
+      "loss": 0.4255,
+      "step": 1995
+    },
+    {
+      "epoch": 2.2,
+      "grad_norm": 1.1115527153015137,
+      "learning_rate": 1.12940141935005e-05,
+      "loss": 0.4976,
+      "step": 2000
+    },
+    {
+      "epoch": 2.21,
+      "grad_norm": 0.8586636781692505,
+      "learning_rate": 1.1161736590852522e-05,
+      "loss": 0.2858,
+      "step": 2005
+    },
+    {
+      "epoch": 2.21,
+      "grad_norm": 2.1333322525024414,
+      "learning_rate": 1.1029777625789982e-05,
+      "loss": 0.3974,
+      "step": 2010
+    },
+    {
+      "epoch": 2.22,
+      "grad_norm": 0.8023208379745483,
+      "learning_rate": 1.0898148253028992e-05,
+      "loss": 0.4768,
+      "step": 2015
+    },
+    {
+      "epoch": 2.22,
+      "grad_norm": 1.3136727809906006,
+      "learning_rate": 1.0766859399924198e-05,
+      "loss": 0.3854,
+      "step": 2020
+    },
+    {
+      "epoch": 2.23,
+      "grad_norm": 1.2942872047424316,
+      "learning_rate": 1.0635921965561644e-05,
+      "loss": 0.6315,
+      "step": 2025
+    },
+    {
+      "epoch": 2.24,
+      "grad_norm": 1.1733945608139038,
+      "learning_rate": 1.0505346819853966e-05,
+      "loss": 0.2202,
+      "step": 2030
+    },
+    {
+      "epoch": 2.24,
+      "grad_norm": 1.2896422147750854,
+      "learning_rate": 1.0375144802638011e-05,
+      "loss": 0.3692,
+      "step": 2035
+    },
+    {
+      "epoch": 2.25,
+      "grad_norm": 1.188212275505066,
+      "learning_rate": 1.024532672277497e-05,
+      "loss": 0.4869,
+      "step": 2040
+    },
+    {
+      "epoch": 2.25,
+      "grad_norm": 1.2566425800323486,
+      "learning_rate": 1.0115903357253056e-05,
+      "loss": 0.3254,
+      "step": 2045
+    },
+    {
+      "epoch": 2.26,
+      "grad_norm": 1.7713689804077148,
+      "learning_rate": 9.986885450292837e-06,
+      "loss": 0.3681,
+      "step": 2050
+    },
+    {
+      "epoch": 2.26,
+      "grad_norm": 1.208200454711914,
+      "learning_rate": 9.858283712455311e-06,
+      "loss": 0.2643,
+      "step": 2055
+    },
+    {
+      "epoch": 2.27,
+      "grad_norm": 0.957243025302887,
+      "learning_rate": 9.730108819752737e-06,
+      "loss": 0.2966,
+      "step": 2060
+    },
+    {
+      "epoch": 2.27,
+      "grad_norm": 0.9913397431373596,
+      "learning_rate": 9.602371412762371e-06,
+      "loss": 0.3375,
+      "step": 2065
+    },
+    {
+      "epoch": 2.28,
+      "grad_norm": 1.0104014873504639,
+      "learning_rate": 9.475082095743123e-06,
+      "loss": 0.3598,
+      "step": 2070
+    },
+    {
+      "epoch": 2.28,
+      "grad_norm": 0.6341925859451294,
+      "learning_rate": 9.348251435755213e-06,
+      "loss": 0.41,
+      "step": 2075
+    },
+    {
+      "epoch": 2.29,
+      "grad_norm": 0.6854486465454102,
+      "learning_rate": 9.221889961782971e-06,
+      "loss": 0.3253,
+      "step": 2080
+    },
+    {
+      "epoch": 2.3,
+      "grad_norm": 1.3789153099060059,
+      "learning_rate": 9.096008163860735e-06,
+      "loss": 0.4887,
+      "step": 2085
+    },
+    {
+      "epoch": 2.3,
+      "grad_norm": 0.9915759563446045,
+      "learning_rate": 8.97061649220201e-06,
+      "loss": 0.3067,
+      "step": 2090
+    },
+    {
+      "epoch": 2.31,
+      "grad_norm": 0.6337260603904724,
+      "learning_rate": 8.84572535633195e-06,
+      "loss": 0.4107,
+      "step": 2095
+    },
+    {
+      "epoch": 2.31,
+      "grad_norm": 1.3694593906402588,
+      "learning_rate": 8.721345124223193e-06,
+      "loss": 0.2412,
+      "step": 2100
+    },
+    {
+      "epoch": 2.32,
+      "grad_norm": 1.0332863330841064,
+      "learning_rate": 8.597486121435126e-06,
+      "loss": 0.5425,
+      "step": 2105
+    },
+    {
+      "epoch": 2.32,
+      "grad_norm": 0.7652474641799927,
+      "learning_rate": 8.474158630256755e-06,
+      "loss": 0.5445,
+      "step": 2110
+    },
+    {
+      "epoch": 2.33,
+      "grad_norm": 1.368090271949768,
+      "learning_rate": 8.351372888853045e-06,
+      "loss": 0.4415,
+      "step": 2115
+    },
+    {
+      "epoch": 2.33,
+      "grad_norm": 1.044106125831604,
+      "learning_rate": 8.229139090415035e-06,
+      "loss": 0.4247,
+      "step": 2120
+    },
+    {
+      "epoch": 2.34,
+      "grad_norm": 0.9483750462532043,
+      "learning_rate": 8.107467382313608e-06,
+      "loss": 0.3865,
+      "step": 2125
+    },
+    {
+      "epoch": 2.35,
+      "grad_norm": 0.7410837411880493,
+      "learning_rate": 7.986367865257129e-06,
+      "loss": 0.3543,
+      "step": 2130
+    },
+    {
+      "epoch": 2.35,
+      "grad_norm": 0.6833118200302124,
+      "learning_rate": 7.865850592452894e-06,
+      "loss": 0.1985,
+      "step": 2135
+    },
+    {
+      "epoch": 2.36,
+      "grad_norm": 0.7763660550117493,
+      "learning_rate": 7.745925568772554e-06,
+      "loss": 0.4385,
+      "step": 2140
+    },
+    {
+      "epoch": 2.36,
+      "grad_norm": 1.189669132232666,
+      "learning_rate": 7.626602749921589e-06,
+      "loss": 0.4026,
+      "step": 2145
+    },
+    {
+      "epoch": 2.37,
+      "grad_norm": 1.6189225912094116,
+      "learning_rate": 7.5078920416127495e-06,
+      "loss": 0.3927,
+      "step": 2150
+    },
+    {
+      "epoch": 2.37,
+      "grad_norm": 1.7551486492156982,
+      "learning_rate": 7.389803298743809e-06,
+      "loss": 0.5965,
+      "step": 2155
+    },
+    {
+      "epoch": 2.38,
+      "grad_norm": 1.2080819606781006,
+      "learning_rate": 7.272346324579388e-06,
+      "loss": 0.2134,
+      "step": 2160
+    },
+    {
+      "epoch": 2.38,
+      "grad_norm": 1.190613865852356,
+      "learning_rate": 7.1555308699371416e-06,
+      "loss": 0.4577,
+      "step": 2165
+    },
+    {
+      "epoch": 2.39,
+      "grad_norm": 1.435311198234558,
+      "learning_rate": 7.039366632378304e-06,
+      "loss": 0.2424,
+      "step": 2170
+    },
+    {
+      "epoch": 2.39,
+      "grad_norm": 0.7552112936973572,
+      "learning_rate": 6.9238632554026084e-06,
+      "loss": 0.2778,
+      "step": 2175
+    },
+    {
+      "epoch": 2.4,
+      "grad_norm": 0.6694565415382385,
+      "learning_rate": 6.809030327647728e-06,
+      "loss": 0.4163,
+      "step": 2180
+    },
+    {
+      "epoch": 2.41,
+      "grad_norm": 1.4150924682617188,
+      "learning_rate": 6.694877382093284e-06,
+      "loss": 0.278,
+      "step": 2185
+    },
+    {
+      "epoch": 2.41,
+      "grad_norm": 1.9875400066375732,
+      "learning_rate": 6.5814138952694305e-06,
+      "loss": 0.4271,
+      "step": 2190
+    },
+    {
+      "epoch": 2.42,
+      "grad_norm": 1.8704239130020142,
+      "learning_rate": 6.468649286470153e-06,
+      "loss": 0.3261,
+      "step": 2195
+    },
+    {
+      "epoch": 2.42,
+      "grad_norm": 0.6322317123413086,
+      "learning_rate": 6.356592916971343e-06,
+      "loss": 0.4293,
+      "step": 2200
+    },
+    {
+      "epoch": 2.43,
+      "grad_norm": 0.6502904891967773,
+      "learning_rate": 6.2452540892536245e-06,
+      "loss": 0.3743,
+      "step": 2205
+    },
+    {
+      "epoch": 2.43,
+      "grad_norm": 3.024371385574341,
+      "learning_rate": 6.1346420462301125e-06,
+      "loss": 0.4385,
+      "step": 2210
+    },
+    {
+      "epoch": 2.44,
+      "grad_norm": 0.7644482254981995,
+      "learning_rate": 6.024765970479122e-06,
+      "loss": 0.2991,
+      "step": 2215
+    },
+    {
+      "epoch": 2.44,
+      "grad_norm": 0.7494433522224426,
+      "learning_rate": 5.915634983481835e-06,
+      "loss": 0.4962,
+      "step": 2220
+    },
+    {
+      "epoch": 2.45,
+      "grad_norm": 2.201181411743164,
+      "learning_rate": 5.807258144865086e-06,
+      "loss": 0.4794,
+      "step": 2225
+    },
+    {
+      "epoch": 2.46,
+      "grad_norm": 0.9114490151405334,
+      "learning_rate": 5.699644451649277e-06,
+      "loss": 0.2898,
+      "step": 2230
+    },
+    {
+      "epoch": 2.46,
+      "grad_norm": 1.601459264755249,
+      "learning_rate": 5.59280283750146e-06,
+      "loss": 0.3268,
+      "step": 2235
+    },
+    {
+      "epoch": 2.47,
+      "grad_norm": 1.5998916625976562,
+      "learning_rate": 5.486742171993705e-06,
+      "loss": 0.4436,
+      "step": 2240
+    },
+    {
+      "epoch": 2.47,
+      "grad_norm": 0.8929490447044373,
+      "learning_rate": 5.381471259866807e-06,
+      "loss": 0.359,
+      "step": 2245
+    },
+    {
+      "epoch": 2.48,
+      "grad_norm": 0.9474959969520569,
+      "learning_rate": 5.276998840299308e-06,
+      "loss": 0.484,
+      "step": 2250
+    },
+    {
+      "epoch": 2.48,
+      "grad_norm": 1.7978712320327759,
+      "learning_rate": 5.1733335861820305e-06,
+      "loss": 0.2033,
+      "step": 2255
+    },
+    {
+      "epoch": 2.49,
+      "grad_norm": 0.9559624195098877,
+      "learning_rate": 5.070484103398103e-06,
+      "loss": 0.3781,
+      "step": 2260
+    },
+    {
+      "epoch": 2.49,
+      "grad_norm": 0.7897327542304993,
+      "learning_rate": 4.968458930108495e-06,
+      "loss": 0.3866,
+      "step": 2265
+    },
+    {
+      "epoch": 2.5,
+      "grad_norm": 1.8434853553771973,
+      "learning_rate": 4.867266536043235e-06,
+      "loss": 0.487,
+      "step": 2270
+    },
+    {
+      "epoch": 2.5,
+      "grad_norm": 1.959332823753357,
+      "learning_rate": 4.766915321798297e-06,
+      "loss": 0.5033,
+      "step": 2275
+    },
+    {
+      "epoch": 2.51,
+      "grad_norm": 1.3424344062805176,
+      "learning_rate": 4.667413618138192e-06,
+      "loss": 0.6175,
+      "step": 2280
+    },
+    {
+      "epoch": 2.52,
+      "grad_norm": 1.1145857572555542,
+      "learning_rate": 4.568769685304388e-06,
+      "loss": 0.2437,
+      "step": 2285
+    },
+    {
+      "epoch": 2.52,
+      "grad_norm": 1.2104226350784302,
+      "learning_rate": 4.470991712329597e-06,
+      "loss": 0.4342,
+      "step": 2290
+    },
+    {
+      "epoch": 2.53,
+      "grad_norm": 0.958635687828064,
+      "learning_rate": 4.374087816357923e-06,
+      "loss": 0.3341,
+      "step": 2295
+    },
+    {
+      "epoch": 2.53,
+      "grad_norm": 1.011942982673645,
+      "learning_rate": 4.278066041971024e-06,
+      "loss": 0.573,
+      "step": 2300
+    },
+    {
+      "epoch": 2.54,
+      "grad_norm": 0.9302467107772827,
+      "learning_rate": 4.1829343605202895e-06,
+      "loss": 0.4891,
+      "step": 2305
+    },
+    {
+      "epoch": 2.54,
+      "grad_norm": 0.598698079586029,
+      "learning_rate": 4.088700669465074e-06,
+      "loss": 0.2921,
+      "step": 2310
+    },
+    {
+      "epoch": 2.55,
+      "grad_norm": 1.0277636051177979,
+      "learning_rate": 3.995372791717092e-06,
+      "loss": 0.3742,
+      "step": 2315
+    },
+    {
+      "epoch": 2.55,
+      "grad_norm": 1.5416873693466187,
+      "learning_rate": 3.902958474990987e-06,
+      "loss": 0.2199,
+      "step": 2320
+    },
+    {
+      "epoch": 2.56,
+      "grad_norm": 1.640619158744812,
+      "learning_rate": 3.8114653911611452e-06,
+      "loss": 0.3463,
+      "step": 2325
+    },
+    {
+      "epoch": 2.57,
+      "grad_norm": 0.802915632724762,
+      "learning_rate": 3.720901135624807e-06,
+      "loss": 0.3876,
+      "step": 2330
+    },
+    {
+      "epoch": 2.57,
+      "grad_norm": 0.6372286677360535,
+      "learning_rate": 3.6312732266715374e-06,
+      "loss": 0.3851,
+      "step": 2335
+    },
+    {
+      "epoch": 2.58,
+      "grad_norm": 1.0239226818084717,
+      "learning_rate": 3.5425891048590693e-06,
+      "loss": 0.5176,
+      "step": 2340
+    },
+    {
+      "epoch": 2.58,
+      "grad_norm": 1.344281792640686,
+      "learning_rate": 3.454856132395623e-06,
+      "loss": 0.4344,
+      "step": 2345
+    },
+    {
+      "epoch": 2.59,
+      "grad_norm": 1.4836249351501465,
+      "learning_rate": 3.368081592528739e-06,
+      "loss": 0.3258,
+      "step": 2350
+    },
+    {
+      "epoch": 2.59,
+      "grad_norm": 0.48362186551094055,
+      "learning_rate": 3.2822726889406296e-06,
+      "loss": 0.3484,
+      "step": 2355
+    },
+    {
+      "epoch": 2.6,
+      "grad_norm": 1.674930214881897,
+      "learning_rate": 3.197436545150168e-06,
+      "loss": 0.2531,
+      "step": 2360
+    },
+    {
+      "epoch": 2.6,
+      "grad_norm": 1.283719539642334,
+      "learning_rate": 3.113580203921533e-06,
+      "loss": 0.4796,
+      "step": 2365
+    },
+    {
+      "epoch": 2.61,
+      "grad_norm": 1.3754234313964844,
+      "learning_rate": 3.030710626679522e-06,
+      "loss": 0.3232,
+      "step": 2370
+    },
+    {
+      "epoch": 2.61,
+      "grad_norm": 1.8028233051300049,
+      "learning_rate": 2.9488346929316546e-06,
+      "loss": 0.4161,
+      "step": 2375
+    },
+    {
+      "epoch": 2.62,
+      "grad_norm": 1.4452279806137085,
+      "learning_rate": 2.8679591996970728e-06,
+      "loss": 0.2511,
+      "step": 2380
+    },
+    {
+      "epoch": 2.63,
+      "grad_norm": 0.9447839856147766,
+      "learning_rate": 2.7880908609422563e-06,
+      "loss": 0.2869,
+      "step": 2385
+    },
+    {
+      "epoch": 2.63,
+      "grad_norm": 0.5915132164955139,
+      "learning_rate": 2.7092363070236686e-06,
+      "loss": 0.3261,
+      "step": 2390
+    },
+    {
+      "epoch": 2.64,
+      "grad_norm": 1.6882514953613281,
+      "learning_rate": 2.631402084137337e-06,
+      "loss": 0.2105,
+      "step": 2395
+    },
+    {
+      "epoch": 2.64,
+      "grad_norm": 1.2579067945480347,
+      "learning_rate": 2.5545946537753994e-06,
+      "loss": 0.2521,
+      "step": 2400
+    },
+    {
+      "epoch": 2.65,
+      "grad_norm": 2.2939579486846924,
+      "learning_rate": 2.478820392189699e-06,
+      "loss": 0.3847,
+      "step": 2405
+    },
+    {
+      "epoch": 2.65,
+      "grad_norm": 1.4380046129226685,
+      "learning_rate": 2.4040855898624677e-06,
+      "loss": 0.3619,
+      "step": 2410
+    },
+    {
+      "epoch": 2.66,
+      "grad_norm": 2.006512403488159,
+      "learning_rate": 2.3303964509840916e-06,
+      "loss": 0.5343,
+      "step": 2415
+    },
+    {
+      "epoch": 2.66,
+      "grad_norm": 0.8354169726371765,
+      "learning_rate": 2.25775909293808e-06,
+      "loss": 0.2209,
+      "step": 2420
+    },
+    {
+      "epoch": 2.67,
+      "grad_norm": 2.4486894607543945,
+      "learning_rate": 2.1861795457932236e-06,
+      "loss": 0.4199,
+      "step": 2425
+    },
+    {
+      "epoch": 2.68,
+      "grad_norm": 2.1726739406585693,
+      "learning_rate": 2.1156637518029943e-06,
+      "loss": 0.3883,
+      "step": 2430
+    },
+    {
+      "epoch": 2.68,
+      "grad_norm": 1.3366342782974243,
+      "learning_rate": 2.0462175649122407e-06,
+      "loss": 0.4508,
+      "step": 2435
+    },
+    {
+      "epoch": 2.69,
+      "grad_norm": 1.2342681884765625,
+      "learning_rate": 1.9778467502712334e-06,
+      "loss": 0.4942,
+      "step": 2440
+    },
+    {
+      "epoch": 2.69,
+      "grad_norm": 0.9087796211242676,
+      "learning_rate": 1.9105569837570457e-06,
+      "loss": 0.3828,
+      "step": 2445
+    },
+    {
+      "epoch": 2.7,
+      "grad_norm": 1.1849201917648315,
+      "learning_rate": 1.844353851502371e-06,
+      "loss": 0.3724,
+      "step": 2450
+    },
+    {
+      "epoch": 2.7,
+      "grad_norm": 1.1934826374053955,
+      "learning_rate": 1.779242849431793e-06,
+      "loss": 0.376,
+      "step": 2455
+    },
+    {
+      "epoch": 2.71,
+      "grad_norm": 1.01549232006073,
+      "learning_rate": 1.7152293828055194e-06,
+      "loss": 0.4463,
+      "step": 2460
+    },
+    {
+      "epoch": 2.71,
+      "grad_norm": 1.314704179763794,
+      "learning_rate": 1.6523187657706672e-06,
+      "loss": 0.4637,
+      "step": 2465
+    },
+    {
+      "epoch": 2.72,
+      "grad_norm": 1.545615553855896,
+      "learning_rate": 1.5905162209201135e-06,
+      "loss": 0.3411,
+      "step": 2470
+    },
+    {
+      "epoch": 2.73,
+      "grad_norm": 0.7735347747802734,
+      "learning_rate": 1.529826878858912e-06,
+      "loss": 0.3245,
+      "step": 2475
+    },
+    {
+      "epoch": 2.73,
+      "grad_norm": 1.3003946542739868,
+      "learning_rate": 1.4702557777783904e-06,
+      "loss": 0.3665,
+      "step": 2480
+    },
+    {
+      "epoch": 2.74,
+      "grad_norm": 1.4502391815185547,
+      "learning_rate": 1.4118078630378906e-06,
+      "loss": 0.3318,
+      "step": 2485
+    },
+    {
+      "epoch": 2.74,
+      "grad_norm": 0.8701452016830444,
+      "learning_rate": 1.3544879867542275e-06,
+      "loss": 0.2807,
+      "step": 2490
+    },
+    {
+      "epoch": 2.75,
+      "grad_norm": 1.2284448146820068,
+      "learning_rate": 1.2983009073988832e-06,
+      "loss": 0.3733,
+      "step": 2495
+    },
+    {
+      "epoch": 2.75,
+      "grad_norm": 0.8479854464530945,
+      "learning_rate": 1.2432512894029823e-06,
+      "loss": 0.3448,
+      "step": 2500
+    },
+    {
+      "epoch": 2.76,
+      "grad_norm": 0.9034194350242615,
+      "learning_rate": 1.1893437027700604e-06,
+      "loss": 0.3248,
+      "step": 2505
+    },
+    {
+      "epoch": 2.76,
+      "grad_norm": 0.9006950259208679,
+      "learning_rate": 1.136582622696684e-06,
+      "loss": 0.389,
+      "step": 2510
+    },
+    {
+      "epoch": 2.77,
+      "grad_norm": 0.8429726362228394,
+      "learning_rate": 1.0849724292009467e-06,
+      "loss": 0.3469,
+      "step": 2515
+    },
+    {
+      "epoch": 2.77,
+      "grad_norm": 1.1029099225997925,
+      "learning_rate": 1.034517406758842e-06,
+      "loss": 0.2766,
+      "step": 2520
+    },
+    {
+      "epoch": 2.78,
+      "grad_norm": 1.1907734870910645,
+      "learning_rate": 9.852217439485893e-07,
+      "loss": 0.341,
+      "step": 2525
+    },
+    {
+      "epoch": 2.79,
+      "grad_norm": 0.5092141628265381,
+      "learning_rate": 9.3708953310292e-07,
+      "loss": 0.6436,
+      "step": 2530
+    },
+    {
+      "epoch": 2.79,
+      "grad_norm": 1.428278923034668,
+      "learning_rate": 8.901247699693399e-07,
+      "loss": 0.2796,
+      "step": 2535
+    },
+    {
+      "epoch": 2.8,
+      "grad_norm": 0.8995442986488342,
+      "learning_rate": 8.443313533784175e-07,
+      "loss": 0.4055,
+      "step": 2540
+    },
+    {
+      "epoch": 2.8,
+      "grad_norm": 0.4836633801460266,
+      "learning_rate": 7.997130849201323e-07,
+      "loss": 0.3853,
+      "step": 2545
+    },
+    {
+      "epoch": 2.81,
+      "grad_norm": 2.5244317054748535,
+      "learning_rate": 7.56273668628264e-07,
+      "loss": 0.5473,
+      "step": 2550
+    },
+    {
+      "epoch": 2.81,
+      "grad_norm": 2.398171901702881,
+      "learning_rate": 7.140167106729017e-07,
+      "loss": 0.4997,
+      "step": 2555
+    },
+    {
+      "epoch": 2.82,
+      "grad_norm": 0.6999978423118591,
+      "learning_rate": 6.729457190610888e-07,
+      "loss": 0.3492,
+      "step": 2560
+    },
+    {
+      "epoch": 2.82,
+      "grad_norm": 1.3380895853042603,
+      "learning_rate": 6.330641033455837e-07,
+      "loss": 0.3369,
+      "step": 2565
+    },
+    {
+      "epoch": 2.83,
+      "grad_norm": 1.4348106384277344,
+      "learning_rate": 5.943751743418179e-07,
+      "loss": 0.2401,
+      "step": 2570
+    },
+    {
+      "epoch": 2.84,
+      "grad_norm": 1.0490989685058594,
+      "learning_rate": 5.568821438530519e-07,
+      "loss": 0.2856,
+      "step": 2575
+    },
+    {
+      "epoch": 2.84,
+      "grad_norm": 2.1060283184051514,
+      "learning_rate": 5.205881244037352e-07,
+      "loss": 0.2513,
+      "step": 2580
+    },
+    {
+      "epoch": 2.85,
+      "grad_norm": 1.4565911293029785,
+      "learning_rate": 4.85496128981116e-07,
+      "loss": 0.545,
+      "step": 2585
+    },
+    {
+      "epoch": 2.85,
+      "grad_norm": 0.6282710433006287,
+      "learning_rate": 4.5160907078512894e-07,
+      "loss": 0.4935,
+      "step": 2590
+    },
+    {
+      "epoch": 2.86,
+      "grad_norm": 1.2574007511138916,
+      "learning_rate": 4.189297629865335e-07,
+      "loss": 0.5524,
+      "step": 2595
+    },
+    {
+      "epoch": 2.86,
+      "grad_norm": 2.5676729679107666,
+      "learning_rate": 3.874609184933847e-07,
+      "loss": 0.3217,
+      "step": 2600
+    },
+    {
+      "epoch": 2.87,
+      "grad_norm": 1.900389552116394,
+      "learning_rate": 3.5720514972582776e-07,
+      "loss": 0.203,
+      "step": 2605
+    },
+    {
+      "epoch": 2.87,
+      "grad_norm": 1.0852973461151123,
+      "learning_rate": 3.281649683992033e-07,
+      "loss": 0.3713,
+      "step": 2610
+    },
+    {
+      "epoch": 2.88,
+      "grad_norm": 1.8293077945709229,
+      "learning_rate": 3.0034278531555105e-07,
+      "loss": 0.426,
+      "step": 2615
+    },
+    {
+      "epoch": 2.88,
+      "grad_norm": 1.1292721033096313,
+      "learning_rate": 2.737409101634747e-07,
+      "loss": 0.4578,
+      "step": 2620
+    },
+    {
+      "epoch": 2.89,
+      "grad_norm": 0.7649122476577759,
+      "learning_rate": 2.4836155132638874e-07,
+      "loss": 0.2586,
+      "step": 2625
+    },
+    {
+      "epoch": 2.9,
+      "grad_norm": 1.4662760496139526,
+      "learning_rate": 2.2420681569919998e-07,
+      "loss": 0.4307,
+      "step": 2630
+    },
+    {
+      "epoch": 2.9,
+      "grad_norm": 1.294457197189331,
+      "learning_rate": 2.012787085133927e-07,
+      "loss": 0.4343,
+      "step": 2635
+    },
+    {
+      "epoch": 2.91,
+      "grad_norm": 0.9145516157150269,
+      "learning_rate": 1.7957913317056817e-07,
+      "loss": 0.4964,
+      "step": 2640
+    },
+    {
+      "epoch": 2.91,
+      "grad_norm": 0.9060572981834412,
+      "learning_rate": 1.5910989108442632e-07,
+      "loss": 0.3721,
+      "step": 2645
+    },
+    {
+      "epoch": 2.92,
+      "grad_norm": 1.0683475732803345,
+      "learning_rate": 1.398726815312218e-07,
+      "loss": 0.5016,
+      "step": 2650
+    },
+    {
+      "epoch": 2.92,
+      "grad_norm": 0.8625506162643433,
+      "learning_rate": 1.2186910150869867e-07,
+      "loss": 0.4209,
+      "step": 2655
+    },
+    {
+      "epoch": 2.93,
+      "grad_norm": 1.0874123573303223,
+      "learning_rate": 1.0510064560350974e-07,
+      "loss": 0.44,
+      "step": 2660
+    },
+    {
+      "epoch": 2.93,
+      "grad_norm": 1.2623063325881958,
+      "learning_rate": 8.956870586714739e-08,
+      "loss": 0.3736,
+      "step": 2665
+    },
+    {
+      "epoch": 2.94,
+      "grad_norm": 1.6609416007995605,
+      "learning_rate": 7.527457170037776e-08,
+      "loss": 0.5932,
+      "step": 2670
+    },
+    {
+      "epoch": 2.95,
+      "grad_norm": 1.819159984588623,
+      "learning_rate": 6.221942974619476e-08,
+      "loss": 0.4801,
+      "step": 2675
+    },
+    {
+      "epoch": 2.95,
+      "grad_norm": 1.391512155532837,
+      "learning_rate": 5.0404363791322074e-08,
+      "loss": 0.4572,
+      "step": 2680
+    },
+    {
+      "epoch": 2.96,
+      "grad_norm": 1.40304696559906,
+      "learning_rate": 3.983035467622853e-08,
+      "loss": 0.2913,
+      "step": 2685
+    },
+    {
+      "epoch": 2.96,
+      "grad_norm": 2.047868013381958,
+      "learning_rate": 3.049828021370826e-08,
+      "loss": 0.2785,
+      "step": 2690
+    },
+    {
+      "epoch": 2.97,
+      "grad_norm": 1.6269519329071045,
+      "learning_rate": 2.2408915116008956e-08,
+      "loss": 0.1878,
+      "step": 2695
+    },
+    {
+      "epoch": 2.97,
+      "grad_norm": 1.4585676193237305,
+      "learning_rate": 1.5562930930519968e-08,
+      "loss": 0.697,
+      "step": 2700
+    },
+    {
+      "epoch": 2.98,
+      "grad_norm": 1.2848623991012573,
+      "learning_rate": 9.960895984016949e-09,
+      "loss": 0.3595,
+      "step": 2705
+    },
+    {
+      "epoch": 2.98,
+      "grad_norm": 1.122453212738037,
+      "learning_rate": 5.603275335484548e-09,
+      "loss": 0.3628,
+      "step": 2710
+    },
+    {
+      "epoch": 2.99,
+      "grad_norm": 1.7388654947280884,
+      "learning_rate": 2.490430737512317e-09,
+      "loss": 0.2751,
+      "step": 2715
+    },
+    {
+      "epoch": 2.99,
+      "grad_norm": 0.46183499693870544,
+      "learning_rate": 6.22620606258728e-10,
+      "loss": 0.357,
+      "step": 2720
+    },
+    {
+      "epoch": 3.0,
+      "step": 2724,
+      "total_flos": 1.004173813492482e+18,
+      "train_loss": 0.5588170812518586,
+      "train_runtime": 28902.9641,
+      "train_samples_per_second": 1.508,
+      "train_steps_per_second": 0.094
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 2724,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 3,
+  "save_steps": 100,
+  "total_flos": 1.004173813492482e+18,
+  "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:fe24035ec61535d0309492249f1ccf018912d6ab37208cbd8964da8cc71b783b
+size 5112