Upload folder using huggingface_hub

added_tokens.json

@@ -1,4 +1,5 @@
 {
-  "</s>": 64932,
-  "<s>": 64931
+  "<|reserved000|>": 64931,
+  "<|reserved001|>": 64932,
+  "<|reserved002|>": 64933
 }

all_results.json

@@ -0,0 +1,8 @@
+{
+    "epoch": 2.9970883209317374,
+    "total_flos": 2.5609799507175014e+18,
+    "train_loss": 0.8316220592066139,
+    "train_runtime": 17540.1657,
+    "train_samples_per_second": 6.344,
+    "train_steps_per_second": 0.099
+}

config.json

@@ -1,5 +1,5 @@
 {
-  "_name_or_path": "/notebooks/cantonese/llm/models/CantonesellmChat-v0.5",
+  "_name_or_path": "hon9kon9ize/CantoneseLLM-v0.5",
   "architectures": [
     "LlamaForCausalLM"
   ],
@@ -23,8 +23,8 @@
   "rope_scaling": null,
   "rope_theta": 5000000.0,
   "tie_word_embeddings": false,
-  "torch_dtype": "bfloat16",
-  "transformers_version": "4.41.2",
+  "torch_dtype": "float32",
+  "transformers_version": "4.42.3",
   "use_cache": false,
   "vocab_size": 64960
 }

generation_config.json

@@ -1,8 +1,7 @@
 {
   "_from_model_config": true,
   "bos_token_id": 1,
-  "bot_token_id": 6,
-  "eos_token_id": 7,
+  "eos_token_id": 2,
   "pad_token_id": 0,
-  "transformers_version": "4.41.2"
+  "transformers_version": "4.42.3"
 }

model-00001-of-00005.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8ff4b128a4d342d43910787b8330c8226f26f45fa0be2a7ec5e5bf6424cabc20
+size 4856125168

model-00002-of-00005.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5595fbedc5adc1b6caa98138f322933fa5af99dd8f67cddd3cbdb16191df831b
+size 4844657744

model-00003-of-00005.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8361fe259dcab2e71608ec0f950836f4ce5c481e9bcf708649fb3b3405e25323
+size 4844657784

model-00004-of-00005.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:26e2fc8ff1b98f64052df273bbda6f21084c4b4fba3d10e349ba9283ab88dd3a
+size 4844657784

model-00005-of-00005.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4567b90aba28e189a2d01ec4a144270c17273e23fdec87c7b5a0b05c8d659758
+size 4885534456

model.safetensors.index.json

@@ -1,298 +1,298 @@
 {
   "metadata": {
-    "total_size": 12137799680
+    "total_size": 24275599360
   },
   "weight_map": {
-    "lm_head.weight": "model-00003-of-00003.safetensors",
-    "model.embed_tokens.weight": "model-00001-of-00003.safetensors",
-    "model.layers.0.input_layernorm.weight": "model-00001-of-00003.safetensors",
-    "model.layers.0.mlp.down_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.0.mlp.gate_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.0.mlp.up_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.0.post_attention_layernorm.weight": "model-00001-of-00003.safetensors",
-    "model.layers.0.self_attn.k_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.0.self_attn.o_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.0.self_attn.q_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.0.self_attn.v_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.1.input_layernorm.weight": "model-00001-of-00003.safetensors",
-    "model.layers.1.mlp.down_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.1.mlp.gate_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.1.mlp.up_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.1.post_attention_layernorm.weight": "model-00001-of-00003.safetensors",
-    "model.layers.1.self_attn.k_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.1.self_attn.o_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.1.self_attn.q_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.1.self_attn.v_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.10.input_layernorm.weight": "model-00001-of-00003.safetensors",
-    "model.layers.10.mlp.down_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.10.mlp.gate_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.10.mlp.up_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.10.post_attention_layernorm.weight": "model-00001-of-00003.safetensors",
-    "model.layers.10.self_attn.k_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.10.self_attn.o_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.10.self_attn.q_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.10.self_attn.v_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.11.input_layernorm.weight": "model-00001-of-00003.safetensors",
-    "model.layers.11.mlp.down_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.11.mlp.gate_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.11.mlp.up_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.11.post_attention_layernorm.weight": "model-00001-of-00003.safetensors",
-    "model.layers.11.self_attn.k_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.11.self_attn.o_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.11.self_attn.q_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.11.self_attn.v_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.12.input_layernorm.weight": "model-00002-of-00003.safetensors",
-    "model.layers.12.mlp.down_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.12.mlp.gate_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.12.mlp.up_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.12.post_attention_layernorm.weight": "model-00002-of-00003.safetensors",
-    "model.layers.12.self_attn.k_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.12.self_attn.o_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.12.self_attn.q_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.12.self_attn.v_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.13.input_layernorm.weight": "model-00002-of-00003.safetensors",
-    "model.layers.13.mlp.down_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.13.mlp.gate_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.13.mlp.up_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.13.post_attention_layernorm.weight": "model-00002-of-00003.safetensors",
-    "model.layers.13.self_attn.k_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.13.self_attn.o_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.13.self_attn.q_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.13.self_attn.v_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.14.input_layernorm.weight": "model-00002-of-00003.safetensors",
-    "model.layers.14.mlp.down_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.14.mlp.gate_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.14.mlp.up_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.14.post_attention_layernorm.weight": "model-00002-of-00003.safetensors",
-    "model.layers.14.self_attn.k_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.14.self_attn.o_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.14.self_attn.q_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.14.self_attn.v_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.15.input_layernorm.weight": "model-00002-of-00003.safetensors",
-    "model.layers.15.mlp.down_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.15.mlp.gate_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.15.mlp.up_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.15.post_attention_layernorm.weight": "model-00002-of-00003.safetensors",
-    "model.layers.15.self_attn.k_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.15.self_attn.o_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.15.self_attn.q_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.15.self_attn.v_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.16.input_layernorm.weight": "model-00002-of-00003.safetensors",
-    "model.layers.16.mlp.down_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.16.mlp.gate_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.16.mlp.up_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.16.post_attention_layernorm.weight": "model-00002-of-00003.safetensors",
-    "model.layers.16.self_attn.k_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.16.self_attn.o_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.16.self_attn.q_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.16.self_attn.v_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.17.input_layernorm.weight": "model-00002-of-00003.safetensors",
-    "model.layers.17.mlp.down_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.17.mlp.gate_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.17.mlp.up_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.17.post_attention_layernorm.weight": "model-00002-of-00003.safetensors",
-    "model.layers.17.self_attn.k_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.17.self_attn.o_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.17.self_attn.q_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.17.self_attn.v_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.18.input_layernorm.weight": "model-00002-of-00003.safetensors",
-    "model.layers.18.mlp.down_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.18.mlp.gate_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.18.mlp.up_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.18.post_attention_layernorm.weight": "model-00002-of-00003.safetensors",
-    "model.layers.18.self_attn.k_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.18.self_attn.o_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.18.self_attn.q_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.18.self_attn.v_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.19.input_layernorm.weight": "model-00002-of-00003.safetensors",
-    "model.layers.19.mlp.down_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.19.mlp.gate_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.19.mlp.up_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.19.post_attention_layernorm.weight": "model-00002-of-00003.safetensors",
-    "model.layers.19.self_attn.k_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.19.self_attn.o_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.19.self_attn.q_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.19.self_attn.v_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.2.input_layernorm.weight": "model-00001-of-00003.safetensors",
-    "model.layers.2.mlp.down_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.2.mlp.gate_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.2.mlp.up_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.2.post_attention_layernorm.weight": "model-00001-of-00003.safetensors",
-    "model.layers.2.self_attn.k_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.2.self_attn.o_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.2.self_attn.q_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.2.self_attn.v_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.20.input_layernorm.weight": "model-00002-of-00003.safetensors",
-    "model.layers.20.mlp.down_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.20.mlp.gate_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.20.mlp.up_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.20.post_attention_layernorm.weight": "model-00002-of-00003.safetensors",
-    "model.layers.20.self_attn.k_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.20.self_attn.o_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.20.self_attn.q_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.20.self_attn.v_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.21.input_layernorm.weight": "model-00002-of-00003.safetensors",
-    "model.layers.21.mlp.down_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.21.mlp.gate_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.21.mlp.up_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.21.post_attention_layernorm.weight": "model-00002-of-00003.safetensors",
-    "model.layers.21.self_attn.k_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.21.self_attn.o_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.21.self_attn.q_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.21.self_attn.v_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.22.input_layernorm.weight": "model-00002-of-00003.safetensors",
-    "model.layers.22.mlp.down_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.22.mlp.gate_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.22.mlp.up_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.22.post_attention_layernorm.weight": "model-00002-of-00003.safetensors",
-    "model.layers.22.self_attn.k_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.22.self_attn.o_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.22.self_attn.q_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.22.self_attn.v_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.23.input_layernorm.weight": "model-00002-of-00003.safetensors",
-    "model.layers.23.mlp.down_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.23.mlp.gate_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.23.mlp.up_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.23.post_attention_layernorm.weight": "model-00002-of-00003.safetensors",
-    "model.layers.23.self_attn.k_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.23.self_attn.o_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.23.self_attn.q_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.23.self_attn.v_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.24.input_layernorm.weight": "model-00002-of-00003.safetensors",
-    "model.layers.24.mlp.down_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.24.mlp.gate_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.24.mlp.up_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.24.post_attention_layernorm.weight": "model-00002-of-00003.safetensors",
-    "model.layers.24.self_attn.k_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.24.self_attn.o_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.24.self_attn.q_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.24.self_attn.v_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.25.input_layernorm.weight": "model-00002-of-00003.safetensors",
-    "model.layers.25.mlp.down_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.25.mlp.gate_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.25.mlp.up_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.25.post_attention_layernorm.weight": "model-00002-of-00003.safetensors",
-    "model.layers.25.self_attn.k_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.25.self_attn.o_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.25.self_attn.q_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.25.self_attn.v_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.26.input_layernorm.weight": "model-00002-of-00003.safetensors",
-    "model.layers.26.mlp.down_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.26.mlp.gate_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.26.mlp.up_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.26.post_attention_layernorm.weight": "model-00002-of-00003.safetensors",
-    "model.layers.26.self_attn.k_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.26.self_attn.o_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.26.self_attn.q_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.26.self_attn.v_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.27.input_layernorm.weight": "model-00003-of-00003.safetensors",
-    "model.layers.27.mlp.down_proj.weight": "model-00003-of-00003.safetensors",
-    "model.layers.27.mlp.gate_proj.weight": "model-00003-of-00003.safetensors",
-    "model.layers.27.mlp.up_proj.weight": "model-00003-of-00003.safetensors",
-    "model.layers.27.post_attention_layernorm.weight": "model-00003-of-00003.safetensors",
-    "model.layers.27.self_attn.k_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.27.self_attn.o_proj.weight": "model-00003-of-00003.safetensors",
-    "model.layers.27.self_attn.q_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.27.self_attn.v_proj.weight": "model-00002-of-00003.safetensors",
-    "model.layers.28.input_layernorm.weight": "model-00003-of-00003.safetensors",
-    "model.layers.28.mlp.down_proj.weight": "model-00003-of-00003.safetensors",
-    "model.layers.28.mlp.gate_proj.weight": "model-00003-of-00003.safetensors",
-    "model.layers.28.mlp.up_proj.weight": "model-00003-of-00003.safetensors",
-    "model.layers.28.post_attention_layernorm.weight": "model-00003-of-00003.safetensors",
-    "model.layers.28.self_attn.k_proj.weight": "model-00003-of-00003.safetensors",
-    "model.layers.28.self_attn.o_proj.weight": "model-00003-of-00003.safetensors",
-    "model.layers.28.self_attn.q_proj.weight": "model-00003-of-00003.safetensors",
-    "model.layers.28.self_attn.v_proj.weight": "model-00003-of-00003.safetensors",
-    "model.layers.29.input_layernorm.weight": "model-00003-of-00003.safetensors",
-    "model.layers.29.mlp.down_proj.weight": "model-00003-of-00003.safetensors",
-    "model.layers.29.mlp.gate_proj.weight": "model-00003-of-00003.safetensors",
-    "model.layers.29.mlp.up_proj.weight": "model-00003-of-00003.safetensors",
-    "model.layers.29.post_attention_layernorm.weight": "model-00003-of-00003.safetensors",
-    "model.layers.29.self_attn.k_proj.weight": "model-00003-of-00003.safetensors",
-    "model.layers.29.self_attn.o_proj.weight": "model-00003-of-00003.safetensors",
-    "model.layers.29.self_attn.q_proj.weight": "model-00003-of-00003.safetensors",
-    "model.layers.29.self_attn.v_proj.weight": "model-00003-of-00003.safetensors",
-    "model.layers.3.input_layernorm.weight": "model-00001-of-00003.safetensors",
-    "model.layers.3.mlp.down_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.3.mlp.gate_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.3.mlp.up_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.3.post_attention_layernorm.weight": "model-00001-of-00003.safetensors",
-    "model.layers.3.self_attn.k_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.3.self_attn.o_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.3.self_attn.q_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.3.self_attn.v_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.30.input_layernorm.weight": "model-00003-of-00003.safetensors",
-    "model.layers.30.mlp.down_proj.weight": "model-00003-of-00003.safetensors",
-    "model.layers.30.mlp.gate_proj.weight": "model-00003-of-00003.safetensors",
-    "model.layers.30.mlp.up_proj.weight": "model-00003-of-00003.safetensors",
-    "model.layers.30.post_attention_layernorm.weight": "model-00003-of-00003.safetensors",
-    "model.layers.30.self_attn.k_proj.weight": "model-00003-of-00003.safetensors",
-    "model.layers.30.self_attn.o_proj.weight": "model-00003-of-00003.safetensors",
-    "model.layers.30.self_attn.q_proj.weight": "model-00003-of-00003.safetensors",
-    "model.layers.30.self_attn.v_proj.weight": "model-00003-of-00003.safetensors",
-    "model.layers.31.input_layernorm.weight": "model-00003-of-00003.safetensors",
-    "model.layers.31.mlp.down_proj.weight": "model-00003-of-00003.safetensors",
-    "model.layers.31.mlp.gate_proj.weight": "model-00003-of-00003.safetensors",
-    "model.layers.31.mlp.up_proj.weight": "model-00003-of-00003.safetensors",
-    "model.layers.31.post_attention_layernorm.weight": "model-00003-of-00003.safetensors",
-    "model.layers.31.self_attn.k_proj.weight": "model-00003-of-00003.safetensors",
-    "model.layers.31.self_attn.o_proj.weight": "model-00003-of-00003.safetensors",
-    "model.layers.31.self_attn.q_proj.weight": "model-00003-of-00003.safetensors",
-    "model.layers.31.self_attn.v_proj.weight": "model-00003-of-00003.safetensors",
-    "model.layers.4.input_layernorm.weight": "model-00001-of-00003.safetensors",
-    "model.layers.4.mlp.down_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.4.mlp.gate_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.4.mlp.up_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.4.post_attention_layernorm.weight": "model-00001-of-00003.safetensors",
-    "model.layers.4.self_attn.k_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.4.self_attn.o_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.4.self_attn.q_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.4.self_attn.v_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.5.input_layernorm.weight": "model-00001-of-00003.safetensors",
-    "model.layers.5.mlp.down_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.5.mlp.gate_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.5.mlp.up_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.5.post_attention_layernorm.weight": "model-00001-of-00003.safetensors",
-    "model.layers.5.self_attn.k_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.5.self_attn.o_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.5.self_attn.q_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.5.self_attn.v_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.6.input_layernorm.weight": "model-00001-of-00003.safetensors",
-    "model.layers.6.mlp.down_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.6.mlp.gate_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.6.mlp.up_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.6.post_attention_layernorm.weight": "model-00001-of-00003.safetensors",
-    "model.layers.6.self_attn.k_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.6.self_attn.o_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.6.self_attn.q_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.6.self_attn.v_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.7.input_layernorm.weight": "model-00001-of-00003.safetensors",
-    "model.layers.7.mlp.down_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.7.mlp.gate_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.7.mlp.up_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.7.post_attention_layernorm.weight": "model-00001-of-00003.safetensors",
-    "model.layers.7.self_attn.k_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.7.self_attn.o_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.7.self_attn.q_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.7.self_attn.v_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.8.input_layernorm.weight": "model-00001-of-00003.safetensors",
-    "model.layers.8.mlp.down_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.8.mlp.gate_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.8.mlp.up_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.8.post_attention_layernorm.weight": "model-00001-of-00003.safetensors",
-    "model.layers.8.self_attn.k_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.8.self_attn.o_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.8.self_attn.q_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.8.self_attn.v_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.9.input_layernorm.weight": "model-00001-of-00003.safetensors",
-    "model.layers.9.mlp.down_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.9.mlp.gate_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.9.mlp.up_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.9.post_attention_layernorm.weight": "model-00001-of-00003.safetensors",
-    "model.layers.9.self_attn.k_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.9.self_attn.o_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.9.self_attn.q_proj.weight": "model-00001-of-00003.safetensors",
-    "model.layers.9.self_attn.v_proj.weight": "model-00001-of-00003.safetensors",
-    "model.norm.weight": "model-00003-of-00003.safetensors"
+    "lm_head.weight": "model-00005-of-00005.safetensors",
+    "model.embed_tokens.weight": "model-00001-of-00005.safetensors",
+    "model.layers.0.input_layernorm.weight": "model-00001-of-00005.safetensors",
+    "model.layers.0.mlp.down_proj.weight": "model-00001-of-00005.safetensors",
+    "model.layers.0.mlp.gate_proj.weight": "model-00001-of-00005.safetensors",
+    "model.layers.0.mlp.up_proj.weight": "model-00001-of-00005.safetensors",
+    "model.layers.0.post_attention_layernorm.weight": "model-00001-of-00005.safetensors",
+    "model.layers.0.self_attn.k_proj.weight": "model-00001-of-00005.safetensors",
+    "model.layers.0.self_attn.o_proj.weight": "model-00001-of-00005.safetensors",
+    "model.layers.0.self_attn.q_proj.weight": "model-00001-of-00005.safetensors",
+    "model.layers.0.self_attn.v_proj.weight": "model-00001-of-00005.safetensors",
+    "model.layers.1.input_layernorm.weight": "model-00001-of-00005.safetensors",
+    "model.layers.1.mlp.down_proj.weight": "model-00001-of-00005.safetensors",
+    "model.layers.1.mlp.gate_proj.weight": "model-00001-of-00005.safetensors",
+    "model.layers.1.mlp.up_proj.weight": "model-00001-of-00005.safetensors",
+    "model.layers.1.post_attention_layernorm.weight": "model-00001-of-00005.safetensors",
+    "model.layers.1.self_attn.k_proj.weight": "model-00001-of-00005.safetensors",
+    "model.layers.1.self_attn.o_proj.weight": "model-00001-of-00005.safetensors",
+    "model.layers.1.self_attn.q_proj.weight": "model-00001-of-00005.safetensors",
+    "model.layers.1.self_attn.v_proj.weight": "model-00001-of-00005.safetensors",
+    "model.layers.10.input_layernorm.weight": "model-00002-of-00005.safetensors",
+    "model.layers.10.mlp.down_proj.weight": "model-00002-of-00005.safetensors",
+    "model.layers.10.mlp.gate_proj.weight": "model-00002-of-00005.safetensors",
+    "model.layers.10.mlp.up_proj.weight": "model-00002-of-00005.safetensors",
+    "model.layers.10.post_attention_layernorm.weight": "model-00002-of-00005.safetensors",
+    "model.layers.10.self_attn.k_proj.weight": "model-00002-of-00005.safetensors",
+    "model.layers.10.self_attn.o_proj.weight": "model-00002-of-00005.safetensors",
+    "model.layers.10.self_attn.q_proj.weight": "model-00002-of-00005.safetensors",
+    "model.layers.10.self_attn.v_proj.weight": "model-00002-of-00005.safetensors",
+    "model.layers.11.input_layernorm.weight": "model-00002-of-00005.safetensors",
+    "model.layers.11.mlp.down_proj.weight": "model-00002-of-00005.safetensors",
+    "model.layers.11.mlp.gate_proj.weight": "model-00002-of-00005.safetensors",
+    "model.layers.11.mlp.up_proj.weight": "model-00002-of-00005.safetensors",
+    "model.layers.11.post_attention_layernorm.weight": "model-00002-of-00005.safetensors",
+    "model.layers.11.self_attn.k_proj.weight": "model-00002-of-00005.safetensors",
+    "model.layers.11.self_attn.o_proj.weight": "model-00002-of-00005.safetensors",
+    "model.layers.11.self_attn.q_proj.weight": "model-00002-of-00005.safetensors",
+    "model.layers.11.self_attn.v_proj.weight": "model-00002-of-00005.safetensors",
+    "model.layers.12.input_layernorm.weight": "model-00003-of-00005.safetensors",
+    "model.layers.12.mlp.down_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.12.mlp.gate_proj.weight": "model-00002-of-00005.safetensors",
+    "model.layers.12.mlp.up_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.12.post_attention_layernorm.weight": "model-00003-of-00005.safetensors",
+    "model.layers.12.self_attn.k_proj.weight": "model-00002-of-00005.safetensors",
+    "model.layers.12.self_attn.o_proj.weight": "model-00002-of-00005.safetensors",
+    "model.layers.12.self_attn.q_proj.weight": "model-00002-of-00005.safetensors",
+    "model.layers.12.self_attn.v_proj.weight": "model-00002-of-00005.safetensors",
+    "model.layers.13.input_layernorm.weight": "model-00003-of-00005.safetensors",
+    "model.layers.13.mlp.down_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.13.mlp.gate_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.13.mlp.up_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.13.post_attention_layernorm.weight": "model-00003-of-00005.safetensors",
+    "model.layers.13.self_attn.k_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.13.self_attn.o_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.13.self_attn.q_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.13.self_attn.v_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.14.input_layernorm.weight": "model-00003-of-00005.safetensors",
+    "model.layers.14.mlp.down_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.14.mlp.gate_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.14.mlp.up_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.14.post_attention_layernorm.weight": "model-00003-of-00005.safetensors",
+    "model.layers.14.self_attn.k_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.14.self_attn.o_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.14.self_attn.q_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.14.self_attn.v_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.15.input_layernorm.weight": "model-00003-of-00005.safetensors",
+    "model.layers.15.mlp.down_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.15.mlp.gate_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.15.mlp.up_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.15.post_attention_layernorm.weight": "model-00003-of-00005.safetensors",
+    "model.layers.15.self_attn.k_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.15.self_attn.o_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.15.self_attn.q_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.15.self_attn.v_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.16.input_layernorm.weight": "model-00003-of-00005.safetensors",
+    "model.layers.16.mlp.down_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.16.mlp.gate_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.16.mlp.up_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.16.post_attention_layernorm.weight": "model-00003-of-00005.safetensors",
+    "model.layers.16.self_attn.k_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.16.self_attn.o_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.16.self_attn.q_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.16.self_attn.v_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.17.input_layernorm.weight": "model-00003-of-00005.safetensors",
+    "model.layers.17.mlp.down_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.17.mlp.gate_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.17.mlp.up_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.17.post_attention_layernorm.weight": "model-00003-of-00005.safetensors",
+    "model.layers.17.self_attn.k_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.17.self_attn.o_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.17.self_attn.q_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.17.self_attn.v_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.18.input_layernorm.weight": "model-00003-of-00005.safetensors",
+    "model.layers.18.mlp.down_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.18.mlp.gate_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.18.mlp.up_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.18.post_attention_layernorm.weight": "model-00003-of-00005.safetensors",
+    "model.layers.18.self_attn.k_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.18.self_attn.o_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.18.self_attn.q_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.18.self_attn.v_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.19.input_layernorm.weight": "model-00004-of-00005.safetensors",
+    "model.layers.19.mlp.down_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.19.mlp.gate_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.19.mlp.up_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.19.post_attention_layernorm.weight": "model-00004-of-00005.safetensors",
+    "model.layers.19.self_attn.k_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.19.self_attn.o_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.19.self_attn.q_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.19.self_attn.v_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.2.input_layernorm.weight": "model-00001-of-00005.safetensors",
+    "model.layers.2.mlp.down_proj.weight": "model-00001-of-00005.safetensors",
+    "model.layers.2.mlp.gate_proj.weight": "model-00001-of-00005.safetensors",
+    "model.layers.2.mlp.up_proj.weight": "model-00001-of-00005.safetensors",
+    "model.layers.2.post_attention_layernorm.weight": "model-00001-of-00005.safetensors",
+    "model.layers.2.self_attn.k_proj.weight": "model-00001-of-00005.safetensors",
+    "model.layers.2.self_attn.o_proj.weight": "model-00001-of-00005.safetensors",
+    "model.layers.2.self_attn.q_proj.weight": "model-00001-of-00005.safetensors",
+    "model.layers.2.self_attn.v_proj.weight": "model-00001-of-00005.safetensors",
+    "model.layers.20.input_layernorm.weight": "model-00004-of-00005.safetensors",
+    "model.layers.20.mlp.down_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.20.mlp.gate_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.20.mlp.up_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.20.post_attention_layernorm.weight": "model-00004-of-00005.safetensors",
+    "model.layers.20.self_attn.k_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.20.self_attn.o_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.20.self_attn.q_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.20.self_attn.v_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.21.input_layernorm.weight": "model-00004-of-00005.safetensors",
+    "model.layers.21.mlp.down_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.21.mlp.gate_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.21.mlp.up_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.21.post_attention_layernorm.weight": "model-00004-of-00005.safetensors",
+    "model.layers.21.self_attn.k_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.21.self_attn.o_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.21.self_attn.q_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.21.self_attn.v_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.22.input_layernorm.weight": "model-00004-of-00005.safetensors",
+    "model.layers.22.mlp.down_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.22.mlp.gate_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.22.mlp.up_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.22.post_attention_layernorm.weight": "model-00004-of-00005.safetensors",
+    "model.layers.22.self_attn.k_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.22.self_attn.o_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.22.self_attn.q_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.22.self_attn.v_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.23.input_layernorm.weight": "model-00004-of-00005.safetensors",
+    "model.layers.23.mlp.down_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.23.mlp.gate_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.23.mlp.up_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.23.post_attention_layernorm.weight": "model-00004-of-00005.safetensors",
+    "model.layers.23.self_attn.k_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.23.self_attn.o_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.23.self_attn.q_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.23.self_attn.v_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.24.input_layernorm.weight": "model-00004-of-00005.safetensors",
+    "model.layers.24.mlp.down_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.24.mlp.gate_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.24.mlp.up_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.24.post_attention_layernorm.weight": "model-00004-of-00005.safetensors",
+    "model.layers.24.self_attn.k_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.24.self_attn.o_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.24.self_attn.q_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.24.self_attn.v_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.25.input_layernorm.weight": "model-00004-of-00005.safetensors",
+    "model.layers.25.mlp.down_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.25.mlp.gate_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.25.mlp.up_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.25.post_attention_layernorm.weight": "model-00004-of-00005.safetensors",
+    "model.layers.25.self_attn.k_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.25.self_attn.o_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.25.self_attn.q_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.25.self_attn.v_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.26.input_layernorm.weight": "model-00005-of-00005.safetensors",
+    "model.layers.26.mlp.down_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.26.mlp.gate_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.26.mlp.up_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.26.post_attention_layernorm.weight": "model-00005-of-00005.safetensors",
+    "model.layers.26.self_attn.k_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.26.self_attn.o_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.26.self_attn.q_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.26.self_attn.v_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.27.input_layernorm.weight": "model-00005-of-00005.safetensors",
+    "model.layers.27.mlp.down_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.27.mlp.gate_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.27.mlp.up_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.27.post_attention_layernorm.weight": "model-00005-of-00005.safetensors",
+    "model.layers.27.self_attn.k_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.27.self_attn.o_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.27.self_attn.q_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.27.self_attn.v_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.28.input_layernorm.weight": "model-00005-of-00005.safetensors",
+    "model.layers.28.mlp.down_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.28.mlp.gate_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.28.mlp.up_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.28.post_attention_layernorm.weight": "model-00005-of-00005.safetensors",
+    "model.layers.28.self_attn.k_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.28.self_attn.o_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.28.self_attn.q_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.28.self_attn.v_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.29.input_layernorm.weight": "model-00005-of-00005.safetensors",
+    "model.layers.29.mlp.down_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.29.mlp.gate_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.29.mlp.up_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.29.post_attention_layernorm.weight": "model-00005-of-00005.safetensors",
+    "model.layers.29.self_attn.k_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.29.self_attn.o_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.29.self_attn.q_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.29.self_attn.v_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.3.input_layernorm.weight": "model-00001-of-00005.safetensors",
+    "model.layers.3.mlp.down_proj.weight": "model-00001-of-00005.safetensors",
+    "model.layers.3.mlp.gate_proj.weight": "model-00001-of-00005.safetensors",
+    "model.layers.3.mlp.up_proj.weight": "model-00001-of-00005.safetensors",
+    "model.layers.3.post_attention_layernorm.weight": "model-00001-of-00005.safetensors",
+    "model.layers.3.self_attn.k_proj.weight": "model-00001-of-00005.safetensors",
+    "model.layers.3.self_attn.o_proj.weight": "model-00001-of-00005.safetensors",
+    "model.layers.3.self_attn.q_proj.weight": "model-00001-of-00005.safetensors",
+    "model.layers.3.self_attn.v_proj.weight": "model-00001-of-00005.safetensors",
+    "model.layers.30.input_layernorm.weight": "model-00005-of-00005.safetensors",
+    "model.layers.30.mlp.down_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.30.mlp.gate_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.30.mlp.up_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.30.post_attention_layernorm.weight": "model-00005-of-00005.safetensors",
+    "model.layers.30.self_attn.k_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.30.self_attn.o_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.30.self_attn.q_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.30.self_attn.v_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.31.input_layernorm.weight": "model-00005-of-00005.safetensors",
+    "model.layers.31.mlp.down_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.31.mlp.gate_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.31.mlp.up_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.31.post_attention_layernorm.weight": "model-00005-of-00005.safetensors",
+    "model.layers.31.self_attn.k_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.31.self_attn.o_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.31.self_attn.q_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.31.self_attn.v_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.4.input_layernorm.weight": "model-00001-of-00005.safetensors",
+    "model.layers.4.mlp.down_proj.weight": "model-00001-of-00005.safetensors",
+    "model.layers.4.mlp.gate_proj.weight": "model-00001-of-00005.safetensors",
+    "model.layers.4.mlp.up_proj.weight": "model-00001-of-00005.safetensors",
+    "model.layers.4.post_attention_layernorm.weight": "model-00001-of-00005.safetensors",
+    "model.layers.4.self_attn.k_proj.weight": "model-00001-of-00005.safetensors",
+    "model.layers.4.self_attn.o_proj.weight": "model-00001-of-00005.safetensors",
+    "model.layers.4.self_attn.q_proj.weight": "model-00001-of-00005.safetensors",
+    "model.layers.4.self_attn.v_proj.weight": "model-00001-of-00005.safetensors",
+    "model.layers.5.input_layernorm.weight": "model-00002-of-00005.safetensors",
+    "model.layers.5.mlp.down_proj.weight": "model-00002-of-00005.safetensors",
+    "model.layers.5.mlp.gate_proj.weight": "model-00001-of-00005.safetensors",
+    "model.layers.5.mlp.up_proj.weight": "model-00002-of-00005.safetensors",
+    "model.layers.5.post_attention_layernorm.weight": "model-00002-of-00005.safetensors",
+    "model.layers.5.self_attn.k_proj.weight": "model-00001-of-00005.safetensors",
+    "model.layers.5.self_attn.o_proj.weight": "model-00001-of-00005.safetensors",
+    "model.layers.5.self_attn.q_proj.weight": "model-00001-of-00005.safetensors",
+    "model.layers.5.self_attn.v_proj.weight": "model-00001-of-00005.safetensors",
+    "model.layers.6.input_layernorm.weight": "model-00002-of-00005.safetensors",
+    "model.layers.6.mlp.down_proj.weight": "model-00002-of-00005.safetensors",
+    "model.layers.6.mlp.gate_proj.weight": "model-00002-of-00005.safetensors",
+    "model.layers.6.mlp.up_proj.weight": "model-00002-of-00005.safetensors",
+    "model.layers.6.post_attention_layernorm.weight": "model-00002-of-00005.safetensors",
+    "model.layers.6.self_attn.k_proj.weight": "model-00002-of-00005.safetensors",
+    "model.layers.6.self_attn.o_proj.weight": "model-00002-of-00005.safetensors",
+    "model.layers.6.self_attn.q_proj.weight": "model-00002-of-00005.safetensors",
+    "model.layers.6.self_attn.v_proj.weight": "model-00002-of-00005.safetensors",
+    "model.layers.7.input_layernorm.weight": "model-00002-of-00005.safetensors",
+    "model.layers.7.mlp.down_proj.weight": "model-00002-of-00005.safetensors",
+    "model.layers.7.mlp.gate_proj.weight": "model-00002-of-00005.safetensors",
+    "model.layers.7.mlp.up_proj.weight": "model-00002-of-00005.safetensors",
+    "model.layers.7.post_attention_layernorm.weight": "model-00002-of-00005.safetensors",
+    "model.layers.7.self_attn.k_proj.weight": "model-00002-of-00005.safetensors",
+    "model.layers.7.self_attn.o_proj.weight": "model-00002-of-00005.safetensors",
+    "model.layers.7.self_attn.q_proj.weight": "model-00002-of-00005.safetensors",
+    "model.layers.7.self_attn.v_proj.weight": "model-00002-of-00005.safetensors",
+    "model.layers.8.input_layernorm.weight": "model-00002-of-00005.safetensors",
+    "model.layers.8.mlp.down_proj.weight": "model-00002-of-00005.safetensors",
+    "model.layers.8.mlp.gate_proj.weight": "model-00002-of-00005.safetensors",
+    "model.layers.8.mlp.up_proj.weight": "model-00002-of-00005.safetensors",
+    "model.layers.8.post_attention_layernorm.weight": "model-00002-of-00005.safetensors",
+    "model.layers.8.self_attn.k_proj.weight": "model-00002-of-00005.safetensors",
+    "model.layers.8.self_attn.o_proj.weight": "model-00002-of-00005.safetensors",
+    "model.layers.8.self_attn.q_proj.weight": "model-00002-of-00005.safetensors",
+    "model.layers.8.self_attn.v_proj.weight": "model-00002-of-00005.safetensors",
+    "model.layers.9.input_layernorm.weight": "model-00002-of-00005.safetensors",
+    "model.layers.9.mlp.down_proj.weight": "model-00002-of-00005.safetensors",
+    "model.layers.9.mlp.gate_proj.weight": "model-00002-of-00005.safetensors",
+    "model.layers.9.mlp.up_proj.weight": "model-00002-of-00005.safetensors",
+    "model.layers.9.post_attention_layernorm.weight": "model-00002-of-00005.safetensors",
+    "model.layers.9.self_attn.k_proj.weight": "model-00002-of-00005.safetensors",
+    "model.layers.9.self_attn.o_proj.weight": "model-00002-of-00005.safetensors",
+    "model.layers.9.self_attn.q_proj.weight": "model-00002-of-00005.safetensors",
+    "model.layers.9.self_attn.v_proj.weight": "model-00002-of-00005.safetensors",
+    "model.norm.weight": "model-00005-of-00005.safetensors"
   }
 }

tokenizer.json

@@ -74,24 +74,6 @@
       "rstrip": false,
       "normalized": false,
       "special": true
-    },
-    {
-      "id": 64931,
-      "content": "<s>",
-      "single_word": false,
-      "lstrip": false,
-      "rstrip": false,
-      "normalized": false,
-      "special": true
-    },
-    {
-      "id": 64932,
-      "content": "</s>",
-      "single_word": false,
-      "lstrip": false,
-      "rstrip": false,
-      "normalized": false,
-      "special": true
     }
   ],
   "normalizer": {

tokenizer_config.json

@@ -1,7 +1,7 @@
 {
   "add_bos_token": false,
   "add_eos_token": false,
-  "add_prefix_space": true,
+  "add_prefix_space": null,
   "added_tokens_decoder": {
     "0": {
       "content": "<unk>",
@@ -66,22 +66,6 @@
       "rstrip": false,
       "single_word": false,
       "special": true
-    },
-    "64931": {
-      "content": "<s>",
-      "lstrip": false,
-      "normalized": false,
-      "rstrip": false,
-      "single_word": false,
-      "special": true
-    },
-    "64932": {
-      "content": "</s>",
-      "lstrip": false,
-      "normalized": false,
-      "rstrip": false,
-      "single_word": false,
-      "special": true
     }
   },
   "additional_special_tokens": [
@@ -100,7 +84,6 @@
   "pad_token": "<unk>",
   "padding_side": "right",
   "sp_model_kwargs": {},
-  "spaces_between_special_tokens": false,
   "split_special_tokens": false,
   "tokenizer_class": "LlamaTokenizer",
   "unk_token": "<unk>",

train_results.json

@@ -0,0 +1,8 @@
+{
+    "epoch": 2.9970883209317374,
+    "total_flos": 2.5609799507175014e+18,
+    "train_loss": 0.8316220592066139,
+    "train_runtime": 17540.1657,
+    "train_samples_per_second": 6.344,
+    "train_steps_per_second": 0.099
+}

trainer_log.jsonl

@@ -0,0 +1,348 @@
+{"current_steps": 5, "total_steps": 1737, "loss": 1.3814, "learning_rate": 1e-05, "epoch": 0.008627197239296884, "percentage": 0.29, "elapsed_time": "0:00:54", "remaining_time": "5:16:42", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 10, "total_steps": 1737, "loss": 1.3848, "learning_rate": 1e-05, "epoch": 0.017254394478593768, "percentage": 0.58, "elapsed_time": "0:01:39", "remaining_time": "4:45:57", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 15, "total_steps": 1737, "loss": 1.3517, "learning_rate": 1e-05, "epoch": 0.02588159171789065, "percentage": 0.86, "elapsed_time": "0:02:28", "remaining_time": "4:43:15", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 20, "total_steps": 1737, "loss": 1.4088, "learning_rate": 1e-05, "epoch": 0.034508788957187536, "percentage": 1.15, "elapsed_time": "0:03:11", "remaining_time": "4:34:06", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 25, "total_steps": 1737, "loss": 1.3542, "learning_rate": 1e-05, "epoch": 0.04313598619648442, "percentage": 1.44, "elapsed_time": "0:03:55", "remaining_time": "4:28:52", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 30, "total_steps": 1737, "loss": 1.3434, "learning_rate": 1e-05, "epoch": 0.0517631834357813, "percentage": 1.73, "elapsed_time": "0:04:45", "remaining_time": "4:30:54", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 35, "total_steps": 1737, "loss": 1.3926, "learning_rate": 1e-05, "epoch": 0.06039038067507818, "percentage": 2.01, "elapsed_time": "0:05:34", "remaining_time": "4:30:48", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 40, "total_steps": 1737, "loss": 1.2819, "learning_rate": 1e-05, "epoch": 0.06901757791437507, "percentage": 2.3, "elapsed_time": "0:06:26", "remaining_time": "4:33:18", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 45, "total_steps": 1737, "loss": 1.3482, "learning_rate": 1e-05, "epoch": 0.07764477515367195, "percentage": 2.59, "elapsed_time": "0:07:22", "remaining_time": "4:37:16", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 50, "total_steps": 1737, "loss": 1.2621, "learning_rate": 1e-05, "epoch": 0.08627197239296884, "percentage": 2.88, "elapsed_time": "0:08:09", "remaining_time": "4:35:14", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 55, "total_steps": 1737, "loss": 1.2695, "learning_rate": 1e-05, "epoch": 0.09489916963226572, "percentage": 3.17, "elapsed_time": "0:08:54", "remaining_time": "4:32:34", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 60, "total_steps": 1737, "loss": 1.2302, "learning_rate": 1e-05, "epoch": 0.1035263668715626, "percentage": 3.45, "elapsed_time": "0:09:45", "remaining_time": "4:32:50", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 65, "total_steps": 1737, "loss": 1.242, "learning_rate": 1e-05, "epoch": 0.11215356411085949, "percentage": 3.74, "elapsed_time": "0:10:34", "remaining_time": "4:31:55", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 70, "total_steps": 1737, "loss": 1.2433, "learning_rate": 1e-05, "epoch": 0.12078076135015636, "percentage": 4.03, "elapsed_time": "0:11:20", "remaining_time": "4:30:00", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 75, "total_steps": 1737, "loss": 1.3095, "learning_rate": 1e-05, "epoch": 0.12940795858945325, "percentage": 4.32, "elapsed_time": "0:12:03", "remaining_time": "4:27:19", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 80, "total_steps": 1737, "loss": 1.3006, "learning_rate": 1e-05, "epoch": 0.13803515582875014, "percentage": 4.61, "elapsed_time": "0:12:50", "remaining_time": "4:25:57", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 85, "total_steps": 1737, "loss": 1.2698, "learning_rate": 1e-05, "epoch": 0.146662353068047, "percentage": 4.89, "elapsed_time": "0:13:36", "remaining_time": "4:24:24", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 90, "total_steps": 1737, "loss": 1.3346, "learning_rate": 1e-05, "epoch": 0.1552895503073439, "percentage": 5.18, "elapsed_time": "0:14:23", "remaining_time": "4:23:26", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 95, "total_steps": 1737, "loss": 1.3634, "learning_rate": 1e-05, "epoch": 0.1639167475466408, "percentage": 5.47, "elapsed_time": "0:15:09", "remaining_time": "4:22:06", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 100, "total_steps": 1737, "loss": 1.2645, "learning_rate": 1e-05, "epoch": 0.17254394478593768, "percentage": 5.76, "elapsed_time": "0:15:58", "remaining_time": "4:21:31", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 105, "total_steps": 1737, "loss": 1.3625, "learning_rate": 1e-05, "epoch": 0.18117114202523454, "percentage": 6.04, "elapsed_time": "0:16:46", "remaining_time": "4:20:41", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 110, "total_steps": 1737, "loss": 1.2592, "learning_rate": 1e-05, "epoch": 0.18979833926453143, "percentage": 6.33, "elapsed_time": "0:17:31", "remaining_time": "4:19:10", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 115, "total_steps": 1737, "loss": 1.2872, "learning_rate": 1e-05, "epoch": 0.19842553650382833, "percentage": 6.62, "elapsed_time": "0:18:21", "remaining_time": "4:18:50", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 120, "total_steps": 1737, "loss": 1.31, "learning_rate": 1e-05, "epoch": 0.2070527337431252, "percentage": 6.91, "elapsed_time": "0:19:11", "remaining_time": "4:18:33", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 125, "total_steps": 1737, "loss": 1.2627, "learning_rate": 1e-05, "epoch": 0.21567993098242208, "percentage": 7.2, "elapsed_time": "0:20:00", "remaining_time": "4:17:58", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 130, "total_steps": 1737, "loss": 1.206, "learning_rate": 1e-05, "epoch": 0.22430712822171897, "percentage": 7.48, "elapsed_time": "0:20:51", "remaining_time": "4:17:54", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 135, "total_steps": 1737, "loss": 1.3192, "learning_rate": 1e-05, "epoch": 0.23293432546101586, "percentage": 7.77, "elapsed_time": "0:21:39", "remaining_time": "4:17:02", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 140, "total_steps": 1737, "loss": 1.2825, "learning_rate": 1e-05, "epoch": 0.24156152270031273, "percentage": 8.06, "elapsed_time": "0:22:26", "remaining_time": "4:16:01", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 145, "total_steps": 1737, "loss": 1.3421, "learning_rate": 1e-05, "epoch": 0.2501887199396096, "percentage": 8.35, "elapsed_time": "0:23:14", "remaining_time": "4:15:05", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 150, "total_steps": 1737, "loss": 1.3141, "learning_rate": 1e-05, "epoch": 0.2588159171789065, "percentage": 8.64, "elapsed_time": "0:24:06", "remaining_time": "4:15:02", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 155, "total_steps": 1737, "loss": 1.298, "learning_rate": 1e-05, "epoch": 0.2674431144182034, "percentage": 8.92, "elapsed_time": "0:24:53", "remaining_time": "4:14:08", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 160, "total_steps": 1737, "loss": 1.3034, "learning_rate": 1e-05, "epoch": 0.2760703116575003, "percentage": 9.21, "elapsed_time": "0:25:48", "remaining_time": "4:14:26", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 165, "total_steps": 1737, "loss": 1.2532, "learning_rate": 1e-05, "epoch": 0.2846975088967972, "percentage": 9.5, "elapsed_time": "0:26:34", "remaining_time": "4:13:13", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 170, "total_steps": 1737, "loss": 1.2745, "learning_rate": 1e-05, "epoch": 0.293324706136094, "percentage": 9.79, "elapsed_time": "0:27:28", "remaining_time": "4:13:13", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 175, "total_steps": 1737, "loss": 1.3067, "learning_rate": 1e-05, "epoch": 0.3019519033753909, "percentage": 10.07, "elapsed_time": "0:28:10", "remaining_time": "4:11:33", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 180, "total_steps": 1737, "loss": 1.2628, "learning_rate": 1e-05, "epoch": 0.3105791006146878, "percentage": 10.36, "elapsed_time": "0:29:08", "remaining_time": "4:12:04", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 185, "total_steps": 1737, "loss": 1.2514, "learning_rate": 1e-05, "epoch": 0.3192062978539847, "percentage": 10.65, "elapsed_time": "0:30:07", "remaining_time": "4:12:40", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 190, "total_steps": 1737, "loss": 1.2122, "learning_rate": 1e-05, "epoch": 0.3278334950932816, "percentage": 10.94, "elapsed_time": "0:30:59", "remaining_time": "4:12:16", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 195, "total_steps": 1737, "loss": 1.2825, "learning_rate": 1e-05, "epoch": 0.33646069233257847, "percentage": 11.23, "elapsed_time": "0:31:41", "remaining_time": "4:10:39", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 200, "total_steps": 1737, "loss": 1.2752, "learning_rate": 1e-05, "epoch": 0.34508788957187536, "percentage": 11.51, "elapsed_time": "0:32:34", "remaining_time": "4:10:22", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 205, "total_steps": 1737, "loss": 1.3145, "learning_rate": 1e-05, "epoch": 0.3537150868111722, "percentage": 11.8, "elapsed_time": "0:34:44", "remaining_time": "4:19:40", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 210, "total_steps": 1737, "loss": 1.2694, "learning_rate": 1e-05, "epoch": 0.3623422840504691, "percentage": 12.09, "elapsed_time": "0:35:25", "remaining_time": "4:17:36", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 215, "total_steps": 1737, "loss": 1.3507, "learning_rate": 1e-05, "epoch": 0.370969481289766, "percentage": 12.38, "elapsed_time": "0:36:21", "remaining_time": "4:17:23", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 220, "total_steps": 1737, "loss": 1.3843, "learning_rate": 1e-05, "epoch": 0.37959667852906287, "percentage": 12.67, "elapsed_time": "0:37:12", "remaining_time": "4:16:31", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 225, "total_steps": 1737, "loss": 1.2553, "learning_rate": 1e-05, "epoch": 0.38822387576835976, "percentage": 12.95, "elapsed_time": "0:37:56", "remaining_time": "4:14:58", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 230, "total_steps": 1737, "loss": 1.3664, "learning_rate": 1e-05, "epoch": 0.39685107300765665, "percentage": 13.24, "elapsed_time": "0:38:47", "remaining_time": "4:14:09", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 235, "total_steps": 1737, "loss": 1.3083, "learning_rate": 1e-05, "epoch": 0.40547827024695354, "percentage": 13.53, "elapsed_time": "0:39:35", "remaining_time": "4:13:00", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 240, "total_steps": 1737, "loss": 1.2945, "learning_rate": 1e-05, "epoch": 0.4141054674862504, "percentage": 13.82, "elapsed_time": "0:40:22", "remaining_time": "4:11:48", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 245, "total_steps": 1737, "loss": 1.2938, "learning_rate": 1e-05, "epoch": 0.42273266472554727, "percentage": 14.1, "elapsed_time": "0:41:12", "remaining_time": "4:10:57", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 250, "total_steps": 1737, "loss": 1.2231, "learning_rate": 1e-05, "epoch": 0.43135986196484416, "percentage": 14.39, "elapsed_time": "0:42:06", "remaining_time": "4:10:28", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 255, "total_steps": 1737, "loss": 1.2237, "learning_rate": 1e-05, "epoch": 0.43998705920414105, "percentage": 14.68, "elapsed_time": "0:42:58", "remaining_time": "4:09:43", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 260, "total_steps": 1737, "loss": 1.1541, "learning_rate": 1e-05, "epoch": 0.44861425644343794, "percentage": 14.97, "elapsed_time": "0:43:44", "remaining_time": "4:08:27", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 265, "total_steps": 1737, "loss": 1.2629, "learning_rate": 1e-05, "epoch": 0.45724145368273483, "percentage": 15.26, "elapsed_time": "0:44:30", "remaining_time": "4:07:14", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 270, "total_steps": 1737, "loss": 1.1911, "learning_rate": 1e-05, "epoch": 0.4658686509220317, "percentage": 15.54, "elapsed_time": "0:45:17", "remaining_time": "4:06:07", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 275, "total_steps": 1737, "loss": 1.4012, "learning_rate": 1e-05, "epoch": 0.4744958481613286, "percentage": 15.83, "elapsed_time": "0:46:03", "remaining_time": "4:04:51", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 280, "total_steps": 1737, "loss": 1.2188, "learning_rate": 1e-05, "epoch": 0.48312304540062545, "percentage": 16.12, "elapsed_time": "0:46:53", "remaining_time": "4:03:58", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 285, "total_steps": 1737, "loss": 1.354, "learning_rate": 1e-05, "epoch": 0.49175024263992234, "percentage": 16.41, "elapsed_time": "0:47:38", "remaining_time": "4:02:44", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 290, "total_steps": 1737, "loss": 1.3322, "learning_rate": 1e-05, "epoch": 0.5003774398792192, "percentage": 16.7, "elapsed_time": "0:48:23", "remaining_time": "4:01:29", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 295, "total_steps": 1737, "loss": 1.3385, "learning_rate": 1e-05, "epoch": 0.5090046371185161, "percentage": 16.98, "elapsed_time": "0:49:15", "remaining_time": "4:00:46", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 300, "total_steps": 1737, "loss": 1.2541, "learning_rate": 1e-05, "epoch": 0.517631834357813, "percentage": 17.27, "elapsed_time": "0:50:00", "remaining_time": "3:59:30", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 305, "total_steps": 1737, "loss": 1.3451, "learning_rate": 1e-05, "epoch": 0.5262590315971098, "percentage": 17.56, "elapsed_time": "0:50:52", "remaining_time": "3:58:53", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 310, "total_steps": 1737, "loss": 1.3269, "learning_rate": 1e-05, "epoch": 0.5348862288364068, "percentage": 17.85, "elapsed_time": "0:51:43", "remaining_time": "3:58:07", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 315, "total_steps": 1737, "loss": 1.2654, "learning_rate": 1e-05, "epoch": 0.5435134260757036, "percentage": 18.13, "elapsed_time": "0:52:34", "remaining_time": "3:57:18", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 320, "total_steps": 1737, "loss": 1.3162, "learning_rate": 1e-05, "epoch": 0.5521406233150006, "percentage": 18.42, "elapsed_time": "0:53:21", "remaining_time": "3:56:17", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 325, "total_steps": 1737, "loss": 1.2839, "learning_rate": 1e-05, "epoch": 0.5607678205542974, "percentage": 18.71, "elapsed_time": "0:54:14", "remaining_time": "3:55:40", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 330, "total_steps": 1737, "loss": 1.3095, "learning_rate": 1e-05, "epoch": 0.5693950177935944, "percentage": 19.0, "elapsed_time": "0:55:04", "remaining_time": "3:54:47", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 335, "total_steps": 1737, "loss": 1.285, "learning_rate": 1e-05, "epoch": 0.5780222150328912, "percentage": 19.29, "elapsed_time": "0:55:50", "remaining_time": "3:53:42", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 340, "total_steps": 1737, "loss": 1.2419, "learning_rate": 1e-05, "epoch": 0.586649412272188, "percentage": 19.57, "elapsed_time": "0:56:35", "remaining_time": "3:52:30", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 345, "total_steps": 1737, "loss": 1.2875, "learning_rate": 1e-05, "epoch": 0.595276609511485, "percentage": 19.86, "elapsed_time": "0:57:22", "remaining_time": "3:51:29", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 350, "total_steps": 1737, "loss": 1.3055, "learning_rate": 1e-05, "epoch": 0.6039038067507818, "percentage": 20.15, "elapsed_time": "0:58:09", "remaining_time": "3:50:29", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 355, "total_steps": 1737, "loss": 1.33, "learning_rate": 1e-05, "epoch": 0.6125310039900788, "percentage": 20.44, "elapsed_time": "0:58:56", "remaining_time": "3:49:27", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 360, "total_steps": 1737, "loss": 1.2003, "learning_rate": 1e-05, "epoch": 0.6211582012293756, "percentage": 20.73, "elapsed_time": "0:59:44", "remaining_time": "3:48:30", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 365, "total_steps": 1737, "loss": 1.2385, "learning_rate": 1e-05, "epoch": 0.6297853984686725, "percentage": 21.01, "elapsed_time": "1:00:33", "remaining_time": "3:47:37", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 370, "total_steps": 1737, "loss": 1.2198, "learning_rate": 1e-05, "epoch": 0.6384125957079694, "percentage": 21.3, "elapsed_time": "1:01:27", "remaining_time": "3:47:04", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 375, "total_steps": 1737, "loss": 1.3135, "learning_rate": 1e-05, "epoch": 0.6470397929472662, "percentage": 21.59, "elapsed_time": "1:02:12", "remaining_time": "3:45:55", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 380, "total_steps": 1737, "loss": 1.3234, "learning_rate": 1e-05, "epoch": 0.6556669901865632, "percentage": 21.88, "elapsed_time": "1:02:57", "remaining_time": "3:44:51", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 385, "total_steps": 1737, "loss": 1.3376, "learning_rate": 1e-05, "epoch": 0.66429418742586, "percentage": 22.16, "elapsed_time": "1:03:47", "remaining_time": "3:44:00", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 390, "total_steps": 1737, "loss": 1.2237, "learning_rate": 1e-05, "epoch": 0.6729213846651569, "percentage": 22.45, "elapsed_time": "1:04:35", "remaining_time": "3:43:04", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 395, "total_steps": 1737, "loss": 1.2465, "learning_rate": 1e-05, "epoch": 0.6815485819044538, "percentage": 22.74, "elapsed_time": "1:05:24", "remaining_time": "3:42:13", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 400, "total_steps": 1737, "loss": 1.2188, "learning_rate": 1e-05, "epoch": 0.6901757791437507, "percentage": 23.03, "elapsed_time": "1:06:12", "remaining_time": "3:41:16", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 405, "total_steps": 1737, "loss": 1.2663, "learning_rate": 1e-05, "epoch": 0.6988029763830476, "percentage": 23.32, "elapsed_time": "1:08:26", "remaining_time": "3:45:04", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 410, "total_steps": 1737, "loss": 1.3354, "learning_rate": 1e-05, "epoch": 0.7074301736223444, "percentage": 23.6, "elapsed_time": "1:09:13", "remaining_time": "3:44:01", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 415, "total_steps": 1737, "loss": 1.3112, "learning_rate": 1e-05, "epoch": 0.7160573708616413, "percentage": 23.89, "elapsed_time": "1:10:01", "remaining_time": "3:43:03", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 420, "total_steps": 1737, "loss": 1.2335, "learning_rate": 1e-05, "epoch": 0.7246845681009382, "percentage": 24.18, "elapsed_time": "1:10:49", "remaining_time": "3:42:04", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 425, "total_steps": 1737, "loss": 1.2291, "learning_rate": 1e-05, "epoch": 0.7333117653402351, "percentage": 24.47, "elapsed_time": "1:11:34", "remaining_time": "3:40:56", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 430, "total_steps": 1737, "loss": 1.1676, "learning_rate": 1e-05, "epoch": 0.741938962579532, "percentage": 24.76, "elapsed_time": "1:12:17", "remaining_time": "3:39:44", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 435, "total_steps": 1737, "loss": 1.2539, "learning_rate": 1e-05, "epoch": 0.7505661598188289, "percentage": 25.04, "elapsed_time": "1:13:11", "remaining_time": "3:39:04", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 440, "total_steps": 1737, "loss": 1.2994, "learning_rate": 1e-05, "epoch": 0.7591933570581257, "percentage": 25.33, "elapsed_time": "1:13:59", "remaining_time": "3:38:07", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 445, "total_steps": 1737, "loss": 1.2304, "learning_rate": 1e-05, "epoch": 0.7678205542974226, "percentage": 25.62, "elapsed_time": "1:14:44", "remaining_time": "3:36:59", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 450, "total_steps": 1737, "loss": 1.2871, "learning_rate": 1e-05, "epoch": 0.7764477515367195, "percentage": 25.91, "elapsed_time": "1:15:32", "remaining_time": "3:36:02", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 455, "total_steps": 1737, "loss": 1.3053, "learning_rate": 1e-05, "epoch": 0.7850749487760164, "percentage": 26.19, "elapsed_time": "1:16:22", "remaining_time": "3:35:12", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 460, "total_steps": 1737, "loss": 1.2691, "learning_rate": 1e-05, "epoch": 0.7937021460153133, "percentage": 26.48, "elapsed_time": "1:17:11", "remaining_time": "3:34:16", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 465, "total_steps": 1737, "loss": 1.2744, "learning_rate": 1e-05, "epoch": 0.8023293432546101, "percentage": 26.77, "elapsed_time": "1:17:54", "remaining_time": "3:33:07", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 470, "total_steps": 1737, "loss": 1.2016, "learning_rate": 1e-05, "epoch": 0.8109565404939071, "percentage": 27.06, "elapsed_time": "1:18:42", "remaining_time": "3:32:10", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 475, "total_steps": 1737, "loss": 1.2787, "learning_rate": 1e-05, "epoch": 0.8195837377332039, "percentage": 27.35, "elapsed_time": "1:19:30", "remaining_time": "3:31:13", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 480, "total_steps": 1737, "loss": 1.3095, "learning_rate": 1e-05, "epoch": 0.8282109349725008, "percentage": 27.63, "elapsed_time": "1:20:16", "remaining_time": "3:30:13", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 485, "total_steps": 1737, "loss": 1.2503, "learning_rate": 1e-05, "epoch": 0.8368381322117977, "percentage": 27.92, "elapsed_time": "1:21:05", "remaining_time": "3:29:20", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 490, "total_steps": 1737, "loss": 1.2598, "learning_rate": 1e-05, "epoch": 0.8454653294510945, "percentage": 28.21, "elapsed_time": "1:21:50", "remaining_time": "3:28:17", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 495, "total_steps": 1737, "loss": 1.2914, "learning_rate": 1e-05, "epoch": 0.8540925266903915, "percentage": 28.5, "elapsed_time": "1:22:40", "remaining_time": "3:27:26", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 500, "total_steps": 1737, "loss": 1.2744, "learning_rate": 1e-05, "epoch": 0.8627197239296883, "percentage": 28.79, "elapsed_time": "1:23:25", "remaining_time": "3:26:23", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 505, "total_steps": 1737, "loss": 1.3817, "learning_rate": 1e-05, "epoch": 0.8713469211689853, "percentage": 29.07, "elapsed_time": "1:24:12", "remaining_time": "3:25:26", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 510, "total_steps": 1737, "loss": 1.1628, "learning_rate": 1e-05, "epoch": 0.8799741184082821, "percentage": 29.36, "elapsed_time": "1:25:01", "remaining_time": "3:24:33", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 515, "total_steps": 1737, "loss": 1.2602, "learning_rate": 1e-05, "epoch": 0.8886013156475789, "percentage": 29.65, "elapsed_time": "1:25:46", "remaining_time": "3:23:31", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 520, "total_steps": 1737, "loss": 1.2036, "learning_rate": 1e-05, "epoch": 0.8972285128868759, "percentage": 29.94, "elapsed_time": "1:26:37", "remaining_time": "3:22:44", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 525, "total_steps": 1737, "loss": 1.2575, "learning_rate": 1e-05, "epoch": 0.9058557101261727, "percentage": 30.22, "elapsed_time": "1:27:32", "remaining_time": "3:22:05", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 530, "total_steps": 1737, "loss": 1.2762, "learning_rate": 1e-05, "epoch": 0.9144829073654697, "percentage": 30.51, "elapsed_time": "1:28:22", "remaining_time": "3:21:16", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 535, "total_steps": 1737, "loss": 1.299, "learning_rate": 1e-05, "epoch": 0.9231101046047665, "percentage": 30.8, "elapsed_time": "1:29:09", "remaining_time": "3:20:18", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 540, "total_steps": 1737, "loss": 1.2246, "learning_rate": 1e-05, "epoch": 0.9317373018440634, "percentage": 31.09, "elapsed_time": "1:29:54", "remaining_time": "3:19:17", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 545, "total_steps": 1737, "loss": 1.2069, "learning_rate": 1e-05, "epoch": 0.9403644990833603, "percentage": 31.38, "elapsed_time": "1:30:41", "remaining_time": "3:18:20", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 550, "total_steps": 1737, "loss": 1.244, "learning_rate": 1e-05, "epoch": 0.9489916963226572, "percentage": 31.66, "elapsed_time": "1:31:26", "remaining_time": "3:17:19", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 555, "total_steps": 1737, "loss": 1.2785, "learning_rate": 1e-05, "epoch": 0.9576188935619541, "percentage": 31.95, "elapsed_time": "1:32:12", "remaining_time": "3:16:21", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 560, "total_steps": 1737, "loss": 1.2542, "learning_rate": 1e-05, "epoch": 0.9662460908012509, "percentage": 32.24, "elapsed_time": "1:32:59", "remaining_time": "3:15:26", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 565, "total_steps": 1737, "loss": 1.2649, "learning_rate": 1e-05, "epoch": 0.9748732880405478, "percentage": 32.53, "elapsed_time": "1:33:42", "remaining_time": "3:14:23", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 570, "total_steps": 1737, "loss": 1.2963, "learning_rate": 1e-05, "epoch": 0.9835004852798447, "percentage": 32.82, "elapsed_time": "1:34:28", "remaining_time": "3:13:25", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 575, "total_steps": 1737, "loss": 1.2534, "learning_rate": 1e-05, "epoch": 0.9921276825191416, "percentage": 33.1, "elapsed_time": "1:35:14", "remaining_time": "3:12:28", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 580, "total_steps": 1737, "loss": 1.1385, "learning_rate": 1e-05, "epoch": 1.0007548797584385, "percentage": 33.39, "elapsed_time": "1:36:02", "remaining_time": "3:11:36", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 585, "total_steps": 1737, "loss": 0.8563, "learning_rate": 1e-05, "epoch": 1.0093820769977353, "percentage": 33.68, "elapsed_time": "1:36:52", "remaining_time": "3:10:46", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 590, "total_steps": 1737, "loss": 0.9018, "learning_rate": 1e-05, "epoch": 1.0180092742370321, "percentage": 33.97, "elapsed_time": "1:37:45", "remaining_time": "3:10:03", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 595, "total_steps": 1737, "loss": 0.8435, "learning_rate": 1e-05, "epoch": 1.0266364714763292, "percentage": 34.25, "elapsed_time": "1:38:36", "remaining_time": "3:09:14", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 600, "total_steps": 1737, "loss": 0.7686, "learning_rate": 1e-05, "epoch": 1.035263668715626, "percentage": 34.54, "elapsed_time": "1:39:21", "remaining_time": "3:08:16", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 605, "total_steps": 1737, "loss": 0.7134, "learning_rate": 1e-05, "epoch": 1.0438908659549229, "percentage": 34.83, "elapsed_time": "1:41:33", "remaining_time": "3:10:00", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 610, "total_steps": 1737, "loss": 0.7836, "learning_rate": 1e-05, "epoch": 1.0525180631942197, "percentage": 35.12, "elapsed_time": "1:42:23", "remaining_time": "3:09:09", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 615, "total_steps": 1737, "loss": 0.8263, "learning_rate": 1e-05, "epoch": 1.0611452604335168, "percentage": 35.41, "elapsed_time": "1:43:06", "remaining_time": "3:08:06", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 620, "total_steps": 1737, "loss": 0.8028, "learning_rate": 1e-05, "epoch": 1.0697724576728136, "percentage": 35.69, "elapsed_time": "1:43:51", "remaining_time": "3:07:06", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 625, "total_steps": 1737, "loss": 0.7552, "learning_rate": 1e-05, "epoch": 1.0783996549121104, "percentage": 35.98, "elapsed_time": "1:44:33", "remaining_time": "3:06:02", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 630, "total_steps": 1737, "loss": 0.838, "learning_rate": 1e-05, "epoch": 1.0870268521514073, "percentage": 36.27, "elapsed_time": "1:45:17", "remaining_time": "3:05:00", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 635, "total_steps": 1737, "loss": 0.7915, "learning_rate": 1e-05, "epoch": 1.095654049390704, "percentage": 36.56, "elapsed_time": "1:46:04", "remaining_time": "3:04:04", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 640, "total_steps": 1737, "loss": 0.7522, "learning_rate": 1e-05, "epoch": 1.1042812466300012, "percentage": 36.85, "elapsed_time": "1:46:53", "remaining_time": "3:03:13", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 645, "total_steps": 1737, "loss": 0.7777, "learning_rate": 1e-05, "epoch": 1.112908443869298, "percentage": 37.13, "elapsed_time": "1:47:46", "remaining_time": "3:02:28", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 650, "total_steps": 1737, "loss": 0.8388, "learning_rate": 1e-05, "epoch": 1.1215356411085948, "percentage": 37.42, "elapsed_time": "1:48:35", "remaining_time": "3:01:36", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 655, "total_steps": 1737, "loss": 0.8455, "learning_rate": 1e-05, "epoch": 1.1301628383478917, "percentage": 37.71, "elapsed_time": "1:49:24", "remaining_time": "3:00:44", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 660, "total_steps": 1737, "loss": 0.8048, "learning_rate": 1e-05, "epoch": 1.1387900355871885, "percentage": 38.0, "elapsed_time": "1:50:07", "remaining_time": "2:59:42", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 665, "total_steps": 1737, "loss": 0.7762, "learning_rate": 1e-05, "epoch": 1.1474172328264856, "percentage": 38.28, "elapsed_time": "1:51:07", "remaining_time": "2:59:08", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 670, "total_steps": 1737, "loss": 0.7812, "learning_rate": 1e-05, "epoch": 1.1560444300657824, "percentage": 38.57, "elapsed_time": "1:51:59", "remaining_time": "2:58:21", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 675, "total_steps": 1737, "loss": 0.8134, "learning_rate": 1e-05, "epoch": 1.1646716273050792, "percentage": 38.86, "elapsed_time": "1:52:51", "remaining_time": "2:57:34", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 680, "total_steps": 1737, "loss": 0.7821, "learning_rate": 1e-05, "epoch": 1.173298824544376, "percentage": 39.15, "elapsed_time": "1:53:43", "remaining_time": "2:56:46", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 685, "total_steps": 1737, "loss": 0.8034, "learning_rate": 1e-05, "epoch": 1.1819260217836731, "percentage": 39.44, "elapsed_time": "1:54:39", "remaining_time": "2:56:04", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 690, "total_steps": 1737, "loss": 0.7592, "learning_rate": 1e-05, "epoch": 1.19055321902297, "percentage": 39.72, "elapsed_time": "1:55:26", "remaining_time": "2:55:10", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 695, "total_steps": 1737, "loss": 0.7738, "learning_rate": 1e-05, "epoch": 1.1991804162622668, "percentage": 40.01, "elapsed_time": "1:56:16", "remaining_time": "2:54:20", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 700, "total_steps": 1737, "loss": 0.7043, "learning_rate": 1e-05, "epoch": 1.2078076135015636, "percentage": 40.3, "elapsed_time": "1:57:13", "remaining_time": "2:53:40", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 705, "total_steps": 1737, "loss": 0.7969, "learning_rate": 1e-05, "epoch": 1.2164348107408607, "percentage": 40.59, "elapsed_time": "1:57:59", "remaining_time": "2:52:43", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 710, "total_steps": 1737, "loss": 0.7299, "learning_rate": 1e-05, "epoch": 1.2250620079801575, "percentage": 40.88, "elapsed_time": "1:58:47", "remaining_time": "2:51:49", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 715, "total_steps": 1737, "loss": 0.811, "learning_rate": 1e-05, "epoch": 1.2336892052194544, "percentage": 41.16, "elapsed_time": "1:59:41", "remaining_time": "2:51:04", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 720, "total_steps": 1737, "loss": 0.7714, "learning_rate": 1e-05, "epoch": 1.2423164024587512, "percentage": 41.45, "elapsed_time": "2:00:30", "remaining_time": "2:50:13", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 725, "total_steps": 1737, "loss": 0.8388, "learning_rate": 1e-05, "epoch": 1.250943599698048, "percentage": 41.74, "elapsed_time": "2:01:20", "remaining_time": "2:49:22", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 730, "total_steps": 1737, "loss": 0.8244, "learning_rate": 1e-05, "epoch": 1.259570796937345, "percentage": 42.03, "elapsed_time": "2:02:10", "remaining_time": "2:48:31", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 735, "total_steps": 1737, "loss": 0.8326, "learning_rate": 1e-05, "epoch": 1.268197994176642, "percentage": 42.31, "elapsed_time": "2:02:57", "remaining_time": "2:47:38", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 740, "total_steps": 1737, "loss": 0.8302, "learning_rate": 1e-05, "epoch": 1.2768251914159388, "percentage": 42.6, "elapsed_time": "2:03:42", "remaining_time": "2:46:40", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 745, "total_steps": 1737, "loss": 0.7911, "learning_rate": 1e-05, "epoch": 1.2854523886552356, "percentage": 42.89, "elapsed_time": "2:04:30", "remaining_time": "2:45:47", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 750, "total_steps": 1737, "loss": 0.8836, "learning_rate": 1e-05, "epoch": 1.2940795858945324, "percentage": 43.18, "elapsed_time": "2:05:22", "remaining_time": "2:44:59", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 755, "total_steps": 1737, "loss": 0.8023, "learning_rate": 1e-05, "epoch": 1.3027067831338295, "percentage": 43.47, "elapsed_time": "2:06:13", "remaining_time": "2:44:10", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 760, "total_steps": 1737, "loss": 0.8338, "learning_rate": 1e-05, "epoch": 1.3113339803731263, "percentage": 43.75, "elapsed_time": "2:07:03", "remaining_time": "2:43:19", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 765, "total_steps": 1737, "loss": 0.8247, "learning_rate": 1e-05, "epoch": 1.3199611776124232, "percentage": 44.04, "elapsed_time": "2:07:53", "remaining_time": "2:42:29", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 770, "total_steps": 1737, "loss": 0.8781, "learning_rate": 1e-05, "epoch": 1.32858837485172, "percentage": 44.33, "elapsed_time": "2:08:45", "remaining_time": "2:41:42", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 775, "total_steps": 1737, "loss": 0.7967, "learning_rate": 1e-05, "epoch": 1.3372155720910168, "percentage": 44.62, "elapsed_time": "2:09:32", "remaining_time": "2:40:47", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 780, "total_steps": 1737, "loss": 0.7918, "learning_rate": 1e-05, "epoch": 1.3458427693303139, "percentage": 44.91, "elapsed_time": "2:10:30", "remaining_time": "2:40:07", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 785, "total_steps": 1737, "loss": 0.7784, "learning_rate": 1e-05, "epoch": 1.3544699665696107, "percentage": 45.19, "elapsed_time": "2:11:16", "remaining_time": "2:39:12", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 790, "total_steps": 1737, "loss": 0.8017, "learning_rate": 1e-05, "epoch": 1.3630971638089076, "percentage": 45.48, "elapsed_time": "2:12:06", "remaining_time": "2:38:21", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 795, "total_steps": 1737, "loss": 0.7803, "learning_rate": 1e-05, "epoch": 1.3717243610482044, "percentage": 45.77, "elapsed_time": "2:13:07", "remaining_time": "2:37:44", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 800, "total_steps": 1737, "loss": 0.8341, "learning_rate": 1e-05, "epoch": 1.3803515582875012, "percentage": 46.06, "elapsed_time": "2:13:51", "remaining_time": "2:36:47", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 805, "total_steps": 1737, "loss": 0.7977, "learning_rate": 1e-05, "epoch": 1.3889787555267983, "percentage": 46.34, "elapsed_time": "2:15:57", "remaining_time": "2:37:25", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 810, "total_steps": 1737, "loss": 0.778, "learning_rate": 1e-05, "epoch": 1.3976059527660951, "percentage": 46.63, "elapsed_time": "2:16:48", "remaining_time": "2:36:33", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 815, "total_steps": 1737, "loss": 0.8107, "learning_rate": 1e-05, "epoch": 1.406233150005392, "percentage": 46.92, "elapsed_time": "2:17:31", "remaining_time": "2:35:35", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 820, "total_steps": 1737, "loss": 0.753, "learning_rate": 1e-05, "epoch": 1.414860347244689, "percentage": 47.21, "elapsed_time": "2:18:23", "remaining_time": "2:34:45", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 825, "total_steps": 1737, "loss": 0.8111, "learning_rate": 1e-05, "epoch": 1.4234875444839858, "percentage": 47.5, "elapsed_time": "2:19:10", "remaining_time": "2:33:51", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 830, "total_steps": 1737, "loss": 0.7559, "learning_rate": 1e-05, "epoch": 1.4321147417232827, "percentage": 47.78, "elapsed_time": "2:19:56", "remaining_time": "2:32:55", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 835, "total_steps": 1737, "loss": 0.8213, "learning_rate": 1e-05, "epoch": 1.4407419389625795, "percentage": 48.07, "elapsed_time": "2:20:47", "remaining_time": "2:32:05", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 840, "total_steps": 1737, "loss": 0.9053, "learning_rate": 1e-05, "epoch": 1.4493691362018764, "percentage": 48.36, "elapsed_time": "2:21:32", "remaining_time": "2:31:08", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 845, "total_steps": 1737, "loss": 0.7486, "learning_rate": 1e-05, "epoch": 1.4579963334411734, "percentage": 48.65, "elapsed_time": "2:22:25", "remaining_time": "2:30:20", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 850, "total_steps": 1737, "loss": 0.8385, "learning_rate": 1e-05, "epoch": 1.4666235306804702, "percentage": 48.93, "elapsed_time": "2:23:08", "remaining_time": "2:29:22", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 855, "total_steps": 1737, "loss": 0.8994, "learning_rate": 1e-05, "epoch": 1.475250727919767, "percentage": 49.22, "elapsed_time": "2:23:57", "remaining_time": "2:28:30", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 860, "total_steps": 1737, "loss": 0.7993, "learning_rate": 1e-05, "epoch": 1.483877925159064, "percentage": 49.51, "elapsed_time": "2:24:48", "remaining_time": "2:27:40", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 865, "total_steps": 1737, "loss": 0.8415, "learning_rate": 1e-05, "epoch": 1.4925051223983608, "percentage": 49.8, "elapsed_time": "2:25:33", "remaining_time": "2:26:44", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 870, "total_steps": 1737, "loss": 0.7544, "learning_rate": 1e-05, "epoch": 1.5011323196376578, "percentage": 50.09, "elapsed_time": "2:26:21", "remaining_time": "2:25:51", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 875, "total_steps": 1737, "loss": 0.8997, "learning_rate": 1e-05, "epoch": 1.5097595168769546, "percentage": 50.37, "elapsed_time": "2:27:09", "remaining_time": "2:24:57", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 880, "total_steps": 1737, "loss": 0.7673, "learning_rate": 1e-05, "epoch": 1.5183867141162515, "percentage": 50.66, "elapsed_time": "2:27:55", "remaining_time": "2:24:03", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 885, "total_steps": 1737, "loss": 0.8582, "learning_rate": 1e-05, "epoch": 1.5270139113555483, "percentage": 50.95, "elapsed_time": "2:28:42", "remaining_time": "2:23:10", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 890, "total_steps": 1737, "loss": 0.7828, "learning_rate": 1e-05, "epoch": 1.5356411085948452, "percentage": 51.24, "elapsed_time": "2:29:30", "remaining_time": "2:22:16", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 895, "total_steps": 1737, "loss": 0.8359, "learning_rate": 1e-05, "epoch": 1.5442683058341422, "percentage": 51.53, "elapsed_time": "2:30:13", "remaining_time": "2:21:19", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 900, "total_steps": 1737, "loss": 0.8226, "learning_rate": 1e-05, "epoch": 1.552895503073439, "percentage": 51.81, "elapsed_time": "2:31:03", "remaining_time": "2:20:29", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 905, "total_steps": 1737, "loss": 0.8219, "learning_rate": 1e-05, "epoch": 1.5615227003127359, "percentage": 52.1, "elapsed_time": "2:31:51", "remaining_time": "2:19:36", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 910, "total_steps": 1737, "loss": 0.8212, "learning_rate": 1e-05, "epoch": 1.570149897552033, "percentage": 52.39, "elapsed_time": "2:32:39", "remaining_time": "2:18:43", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 915, "total_steps": 1737, "loss": 0.8036, "learning_rate": 1e-05, "epoch": 1.5787770947913295, "percentage": 52.68, "elapsed_time": "2:33:26", "remaining_time": "2:17:50", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 920, "total_steps": 1737, "loss": 0.7633, "learning_rate": 1e-05, "epoch": 1.5874042920306266, "percentage": 52.96, "elapsed_time": "2:34:14", "remaining_time": "2:16:58", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 925, "total_steps": 1737, "loss": 0.7673, "learning_rate": 1e-05, "epoch": 1.5960314892699234, "percentage": 53.25, "elapsed_time": "2:35:00", "remaining_time": "2:16:04", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 930, "total_steps": 1737, "loss": 0.7732, "learning_rate": 1e-05, "epoch": 1.6046586865092203, "percentage": 53.54, "elapsed_time": "2:35:47", "remaining_time": "2:15:10", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 935, "total_steps": 1737, "loss": 0.7941, "learning_rate": 1e-05, "epoch": 1.6132858837485173, "percentage": 53.83, "elapsed_time": "2:36:34", "remaining_time": "2:14:18", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 940, "total_steps": 1737, "loss": 0.8184, "learning_rate": 1e-05, "epoch": 1.621913080987814, "percentage": 54.12, "elapsed_time": "2:37:25", "remaining_time": "2:13:28", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 945, "total_steps": 1737, "loss": 0.8496, "learning_rate": 1e-05, "epoch": 1.630540278227111, "percentage": 54.4, "elapsed_time": "2:38:10", "remaining_time": "2:12:34", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 950, "total_steps": 1737, "loss": 0.888, "learning_rate": 1e-05, "epoch": 1.6391674754664078, "percentage": 54.69, "elapsed_time": "2:38:51", "remaining_time": "2:11:36", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 955, "total_steps": 1737, "loss": 0.8551, "learning_rate": 1e-05, "epoch": 1.6477946727057047, "percentage": 54.98, "elapsed_time": "2:39:37", "remaining_time": "2:10:42", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 960, "total_steps": 1737, "loss": 0.8184, "learning_rate": 1e-05, "epoch": 1.6564218699450017, "percentage": 55.27, "elapsed_time": "2:40:32", "remaining_time": "2:09:56", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 965, "total_steps": 1737, "loss": 0.8045, "learning_rate": 1e-05, "epoch": 1.6650490671842983, "percentage": 55.56, "elapsed_time": "2:41:37", "remaining_time": "2:09:17", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 970, "total_steps": 1737, "loss": 0.7489, "learning_rate": 1e-05, "epoch": 1.6736762644235954, "percentage": 55.84, "elapsed_time": "2:42:27", "remaining_time": "2:08:27", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 975, "total_steps": 1737, "loss": 0.7816, "learning_rate": 1e-05, "epoch": 1.6823034616628922, "percentage": 56.13, "elapsed_time": "2:43:15", "remaining_time": "2:07:35", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 980, "total_steps": 1737, "loss": 0.884, "learning_rate": 1e-05, "epoch": 1.690930658902189, "percentage": 56.42, "elapsed_time": "2:44:02", "remaining_time": "2:06:42", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 985, "total_steps": 1737, "loss": 0.9077, "learning_rate": 1e-05, "epoch": 1.6995578561414861, "percentage": 56.71, "elapsed_time": "2:44:48", "remaining_time": "2:05:49", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 990, "total_steps": 1737, "loss": 0.8371, "learning_rate": 1e-05, "epoch": 1.708185053380783, "percentage": 56.99, "elapsed_time": "2:45:36", "remaining_time": "2:04:57", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 995, "total_steps": 1737, "loss": 0.8028, "learning_rate": 1e-05, "epoch": 1.7168122506200798, "percentage": 57.28, "elapsed_time": "2:46:21", "remaining_time": "2:04:03", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1000, "total_steps": 1737, "loss": 0.7969, "learning_rate": 1e-05, "epoch": 1.7254394478593766, "percentage": 57.57, "elapsed_time": "2:47:07", "remaining_time": "2:03:09", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1005, "total_steps": 1737, "loss": 0.8557, "learning_rate": 1e-05, "epoch": 1.7340666450986735, "percentage": 57.86, "elapsed_time": "2:49:21", "remaining_time": "2:03:21", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1010, "total_steps": 1737, "loss": 0.7524, "learning_rate": 1e-05, "epoch": 1.7426938423379705, "percentage": 58.15, "elapsed_time": "2:50:12", "remaining_time": "2:02:30", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1015, "total_steps": 1737, "loss": 0.8103, "learning_rate": 1e-05, "epoch": 1.7513210395772674, "percentage": 58.43, "elapsed_time": "2:51:10", "remaining_time": "2:01:45", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1020, "total_steps": 1737, "loss": 0.7795, "learning_rate": 1e-05, "epoch": 1.7599482368165642, "percentage": 58.72, "elapsed_time": "2:51:59", "remaining_time": "2:00:54", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1025, "total_steps": 1737, "loss": 0.8534, "learning_rate": 1e-05, "epoch": 1.7685754340558613, "percentage": 59.01, "elapsed_time": "2:52:47", "remaining_time": "2:00:01", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1030, "total_steps": 1737, "loss": 0.8371, "learning_rate": 1e-05, "epoch": 1.7772026312951579, "percentage": 59.3, "elapsed_time": "2:53:31", "remaining_time": "1:59:06", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1035, "total_steps": 1737, "loss": 0.7688, "learning_rate": 1e-05, "epoch": 1.785829828534455, "percentage": 59.59, "elapsed_time": "2:54:15", "remaining_time": "1:58:11", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1040, "total_steps": 1737, "loss": 0.7823, "learning_rate": 1e-05, "epoch": 1.7944570257737518, "percentage": 59.87, "elapsed_time": "2:55:06", "remaining_time": "1:57:21", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1045, "total_steps": 1737, "loss": 0.8135, "learning_rate": 1e-05, "epoch": 1.8030842230130486, "percentage": 60.16, "elapsed_time": "2:55:53", "remaining_time": "1:56:28", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1050, "total_steps": 1737, "loss": 0.8169, "learning_rate": 1e-05, "epoch": 1.8117114202523457, "percentage": 60.45, "elapsed_time": "2:56:40", "remaining_time": "1:55:35", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1055, "total_steps": 1737, "loss": 0.846, "learning_rate": 1e-05, "epoch": 1.8203386174916423, "percentage": 60.74, "elapsed_time": "2:57:27", "remaining_time": "1:54:43", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1060, "total_steps": 1737, "loss": 0.8055, "learning_rate": 1e-05, "epoch": 1.8289658147309393, "percentage": 61.02, "elapsed_time": "2:58:10", "remaining_time": "1:53:47", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1065, "total_steps": 1737, "loss": 0.8216, "learning_rate": 1e-05, "epoch": 1.8375930119702362, "percentage": 61.31, "elapsed_time": "2:58:59", "remaining_time": "1:52:56", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1070, "total_steps": 1737, "loss": 0.8663, "learning_rate": 1e-05, "epoch": 1.846220209209533, "percentage": 61.6, "elapsed_time": "2:59:46", "remaining_time": "1:52:04", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1075, "total_steps": 1737, "loss": 0.8332, "learning_rate": 1e-05, "epoch": 1.85484740644883, "percentage": 61.89, "elapsed_time": "3:00:33", "remaining_time": "1:51:11", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1080, "total_steps": 1737, "loss": 0.8408, "learning_rate": 1e-05, "epoch": 1.8634746036881267, "percentage": 62.18, "elapsed_time": "3:01:22", "remaining_time": "1:50:19", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1085, "total_steps": 1737, "loss": 0.828, "learning_rate": 1e-05, "epoch": 1.8721018009274237, "percentage": 62.46, "elapsed_time": "3:02:06", "remaining_time": "1:49:26", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1090, "total_steps": 1737, "loss": 0.8127, "learning_rate": 1e-05, "epoch": 1.8807289981667206, "percentage": 62.75, "elapsed_time": "3:02:51", "remaining_time": "1:48:32", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1095, "total_steps": 1737, "loss": 0.8607, "learning_rate": 1e-05, "epoch": 1.8893561954060174, "percentage": 63.04, "elapsed_time": "3:03:39", "remaining_time": "1:47:40", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1100, "total_steps": 1737, "loss": 0.9251, "learning_rate": 1e-05, "epoch": 1.8979833926453145, "percentage": 63.33, "elapsed_time": "3:04:28", "remaining_time": "1:46:49", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1105, "total_steps": 1737, "loss": 0.8388, "learning_rate": 1e-05, "epoch": 1.906610589884611, "percentage": 63.62, "elapsed_time": "3:05:17", "remaining_time": "1:45:58", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1110, "total_steps": 1737, "loss": 0.8036, "learning_rate": 1e-05, "epoch": 1.9152377871239081, "percentage": 63.9, "elapsed_time": "3:06:02", "remaining_time": "1:45:05", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1115, "total_steps": 1737, "loss": 0.8001, "learning_rate": 1e-05, "epoch": 1.923864984363205, "percentage": 64.19, "elapsed_time": "3:06:53", "remaining_time": "1:44:15", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1120, "total_steps": 1737, "loss": 0.775, "learning_rate": 1e-05, "epoch": 1.9324921816025018, "percentage": 64.48, "elapsed_time": "3:07:41", "remaining_time": "1:43:23", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1125, "total_steps": 1737, "loss": 0.8639, "learning_rate": 1e-05, "epoch": 1.9411193788417989, "percentage": 64.77, "elapsed_time": "3:08:25", "remaining_time": "1:42:30", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1130, "total_steps": 1737, "loss": 0.8513, "learning_rate": 1e-05, "epoch": 1.9497465760810957, "percentage": 65.05, "elapsed_time": "3:09:12", "remaining_time": "1:41:38", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1135, "total_steps": 1737, "loss": 0.77, "learning_rate": 1e-05, "epoch": 1.9583737733203925, "percentage": 65.34, "elapsed_time": "3:10:04", "remaining_time": "1:40:48", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1140, "total_steps": 1737, "loss": 0.732, "learning_rate": 1e-05, "epoch": 1.9670009705596894, "percentage": 65.63, "elapsed_time": "3:10:48", "remaining_time": "1:39:55", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1145, "total_steps": 1737, "loss": 0.8029, "learning_rate": 1e-05, "epoch": 1.9756281677989862, "percentage": 65.92, "elapsed_time": "3:11:35", "remaining_time": "1:39:03", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1150, "total_steps": 1737, "loss": 0.8205, "learning_rate": 1e-05, "epoch": 1.9842553650382833, "percentage": 66.21, "elapsed_time": "3:12:21", "remaining_time": "1:38:11", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1155, "total_steps": 1737, "loss": 0.7995, "learning_rate": 1e-05, "epoch": 1.99288256227758, "percentage": 66.49, "elapsed_time": "3:13:10", "remaining_time": "1:37:20", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1160, "total_steps": 1737, "loss": 0.7343, "learning_rate": 1e-05, "epoch": 2.001509759516877, "percentage": 66.78, "elapsed_time": "3:14:25", "remaining_time": "1:36:42", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1165, "total_steps": 1737, "loss": 0.4413, "learning_rate": 1e-05, "epoch": 2.010136956756174, "percentage": 67.07, "elapsed_time": "3:15:22", "remaining_time": "1:35:55", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1170, "total_steps": 1737, "loss": 0.4171, "learning_rate": 1e-05, "epoch": 2.0187641539954706, "percentage": 67.36, "elapsed_time": "3:16:10", "remaining_time": "1:35:04", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1175, "total_steps": 1737, "loss": 0.3829, "learning_rate": 1e-05, "epoch": 2.0273913512347677, "percentage": 67.65, "elapsed_time": "3:16:59", "remaining_time": "1:34:13", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1180, "total_steps": 1737, "loss": 0.3505, "learning_rate": 1e-05, "epoch": 2.0360185484740643, "percentage": 67.93, "elapsed_time": "3:17:48", "remaining_time": "1:33:22", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1185, "total_steps": 1737, "loss": 0.3448, "learning_rate": 1e-05, "epoch": 2.0446457457133613, "percentage": 68.22, "elapsed_time": "3:18:36", "remaining_time": "1:32:30", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1190, "total_steps": 1737, "loss": 0.3737, "learning_rate": 1e-05, "epoch": 2.0532729429526584, "percentage": 68.51, "elapsed_time": "3:19:31", "remaining_time": "1:31:42", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1195, "total_steps": 1737, "loss": 0.3939, "learning_rate": 1e-05, "epoch": 2.061900140191955, "percentage": 68.8, "elapsed_time": "3:20:12", "remaining_time": "1:30:48", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1200, "total_steps": 1737, "loss": 0.3638, "learning_rate": 1e-05, "epoch": 2.070527337431252, "percentage": 69.08, "elapsed_time": "3:21:06", "remaining_time": "1:29:59", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1205, "total_steps": 1737, "loss": 0.3634, "learning_rate": 1e-05, "epoch": 2.079154534670549, "percentage": 69.37, "elapsed_time": "3:23:11", "remaining_time": "1:29:42", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1210, "total_steps": 1737, "loss": 0.4462, "learning_rate": 1e-05, "epoch": 2.0877817319098457, "percentage": 69.66, "elapsed_time": "3:23:59", "remaining_time": "1:28:50", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1215, "total_steps": 1737, "loss": 0.4071, "learning_rate": 1e-05, "epoch": 2.096408929149143, "percentage": 69.95, "elapsed_time": "3:24:54", "remaining_time": "1:28:01", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1220, "total_steps": 1737, "loss": 0.3735, "learning_rate": 1e-05, "epoch": 2.1050361263884394, "percentage": 70.24, "elapsed_time": "3:25:39", "remaining_time": "1:27:09", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1225, "total_steps": 1737, "loss": 0.383, "learning_rate": 1e-05, "epoch": 2.1136633236277365, "percentage": 70.52, "elapsed_time": "3:26:29", "remaining_time": "1:26:18", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1230, "total_steps": 1737, "loss": 0.3077, "learning_rate": 1e-05, "epoch": 2.1222905208670335, "percentage": 70.81, "elapsed_time": "3:27:18", "remaining_time": "1:25:27", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1235, "total_steps": 1737, "loss": 0.3774, "learning_rate": 1e-05, "epoch": 2.13091771810633, "percentage": 71.1, "elapsed_time": "3:28:08", "remaining_time": "1:24:36", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1240, "total_steps": 1737, "loss": 0.3388, "learning_rate": 1e-05, "epoch": 2.139544915345627, "percentage": 71.39, "elapsed_time": "3:29:07", "remaining_time": "1:23:49", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1245, "total_steps": 1737, "loss": 0.3848, "learning_rate": 1e-05, "epoch": 2.148172112584924, "percentage": 71.68, "elapsed_time": "3:29:55", "remaining_time": "1:22:57", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1250, "total_steps": 1737, "loss": 0.4173, "learning_rate": 1e-05, "epoch": 2.156799309824221, "percentage": 71.96, "elapsed_time": "3:30:43", "remaining_time": "1:22:05", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1255, "total_steps": 1737, "loss": 0.3514, "learning_rate": 1e-05, "epoch": 2.165426507063518, "percentage": 72.25, "elapsed_time": "3:31:28", "remaining_time": "1:21:13", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1260, "total_steps": 1737, "loss": 0.393, "learning_rate": 1e-05, "epoch": 2.1740537043028145, "percentage": 72.54, "elapsed_time": "3:32:10", "remaining_time": "1:20:19", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1265, "total_steps": 1737, "loss": 0.3886, "learning_rate": 1e-05, "epoch": 2.1826809015421116, "percentage": 72.83, "elapsed_time": "3:32:56", "remaining_time": "1:19:27", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1270, "total_steps": 1737, "loss": 0.3943, "learning_rate": 1e-05, "epoch": 2.191308098781408, "percentage": 73.11, "elapsed_time": "3:33:46", "remaining_time": "1:18:36", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1275, "total_steps": 1737, "loss": 0.3562, "learning_rate": 1e-05, "epoch": 2.1999352960207053, "percentage": 73.4, "elapsed_time": "3:34:43", "remaining_time": "1:17:48", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1280, "total_steps": 1737, "loss": 0.3609, "learning_rate": 1e-05, "epoch": 2.2085624932600023, "percentage": 73.69, "elapsed_time": "3:35:37", "remaining_time": "1:16:58", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1285, "total_steps": 1737, "loss": 0.4031, "learning_rate": 1e-05, "epoch": 2.217189690499299, "percentage": 73.98, "elapsed_time": "3:36:27", "remaining_time": "1:16:08", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1290, "total_steps": 1737, "loss": 0.3989, "learning_rate": 1e-05, "epoch": 2.225816887738596, "percentage": 74.27, "elapsed_time": "3:37:18", "remaining_time": "1:15:17", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1295, "total_steps": 1737, "loss": 0.4016, "learning_rate": 1e-05, "epoch": 2.234444084977893, "percentage": 74.55, "elapsed_time": "3:38:09", "remaining_time": "1:14:27", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1300, "total_steps": 1737, "loss": 0.3735, "learning_rate": 1e-05, "epoch": 2.2430712822171897, "percentage": 74.84, "elapsed_time": "3:38:54", "remaining_time": "1:13:35", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1305, "total_steps": 1737, "loss": 0.376, "learning_rate": 1e-05, "epoch": 2.2516984794564867, "percentage": 75.13, "elapsed_time": "3:39:38", "remaining_time": "1:12:42", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1310, "total_steps": 1737, "loss": 0.4118, "learning_rate": 1e-05, "epoch": 2.2603256766957833, "percentage": 75.42, "elapsed_time": "3:40:30", "remaining_time": "1:11:52", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1315, "total_steps": 1737, "loss": 0.4483, "learning_rate": 1e-05, "epoch": 2.2689528739350804, "percentage": 75.71, "elapsed_time": "3:41:17", "remaining_time": "1:11:00", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1320, "total_steps": 1737, "loss": 0.3646, "learning_rate": 1e-05, "epoch": 2.277580071174377, "percentage": 75.99, "elapsed_time": "3:42:01", "remaining_time": "1:10:08", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1325, "total_steps": 1737, "loss": 0.3901, "learning_rate": 1e-05, "epoch": 2.286207268413674, "percentage": 76.28, "elapsed_time": "3:42:46", "remaining_time": "1:09:16", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1330, "total_steps": 1737, "loss": 0.3684, "learning_rate": 1e-05, "epoch": 2.294834465652971, "percentage": 76.57, "elapsed_time": "3:43:35", "remaining_time": "1:08:25", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1335, "total_steps": 1737, "loss": 0.3885, "learning_rate": 1e-05, "epoch": 2.3034616628922677, "percentage": 76.86, "elapsed_time": "3:44:27", "remaining_time": "1:07:35", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1340, "total_steps": 1737, "loss": 0.424, "learning_rate": 1e-05, "epoch": 2.312088860131565, "percentage": 77.14, "elapsed_time": "3:45:13", "remaining_time": "1:06:43", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1345, "total_steps": 1737, "loss": 0.4194, "learning_rate": 1e-05, "epoch": 2.320716057370862, "percentage": 77.43, "elapsed_time": "3:45:59", "remaining_time": "1:05:51", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1350, "total_steps": 1737, "loss": 0.3982, "learning_rate": 1e-05, "epoch": 2.3293432546101585, "percentage": 77.72, "elapsed_time": "3:46:43", "remaining_time": "1:04:59", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1355, "total_steps": 1737, "loss": 0.4565, "learning_rate": 1e-05, "epoch": 2.3379704518494555, "percentage": 78.01, "elapsed_time": "3:47:31", "remaining_time": "1:04:08", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1360, "total_steps": 1737, "loss": 0.3995, "learning_rate": 1e-05, "epoch": 2.346597649088752, "percentage": 78.3, "elapsed_time": "3:48:25", "remaining_time": "1:03:19", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1365, "total_steps": 1737, "loss": 0.426, "learning_rate": 1e-05, "epoch": 2.355224846328049, "percentage": 78.58, "elapsed_time": "3:49:09", "remaining_time": "1:02:27", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1370, "total_steps": 1737, "loss": 0.3854, "learning_rate": 1e-05, "epoch": 2.3638520435673462, "percentage": 78.87, "elapsed_time": "3:49:54", "remaining_time": "1:01:35", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1375, "total_steps": 1737, "loss": 0.3808, "learning_rate": 1e-05, "epoch": 2.372479240806643, "percentage": 79.16, "elapsed_time": "3:50:38", "remaining_time": "1:00:43", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1380, "total_steps": 1737, "loss": 0.3598, "learning_rate": 1e-05, "epoch": 2.38110643804594, "percentage": 79.45, "elapsed_time": "3:51:26", "remaining_time": "0:59:52", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1385, "total_steps": 1737, "loss": 0.4018, "learning_rate": 1e-05, "epoch": 2.3897336352852365, "percentage": 79.74, "elapsed_time": "3:52:07", "remaining_time": "0:58:59", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1390, "total_steps": 1737, "loss": 0.4105, "learning_rate": 1e-05, "epoch": 2.3983608325245336, "percentage": 80.02, "elapsed_time": "3:52:54", "remaining_time": "0:58:08", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1395, "total_steps": 1737, "loss": 0.3832, "learning_rate": 1e-05, "epoch": 2.4069880297638306, "percentage": 80.31, "elapsed_time": "3:53:43", "remaining_time": "0:57:17", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1400, "total_steps": 1737, "loss": 0.4288, "learning_rate": 1e-05, "epoch": 2.4156152270031273, "percentage": 80.6, "elapsed_time": "3:54:25", "remaining_time": "0:56:25", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1405, "total_steps": 1737, "loss": 0.3759, "learning_rate": 1e-05, "epoch": 2.4242424242424243, "percentage": 80.89, "elapsed_time": "3:56:31", "remaining_time": "0:55:53", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1410, "total_steps": 1737, "loss": 0.3981, "learning_rate": 1e-05, "epoch": 2.4328696214817214, "percentage": 81.17, "elapsed_time": "3:57:25", "remaining_time": "0:55:03", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1415, "total_steps": 1737, "loss": 0.3461, "learning_rate": 1e-05, "epoch": 2.441496818721018, "percentage": 81.46, "elapsed_time": "3:58:12", "remaining_time": "0:54:12", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1420, "total_steps": 1737, "loss": 0.3898, "learning_rate": 1e-05, "epoch": 2.450124015960315, "percentage": 81.75, "elapsed_time": "3:59:00", "remaining_time": "0:53:21", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1425, "total_steps": 1737, "loss": 0.36, "learning_rate": 1e-05, "epoch": 2.4587512131996117, "percentage": 82.04, "elapsed_time": "3:59:52", "remaining_time": "0:52:31", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1430, "total_steps": 1737, "loss": 0.4348, "learning_rate": 1e-05, "epoch": 2.4673784104389087, "percentage": 82.33, "elapsed_time": "4:00:40", "remaining_time": "0:51:40", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1435, "total_steps": 1737, "loss": 0.3582, "learning_rate": 1e-05, "epoch": 2.4760056076782053, "percentage": 82.61, "elapsed_time": "4:01:30", "remaining_time": "0:50:49", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1440, "total_steps": 1737, "loss": 0.4337, "learning_rate": 1e-05, "epoch": 2.4846328049175024, "percentage": 82.9, "elapsed_time": "4:02:17", "remaining_time": "0:49:58", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1445, "total_steps": 1737, "loss": 0.4196, "learning_rate": 1e-05, "epoch": 2.4932600021567994, "percentage": 83.19, "elapsed_time": "4:02:58", "remaining_time": "0:49:05", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1450, "total_steps": 1737, "loss": 0.3901, "learning_rate": 1e-05, "epoch": 2.501887199396096, "percentage": 83.48, "elapsed_time": "4:03:45", "remaining_time": "0:48:14", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1455, "total_steps": 1737, "loss": 0.3864, "learning_rate": 1e-05, "epoch": 2.510514396635393, "percentage": 83.77, "elapsed_time": "4:04:33", "remaining_time": "0:47:23", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1460, "total_steps": 1737, "loss": 0.3911, "learning_rate": 1e-05, "epoch": 2.51914159387469, "percentage": 84.05, "elapsed_time": "4:05:20", "remaining_time": "0:46:32", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1465, "total_steps": 1737, "loss": 0.4008, "learning_rate": 1e-05, "epoch": 2.527768791113987, "percentage": 84.34, "elapsed_time": "4:06:14", "remaining_time": "0:45:43", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1470, "total_steps": 1737, "loss": 0.3738, "learning_rate": 1e-05, "epoch": 2.536395988353284, "percentage": 84.63, "elapsed_time": "4:06:58", "remaining_time": "0:44:51", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1475, "total_steps": 1737, "loss": 0.3892, "learning_rate": 1e-05, "epoch": 2.5450231855925805, "percentage": 84.92, "elapsed_time": "4:07:42", "remaining_time": "0:44:00", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1480, "total_steps": 1737, "loss": 0.4022, "learning_rate": 1e-05, "epoch": 2.5536503828318775, "percentage": 85.2, "elapsed_time": "4:08:27", "remaining_time": "0:43:08", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1485, "total_steps": 1737, "loss": 0.436, "learning_rate": 1e-05, "epoch": 2.562277580071174, "percentage": 85.49, "elapsed_time": "4:09:15", "remaining_time": "0:42:17", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1490, "total_steps": 1737, "loss": 0.4164, "learning_rate": 1e-05, "epoch": 2.570904777310471, "percentage": 85.78, "elapsed_time": "4:10:02", "remaining_time": "0:41:27", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1495, "total_steps": 1737, "loss": 0.3914, "learning_rate": 1e-05, "epoch": 2.5795319745497682, "percentage": 86.07, "elapsed_time": "4:10:52", "remaining_time": "0:40:36", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1500, "total_steps": 1737, "loss": 0.3878, "learning_rate": 1e-05, "epoch": 2.588159171789065, "percentage": 86.36, "elapsed_time": "4:11:40", "remaining_time": "0:39:45", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1505, "total_steps": 1737, "loss": 0.39, "learning_rate": 1e-05, "epoch": 2.596786369028362, "percentage": 86.64, "elapsed_time": "4:12:20", "remaining_time": "0:38:53", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1510, "total_steps": 1737, "loss": 0.3989, "learning_rate": 1e-05, "epoch": 2.605413566267659, "percentage": 86.93, "elapsed_time": "4:13:03", "remaining_time": "0:38:02", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1515, "total_steps": 1737, "loss": 0.3613, "learning_rate": 1e-05, "epoch": 2.6140407635069556, "percentage": 87.22, "elapsed_time": "4:13:47", "remaining_time": "0:37:11", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1520, "total_steps": 1737, "loss": 0.3995, "learning_rate": 1e-05, "epoch": 2.6226679607462526, "percentage": 87.51, "elapsed_time": "4:14:37", "remaining_time": "0:36:21", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1525, "total_steps": 1737, "loss": 0.3376, "learning_rate": 1e-05, "epoch": 2.6312951579855497, "percentage": 87.8, "elapsed_time": "4:15:20", "remaining_time": "0:35:29", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1530, "total_steps": 1737, "loss": 0.3861, "learning_rate": 1e-05, "epoch": 2.6399223552248463, "percentage": 88.08, "elapsed_time": "4:16:03", "remaining_time": "0:34:38", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1535, "total_steps": 1737, "loss": 0.4193, "learning_rate": 1e-05, "epoch": 2.6485495524641434, "percentage": 88.37, "elapsed_time": "4:16:52", "remaining_time": "0:33:48", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1540, "total_steps": 1737, "loss": 0.4235, "learning_rate": 1e-05, "epoch": 2.65717674970344, "percentage": 88.66, "elapsed_time": "4:17:43", "remaining_time": "0:32:58", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1545, "total_steps": 1737, "loss": 0.39, "learning_rate": 1e-05, "epoch": 2.665803946942737, "percentage": 88.95, "elapsed_time": "4:18:38", "remaining_time": "0:32:08", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1550, "total_steps": 1737, "loss": 0.4395, "learning_rate": 1e-05, "epoch": 2.6744311441820336, "percentage": 89.23, "elapsed_time": "4:19:30", "remaining_time": "0:31:18", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1555, "total_steps": 1737, "loss": 0.3956, "learning_rate": 1e-05, "epoch": 2.6830583414213307, "percentage": 89.52, "elapsed_time": "4:20:20", "remaining_time": "0:30:28", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1560, "total_steps": 1737, "loss": 0.4399, "learning_rate": 1e-05, "epoch": 2.6916855386606278, "percentage": 89.81, "elapsed_time": "4:21:06", "remaining_time": "0:29:37", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1565, "total_steps": 1737, "loss": 0.4161, "learning_rate": 1e-05, "epoch": 2.7003127358999244, "percentage": 90.1, "elapsed_time": "4:21:57", "remaining_time": "0:28:47", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1570, "total_steps": 1737, "loss": 0.4128, "learning_rate": 1e-05, "epoch": 2.7089399331392214, "percentage": 90.39, "elapsed_time": "4:22:51", "remaining_time": "0:27:57", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1575, "total_steps": 1737, "loss": 0.4291, "learning_rate": 1e-05, "epoch": 2.7175671303785185, "percentage": 90.67, "elapsed_time": "4:23:36", "remaining_time": "0:27:06", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1580, "total_steps": 1737, "loss": 0.4198, "learning_rate": 1e-05, "epoch": 2.726194327617815, "percentage": 90.96, "elapsed_time": "4:24:19", "remaining_time": "0:26:15", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1585, "total_steps": 1737, "loss": 0.3954, "learning_rate": 1e-05, "epoch": 2.734821524857112, "percentage": 91.25, "elapsed_time": "4:25:05", "remaining_time": "0:25:25", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1590, "total_steps": 1737, "loss": 0.3718, "learning_rate": 1e-05, "epoch": 2.7434487220964088, "percentage": 91.54, "elapsed_time": "4:25:47", "remaining_time": "0:24:34", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1595, "total_steps": 1737, "loss": 0.3488, "learning_rate": 1e-05, "epoch": 2.752075919335706, "percentage": 91.82, "elapsed_time": "4:26:39", "remaining_time": "0:23:44", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1600, "total_steps": 1737, "loss": 0.4581, "learning_rate": 1e-05, "epoch": 2.7607031165750024, "percentage": 92.11, "elapsed_time": "4:27:28", "remaining_time": "0:22:54", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1605, "total_steps": 1737, "loss": 0.4455, "learning_rate": 1e-05, "epoch": 2.7693303138142995, "percentage": 92.4, "elapsed_time": "4:29:35", "remaining_time": "0:22:10", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1610, "total_steps": 1737, "loss": 0.3918, "learning_rate": 1e-05, "epoch": 2.7779575110535966, "percentage": 92.69, "elapsed_time": "4:30:24", "remaining_time": "0:21:19", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1615, "total_steps": 1737, "loss": 0.4067, "learning_rate": 1e-05, "epoch": 2.786584708292893, "percentage": 92.98, "elapsed_time": "4:31:10", "remaining_time": "0:20:29", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1620, "total_steps": 1737, "loss": 0.4237, "learning_rate": 1e-05, "epoch": 2.7952119055321902, "percentage": 93.26, "elapsed_time": "4:31:53", "remaining_time": "0:19:38", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1625, "total_steps": 1737, "loss": 0.3939, "learning_rate": 1e-05, "epoch": 2.8038391027714873, "percentage": 93.55, "elapsed_time": "4:32:45", "remaining_time": "0:18:47", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1630, "total_steps": 1737, "loss": 0.4385, "learning_rate": 1e-05, "epoch": 2.812466300010784, "percentage": 93.84, "elapsed_time": "4:33:33", "remaining_time": "0:17:57", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1635, "total_steps": 1737, "loss": 0.4109, "learning_rate": 1e-05, "epoch": 2.821093497250081, "percentage": 94.13, "elapsed_time": "4:34:20", "remaining_time": "0:17:06", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1640, "total_steps": 1737, "loss": 0.4402, "learning_rate": 1e-05, "epoch": 2.829720694489378, "percentage": 94.42, "elapsed_time": "4:35:11", "remaining_time": "0:16:16", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1645, "total_steps": 1737, "loss": 0.444, "learning_rate": 1e-05, "epoch": 2.8383478917286746, "percentage": 94.7, "elapsed_time": "4:36:07", "remaining_time": "0:15:26", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1650, "total_steps": 1737, "loss": 0.413, "learning_rate": 1e-05, "epoch": 2.8469750889679717, "percentage": 94.99, "elapsed_time": "4:36:59", "remaining_time": "0:14:36", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1655, "total_steps": 1737, "loss": 0.4245, "learning_rate": 1e-05, "epoch": 2.8556022862072683, "percentage": 95.28, "elapsed_time": "4:37:47", "remaining_time": "0:13:45", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1660, "total_steps": 1737, "loss": 0.4091, "learning_rate": 1e-05, "epoch": 2.8642294834465654, "percentage": 95.57, "elapsed_time": "4:38:33", "remaining_time": "0:12:55", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1665, "total_steps": 1737, "loss": 0.4206, "learning_rate": 1e-05, "epoch": 2.872856680685862, "percentage": 95.85, "elapsed_time": "4:39:14", "remaining_time": "0:12:04", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1670, "total_steps": 1737, "loss": 0.3868, "learning_rate": 1e-05, "epoch": 2.881483877925159, "percentage": 96.14, "elapsed_time": "4:40:06", "remaining_time": "0:11:14", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1675, "total_steps": 1737, "loss": 0.4102, "learning_rate": 1e-05, "epoch": 2.890111075164456, "percentage": 96.43, "elapsed_time": "4:40:52", "remaining_time": "0:10:23", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1680, "total_steps": 1737, "loss": 0.4349, "learning_rate": 1e-05, "epoch": 2.8987382724037527, "percentage": 96.72, "elapsed_time": "4:41:40", "remaining_time": "0:09:33", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1685, "total_steps": 1737, "loss": 0.4298, "learning_rate": 1e-05, "epoch": 2.9073654696430498, "percentage": 97.01, "elapsed_time": "4:42:34", "remaining_time": "0:08:43", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1690, "total_steps": 1737, "loss": 0.431, "learning_rate": 1e-05, "epoch": 2.915992666882347, "percentage": 97.29, "elapsed_time": "4:43:21", "remaining_time": "0:07:52", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1695, "total_steps": 1737, "loss": 0.4373, "learning_rate": 1e-05, "epoch": 2.9246198641216434, "percentage": 97.58, "elapsed_time": "4:44:20", "remaining_time": "0:07:02", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1700, "total_steps": 1737, "loss": 0.4239, "learning_rate": 1e-05, "epoch": 2.9332470613609405, "percentage": 97.87, "elapsed_time": "4:45:05", "remaining_time": "0:06:12", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1705, "total_steps": 1737, "loss": 0.4277, "learning_rate": 1e-05, "epoch": 2.941874258600237, "percentage": 98.16, "elapsed_time": "4:45:54", "remaining_time": "0:05:21", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1710, "total_steps": 1737, "loss": 0.4302, "learning_rate": 1e-05, "epoch": 2.950501455839534, "percentage": 98.45, "elapsed_time": "4:46:39", "remaining_time": "0:04:31", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1715, "total_steps": 1737, "loss": 0.3822, "learning_rate": 1e-05, "epoch": 2.9591286530788308, "percentage": 98.73, "elapsed_time": "4:47:27", "remaining_time": "0:03:41", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1720, "total_steps": 1737, "loss": 0.4402, "learning_rate": 1e-05, "epoch": 2.967755850318128, "percentage": 99.02, "elapsed_time": "4:48:11", "remaining_time": "0:02:50", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1725, "total_steps": 1737, "loss": 0.3803, "learning_rate": 1e-05, "epoch": 2.976383047557425, "percentage": 99.31, "elapsed_time": "4:49:02", "remaining_time": "0:02:00", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1730, "total_steps": 1737, "loss": 0.4109, "learning_rate": 1e-05, "epoch": 2.9850102447967215, "percentage": 99.6, "elapsed_time": "4:49:49", "remaining_time": "0:01:10", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1735, "total_steps": 1737, "loss": 0.4232, "learning_rate": 1e-05, "epoch": 2.9936374420360186, "percentage": 99.88, "elapsed_time": "4:50:32", "remaining_time": "0:00:20", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 1737, "total_steps": 1737, "epoch": 2.9970883209317374, "percentage": 100.0, "elapsed_time": "4:52:15", "remaining_time": "0:00:00", "throughput": "0.00", "total_tokens": 0}

trainer_state.json

@@ -0,0 +1,2471 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.9970883209317374,
+  "eval_steps": 500,
+  "global_step": 1737,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.008627197239296884,
+      "grad_norm": 5.919341564178467,
+      "learning_rate": 1e-05,
+      "loss": 1.3814,
+      "step": 5
+    },
+    {
+      "epoch": 0.017254394478593768,
+      "grad_norm": 4.066539287567139,
+      "learning_rate": 1e-05,
+      "loss": 1.3848,
+      "step": 10
+    },
+    {
+      "epoch": 0.02588159171789065,
+      "grad_norm": 4.004323959350586,
+      "learning_rate": 1e-05,
+      "loss": 1.3517,
+      "step": 15
+    },
+    {
+      "epoch": 0.034508788957187536,
+      "grad_norm": 4.424528121948242,
+      "learning_rate": 1e-05,
+      "loss": 1.4088,
+      "step": 20
+    },
+    {
+      "epoch": 0.04313598619648442,
+      "grad_norm": 4.122994899749756,
+      "learning_rate": 1e-05,
+      "loss": 1.3542,
+      "step": 25
+    },
+    {
+      "epoch": 0.0517631834357813,
+      "grad_norm": 3.9815385341644287,
+      "learning_rate": 1e-05,
+      "loss": 1.3434,
+      "step": 30
+    },
+    {
+      "epoch": 0.06039038067507818,
+      "grad_norm": 3.663285255432129,
+      "learning_rate": 1e-05,
+      "loss": 1.3926,
+      "step": 35
+    },
+    {
+      "epoch": 0.06901757791437507,
+      "grad_norm": 3.6322672367095947,
+      "learning_rate": 1e-05,
+      "loss": 1.2819,
+      "step": 40
+    },
+    {
+      "epoch": 0.07764477515367195,
+      "grad_norm": 4.1651387214660645,
+      "learning_rate": 1e-05,
+      "loss": 1.3482,
+      "step": 45
+    },
+    {
+      "epoch": 0.08627197239296884,
+      "grad_norm": 3.8225786685943604,
+      "learning_rate": 1e-05,
+      "loss": 1.2621,
+      "step": 50
+    },
+    {
+      "epoch": 0.09489916963226572,
+      "grad_norm": 3.9319000244140625,
+      "learning_rate": 1e-05,
+      "loss": 1.2695,
+      "step": 55
+    },
+    {
+      "epoch": 0.1035263668715626,
+      "grad_norm": 4.2429633140563965,
+      "learning_rate": 1e-05,
+      "loss": 1.2302,
+      "step": 60
+    },
+    {
+      "epoch": 0.11215356411085949,
+      "grad_norm": 3.836970090866089,
+      "learning_rate": 1e-05,
+      "loss": 1.242,
+      "step": 65
+    },
+    {
+      "epoch": 0.12078076135015636,
+      "grad_norm": 3.622291326522827,
+      "learning_rate": 1e-05,
+      "loss": 1.2433,
+      "step": 70
+    },
+    {
+      "epoch": 0.12940795858945325,
+      "grad_norm": 3.876661539077759,
+      "learning_rate": 1e-05,
+      "loss": 1.3095,
+      "step": 75
+    },
+    {
+      "epoch": 0.13803515582875014,
+      "grad_norm": 3.3505313396453857,
+      "learning_rate": 1e-05,
+      "loss": 1.3006,
+      "step": 80
+    },
+    {
+      "epoch": 0.146662353068047,
+      "grad_norm": 4.024404048919678,
+      "learning_rate": 1e-05,
+      "loss": 1.2698,
+      "step": 85
+    },
+    {
+      "epoch": 0.1552895503073439,
+      "grad_norm": 3.4006166458129883,
+      "learning_rate": 1e-05,
+      "loss": 1.3346,
+      "step": 90
+    },
+    {
+      "epoch": 0.1639167475466408,
+      "grad_norm": 3.5438950061798096,
+      "learning_rate": 1e-05,
+      "loss": 1.3634,
+      "step": 95
+    },
+    {
+      "epoch": 0.17254394478593768,
+      "grad_norm": 4.186426162719727,
+      "learning_rate": 1e-05,
+      "loss": 1.2645,
+      "step": 100
+    },
+    {
+      "epoch": 0.18117114202523454,
+      "grad_norm": 3.5808355808258057,
+      "learning_rate": 1e-05,
+      "loss": 1.3625,
+      "step": 105
+    },
+    {
+      "epoch": 0.18979833926453143,
+      "grad_norm": 4.477747917175293,
+      "learning_rate": 1e-05,
+      "loss": 1.2592,
+      "step": 110
+    },
+    {
+      "epoch": 0.19842553650382833,
+      "grad_norm": 3.461580753326416,
+      "learning_rate": 1e-05,
+      "loss": 1.2872,
+      "step": 115
+    },
+    {
+      "epoch": 0.2070527337431252,
+      "grad_norm": 4.153059005737305,
+      "learning_rate": 1e-05,
+      "loss": 1.31,
+      "step": 120
+    },
+    {
+      "epoch": 0.21567993098242208,
+      "grad_norm": 4.117781639099121,
+      "learning_rate": 1e-05,
+      "loss": 1.2627,
+      "step": 125
+    },
+    {
+      "epoch": 0.22430712822171897,
+      "grad_norm": 4.274353504180908,
+      "learning_rate": 1e-05,
+      "loss": 1.206,
+      "step": 130
+    },
+    {
+      "epoch": 0.23293432546101586,
+      "grad_norm": 3.5153017044067383,
+      "learning_rate": 1e-05,
+      "loss": 1.3192,
+      "step": 135
+    },
+    {
+      "epoch": 0.24156152270031273,
+      "grad_norm": 3.2502472400665283,
+      "learning_rate": 1e-05,
+      "loss": 1.2825,
+      "step": 140
+    },
+    {
+      "epoch": 0.2501887199396096,
+      "grad_norm": 3.598522186279297,
+      "learning_rate": 1e-05,
+      "loss": 1.3421,
+      "step": 145
+    },
+    {
+      "epoch": 0.2588159171789065,
+      "grad_norm": 3.85213565826416,
+      "learning_rate": 1e-05,
+      "loss": 1.3141,
+      "step": 150
+    },
+    {
+      "epoch": 0.2674431144182034,
+      "grad_norm": 4.03485107421875,
+      "learning_rate": 1e-05,
+      "loss": 1.298,
+      "step": 155
+    },
+    {
+      "epoch": 0.2760703116575003,
+      "grad_norm": 3.352916717529297,
+      "learning_rate": 1e-05,
+      "loss": 1.3034,
+      "step": 160
+    },
+    {
+      "epoch": 0.2846975088967972,
+      "grad_norm": 3.504659652709961,
+      "learning_rate": 1e-05,
+      "loss": 1.2532,
+      "step": 165
+    },
+    {
+      "epoch": 0.293324706136094,
+      "grad_norm": 3.5304458141326904,
+      "learning_rate": 1e-05,
+      "loss": 1.2745,
+      "step": 170
+    },
+    {
+      "epoch": 0.3019519033753909,
+      "grad_norm": 3.2286152839660645,
+      "learning_rate": 1e-05,
+      "loss": 1.3067,
+      "step": 175
+    },
+    {
+      "epoch": 0.3105791006146878,
+      "grad_norm": 3.1940603256225586,
+      "learning_rate": 1e-05,
+      "loss": 1.2628,
+      "step": 180
+    },
+    {
+      "epoch": 0.3192062978539847,
+      "grad_norm": 3.7485692501068115,
+      "learning_rate": 1e-05,
+      "loss": 1.2514,
+      "step": 185
+    },
+    {
+      "epoch": 0.3278334950932816,
+      "grad_norm": 3.438934326171875,
+      "learning_rate": 1e-05,
+      "loss": 1.2122,
+      "step": 190
+    },
+    {
+      "epoch": 0.33646069233257847,
+      "grad_norm": 4.063304424285889,
+      "learning_rate": 1e-05,
+      "loss": 1.2825,
+      "step": 195
+    },
+    {
+      "epoch": 0.34508788957187536,
+      "grad_norm": 3.3390140533447266,
+      "learning_rate": 1e-05,
+      "loss": 1.2752,
+      "step": 200
+    },
+    {
+      "epoch": 0.3537150868111722,
+      "grad_norm": 3.5218873023986816,
+      "learning_rate": 1e-05,
+      "loss": 1.3145,
+      "step": 205
+    },
+    {
+      "epoch": 0.3623422840504691,
+      "grad_norm": 4.008931636810303,
+      "learning_rate": 1e-05,
+      "loss": 1.2694,
+      "step": 210
+    },
+    {
+      "epoch": 0.370969481289766,
+      "grad_norm": 3.844992160797119,
+      "learning_rate": 1e-05,
+      "loss": 1.3507,
+      "step": 215
+    },
+    {
+      "epoch": 0.37959667852906287,
+      "grad_norm": 3.2494778633117676,
+      "learning_rate": 1e-05,
+      "loss": 1.3843,
+      "step": 220
+    },
+    {
+      "epoch": 0.38822387576835976,
+      "grad_norm": 3.2193210124969482,
+      "learning_rate": 1e-05,
+      "loss": 1.2553,
+      "step": 225
+    },
+    {
+      "epoch": 0.39685107300765665,
+      "grad_norm": 3.6562159061431885,
+      "learning_rate": 1e-05,
+      "loss": 1.3664,
+      "step": 230
+    },
+    {
+      "epoch": 0.40547827024695354,
+      "grad_norm": 3.219736099243164,
+      "learning_rate": 1e-05,
+      "loss": 1.3083,
+      "step": 235
+    },
+    {
+      "epoch": 0.4141054674862504,
+      "grad_norm": 3.986835241317749,
+      "learning_rate": 1e-05,
+      "loss": 1.2945,
+      "step": 240
+    },
+    {
+      "epoch": 0.42273266472554727,
+      "grad_norm": 3.874375581741333,
+      "learning_rate": 1e-05,
+      "loss": 1.2938,
+      "step": 245
+    },
+    {
+      "epoch": 0.43135986196484416,
+      "grad_norm": 3.1514320373535156,
+      "learning_rate": 1e-05,
+      "loss": 1.2231,
+      "step": 250
+    },
+    {
+      "epoch": 0.43998705920414105,
+      "grad_norm": 3.3543567657470703,
+      "learning_rate": 1e-05,
+      "loss": 1.2237,
+      "step": 255
+    },
+    {
+      "epoch": 0.44861425644343794,
+      "grad_norm": 4.068399906158447,
+      "learning_rate": 1e-05,
+      "loss": 1.1541,
+      "step": 260
+    },
+    {
+      "epoch": 0.45724145368273483,
+      "grad_norm": 3.7400364875793457,
+      "learning_rate": 1e-05,
+      "loss": 1.2629,
+      "step": 265
+    },
+    {
+      "epoch": 0.4658686509220317,
+      "grad_norm": 3.091808557510376,
+      "learning_rate": 1e-05,
+      "loss": 1.1911,
+      "step": 270
+    },
+    {
+      "epoch": 0.4744958481613286,
+      "grad_norm": 3.622913360595703,
+      "learning_rate": 1e-05,
+      "loss": 1.4012,
+      "step": 275
+    },
+    {
+      "epoch": 0.48312304540062545,
+      "grad_norm": 3.368582010269165,
+      "learning_rate": 1e-05,
+      "loss": 1.2188,
+      "step": 280
+    },
+    {
+      "epoch": 0.49175024263992234,
+      "grad_norm": 3.4569334983825684,
+      "learning_rate": 1e-05,
+      "loss": 1.354,
+      "step": 285
+    },
+    {
+      "epoch": 0.5003774398792192,
+      "grad_norm": 3.7725794315338135,
+      "learning_rate": 1e-05,
+      "loss": 1.3322,
+      "step": 290
+    },
+    {
+      "epoch": 0.5090046371185161,
+      "grad_norm": 3.446682929992676,
+      "learning_rate": 1e-05,
+      "loss": 1.3385,
+      "step": 295
+    },
+    {
+      "epoch": 0.517631834357813,
+      "grad_norm": 3.31061053276062,
+      "learning_rate": 1e-05,
+      "loss": 1.2541,
+      "step": 300
+    },
+    {
+      "epoch": 0.5262590315971098,
+      "grad_norm": 3.3477659225463867,
+      "learning_rate": 1e-05,
+      "loss": 1.3451,
+      "step": 305
+    },
+    {
+      "epoch": 0.5348862288364068,
+      "grad_norm": 4.0511322021484375,
+      "learning_rate": 1e-05,
+      "loss": 1.3269,
+      "step": 310
+    },
+    {
+      "epoch": 0.5435134260757036,
+      "grad_norm": 3.5980873107910156,
+      "learning_rate": 1e-05,
+      "loss": 1.2654,
+      "step": 315
+    },
+    {
+      "epoch": 0.5521406233150006,
+      "grad_norm": 4.6503119468688965,
+      "learning_rate": 1e-05,
+      "loss": 1.3162,
+      "step": 320
+    },
+    {
+      "epoch": 0.5607678205542974,
+      "grad_norm": 3.890101432800293,
+      "learning_rate": 1e-05,
+      "loss": 1.2839,
+      "step": 325
+    },
+    {
+      "epoch": 0.5693950177935944,
+      "grad_norm": 3.019289016723633,
+      "learning_rate": 1e-05,
+      "loss": 1.3095,
+      "step": 330
+    },
+    {
+      "epoch": 0.5780222150328912,
+      "grad_norm": 3.544351100921631,
+      "learning_rate": 1e-05,
+      "loss": 1.285,
+      "step": 335
+    },
+    {
+      "epoch": 0.586649412272188,
+      "grad_norm": 4.081313133239746,
+      "learning_rate": 1e-05,
+      "loss": 1.2419,
+      "step": 340
+    },
+    {
+      "epoch": 0.595276609511485,
+      "grad_norm": 3.010798454284668,
+      "learning_rate": 1e-05,
+      "loss": 1.2875,
+      "step": 345
+    },
+    {
+      "epoch": 0.6039038067507818,
+      "grad_norm": 3.469045639038086,
+      "learning_rate": 1e-05,
+      "loss": 1.3055,
+      "step": 350
+    },
+    {
+      "epoch": 0.6125310039900788,
+      "grad_norm": 3.8424911499023438,
+      "learning_rate": 1e-05,
+      "loss": 1.33,
+      "step": 355
+    },
+    {
+      "epoch": 0.6211582012293756,
+      "grad_norm": 3.3473572731018066,
+      "learning_rate": 1e-05,
+      "loss": 1.2003,
+      "step": 360
+    },
+    {
+      "epoch": 0.6297853984686725,
+      "grad_norm": 3.4214556217193604,
+      "learning_rate": 1e-05,
+      "loss": 1.2385,
+      "step": 365
+    },
+    {
+      "epoch": 0.6384125957079694,
+      "grad_norm": 3.459791898727417,
+      "learning_rate": 1e-05,
+      "loss": 1.2198,
+      "step": 370
+    },
+    {
+      "epoch": 0.6470397929472662,
+      "grad_norm": 3.77476167678833,
+      "learning_rate": 1e-05,
+      "loss": 1.3135,
+      "step": 375
+    },
+    {
+      "epoch": 0.6556669901865632,
+      "grad_norm": 3.092871904373169,
+      "learning_rate": 1e-05,
+      "loss": 1.3234,
+      "step": 380
+    },
+    {
+      "epoch": 0.66429418742586,
+      "grad_norm": 3.2897961139678955,
+      "learning_rate": 1e-05,
+      "loss": 1.3376,
+      "step": 385
+    },
+    {
+      "epoch": 0.6729213846651569,
+      "grad_norm": 3.227506399154663,
+      "learning_rate": 1e-05,
+      "loss": 1.2237,
+      "step": 390
+    },
+    {
+      "epoch": 0.6815485819044538,
+      "grad_norm": 3.3434855937957764,
+      "learning_rate": 1e-05,
+      "loss": 1.2465,
+      "step": 395
+    },
+    {
+      "epoch": 0.6901757791437507,
+      "grad_norm": 3.5809898376464844,
+      "learning_rate": 1e-05,
+      "loss": 1.2188,
+      "step": 400
+    },
+    {
+      "epoch": 0.6988029763830476,
+      "grad_norm": 3.5106072425842285,
+      "learning_rate": 1e-05,
+      "loss": 1.2663,
+      "step": 405
+    },
+    {
+      "epoch": 0.7074301736223444,
+      "grad_norm": 3.0715136528015137,
+      "learning_rate": 1e-05,
+      "loss": 1.3354,
+      "step": 410
+    },
+    {
+      "epoch": 0.7160573708616413,
+      "grad_norm": 3.441499948501587,
+      "learning_rate": 1e-05,
+      "loss": 1.3112,
+      "step": 415
+    },
+    {
+      "epoch": 0.7246845681009382,
+      "grad_norm": 2.81852388381958,
+      "learning_rate": 1e-05,
+      "loss": 1.2335,
+      "step": 420
+    },
+    {
+      "epoch": 0.7333117653402351,
+      "grad_norm": 4.062169075012207,
+      "learning_rate": 1e-05,
+      "loss": 1.2291,
+      "step": 425
+    },
+    {
+      "epoch": 0.741938962579532,
+      "grad_norm": 3.453822612762451,
+      "learning_rate": 1e-05,
+      "loss": 1.1676,
+      "step": 430
+    },
+    {
+      "epoch": 0.7505661598188289,
+      "grad_norm": 3.2253923416137695,
+      "learning_rate": 1e-05,
+      "loss": 1.2539,
+      "step": 435
+    },
+    {
+      "epoch": 0.7591933570581257,
+      "grad_norm": 3.092754364013672,
+      "learning_rate": 1e-05,
+      "loss": 1.2994,
+      "step": 440
+    },
+    {
+      "epoch": 0.7678205542974226,
+      "grad_norm": 3.07718825340271,
+      "learning_rate": 1e-05,
+      "loss": 1.2304,
+      "step": 445
+    },
+    {
+      "epoch": 0.7764477515367195,
+      "grad_norm": 3.2939090728759766,
+      "learning_rate": 1e-05,
+      "loss": 1.2871,
+      "step": 450
+    },
+    {
+      "epoch": 0.7850749487760164,
+      "grad_norm": 3.8047966957092285,
+      "learning_rate": 1e-05,
+      "loss": 1.3053,
+      "step": 455
+    },
+    {
+      "epoch": 0.7937021460153133,
+      "grad_norm": 4.2863969802856445,
+      "learning_rate": 1e-05,
+      "loss": 1.2691,
+      "step": 460
+    },
+    {
+      "epoch": 0.8023293432546101,
+      "grad_norm": 3.9592127799987793,
+      "learning_rate": 1e-05,
+      "loss": 1.2744,
+      "step": 465
+    },
+    {
+      "epoch": 0.8109565404939071,
+      "grad_norm": 4.489944934844971,
+      "learning_rate": 1e-05,
+      "loss": 1.2016,
+      "step": 470
+    },
+    {
+      "epoch": 0.8195837377332039,
+      "grad_norm": 3.259884834289551,
+      "learning_rate": 1e-05,
+      "loss": 1.2787,
+      "step": 475
+    },
+    {
+      "epoch": 0.8282109349725008,
+      "grad_norm": 3.40492582321167,
+      "learning_rate": 1e-05,
+      "loss": 1.3095,
+      "step": 480
+    },
+    {
+      "epoch": 0.8368381322117977,
+      "grad_norm": 3.3192970752716064,
+      "learning_rate": 1e-05,
+      "loss": 1.2503,
+      "step": 485
+    },
+    {
+      "epoch": 0.8454653294510945,
+      "grad_norm": 3.331552743911743,
+      "learning_rate": 1e-05,
+      "loss": 1.2598,
+      "step": 490
+    },
+    {
+      "epoch": 0.8540925266903915,
+      "grad_norm": 3.073087692260742,
+      "learning_rate": 1e-05,
+      "loss": 1.2914,
+      "step": 495
+    },
+    {
+      "epoch": 0.8627197239296883,
+      "grad_norm": 3.344917058944702,
+      "learning_rate": 1e-05,
+      "loss": 1.2744,
+      "step": 500
+    },
+    {
+      "epoch": 0.8713469211689853,
+      "grad_norm": 2.9080522060394287,
+      "learning_rate": 1e-05,
+      "loss": 1.3817,
+      "step": 505
+    },
+    {
+      "epoch": 0.8799741184082821,
+      "grad_norm": 3.2660365104675293,
+      "learning_rate": 1e-05,
+      "loss": 1.1628,
+      "step": 510
+    },
+    {
+      "epoch": 0.8886013156475789,
+      "grad_norm": 3.2718842029571533,
+      "learning_rate": 1e-05,
+      "loss": 1.2602,
+      "step": 515
+    },
+    {
+      "epoch": 0.8972285128868759,
+      "grad_norm": 3.0166728496551514,
+      "learning_rate": 1e-05,
+      "loss": 1.2036,
+      "step": 520
+    },
+    {
+      "epoch": 0.9058557101261727,
+      "grad_norm": 3.566046953201294,
+      "learning_rate": 1e-05,
+      "loss": 1.2575,
+      "step": 525
+    },
+    {
+      "epoch": 0.9144829073654697,
+      "grad_norm": 3.3579084873199463,
+      "learning_rate": 1e-05,
+      "loss": 1.2762,
+      "step": 530
+    },
+    {
+      "epoch": 0.9231101046047665,
+      "grad_norm": 3.251415252685547,
+      "learning_rate": 1e-05,
+      "loss": 1.299,
+      "step": 535
+    },
+    {
+      "epoch": 0.9317373018440634,
+      "grad_norm": 3.4540982246398926,
+      "learning_rate": 1e-05,
+      "loss": 1.2246,
+      "step": 540
+    },
+    {
+      "epoch": 0.9403644990833603,
+      "grad_norm": 3.0218403339385986,
+      "learning_rate": 1e-05,
+      "loss": 1.2069,
+      "step": 545
+    },
+    {
+      "epoch": 0.9489916963226572,
+      "grad_norm": 3.457592248916626,
+      "learning_rate": 1e-05,
+      "loss": 1.244,
+      "step": 550
+    },
+    {
+      "epoch": 0.9576188935619541,
+      "grad_norm": 2.8288912773132324,
+      "learning_rate": 1e-05,
+      "loss": 1.2785,
+      "step": 555
+    },
+    {
+      "epoch": 0.9662460908012509,
+      "grad_norm": 3.9383344650268555,
+      "learning_rate": 1e-05,
+      "loss": 1.2542,
+      "step": 560
+    },
+    {
+      "epoch": 0.9748732880405478,
+      "grad_norm": 3.215847969055176,
+      "learning_rate": 1e-05,
+      "loss": 1.2649,
+      "step": 565
+    },
+    {
+      "epoch": 0.9835004852798447,
+      "grad_norm": 3.3479530811309814,
+      "learning_rate": 1e-05,
+      "loss": 1.2963,
+      "step": 570
+    },
+    {
+      "epoch": 0.9921276825191416,
+      "grad_norm": 3.1215426921844482,
+      "learning_rate": 1e-05,
+      "loss": 1.2534,
+      "step": 575
+    },
+    {
+      "epoch": 1.0007548797584385,
+      "grad_norm": 3.651639223098755,
+      "learning_rate": 1e-05,
+      "loss": 1.1385,
+      "step": 580
+    },
+    {
+      "epoch": 1.0093820769977353,
+      "grad_norm": 3.4434738159179688,
+      "learning_rate": 1e-05,
+      "loss": 0.8563,
+      "step": 585
+    },
+    {
+      "epoch": 1.0180092742370321,
+      "grad_norm": 4.103019714355469,
+      "learning_rate": 1e-05,
+      "loss": 0.9018,
+      "step": 590
+    },
+    {
+      "epoch": 1.0266364714763292,
+      "grad_norm": 3.472033977508545,
+      "learning_rate": 1e-05,
+      "loss": 0.8435,
+      "step": 595
+    },
+    {
+      "epoch": 1.035263668715626,
+      "grad_norm": 3.869497537612915,
+      "learning_rate": 1e-05,
+      "loss": 0.7686,
+      "step": 600
+    },
+    {
+      "epoch": 1.0438908659549229,
+      "grad_norm": 3.7982940673828125,
+      "learning_rate": 1e-05,
+      "loss": 0.7134,
+      "step": 605
+    },
+    {
+      "epoch": 1.0525180631942197,
+      "grad_norm": 3.038923978805542,
+      "learning_rate": 1e-05,
+      "loss": 0.7836,
+      "step": 610
+    },
+    {
+      "epoch": 1.0611452604335168,
+      "grad_norm": 3.709296464920044,
+      "learning_rate": 1e-05,
+      "loss": 0.8263,
+      "step": 615
+    },
+    {
+      "epoch": 1.0697724576728136,
+      "grad_norm": 3.110173225402832,
+      "learning_rate": 1e-05,
+      "loss": 0.8028,
+      "step": 620
+    },
+    {
+      "epoch": 1.0783996549121104,
+      "grad_norm": 4.099541187286377,
+      "learning_rate": 1e-05,
+      "loss": 0.7552,
+      "step": 625
+    },
+    {
+      "epoch": 1.0870268521514073,
+      "grad_norm": 3.3552350997924805,
+      "learning_rate": 1e-05,
+      "loss": 0.838,
+      "step": 630
+    },
+    {
+      "epoch": 1.095654049390704,
+      "grad_norm": 3.6320431232452393,
+      "learning_rate": 1e-05,
+      "loss": 0.7915,
+      "step": 635
+    },
+    {
+      "epoch": 1.1042812466300012,
+      "grad_norm": 3.646231174468994,
+      "learning_rate": 1e-05,
+      "loss": 0.7522,
+      "step": 640
+    },
+    {
+      "epoch": 1.112908443869298,
+      "grad_norm": 3.390377998352051,
+      "learning_rate": 1e-05,
+      "loss": 0.7777,
+      "step": 645
+    },
+    {
+      "epoch": 1.1215356411085948,
+      "grad_norm": 3.605112075805664,
+      "learning_rate": 1e-05,
+      "loss": 0.8388,
+      "step": 650
+    },
+    {
+      "epoch": 1.1301628383478917,
+      "grad_norm": 3.313898801803589,
+      "learning_rate": 1e-05,
+      "loss": 0.8455,
+      "step": 655
+    },
+    {
+      "epoch": 1.1387900355871885,
+      "grad_norm": 3.3568379878997803,
+      "learning_rate": 1e-05,
+      "loss": 0.8048,
+      "step": 660
+    },
+    {
+      "epoch": 1.1474172328264856,
+      "grad_norm": 3.5753304958343506,
+      "learning_rate": 1e-05,
+      "loss": 0.7762,
+      "step": 665
+    },
+    {
+      "epoch": 1.1560444300657824,
+      "grad_norm": 3.9311811923980713,
+      "learning_rate": 1e-05,
+      "loss": 0.7812,
+      "step": 670
+    },
+    {
+      "epoch": 1.1646716273050792,
+      "grad_norm": 3.399244546890259,
+      "learning_rate": 1e-05,
+      "loss": 0.8134,
+      "step": 675
+    },
+    {
+      "epoch": 1.173298824544376,
+      "grad_norm": 3.5281805992126465,
+      "learning_rate": 1e-05,
+      "loss": 0.7821,
+      "step": 680
+    },
+    {
+      "epoch": 1.1819260217836731,
+      "grad_norm": 3.510946273803711,
+      "learning_rate": 1e-05,
+      "loss": 0.8034,
+      "step": 685
+    },
+    {
+      "epoch": 1.19055321902297,
+      "grad_norm": 3.504866361618042,
+      "learning_rate": 1e-05,
+      "loss": 0.7592,
+      "step": 690
+    },
+    {
+      "epoch": 1.1991804162622668,
+      "grad_norm": 3.339489698410034,
+      "learning_rate": 1e-05,
+      "loss": 0.7738,
+      "step": 695
+    },
+    {
+      "epoch": 1.2078076135015636,
+      "grad_norm": 2.915128231048584,
+      "learning_rate": 1e-05,
+      "loss": 0.7043,
+      "step": 700
+    },
+    {
+      "epoch": 1.2164348107408607,
+      "grad_norm": 3.4995436668395996,
+      "learning_rate": 1e-05,
+      "loss": 0.7969,
+      "step": 705
+    },
+    {
+      "epoch": 1.2250620079801575,
+      "grad_norm": 3.3331191539764404,
+      "learning_rate": 1e-05,
+      "loss": 0.7299,
+      "step": 710
+    },
+    {
+      "epoch": 1.2336892052194544,
+      "grad_norm": 3.898615598678589,
+      "learning_rate": 1e-05,
+      "loss": 0.811,
+      "step": 715
+    },
+    {
+      "epoch": 1.2423164024587512,
+      "grad_norm": 3.1952695846557617,
+      "learning_rate": 1e-05,
+      "loss": 0.7714,
+      "step": 720
+    },
+    {
+      "epoch": 1.250943599698048,
+      "grad_norm": 3.310316801071167,
+      "learning_rate": 1e-05,
+      "loss": 0.8388,
+      "step": 725
+    },
+    {
+      "epoch": 1.259570796937345,
+      "grad_norm": 3.0263495445251465,
+      "learning_rate": 1e-05,
+      "loss": 0.8244,
+      "step": 730
+    },
+    {
+      "epoch": 1.268197994176642,
+      "grad_norm": 3.4661359786987305,
+      "learning_rate": 1e-05,
+      "loss": 0.8326,
+      "step": 735
+    },
+    {
+      "epoch": 1.2768251914159388,
+      "grad_norm": 3.4017932415008545,
+      "learning_rate": 1e-05,
+      "loss": 0.8302,
+      "step": 740
+    },
+    {
+      "epoch": 1.2854523886552356,
+      "grad_norm": 3.6982741355895996,
+      "learning_rate": 1e-05,
+      "loss": 0.7911,
+      "step": 745
+    },
+    {
+      "epoch": 1.2940795858945324,
+      "grad_norm": 3.737339973449707,
+      "learning_rate": 1e-05,
+      "loss": 0.8836,
+      "step": 750
+    },
+    {
+      "epoch": 1.3027067831338295,
+      "grad_norm": 3.491697311401367,
+      "learning_rate": 1e-05,
+      "loss": 0.8023,
+      "step": 755
+    },
+    {
+      "epoch": 1.3113339803731263,
+      "grad_norm": 3.9661755561828613,
+      "learning_rate": 1e-05,
+      "loss": 0.8338,
+      "step": 760
+    },
+    {
+      "epoch": 1.3199611776124232,
+      "grad_norm": 3.4925425052642822,
+      "learning_rate": 1e-05,
+      "loss": 0.8247,
+      "step": 765
+    },
+    {
+      "epoch": 1.32858837485172,
+      "grad_norm": 3.0747172832489014,
+      "learning_rate": 1e-05,
+      "loss": 0.8781,
+      "step": 770
+    },
+    {
+      "epoch": 1.3372155720910168,
+      "grad_norm": 3.2208268642425537,
+      "learning_rate": 1e-05,
+      "loss": 0.7967,
+      "step": 775
+    },
+    {
+      "epoch": 1.3458427693303139,
+      "grad_norm": 3.751079559326172,
+      "learning_rate": 1e-05,
+      "loss": 0.7918,
+      "step": 780
+    },
+    {
+      "epoch": 1.3544699665696107,
+      "grad_norm": 3.1584877967834473,
+      "learning_rate": 1e-05,
+      "loss": 0.7784,
+      "step": 785
+    },
+    {
+      "epoch": 1.3630971638089076,
+      "grad_norm": 3.420969009399414,
+      "learning_rate": 1e-05,
+      "loss": 0.8017,
+      "step": 790
+    },
+    {
+      "epoch": 1.3717243610482044,
+      "grad_norm": 3.333355188369751,
+      "learning_rate": 1e-05,
+      "loss": 0.7803,
+      "step": 795
+    },
+    {
+      "epoch": 1.3803515582875012,
+      "grad_norm": 3.5859718322753906,
+      "learning_rate": 1e-05,
+      "loss": 0.8341,
+      "step": 800
+    },
+    {
+      "epoch": 1.3889787555267983,
+      "grad_norm": 3.3463239669799805,
+      "learning_rate": 1e-05,
+      "loss": 0.7977,
+      "step": 805
+    },
+    {
+      "epoch": 1.3976059527660951,
+      "grad_norm": 3.34783935546875,
+      "learning_rate": 1e-05,
+      "loss": 0.778,
+      "step": 810
+    },
+    {
+      "epoch": 1.406233150005392,
+      "grad_norm": 4.29722261428833,
+      "learning_rate": 1e-05,
+      "loss": 0.8107,
+      "step": 815
+    },
+    {
+      "epoch": 1.414860347244689,
+      "grad_norm": 3.2892956733703613,
+      "learning_rate": 1e-05,
+      "loss": 0.753,
+      "step": 820
+    },
+    {
+      "epoch": 1.4234875444839858,
+      "grad_norm": 3.7883739471435547,
+      "learning_rate": 1e-05,
+      "loss": 0.8111,
+      "step": 825
+    },
+    {
+      "epoch": 1.4321147417232827,
+      "grad_norm": 3.545917510986328,
+      "learning_rate": 1e-05,
+      "loss": 0.7559,
+      "step": 830
+    },
+    {
+      "epoch": 1.4407419389625795,
+      "grad_norm": 3.5111639499664307,
+      "learning_rate": 1e-05,
+      "loss": 0.8213,
+      "step": 835
+    },
+    {
+      "epoch": 1.4493691362018764,
+      "grad_norm": 3.9835708141326904,
+      "learning_rate": 1e-05,
+      "loss": 0.9053,
+      "step": 840
+    },
+    {
+      "epoch": 1.4579963334411734,
+      "grad_norm": 3.5352683067321777,
+      "learning_rate": 1e-05,
+      "loss": 0.7486,
+      "step": 845
+    },
+    {
+      "epoch": 1.4666235306804702,
+      "grad_norm": 3.8274669647216797,
+      "learning_rate": 1e-05,
+      "loss": 0.8385,
+      "step": 850
+    },
+    {
+      "epoch": 1.475250727919767,
+      "grad_norm": 3.425276041030884,
+      "learning_rate": 1e-05,
+      "loss": 0.8994,
+      "step": 855
+    },
+    {
+      "epoch": 1.483877925159064,
+      "grad_norm": 3.498198986053467,
+      "learning_rate": 1e-05,
+      "loss": 0.7993,
+      "step": 860
+    },
+    {
+      "epoch": 1.4925051223983608,
+      "grad_norm": 3.5711140632629395,
+      "learning_rate": 1e-05,
+      "loss": 0.8415,
+      "step": 865
+    },
+    {
+      "epoch": 1.5011323196376578,
+      "grad_norm": 3.628063917160034,
+      "learning_rate": 1e-05,
+      "loss": 0.7544,
+      "step": 870
+    },
+    {
+      "epoch": 1.5097595168769546,
+      "grad_norm": 3.185468912124634,
+      "learning_rate": 1e-05,
+      "loss": 0.8997,
+      "step": 875
+    },
+    {
+      "epoch": 1.5183867141162515,
+      "grad_norm": 3.289321184158325,
+      "learning_rate": 1e-05,
+      "loss": 0.7673,
+      "step": 880
+    },
+    {
+      "epoch": 1.5270139113555483,
+      "grad_norm": 3.4474823474884033,
+      "learning_rate": 1e-05,
+      "loss": 0.8582,
+      "step": 885
+    },
+    {
+      "epoch": 1.5356411085948452,
+      "grad_norm": 4.132356643676758,
+      "learning_rate": 1e-05,
+      "loss": 0.7828,
+      "step": 890
+    },
+    {
+      "epoch": 1.5442683058341422,
+      "grad_norm": 3.1645920276641846,
+      "learning_rate": 1e-05,
+      "loss": 0.8359,
+      "step": 895
+    },
+    {
+      "epoch": 1.552895503073439,
+      "grad_norm": 4.218339443206787,
+      "learning_rate": 1e-05,
+      "loss": 0.8226,
+      "step": 900
+    },
+    {
+      "epoch": 1.5615227003127359,
+      "grad_norm": 3.5735368728637695,
+      "learning_rate": 1e-05,
+      "loss": 0.8219,
+      "step": 905
+    },
+    {
+      "epoch": 1.570149897552033,
+      "grad_norm": 3.3347320556640625,
+      "learning_rate": 1e-05,
+      "loss": 0.8212,
+      "step": 910
+    },
+    {
+      "epoch": 1.5787770947913295,
+      "grad_norm": 3.6230201721191406,
+      "learning_rate": 1e-05,
+      "loss": 0.8036,
+      "step": 915
+    },
+    {
+      "epoch": 1.5874042920306266,
+      "grad_norm": 3.4478774070739746,
+      "learning_rate": 1e-05,
+      "loss": 0.7633,
+      "step": 920
+    },
+    {
+      "epoch": 1.5960314892699234,
+      "grad_norm": 3.934248924255371,
+      "learning_rate": 1e-05,
+      "loss": 0.7673,
+      "step": 925
+    },
+    {
+      "epoch": 1.6046586865092203,
+      "grad_norm": 3.3679585456848145,
+      "learning_rate": 1e-05,
+      "loss": 0.7732,
+      "step": 930
+    },
+    {
+      "epoch": 1.6132858837485173,
+      "grad_norm": 3.6777756214141846,
+      "learning_rate": 1e-05,
+      "loss": 0.7941,
+      "step": 935
+    },
+    {
+      "epoch": 1.621913080987814,
+      "grad_norm": 3.4815165996551514,
+      "learning_rate": 1e-05,
+      "loss": 0.8184,
+      "step": 940
+    },
+    {
+      "epoch": 1.630540278227111,
+      "grad_norm": 3.4345767498016357,
+      "learning_rate": 1e-05,
+      "loss": 0.8496,
+      "step": 945
+    },
+    {
+      "epoch": 1.6391674754664078,
+      "grad_norm": 3.3284542560577393,
+      "learning_rate": 1e-05,
+      "loss": 0.888,
+      "step": 950
+    },
+    {
+      "epoch": 1.6477946727057047,
+      "grad_norm": 3.9973185062408447,
+      "learning_rate": 1e-05,
+      "loss": 0.8551,
+      "step": 955
+    },
+    {
+      "epoch": 1.6564218699450017,
+      "grad_norm": 3.136579751968384,
+      "learning_rate": 1e-05,
+      "loss": 0.8184,
+      "step": 960
+    },
+    {
+      "epoch": 1.6650490671842983,
+      "grad_norm": 3.745621919631958,
+      "learning_rate": 1e-05,
+      "loss": 0.8045,
+      "step": 965
+    },
+    {
+      "epoch": 1.6736762644235954,
+      "grad_norm": 3.2420082092285156,
+      "learning_rate": 1e-05,
+      "loss": 0.7489,
+      "step": 970
+    },
+    {
+      "epoch": 1.6823034616628922,
+      "grad_norm": 3.52974009513855,
+      "learning_rate": 1e-05,
+      "loss": 0.7816,
+      "step": 975
+    },
+    {
+      "epoch": 1.690930658902189,
+      "grad_norm": 3.9357786178588867,
+      "learning_rate": 1e-05,
+      "loss": 0.884,
+      "step": 980
+    },
+    {
+      "epoch": 1.6995578561414861,
+      "grad_norm": 3.807739734649658,
+      "learning_rate": 1e-05,
+      "loss": 0.9077,
+      "step": 985
+    },
+    {
+      "epoch": 1.708185053380783,
+      "grad_norm": 3.0585408210754395,
+      "learning_rate": 1e-05,
+      "loss": 0.8371,
+      "step": 990
+    },
+    {
+      "epoch": 1.7168122506200798,
+      "grad_norm": 3.0159707069396973,
+      "learning_rate": 1e-05,
+      "loss": 0.8028,
+      "step": 995
+    },
+    {
+      "epoch": 1.7254394478593766,
+      "grad_norm": 3.9987428188323975,
+      "learning_rate": 1e-05,
+      "loss": 0.7969,
+      "step": 1000
+    },
+    {
+      "epoch": 1.7340666450986735,
+      "grad_norm": 3.4167709350585938,
+      "learning_rate": 1e-05,
+      "loss": 0.8557,
+      "step": 1005
+    },
+    {
+      "epoch": 1.7426938423379705,
+      "grad_norm": 3.235368251800537,
+      "learning_rate": 1e-05,
+      "loss": 0.7524,
+      "step": 1010
+    },
+    {
+      "epoch": 1.7513210395772674,
+      "grad_norm": 3.3228580951690674,
+      "learning_rate": 1e-05,
+      "loss": 0.8103,
+      "step": 1015
+    },
+    {
+      "epoch": 1.7599482368165642,
+      "grad_norm": 3.2793571949005127,
+      "learning_rate": 1e-05,
+      "loss": 0.7795,
+      "step": 1020
+    },
+    {
+      "epoch": 1.7685754340558613,
+      "grad_norm": 3.4918506145477295,
+      "learning_rate": 1e-05,
+      "loss": 0.8534,
+      "step": 1025
+    },
+    {
+      "epoch": 1.7772026312951579,
+      "grad_norm": 3.5630784034729004,
+      "learning_rate": 1e-05,
+      "loss": 0.8371,
+      "step": 1030
+    },
+    {
+      "epoch": 1.785829828534455,
+      "grad_norm": 3.330843210220337,
+      "learning_rate": 1e-05,
+      "loss": 0.7688,
+      "step": 1035
+    },
+    {
+      "epoch": 1.7944570257737518,
+      "grad_norm": 3.340770721435547,
+      "learning_rate": 1e-05,
+      "loss": 0.7823,
+      "step": 1040
+    },
+    {
+      "epoch": 1.8030842230130486,
+      "grad_norm": 3.48885440826416,
+      "learning_rate": 1e-05,
+      "loss": 0.8135,
+      "step": 1045
+    },
+    {
+      "epoch": 1.8117114202523457,
+      "grad_norm": 3.783395528793335,
+      "learning_rate": 1e-05,
+      "loss": 0.8169,
+      "step": 1050
+    },
+    {
+      "epoch": 1.8203386174916423,
+      "grad_norm": 3.1166722774505615,
+      "learning_rate": 1e-05,
+      "loss": 0.846,
+      "step": 1055
+    },
+    {
+      "epoch": 1.8289658147309393,
+      "grad_norm": 3.627532720565796,
+      "learning_rate": 1e-05,
+      "loss": 0.8055,
+      "step": 1060
+    },
+    {
+      "epoch": 1.8375930119702362,
+      "grad_norm": 3.264894723892212,
+      "learning_rate": 1e-05,
+      "loss": 0.8216,
+      "step": 1065
+    },
+    {
+      "epoch": 1.846220209209533,
+      "grad_norm": 3.652085781097412,
+      "learning_rate": 1e-05,
+      "loss": 0.8663,
+      "step": 1070
+    },
+    {
+      "epoch": 1.85484740644883,
+      "grad_norm": 3.269554376602173,
+      "learning_rate": 1e-05,
+      "loss": 0.8332,
+      "step": 1075
+    },
+    {
+      "epoch": 1.8634746036881267,
+      "grad_norm": 3.3206093311309814,
+      "learning_rate": 1e-05,
+      "loss": 0.8408,
+      "step": 1080
+    },
+    {
+      "epoch": 1.8721018009274237,
+      "grad_norm": 4.427550792694092,
+      "learning_rate": 1e-05,
+      "loss": 0.828,
+      "step": 1085
+    },
+    {
+      "epoch": 1.8807289981667206,
+      "grad_norm": 3.235623359680176,
+      "learning_rate": 1e-05,
+      "loss": 0.8127,
+      "step": 1090
+    },
+    {
+      "epoch": 1.8893561954060174,
+      "grad_norm": 3.4541192054748535,
+      "learning_rate": 1e-05,
+      "loss": 0.8607,
+      "step": 1095
+    },
+    {
+      "epoch": 1.8979833926453145,
+      "grad_norm": 3.4523119926452637,
+      "learning_rate": 1e-05,
+      "loss": 0.9251,
+      "step": 1100
+    },
+    {
+      "epoch": 1.906610589884611,
+      "grad_norm": 3.4309825897216797,
+      "learning_rate": 1e-05,
+      "loss": 0.8388,
+      "step": 1105
+    },
+    {
+      "epoch": 1.9152377871239081,
+      "grad_norm": 3.8286280632019043,
+      "learning_rate": 1e-05,
+      "loss": 0.8036,
+      "step": 1110
+    },
+    {
+      "epoch": 1.923864984363205,
+      "grad_norm": 3.4782304763793945,
+      "learning_rate": 1e-05,
+      "loss": 0.8001,
+      "step": 1115
+    },
+    {
+      "epoch": 1.9324921816025018,
+      "grad_norm": 3.265873432159424,
+      "learning_rate": 1e-05,
+      "loss": 0.775,
+      "step": 1120
+    },
+    {
+      "epoch": 1.9411193788417989,
+      "grad_norm": 3.65558123588562,
+      "learning_rate": 1e-05,
+      "loss": 0.8639,
+      "step": 1125
+    },
+    {
+      "epoch": 1.9497465760810957,
+      "grad_norm": 3.7924466133117676,
+      "learning_rate": 1e-05,
+      "loss": 0.8513,
+      "step": 1130
+    },
+    {
+      "epoch": 1.9583737733203925,
+      "grad_norm": 3.99310564994812,
+      "learning_rate": 1e-05,
+      "loss": 0.77,
+      "step": 1135
+    },
+    {
+      "epoch": 1.9670009705596894,
+      "grad_norm": 3.121508836746216,
+      "learning_rate": 1e-05,
+      "loss": 0.732,
+      "step": 1140
+    },
+    {
+      "epoch": 1.9756281677989862,
+      "grad_norm": 3.812880277633667,
+      "learning_rate": 1e-05,
+      "loss": 0.8029,
+      "step": 1145
+    },
+    {
+      "epoch": 1.9842553650382833,
+      "grad_norm": 3.398724317550659,
+      "learning_rate": 1e-05,
+      "loss": 0.8205,
+      "step": 1150
+    },
+    {
+      "epoch": 1.99288256227758,
+      "grad_norm": 3.3330962657928467,
+      "learning_rate": 1e-05,
+      "loss": 0.7995,
+      "step": 1155
+    },
+    {
+      "epoch": 2.001509759516877,
+      "grad_norm": 4.028820991516113,
+      "learning_rate": 1e-05,
+      "loss": 0.7343,
+      "step": 1160
+    },
+    {
+      "epoch": 2.010136956756174,
+      "grad_norm": 4.329206943511963,
+      "learning_rate": 1e-05,
+      "loss": 0.4413,
+      "step": 1165
+    },
+    {
+      "epoch": 2.0187641539954706,
+      "grad_norm": 7.581143379211426,
+      "learning_rate": 1e-05,
+      "loss": 0.4171,
+      "step": 1170
+    },
+    {
+      "epoch": 2.0273913512347677,
+      "grad_norm": 3.6165997982025146,
+      "learning_rate": 1e-05,
+      "loss": 0.3829,
+      "step": 1175
+    },
+    {
+      "epoch": 2.0360185484740643,
+      "grad_norm": 3.6805825233459473,
+      "learning_rate": 1e-05,
+      "loss": 0.3505,
+      "step": 1180
+    },
+    {
+      "epoch": 2.0446457457133613,
+      "grad_norm": 3.5917325019836426,
+      "learning_rate": 1e-05,
+      "loss": 0.3448,
+      "step": 1185
+    },
+    {
+      "epoch": 2.0532729429526584,
+      "grad_norm": 4.101263046264648,
+      "learning_rate": 1e-05,
+      "loss": 0.3737,
+      "step": 1190
+    },
+    {
+      "epoch": 2.061900140191955,
+      "grad_norm": 4.450289726257324,
+      "learning_rate": 1e-05,
+      "loss": 0.3939,
+      "step": 1195
+    },
+    {
+      "epoch": 2.070527337431252,
+      "grad_norm": 3.6648216247558594,
+      "learning_rate": 1e-05,
+      "loss": 0.3638,
+      "step": 1200
+    },
+    {
+      "epoch": 2.079154534670549,
+      "grad_norm": 4.237855434417725,
+      "learning_rate": 1e-05,
+      "loss": 0.3634,
+      "step": 1205
+    },
+    {
+      "epoch": 2.0877817319098457,
+      "grad_norm": 4.677015781402588,
+      "learning_rate": 1e-05,
+      "loss": 0.4462,
+      "step": 1210
+    },
+    {
+      "epoch": 2.096408929149143,
+      "grad_norm": 3.543541431427002,
+      "learning_rate": 1e-05,
+      "loss": 0.4071,
+      "step": 1215
+    },
+    {
+      "epoch": 2.1050361263884394,
+      "grad_norm": 4.441009521484375,
+      "learning_rate": 1e-05,
+      "loss": 0.3735,
+      "step": 1220
+    },
+    {
+      "epoch": 2.1136633236277365,
+      "grad_norm": 3.5603976249694824,
+      "learning_rate": 1e-05,
+      "loss": 0.383,
+      "step": 1225
+    },
+    {
+      "epoch": 2.1222905208670335,
+      "grad_norm": 3.347618579864502,
+      "learning_rate": 1e-05,
+      "loss": 0.3077,
+      "step": 1230
+    },
+    {
+      "epoch": 2.13091771810633,
+      "grad_norm": 3.57377028465271,
+      "learning_rate": 1e-05,
+      "loss": 0.3774,
+      "step": 1235
+    },
+    {
+      "epoch": 2.139544915345627,
+      "grad_norm": 3.4287397861480713,
+      "learning_rate": 1e-05,
+      "loss": 0.3388,
+      "step": 1240
+    },
+    {
+      "epoch": 2.148172112584924,
+      "grad_norm": 3.9330341815948486,
+      "learning_rate": 1e-05,
+      "loss": 0.3848,
+      "step": 1245
+    },
+    {
+      "epoch": 2.156799309824221,
+      "grad_norm": 3.7925400733947754,
+      "learning_rate": 1e-05,
+      "loss": 0.4173,
+      "step": 1250
+    },
+    {
+      "epoch": 2.165426507063518,
+      "grad_norm": 3.818533420562744,
+      "learning_rate": 1e-05,
+      "loss": 0.3514,
+      "step": 1255
+    },
+    {
+      "epoch": 2.1740537043028145,
+      "grad_norm": 4.525712013244629,
+      "learning_rate": 1e-05,
+      "loss": 0.393,
+      "step": 1260
+    },
+    {
+      "epoch": 2.1826809015421116,
+      "grad_norm": 3.9103636741638184,
+      "learning_rate": 1e-05,
+      "loss": 0.3886,
+      "step": 1265
+    },
+    {
+      "epoch": 2.191308098781408,
+      "grad_norm": 4.36718225479126,
+      "learning_rate": 1e-05,
+      "loss": 0.3943,
+      "step": 1270
+    },
+    {
+      "epoch": 2.1999352960207053,
+      "grad_norm": 4.249504566192627,
+      "learning_rate": 1e-05,
+      "loss": 0.3562,
+      "step": 1275
+    },
+    {
+      "epoch": 2.2085624932600023,
+      "grad_norm": 3.6772570610046387,
+      "learning_rate": 1e-05,
+      "loss": 0.3609,
+      "step": 1280
+    },
+    {
+      "epoch": 2.217189690499299,
+      "grad_norm": 4.188706398010254,
+      "learning_rate": 1e-05,
+      "loss": 0.4031,
+      "step": 1285
+    },
+    {
+      "epoch": 2.225816887738596,
+      "grad_norm": 4.41935920715332,
+      "learning_rate": 1e-05,
+      "loss": 0.3989,
+      "step": 1290
+    },
+    {
+      "epoch": 2.234444084977893,
+      "grad_norm": 3.972947359085083,
+      "learning_rate": 1e-05,
+      "loss": 0.4016,
+      "step": 1295
+    },
+    {
+      "epoch": 2.2430712822171897,
+      "grad_norm": 3.753408193588257,
+      "learning_rate": 1e-05,
+      "loss": 0.3735,
+      "step": 1300
+    },
+    {
+      "epoch": 2.2516984794564867,
+      "grad_norm": 3.882584810256958,
+      "learning_rate": 1e-05,
+      "loss": 0.376,
+      "step": 1305
+    },
+    {
+      "epoch": 2.2603256766957833,
+      "grad_norm": 3.9297213554382324,
+      "learning_rate": 1e-05,
+      "loss": 0.4118,
+      "step": 1310
+    },
+    {
+      "epoch": 2.2689528739350804,
+      "grad_norm": 4.477977752685547,
+      "learning_rate": 1e-05,
+      "loss": 0.4483,
+      "step": 1315
+    },
+    {
+      "epoch": 2.277580071174377,
+      "grad_norm": 4.420772552490234,
+      "learning_rate": 1e-05,
+      "loss": 0.3646,
+      "step": 1320
+    },
+    {
+      "epoch": 2.286207268413674,
+      "grad_norm": 4.0902276039123535,
+      "learning_rate": 1e-05,
+      "loss": 0.3901,
+      "step": 1325
+    },
+    {
+      "epoch": 2.294834465652971,
+      "grad_norm": 3.598499298095703,
+      "learning_rate": 1e-05,
+      "loss": 0.3684,
+      "step": 1330
+    },
+    {
+      "epoch": 2.3034616628922677,
+      "grad_norm": 3.4670894145965576,
+      "learning_rate": 1e-05,
+      "loss": 0.3885,
+      "step": 1335
+    },
+    {
+      "epoch": 2.312088860131565,
+      "grad_norm": 4.802315711975098,
+      "learning_rate": 1e-05,
+      "loss": 0.424,
+      "step": 1340
+    },
+    {
+      "epoch": 2.320716057370862,
+      "grad_norm": 4.187905788421631,
+      "learning_rate": 1e-05,
+      "loss": 0.4194,
+      "step": 1345
+    },
+    {
+      "epoch": 2.3293432546101585,
+      "grad_norm": 3.7030727863311768,
+      "learning_rate": 1e-05,
+      "loss": 0.3982,
+      "step": 1350
+    },
+    {
+      "epoch": 2.3379704518494555,
+      "grad_norm": 4.434927463531494,
+      "learning_rate": 1e-05,
+      "loss": 0.4565,
+      "step": 1355
+    },
+    {
+      "epoch": 2.346597649088752,
+      "grad_norm": 4.65302038192749,
+      "learning_rate": 1e-05,
+      "loss": 0.3995,
+      "step": 1360
+    },
+    {
+      "epoch": 2.355224846328049,
+      "grad_norm": 4.085888862609863,
+      "learning_rate": 1e-05,
+      "loss": 0.426,
+      "step": 1365
+    },
+    {
+      "epoch": 2.3638520435673462,
+      "grad_norm": 4.577415466308594,
+      "learning_rate": 1e-05,
+      "loss": 0.3854,
+      "step": 1370
+    },
+    {
+      "epoch": 2.372479240806643,
+      "grad_norm": 4.255492210388184,
+      "learning_rate": 1e-05,
+      "loss": 0.3808,
+      "step": 1375
+    },
+    {
+      "epoch": 2.38110643804594,
+      "grad_norm": 3.9601423740386963,
+      "learning_rate": 1e-05,
+      "loss": 0.3598,
+      "step": 1380
+    },
+    {
+      "epoch": 2.3897336352852365,
+      "grad_norm": 3.984498977661133,
+      "learning_rate": 1e-05,
+      "loss": 0.4018,
+      "step": 1385
+    },
+    {
+      "epoch": 2.3983608325245336,
+      "grad_norm": 4.054710388183594,
+      "learning_rate": 1e-05,
+      "loss": 0.4105,
+      "step": 1390
+    },
+    {
+      "epoch": 2.4069880297638306,
+      "grad_norm": 3.6106574535369873,
+      "learning_rate": 1e-05,
+      "loss": 0.3832,
+      "step": 1395
+    },
+    {
+      "epoch": 2.4156152270031273,
+      "grad_norm": 3.7792673110961914,
+      "learning_rate": 1e-05,
+      "loss": 0.4288,
+      "step": 1400
+    },
+    {
+      "epoch": 2.4242424242424243,
+      "grad_norm": 4.984143257141113,
+      "learning_rate": 1e-05,
+      "loss": 0.3759,
+      "step": 1405
+    },
+    {
+      "epoch": 2.4328696214817214,
+      "grad_norm": 4.424108028411865,
+      "learning_rate": 1e-05,
+      "loss": 0.3981,
+      "step": 1410
+    },
+    {
+      "epoch": 2.441496818721018,
+      "grad_norm": 3.914560079574585,
+      "learning_rate": 1e-05,
+      "loss": 0.3461,
+      "step": 1415
+    },
+    {
+      "epoch": 2.450124015960315,
+      "grad_norm": 3.667935848236084,
+      "learning_rate": 1e-05,
+      "loss": 0.3898,
+      "step": 1420
+    },
+    {
+      "epoch": 2.4587512131996117,
+      "grad_norm": 3.2983341217041016,
+      "learning_rate": 1e-05,
+      "loss": 0.36,
+      "step": 1425
+    },
+    {
+      "epoch": 2.4673784104389087,
+      "grad_norm": 5.143230438232422,
+      "learning_rate": 1e-05,
+      "loss": 0.4348,
+      "step": 1430
+    },
+    {
+      "epoch": 2.4760056076782053,
+      "grad_norm": 3.818688154220581,
+      "learning_rate": 1e-05,
+      "loss": 0.3582,
+      "step": 1435
+    },
+    {
+      "epoch": 2.4846328049175024,
+      "grad_norm": 4.086068153381348,
+      "learning_rate": 1e-05,
+      "loss": 0.4337,
+      "step": 1440
+    },
+    {
+      "epoch": 2.4932600021567994,
+      "grad_norm": 3.717658281326294,
+      "learning_rate": 1e-05,
+      "loss": 0.4196,
+      "step": 1445
+    },
+    {
+      "epoch": 2.501887199396096,
+      "grad_norm": 3.619304895401001,
+      "learning_rate": 1e-05,
+      "loss": 0.3901,
+      "step": 1450
+    },
+    {
+      "epoch": 2.510514396635393,
+      "grad_norm": 4.011234283447266,
+      "learning_rate": 1e-05,
+      "loss": 0.3864,
+      "step": 1455
+    },
+    {
+      "epoch": 2.51914159387469,
+      "grad_norm": 3.993927478790283,
+      "learning_rate": 1e-05,
+      "loss": 0.3911,
+      "step": 1460
+    },
+    {
+      "epoch": 2.527768791113987,
+      "grad_norm": 4.390587329864502,
+      "learning_rate": 1e-05,
+      "loss": 0.4008,
+      "step": 1465
+    },
+    {
+      "epoch": 2.536395988353284,
+      "grad_norm": 4.767730236053467,
+      "learning_rate": 1e-05,
+      "loss": 0.3738,
+      "step": 1470
+    },
+    {
+      "epoch": 2.5450231855925805,
+      "grad_norm": 3.377415180206299,
+      "learning_rate": 1e-05,
+      "loss": 0.3892,
+      "step": 1475
+    },
+    {
+      "epoch": 2.5536503828318775,
+      "grad_norm": 3.5132322311401367,
+      "learning_rate": 1e-05,
+      "loss": 0.4022,
+      "step": 1480
+    },
+    {
+      "epoch": 2.562277580071174,
+      "grad_norm": 3.9030888080596924,
+      "learning_rate": 1e-05,
+      "loss": 0.436,
+      "step": 1485
+    },
+    {
+      "epoch": 2.570904777310471,
+      "grad_norm": 3.88025164604187,
+      "learning_rate": 1e-05,
+      "loss": 0.4164,
+      "step": 1490
+    },
+    {
+      "epoch": 2.5795319745497682,
+      "grad_norm": 4.036890983581543,
+      "learning_rate": 1e-05,
+      "loss": 0.3914,
+      "step": 1495
+    },
+    {
+      "epoch": 2.588159171789065,
+      "grad_norm": 3.56758189201355,
+      "learning_rate": 1e-05,
+      "loss": 0.3878,
+      "step": 1500
+    },
+    {
+      "epoch": 2.596786369028362,
+      "grad_norm": 3.9577293395996094,
+      "learning_rate": 1e-05,
+      "loss": 0.39,
+      "step": 1505
+    },
+    {
+      "epoch": 2.605413566267659,
+      "grad_norm": 4.392646312713623,
+      "learning_rate": 1e-05,
+      "loss": 0.3989,
+      "step": 1510
+    },
+    {
+      "epoch": 2.6140407635069556,
+      "grad_norm": 4.026916027069092,
+      "learning_rate": 1e-05,
+      "loss": 0.3613,
+      "step": 1515
+    },
+    {
+      "epoch": 2.6226679607462526,
+      "grad_norm": 4.357356548309326,
+      "learning_rate": 1e-05,
+      "loss": 0.3995,
+      "step": 1520
+    },
+    {
+      "epoch": 2.6312951579855497,
+      "grad_norm": 3.703214406967163,
+      "learning_rate": 1e-05,
+      "loss": 0.3376,
+      "step": 1525
+    },
+    {
+      "epoch": 2.6399223552248463,
+      "grad_norm": 3.8401548862457275,
+      "learning_rate": 1e-05,
+      "loss": 0.3861,
+      "step": 1530
+    },
+    {
+      "epoch": 2.6485495524641434,
+      "grad_norm": 3.7734174728393555,
+      "learning_rate": 1e-05,
+      "loss": 0.4193,
+      "step": 1535
+    },
+    {
+      "epoch": 2.65717674970344,
+      "grad_norm": 3.929262161254883,
+      "learning_rate": 1e-05,
+      "loss": 0.4235,
+      "step": 1540
+    },
+    {
+      "epoch": 2.665803946942737,
+      "grad_norm": 4.2804975509643555,
+      "learning_rate": 1e-05,
+      "loss": 0.39,
+      "step": 1545
+    },
+    {
+      "epoch": 2.6744311441820336,
+      "grad_norm": 4.476975440979004,
+      "learning_rate": 1e-05,
+      "loss": 0.4395,
+      "step": 1550
+    },
+    {
+      "epoch": 2.6830583414213307,
+      "grad_norm": 3.710632562637329,
+      "learning_rate": 1e-05,
+      "loss": 0.3956,
+      "step": 1555
+    },
+    {
+      "epoch": 2.6916855386606278,
+      "grad_norm": 5.090831756591797,
+      "learning_rate": 1e-05,
+      "loss": 0.4399,
+      "step": 1560
+    },
+    {
+      "epoch": 2.7003127358999244,
+      "grad_norm": 4.323583602905273,
+      "learning_rate": 1e-05,
+      "loss": 0.4161,
+      "step": 1565
+    },
+    {
+      "epoch": 2.7089399331392214,
+      "grad_norm": 4.236275672912598,
+      "learning_rate": 1e-05,
+      "loss": 0.4128,
+      "step": 1570
+    },
+    {
+      "epoch": 2.7175671303785185,
+      "grad_norm": 4.034735202789307,
+      "learning_rate": 1e-05,
+      "loss": 0.4291,
+      "step": 1575
+    },
+    {
+      "epoch": 2.726194327617815,
+      "grad_norm": 4.452931880950928,
+      "learning_rate": 1e-05,
+      "loss": 0.4198,
+      "step": 1580
+    },
+    {
+      "epoch": 2.734821524857112,
+      "grad_norm": 3.9385485649108887,
+      "learning_rate": 1e-05,
+      "loss": 0.3954,
+      "step": 1585
+    },
+    {
+      "epoch": 2.7434487220964088,
+      "grad_norm": 3.8436594009399414,
+      "learning_rate": 1e-05,
+      "loss": 0.3718,
+      "step": 1590
+    },
+    {
+      "epoch": 2.752075919335706,
+      "grad_norm": 3.962069272994995,
+      "learning_rate": 1e-05,
+      "loss": 0.3488,
+      "step": 1595
+    },
+    {
+      "epoch": 2.7607031165750024,
+      "grad_norm": 4.306760787963867,
+      "learning_rate": 1e-05,
+      "loss": 0.4581,
+      "step": 1600
+    },
+    {
+      "epoch": 2.7693303138142995,
+      "grad_norm": 4.657399654388428,
+      "learning_rate": 1e-05,
+      "loss": 0.4455,
+      "step": 1605
+    },
+    {
+      "epoch": 2.7779575110535966,
+      "grad_norm": 4.039801120758057,
+      "learning_rate": 1e-05,
+      "loss": 0.3918,
+      "step": 1610
+    },
+    {
+      "epoch": 2.786584708292893,
+      "grad_norm": 5.203350067138672,
+      "learning_rate": 1e-05,
+      "loss": 0.4067,
+      "step": 1615
+    },
+    {
+      "epoch": 2.7952119055321902,
+      "grad_norm": 3.9145352840423584,
+      "learning_rate": 1e-05,
+      "loss": 0.4237,
+      "step": 1620
+    },
+    {
+      "epoch": 2.8038391027714873,
+      "grad_norm": 3.5493838787078857,
+      "learning_rate": 1e-05,
+      "loss": 0.3939,
+      "step": 1625
+    },
+    {
+      "epoch": 2.812466300010784,
+      "grad_norm": 4.310389995574951,
+      "learning_rate": 1e-05,
+      "loss": 0.4385,
+      "step": 1630
+    },
+    {
+      "epoch": 2.821093497250081,
+      "grad_norm": 5.43610954284668,
+      "learning_rate": 1e-05,
+      "loss": 0.4109,
+      "step": 1635
+    },
+    {
+      "epoch": 2.829720694489378,
+      "grad_norm": 4.191155433654785,
+      "learning_rate": 1e-05,
+      "loss": 0.4402,
+      "step": 1640
+    },
+    {
+      "epoch": 2.8383478917286746,
+      "grad_norm": 3.9996650218963623,
+      "learning_rate": 1e-05,
+      "loss": 0.444,
+      "step": 1645
+    },
+    {
+      "epoch": 2.8469750889679717,
+      "grad_norm": 10.163007736206055,
+      "learning_rate": 1e-05,
+      "loss": 0.413,
+      "step": 1650
+    },
+    {
+      "epoch": 2.8556022862072683,
+      "grad_norm": 3.8168654441833496,
+      "learning_rate": 1e-05,
+      "loss": 0.4245,
+      "step": 1655
+    },
+    {
+      "epoch": 2.8642294834465654,
+      "grad_norm": 3.9268579483032227,
+      "learning_rate": 1e-05,
+      "loss": 0.4091,
+      "step": 1660
+    },
+    {
+      "epoch": 2.872856680685862,
+      "grad_norm": 4.146000385284424,
+      "learning_rate": 1e-05,
+      "loss": 0.4206,
+      "step": 1665
+    },
+    {
+      "epoch": 2.881483877925159,
+      "grad_norm": 4.102123260498047,
+      "learning_rate": 1e-05,
+      "loss": 0.3868,
+      "step": 1670
+    },
+    {
+      "epoch": 2.890111075164456,
+      "grad_norm": 4.318619728088379,
+      "learning_rate": 1e-05,
+      "loss": 0.4102,
+      "step": 1675
+    },
+    {
+      "epoch": 2.8987382724037527,
+      "grad_norm": 4.203400611877441,
+      "learning_rate": 1e-05,
+      "loss": 0.4349,
+      "step": 1680
+    },
+    {
+      "epoch": 2.9073654696430498,
+      "grad_norm": 4.164772987365723,
+      "learning_rate": 1e-05,
+      "loss": 0.4298,
+      "step": 1685
+    },
+    {
+      "epoch": 2.915992666882347,
+      "grad_norm": 4.088376998901367,
+      "learning_rate": 1e-05,
+      "loss": 0.431,
+      "step": 1690
+    },
+    {
+      "epoch": 2.9246198641216434,
+      "grad_norm": 4.593983173370361,
+      "learning_rate": 1e-05,
+      "loss": 0.4373,
+      "step": 1695
+    },
+    {
+      "epoch": 2.9332470613609405,
+      "grad_norm": 4.246501445770264,
+      "learning_rate": 1e-05,
+      "loss": 0.4239,
+      "step": 1700
+    },
+    {
+      "epoch": 2.941874258600237,
+      "grad_norm": 3.8412177562713623,
+      "learning_rate": 1e-05,
+      "loss": 0.4277,
+      "step": 1705
+    },
+    {
+      "epoch": 2.950501455839534,
+      "grad_norm": 3.7681996822357178,
+      "learning_rate": 1e-05,
+      "loss": 0.4302,
+      "step": 1710
+    },
+    {
+      "epoch": 2.9591286530788308,
+      "grad_norm": 4.009012699127197,
+      "learning_rate": 1e-05,
+      "loss": 0.3822,
+      "step": 1715
+    },
+    {
+      "epoch": 2.967755850318128,
+      "grad_norm": 4.4853434562683105,
+      "learning_rate": 1e-05,
+      "loss": 0.4402,
+      "step": 1720
+    },
+    {
+      "epoch": 2.976383047557425,
+      "grad_norm": 3.743802785873413,
+      "learning_rate": 1e-05,
+      "loss": 0.3803,
+      "step": 1725
+    },
+    {
+      "epoch": 2.9850102447967215,
+      "grad_norm": 4.453946113586426,
+      "learning_rate": 1e-05,
+      "loss": 0.4109,
+      "step": 1730
+    },
+    {
+      "epoch": 2.9936374420360186,
+      "grad_norm": 3.8939476013183594,
+      "learning_rate": 1e-05,
+      "loss": 0.4232,
+      "step": 1735
+    },
+    {
+      "epoch": 2.9970883209317374,
+      "step": 1737,
+      "total_flos": 2.5609799507175014e+18,
+      "train_loss": 0.8316220592066139,
+      "train_runtime": 17540.1657,
+      "train_samples_per_second": 6.344,
+      "train_steps_per_second": 0.099
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 1737,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 3,
+  "save_steps": 200,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.5609799507175014e+18,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e7a9c64e1061d6daa7d8c8efc1ab79ab9efdc9c9da5ca32c4fbd246f0c8ac75b
+size 5368