Upload folder using huggingface_hub

.gitattributes

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

added_tokens.json

@@ -0,0 +1,28 @@
+{
+  "</think>": 151668,
+  "</tool_call>": 151658,
+  "</tool_response>": 151666,
+  "<think>": 151667,
+  "<tool_call>": 151657,
+  "<tool_response>": 151665,
+  "<|box_end|>": 151649,
+  "<|box_start|>": 151648,
+  "<|endoftext|>": 151643,
+  "<|file_sep|>": 151664,
+  "<|fim_middle|>": 151660,
+  "<|fim_pad|>": 151662,
+  "<|fim_prefix|>": 151659,
+  "<|fim_suffix|>": 151661,
+  "<|im_end|>": 151645,
+  "<|im_start|>": 151644,
+  "<|image_pad|>": 151655,
+  "<|object_ref_end|>": 151647,
+  "<|object_ref_start|>": 151646,
+  "<|quad_end|>": 151651,
+  "<|quad_start|>": 151650,
+  "<|repo_name|>": 151663,
+  "<|video_pad|>": 151656,
+  "<|vision_end|>": 151653,
+  "<|vision_pad|>": 151654,
+  "<|vision_start|>": 151652
+}

config.json

@@ -0,0 +1,38 @@
+{
+  "architectures": [
+    "Qwen3MoeForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "bos_token_id": 151643,
+  "decoder_sparse_step": 1,
+  "eos_token_id": 151643,
+  "head_dim": 128,
+  "hidden_act": "silu",
+  "hidden_size": 2048,
+  "initializer_range": 0.02,
+  "intermediate_size": 6144,
+  "max_position_embeddings": 32768,
+  "max_window_layers": 48,
+  "mlp_only_layers": [],
+  "model_type": "qwen3_moe",
+  "moe_intermediate_size": 768,
+  "norm_topk_prob": true,
+  "num_attention_heads": 32,
+  "num_experts": 128,
+  "num_experts_per_tok": 8,
+  "num_hidden_layers": 48,
+  "num_key_value_heads": 4,
+  "output_router_logits": false,
+  "rms_norm_eps": 1e-06,
+  "rope_scaling": null,
+  "rope_theta": 1000000.0,
+  "router_aux_loss_coef": 0.001,
+  "sliding_window": null,
+  "tie_word_embeddings": false,
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.51.3",
+  "use_cache": false,
+  "use_sliding_window": false,
+  "vocab_size": 151936
+}

generation_config.json

@@ -0,0 +1,6 @@
+{
+  "bos_token_id": 151643,
+  "eos_token_id": 151643,
+  "max_new_tokens": 2048,
+  "transformers_version": "4.51.3"
+}

global_step521/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ce24cbd9b5d232185a3a183e55cc78e989c59e41fe53de52a7393b7b376d1354
+size 45798193403

global_step521/bf16_zero_pp_rank_1_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e7572f09c1fdd2fa98cdba1858a336b6b337336ce7c566d891fcc81da15db939
+size 45798193403

global_step521/bf16_zero_pp_rank_2_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c4351750df56a6ca304d6667ffa85e80881e890ca719c554f032147dc63dd79c
+size 45798193403

global_step521/bf16_zero_pp_rank_3_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b308865d5b16fa59e8c6bad82be744c67dfcd7b85ec4c774e743db93f810c89f
+size 45798193403

global_step521/bf16_zero_pp_rank_4_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0f7eb2ccce3c870c962a3a3604f28fe8eb9236529dad92c45a92848dc084c822
+size 45798193403

global_step521/bf16_zero_pp_rank_5_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e094ee088310e18c2bf5c7f896965e061fa077faccbaf0bcd27ef65b5a86d333
+size 45798193403

global_step521/bf16_zero_pp_rank_6_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aa6b599a6196b32685825592ba7a81caf3872143aed960653534ef2fc5c60d55
+size 45798193403

global_step521/bf16_zero_pp_rank_7_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9f474a1b78decd228f7641e41c393fb3d27889fe49d3887c93174e6a4f501545
+size 45798193403

global_step521/zero_pp_rank_0_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f780276e960e42bf712d3318a395e01fa5cf7a712a2f0329b8c1694e73231a80
+size 10415534

global_step521/zero_pp_rank_1_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:148aebf45019375590f40f2fddbe83a1c746f9e9e70452b379a8f19b485c4096
+size 10415534

global_step521/zero_pp_rank_2_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f764b48f0b44d6522106b23d90394ee04a6674f714a6889869c910c86936c2d7
+size 10415534

global_step521/zero_pp_rank_3_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2d1ac904cc4bdf8d0296875adcd24f03ba34ccfbe9652e27b03ad2f37afe1bf1
+size 10415534

global_step521/zero_pp_rank_4_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d37a758d6f1d7adaf082c5e35b0934de8dc24fe2ed1610bd5e2fec23a21e0f5b
+size 10415534

global_step521/zero_pp_rank_5_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d424f76fd9a93a9b6a3e964114785fb7054ceedfd1d5d64913ed2570438e95aa
+size 10415534

global_step521/zero_pp_rank_6_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0eb6d599a6b2bd9c4f460e99cd0aa1b6e146df442472cf594752448c9fd7f65e
+size 10415534

global_step521/zero_pp_rank_7_mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e0fc95cae2175b3b6ddf6a66f4938ac3cda353fbe08f7d77c593d630e4b3c604
+size 10415534

latest

@@ -0,0 +1 @@
+global_step521

model-00001-of-00013.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bf2afe7e4c8ce6c717586ec2b7ef36152ab18431e0328b8b2e552a3c2ffd9d18
+size 4997184968

model-00002-of-00013.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4e5ce043f10d1352af808f9cb213461953c5970df5e5202697e28c0625dcf400
+size 4997741608

model-00003-of-00013.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d27e7f321f12e8c1e229eff580c7d7b7c0961c52175f82c4eb60b72d1c0b9cf1
+size 4997742208

model-00004-of-00013.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7c279168aa758af0fb39ad019654340f45c89983385a7336f98aab2f74f45ced
+size 4997743184

model-00005-of-00013.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:00c24c2fda0b361d273b44925d13deda6ded7ab3c0d0abe5d242a498d95019e3
+size 4997743184

model-00006-of-00013.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:628f6854a6411c55c5407ab31021407c00c6df59705a211df5f3ba1d004c2449
+size 4997743184

model-00007-of-00013.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7e3ea1c1121e2ee01d18865a977a5ef6056b0256d1973aa162c66badf26f6659
+size 4997743184

model-00008-of-00013.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9cb97e82819d1d8c486608104119e8911a16a90b01d35ec0daa7e07c2916edac
+size 4997743184

model-00009-of-00013.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:507d28f56ceace257399694e542ce098e8ad42f1c728ad7373a44d0b83f7a904
+size 4997743184

model-00010-of-00013.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:18fefff3cfbfe9262662958e71b18678baa6516b6534049af6bfa6147735d41e
+size 4997743184

model-00011-of-00013.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f490d6b163fceec3de435d686fe6bdf80a27dcc2a9308faf056b8027217e42f0
+size 4997743184

model-00012-of-00013.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6e416e9145943c45fcddfc32b434ae301bf1c5abb5e359a96357f6c557848269
+size 4997743184

model-00013-of-00013.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:58e9120189365e087520e922fae9098fd0f3916079f6ba3f2bdc46b304421dc2
+size 1094220288

rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:478b41e9f26d338fd8f896e08cad1adab7c423b61f1b45754113bc78d256a3f9
+size 16389

rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ce29a8767a7d907dd24987aa2c3e654d4317f3042fbc13b5b72cadb46d43311a
+size 16389

rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:61a48db011646b4e9a867bf12f4a233cad5dfbfe309686f8996c250196d3783a
+size 16389

rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b9562ee822472a4f01dcd6349ab3d1ef42a48915fe3b92e843a0c37db53c8421
+size 16389

rng_state_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e7d2767d83c3bf27f12db022b0632e2c4f8c164274ba75e380cf18f9d5f21819
+size 16389

rng_state_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:76816358d4e5db8149d60d55234db658d67a13c0c1ce05d7404cf7125a676a5c
+size 16389

rng_state_6.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1562e7520c977d178183d641f70abcf3f57da2489938756cfbebf9b6e6c1a9fd
+size 16389

rng_state_7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a6b6cabaed045c5398cd1b732f7ec48bd363f3b43cd24e0e70e641a42bd00c28
+size 16389

scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c4db2f3374aacaf5fda6fda21f8d36806d0ea19b20646a97be1c8e531b74e70e
+size 1465

special_tokens_map.json

@@ -0,0 +1,31 @@
+{
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "eos_token": {
+    "content": "<|im_end|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aeb13307a71acd8fe81861d94ad54ab689df773318809eed3cbe794b4492dae4
+size 11422654

tokenizer_config.json

@@ -0,0 +1,241 @@
+{
+  "add_bos_token": false,
+  "add_prefix_space": false,
+  "added_tokens_decoder": {
+    "151643": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151644": {
+      "content": "<|im_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151645": {
+      "content": "<|im_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151646": {
+      "content": "<|object_ref_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151647": {
+      "content": "<|object_ref_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151648": {
+      "content": "<|box_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151649": {
+      "content": "<|box_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151650": {
+      "content": "<|quad_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151651": {
+      "content": "<|quad_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151652": {
+      "content": "<|vision_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151653": {
+      "content": "<|vision_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151654": {
+      "content": "<|vision_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151655": {
+      "content": "<|image_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151656": {
+      "content": "<|video_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151657": {
+      "content": "<tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151658": {
+      "content": "</tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151659": {
+      "content": "<|fim_prefix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151660": {
+      "content": "<|fim_middle|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151661": {
+      "content": "<|fim_suffix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151662": {
+      "content": "<|fim_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151663": {
+      "content": "<|repo_name|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151664": {
+      "content": "<|file_sep|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151665": {
+      "content": "<tool_response>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151666": {
+      "content": "</tool_response>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151667": {
+      "content": "<think>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151668": {
+      "content": "</think>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    }
+  },
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "bos_token": null,
+  "chat_template": "{%- if tools %}\n    {{- '<|im_start|>system\\n' }}\n    {%- if messages[0].role == 'system' %}\n        {{- messages[0].content + '\\n\\n' }}\n    {%- endif %}\n    {{- \"# Tools\\n\\nYou may call one or more functions to assist with the user query.\\n\\nYou are provided with function signatures within <tools></tools> XML tags:\\n<tools>\" }}\n    {%- for tool in tools %}\n        {{- \"\\n\" }}\n        {{- tool | tojson }}\n    {%- endfor %}\n    {{- \"\\n</tools>\\n\\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\\n<tool_call>\\n{\\\"name\\\": <function-name>, \\\"arguments\\\": <args-json-object>}\\n</tool_call><|im_end|>\\n\" }}\n{%- else %}\n    {%- if messages[0].role == 'system' %}\n        {{- '<|im_start|>system\\n' + messages[0].content + '<|im_end|>\\n' }}\n    {%- endif %}\n{%- endif %}\n{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}\n{%- for message in messages[::-1] %}\n    {%- set index = (messages|length - 1) - loop.index0 %}\n    {%- if ns.multi_step_tool and message.role == \"user\" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}\n        {%- set ns.multi_step_tool = false %}\n        {%- set ns.last_query_index = index %}\n    {%- endif %}\n{%- endfor %}\n{%- for message in messages %}\n    {%- if (message.role == \"user\") or (message.role == \"system\" and not loop.first) %}\n        {{- '<|im_start|>' + message.role + '\\n' + message.content + '<|im_end|>' + '\\n' }}\n    {%- elif message.role == \"assistant\" %}\n        {%- set content = message.content %}\n        {%- set reasoning_content = '' %}\n        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}\n            {%- set reasoning_content = message.reasoning_content %}\n        {%- else %}\n            {%- if '</think>' in message.content %}\n                {%- set content = message.content.split('</think>')[-1].lstrip('\\n') %}\n                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\\n').split('<think>')[-1].lstrip('\\n') %}\n            {%- endif %}\n        {%- endif %}\n        {%- if loop.index0 > ns.last_query_index %}\n            {%- if loop.last or (not loop.last and reasoning_content) %}\n                {{- '<|im_start|>' + message.role + '\\n<think>\\n' + reasoning_content.strip('\\n') + '\\n</think>\\n\\n' + content.lstrip('\\n') }}\n            {%- else %}\n                {{- '<|im_start|>' + message.role + '\\n' + content }}\n            {%- endif %}\n        {%- else %}\n            {{- '<|im_start|>' + message.role + '\\n' + content }}\n        {%- endif %}\n        {%- if message.tool_calls %}\n            {%- for tool_call in message.tool_calls %}\n                {%- if (loop.first and content) or (not loop.first) %}\n                    {{- '\\n' }}\n                {%- endif %}\n                {%- if tool_call.function %}\n                    {%- set tool_call = tool_call.function %}\n                {%- endif %}\n                {{- '<tool_call>\\n{\"name\": \"' }}\n                {{- tool_call.name }}\n                {{- '\", \"arguments\": ' }}\n                {%- if tool_call.arguments is string %}\n                    {{- tool_call.arguments }}\n                {%- else %}\n                    {{- tool_call.arguments | tojson }}\n                {%- endif %}\n                {{- '}\\n</tool_call>' }}\n            {%- endfor %}\n        {%- endif %}\n        {{- '<|im_end|>\\n' }}\n    {%- elif message.role == \"tool\" %}\n        {%- if loop.first or (messages[loop.index0 - 1].role != \"tool\") %}\n            {{- '<|im_start|>user' }}\n        {%- endif %}\n        {{- '\\n<tool_response>\\n' }}\n        {{- message.content }}\n        {{- '\\n</tool_response>' }}\n        {%- if loop.last or (messages[loop.index0 + 1].role != \"tool\") %}\n            {{- '<|im_end|>\\n' }}\n        {%- endif %}\n    {%- endif %}\n{%- endfor %}\n{%- if add_generation_prompt %}\n    {{- '<|im_start|>assistant\\n' }}\n    {%- if enable_thinking is defined and enable_thinking is false %}\n        {{- '<think>\\n\\n</think>\\n\\n' }}\n    {%- endif %}\n{%- endif %}",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|im_end|>",
+  "errors": "replace",
+  "extra_special_tokens": {},
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "padding_side": "right",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}

trainer_state.json

@@ -0,0 +1,3681 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.33344,
+  "eval_steps": 500,
+  "global_step": 521,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.00064,
+      "grad_norm": 2.353584373604404,
+      "learning_rate": 0.0,
+      "loss": 0.5764,
+      "step": 1
+    },
+    {
+      "epoch": 0.00128,
+      "grad_norm": 2.5677274470706184,
+      "learning_rate": 3.79746835443038e-07,
+      "loss": 0.6154,
+      "step": 2
+    },
+    {
+      "epoch": 0.00192,
+      "grad_norm": 2.434815038353528,
+      "learning_rate": 7.59493670886076e-07,
+      "loss": 0.5762,
+      "step": 3
+    },
+    {
+      "epoch": 0.00256,
+      "grad_norm": 2.4265789349847133,
+      "learning_rate": 1.139240506329114e-06,
+      "loss": 0.5898,
+      "step": 4
+    },
+    {
+      "epoch": 0.0032,
+      "grad_norm": 2.4968640172692007,
+      "learning_rate": 1.518987341772152e-06,
+      "loss": 0.5854,
+      "step": 5
+    },
+    {
+      "epoch": 0.00384,
+      "grad_norm": 2.6270464110646,
+      "learning_rate": 1.8987341772151901e-06,
+      "loss": 0.6063,
+      "step": 6
+    },
+    {
+      "epoch": 0.00448,
+      "grad_norm": 2.3240861061539664,
+      "learning_rate": 2.278481012658228e-06,
+      "loss": 0.5713,
+      "step": 7
+    },
+    {
+      "epoch": 0.00512,
+      "grad_norm": 2.277096457178879,
+      "learning_rate": 2.6582278481012658e-06,
+      "loss": 0.5981,
+      "step": 8
+    },
+    {
+      "epoch": 0.00576,
+      "grad_norm": 2.0556904877285276,
+      "learning_rate": 3.037974683544304e-06,
+      "loss": 0.5783,
+      "step": 9
+    },
+    {
+      "epoch": 0.0064,
+      "grad_norm": 1.8883532047436715,
+      "learning_rate": 3.4177215189873417e-06,
+      "loss": 0.5626,
+      "step": 10
+    },
+    {
+      "epoch": 0.00704,
+      "grad_norm": 1.4054834462317158,
+      "learning_rate": 3.7974683544303802e-06,
+      "loss": 0.5762,
+      "step": 11
+    },
+    {
+      "epoch": 0.00768,
+      "grad_norm": 1.2672259034804543,
+      "learning_rate": 4.1772151898734175e-06,
+      "loss": 0.5447,
+      "step": 12
+    },
+    {
+      "epoch": 0.00832,
+      "grad_norm": 1.1315187874621562,
+      "learning_rate": 4.556962025316456e-06,
+      "loss": 0.5003,
+      "step": 13
+    },
+    {
+      "epoch": 0.00896,
+      "grad_norm": 1.2442008617105749,
+      "learning_rate": 4.936708860759494e-06,
+      "loss": 0.5235,
+      "step": 14
+    },
+    {
+      "epoch": 0.0096,
+      "grad_norm": 1.1708354130739331,
+      "learning_rate": 5.3164556962025316e-06,
+      "loss": 0.4659,
+      "step": 15
+    },
+    {
+      "epoch": 0.01024,
+      "grad_norm": 1.3068048804077694,
+      "learning_rate": 5.69620253164557e-06,
+      "loss": 0.4601,
+      "step": 16
+    },
+    {
+      "epoch": 0.01088,
+      "grad_norm": 1.487196594526998,
+      "learning_rate": 6.075949367088608e-06,
+      "loss": 0.487,
+      "step": 17
+    },
+    {
+      "epoch": 0.01152,
+      "grad_norm": 1.3702547576006188,
+      "learning_rate": 6.455696202531646e-06,
+      "loss": 0.4829,
+      "step": 18
+    },
+    {
+      "epoch": 0.01216,
+      "grad_norm": 0.985500593752308,
+      "learning_rate": 6.835443037974683e-06,
+      "loss": 0.4463,
+      "step": 19
+    },
+    {
+      "epoch": 0.0128,
+      "grad_norm": 0.9909916347896555,
+      "learning_rate": 7.215189873417722e-06,
+      "loss": 0.4554,
+      "step": 20
+    },
+    {
+      "epoch": 0.01344,
+      "grad_norm": 1.1838910835370198,
+      "learning_rate": 7.5949367088607605e-06,
+      "loss": 0.4622,
+      "step": 21
+    },
+    {
+      "epoch": 0.01408,
+      "grad_norm": 1.6876677020755413,
+      "learning_rate": 7.974683544303797e-06,
+      "loss": 0.3952,
+      "step": 22
+    },
+    {
+      "epoch": 0.01472,
+      "grad_norm": 1.0417104358020024,
+      "learning_rate": 8.354430379746835e-06,
+      "loss": 0.4105,
+      "step": 23
+    },
+    {
+      "epoch": 0.01536,
+      "grad_norm": 0.703419627437881,
+      "learning_rate": 8.734177215189873e-06,
+      "loss": 0.3797,
+      "step": 24
+    },
+    {
+      "epoch": 0.016,
+      "grad_norm": 0.7189072691985104,
+      "learning_rate": 9.113924050632912e-06,
+      "loss": 0.4206,
+      "step": 25
+    },
+    {
+      "epoch": 0.01664,
+      "grad_norm": 0.7280445280679197,
+      "learning_rate": 9.49367088607595e-06,
+      "loss": 0.3963,
+      "step": 26
+    },
+    {
+      "epoch": 0.01728,
+      "grad_norm": 0.5914861117650524,
+      "learning_rate": 9.873417721518988e-06,
+      "loss": 0.3928,
+      "step": 27
+    },
+    {
+      "epoch": 0.01792,
+      "grad_norm": 0.6824922709052414,
+      "learning_rate": 1.0253164556962025e-05,
+      "loss": 0.3909,
+      "step": 28
+    },
+    {
+      "epoch": 0.01856,
+      "grad_norm": 0.6157178643639211,
+      "learning_rate": 1.0632911392405063e-05,
+      "loss": 0.3778,
+      "step": 29
+    },
+    {
+      "epoch": 0.0192,
+      "grad_norm": 0.7040224694990468,
+      "learning_rate": 1.1012658227848103e-05,
+      "loss": 0.4034,
+      "step": 30
+    },
+    {
+      "epoch": 0.01984,
+      "grad_norm": 0.6042406871448358,
+      "learning_rate": 1.139240506329114e-05,
+      "loss": 0.3991,
+      "step": 31
+    },
+    {
+      "epoch": 0.02048,
+      "grad_norm": 0.5581052443719693,
+      "learning_rate": 1.1772151898734176e-05,
+      "loss": 0.4011,
+      "step": 32
+    },
+    {
+      "epoch": 0.02112,
+      "grad_norm": 0.4266761178757486,
+      "learning_rate": 1.2151898734177216e-05,
+      "loss": 0.3639,
+      "step": 33
+    },
+    {
+      "epoch": 0.02176,
+      "grad_norm": 0.49600505321594285,
+      "learning_rate": 1.2531645569620253e-05,
+      "loss": 0.3784,
+      "step": 34
+    },
+    {
+      "epoch": 0.0224,
+      "grad_norm": 0.45932753795907266,
+      "learning_rate": 1.2911392405063291e-05,
+      "loss": 0.3751,
+      "step": 35
+    },
+    {
+      "epoch": 0.02304,
+      "grad_norm": 0.5363867843128812,
+      "learning_rate": 1.3291139240506329e-05,
+      "loss": 0.3865,
+      "step": 36
+    },
+    {
+      "epoch": 0.02368,
+      "grad_norm": 0.5488419069165299,
+      "learning_rate": 1.3670886075949367e-05,
+      "loss": 0.3787,
+      "step": 37
+    },
+    {
+      "epoch": 0.02432,
+      "grad_norm": 0.48587043713463085,
+      "learning_rate": 1.4050632911392406e-05,
+      "loss": 0.3712,
+      "step": 38
+    },
+    {
+      "epoch": 0.02496,
+      "grad_norm": 0.4440990518273777,
+      "learning_rate": 1.4430379746835444e-05,
+      "loss": 0.3652,
+      "step": 39
+    },
+    {
+      "epoch": 0.0256,
+      "grad_norm": 0.3751498370487217,
+      "learning_rate": 1.4810126582278482e-05,
+      "loss": 0.3569,
+      "step": 40
+    },
+    {
+      "epoch": 0.02624,
+      "grad_norm": 0.43725343772159886,
+      "learning_rate": 1.5189873417721521e-05,
+      "loss": 0.3158,
+      "step": 41
+    },
+    {
+      "epoch": 0.02688,
+      "grad_norm": 0.4888026864502293,
+      "learning_rate": 1.5569620253164557e-05,
+      "loss": 0.3442,
+      "step": 42
+    },
+    {
+      "epoch": 0.02752,
+      "grad_norm": 0.44586123200596317,
+      "learning_rate": 1.5949367088607595e-05,
+      "loss": 0.3328,
+      "step": 43
+    },
+    {
+      "epoch": 0.02816,
+      "grad_norm": 0.49767798630217713,
+      "learning_rate": 1.6329113924050632e-05,
+      "loss": 0.351,
+      "step": 44
+    },
+    {
+      "epoch": 0.0288,
+      "grad_norm": 0.4336104979866255,
+      "learning_rate": 1.670886075949367e-05,
+      "loss": 0.3612,
+      "step": 45
+    },
+    {
+      "epoch": 0.02944,
+      "grad_norm": 0.4981279303284916,
+      "learning_rate": 1.708860759493671e-05,
+      "loss": 0.3543,
+      "step": 46
+    },
+    {
+      "epoch": 0.03008,
+      "grad_norm": 0.3991975649542343,
+      "learning_rate": 1.7468354430379746e-05,
+      "loss": 0.346,
+      "step": 47
+    },
+    {
+      "epoch": 0.03072,
+      "grad_norm": 0.4340035974965301,
+      "learning_rate": 1.7848101265822783e-05,
+      "loss": 0.3563,
+      "step": 48
+    },
+    {
+      "epoch": 0.03136,
+      "grad_norm": 0.40441725662186223,
+      "learning_rate": 1.8227848101265824e-05,
+      "loss": 0.3378,
+      "step": 49
+    },
+    {
+      "epoch": 0.032,
+      "grad_norm": 0.4015157764097668,
+      "learning_rate": 1.860759493670886e-05,
+      "loss": 0.3477,
+      "step": 50
+    },
+    {
+      "epoch": 0.03264,
+      "grad_norm": 0.3897529365297187,
+      "learning_rate": 1.89873417721519e-05,
+      "loss": 0.3341,
+      "step": 51
+    },
+    {
+      "epoch": 0.03328,
+      "grad_norm": 0.39864724195969276,
+      "learning_rate": 1.9367088607594938e-05,
+      "loss": 0.3322,
+      "step": 52
+    },
+    {
+      "epoch": 0.03392,
+      "grad_norm": 0.365402956200434,
+      "learning_rate": 1.9746835443037975e-05,
+      "loss": 0.297,
+      "step": 53
+    },
+    {
+      "epoch": 0.03456,
+      "grad_norm": 0.4449097185367681,
+      "learning_rate": 2.0126582278481013e-05,
+      "loss": 0.3424,
+      "step": 54
+    },
+    {
+      "epoch": 0.0352,
+      "grad_norm": 0.43181805843495463,
+      "learning_rate": 2.050632911392405e-05,
+      "loss": 0.3351,
+      "step": 55
+    },
+    {
+      "epoch": 0.03584,
+      "grad_norm": 0.392546267080926,
+      "learning_rate": 2.0886075949367092e-05,
+      "loss": 0.334,
+      "step": 56
+    },
+    {
+      "epoch": 0.03648,
+      "grad_norm": 0.38060689632546174,
+      "learning_rate": 2.1265822784810126e-05,
+      "loss": 0.3328,
+      "step": 57
+    },
+    {
+      "epoch": 0.03712,
+      "grad_norm": 0.389736897725578,
+      "learning_rate": 2.1645569620253164e-05,
+      "loss": 0.351,
+      "step": 58
+    },
+    {
+      "epoch": 0.03776,
+      "grad_norm": 0.42262406549661324,
+      "learning_rate": 2.2025316455696205e-05,
+      "loss": 0.3378,
+      "step": 59
+    },
+    {
+      "epoch": 0.0384,
+      "grad_norm": 0.43235384034762964,
+      "learning_rate": 2.240506329113924e-05,
+      "loss": 0.3308,
+      "step": 60
+    },
+    {
+      "epoch": 0.03904,
+      "grad_norm": 0.3945507271925106,
+      "learning_rate": 2.278481012658228e-05,
+      "loss": 0.3388,
+      "step": 61
+    },
+    {
+      "epoch": 0.03968,
+      "grad_norm": 0.40849904164663575,
+      "learning_rate": 2.3164556962025318e-05,
+      "loss": 0.3191,
+      "step": 62
+    },
+    {
+      "epoch": 0.04032,
+      "grad_norm": 0.4559537771952015,
+      "learning_rate": 2.3544303797468353e-05,
+      "loss": 0.3245,
+      "step": 63
+    },
+    {
+      "epoch": 0.04096,
+      "grad_norm": 0.40069809350325963,
+      "learning_rate": 2.3924050632911394e-05,
+      "loss": 0.3566,
+      "step": 64
+    },
+    {
+      "epoch": 0.0416,
+      "grad_norm": 0.37133926498283504,
+      "learning_rate": 2.430379746835443e-05,
+      "loss": 0.3147,
+      "step": 65
+    },
+    {
+      "epoch": 0.04224,
+      "grad_norm": 0.4628360444788038,
+      "learning_rate": 2.468354430379747e-05,
+      "loss": 0.3498,
+      "step": 66
+    },
+    {
+      "epoch": 0.04288,
+      "grad_norm": 0.38282967415212005,
+      "learning_rate": 2.5063291139240507e-05,
+      "loss": 0.3171,
+      "step": 67
+    },
+    {
+      "epoch": 0.04352,
+      "grad_norm": 0.4221420372731314,
+      "learning_rate": 2.5443037974683545e-05,
+      "loss": 0.3288,
+      "step": 68
+    },
+    {
+      "epoch": 0.04416,
+      "grad_norm": 0.4014634593404327,
+      "learning_rate": 2.5822784810126582e-05,
+      "loss": 0.2967,
+      "step": 69
+    },
+    {
+      "epoch": 0.0448,
+      "grad_norm": 0.4226874979052922,
+      "learning_rate": 2.620253164556962e-05,
+      "loss": 0.3311,
+      "step": 70
+    },
+    {
+      "epoch": 0.04544,
+      "grad_norm": 0.3691551141136815,
+      "learning_rate": 2.6582278481012658e-05,
+      "loss": 0.3154,
+      "step": 71
+    },
+    {
+      "epoch": 0.04608,
+      "grad_norm": 0.4930935215298506,
+      "learning_rate": 2.69620253164557e-05,
+      "loss": 0.3248,
+      "step": 72
+    },
+    {
+      "epoch": 0.04672,
+      "grad_norm": 0.40344283789369767,
+      "learning_rate": 2.7341772151898733e-05,
+      "loss": 0.3146,
+      "step": 73
+    },
+    {
+      "epoch": 0.04736,
+      "grad_norm": 0.40136750473831273,
+      "learning_rate": 2.7721518987341774e-05,
+      "loss": 0.3084,
+      "step": 74
+    },
+    {
+      "epoch": 0.048,
+      "grad_norm": 0.433002156207923,
+      "learning_rate": 2.8101265822784812e-05,
+      "loss": 0.2992,
+      "step": 75
+    },
+    {
+      "epoch": 0.04864,
+      "grad_norm": 0.45976406035512724,
+      "learning_rate": 2.8481012658227846e-05,
+      "loss": 0.3203,
+      "step": 76
+    },
+    {
+      "epoch": 0.04928,
+      "grad_norm": 0.46555202230017373,
+      "learning_rate": 2.8860759493670888e-05,
+      "loss": 0.326,
+      "step": 77
+    },
+    {
+      "epoch": 0.04992,
+      "grad_norm": 0.40061810400853354,
+      "learning_rate": 2.9240506329113925e-05,
+      "loss": 0.2902,
+      "step": 78
+    },
+    {
+      "epoch": 0.05056,
+      "grad_norm": 0.44228162263369347,
+      "learning_rate": 2.9620253164556963e-05,
+      "loss": 0.3235,
+      "step": 79
+    },
+    {
+      "epoch": 0.0512,
+      "grad_norm": 0.4587098467220033,
+      "learning_rate": 3e-05,
+      "loss": 0.3135,
+      "step": 80
+    },
+    {
+      "epoch": 0.05184,
+      "grad_norm": 0.42643590142876026,
+      "learning_rate": 2.9999966342756535e-05,
+      "loss": 0.3222,
+      "step": 81
+    },
+    {
+      "epoch": 0.05248,
+      "grad_norm": 0.4312879132975502,
+      "learning_rate": 2.9999865371177178e-05,
+      "loss": 0.3161,
+      "step": 82
+    },
+    {
+      "epoch": 0.05312,
+      "grad_norm": 0.44554806900527943,
+      "learning_rate": 2.9999697085715054e-05,
+      "loss": 0.3156,
+      "step": 83
+    },
+    {
+      "epoch": 0.05376,
+      "grad_norm": 0.4388949943183408,
+      "learning_rate": 2.9999461487125358e-05,
+      "loss": 0.3278,
+      "step": 84
+    },
+    {
+      "epoch": 0.0544,
+      "grad_norm": 0.4830420761548017,
+      "learning_rate": 2.999915857646538e-05,
+      "loss": 0.3242,
+      "step": 85
+    },
+    {
+      "epoch": 0.05504,
+      "grad_norm": 0.42290888303690005,
+      "learning_rate": 2.9998788355094472e-05,
+      "loss": 0.3051,
+      "step": 86
+    },
+    {
+      "epoch": 0.05568,
+      "grad_norm": 0.39931894559101605,
+      "learning_rate": 2.9998350824674046e-05,
+      "loss": 0.3174,
+      "step": 87
+    },
+    {
+      "epoch": 0.05632,
+      "grad_norm": 0.43885002522435895,
+      "learning_rate": 2.999784598716758e-05,
+      "loss": 0.3167,
+      "step": 88
+    },
+    {
+      "epoch": 0.05696,
+      "grad_norm": 0.4385666955310095,
+      "learning_rate": 2.9997273844840597e-05,
+      "loss": 0.32,
+      "step": 89
+    },
+    {
+      "epoch": 0.0576,
+      "grad_norm": 0.4241668561559285,
+      "learning_rate": 2.9996634400260665e-05,
+      "loss": 0.339,
+      "step": 90
+    },
+    {
+      "epoch": 0.05824,
+      "grad_norm": 0.49455156274720913,
+      "learning_rate": 2.9995927656297376e-05,
+      "loss": 0.3135,
+      "step": 91
+    },
+    {
+      "epoch": 0.05888,
+      "grad_norm": 0.3990302088629719,
+      "learning_rate": 2.9995153616122335e-05,
+      "loss": 0.3026,
+      "step": 92
+    },
+    {
+      "epoch": 0.05952,
+      "grad_norm": 0.38959449381416456,
+      "learning_rate": 2.9994312283209147e-05,
+      "loss": 0.2898,
+      "step": 93
+    },
+    {
+      "epoch": 0.06016,
+      "grad_norm": 0.42585557119259293,
+      "learning_rate": 2.9993403661333413e-05,
+      "loss": 0.3021,
+      "step": 94
+    },
+    {
+      "epoch": 0.0608,
+      "grad_norm": 0.37467889420763917,
+      "learning_rate": 2.9992427754572692e-05,
+      "loss": 0.273,
+      "step": 95
+    },
+    {
+      "epoch": 0.06144,
+      "grad_norm": 0.4466178048330342,
+      "learning_rate": 2.9991384567306485e-05,
+      "loss": 0.3116,
+      "step": 96
+    },
+    {
+      "epoch": 0.06208,
+      "grad_norm": 0.4133406435053229,
+      "learning_rate": 2.9990274104216246e-05,
+      "loss": 0.3066,
+      "step": 97
+    },
+    {
+      "epoch": 0.06272,
+      "grad_norm": 0.42703847116501487,
+      "learning_rate": 2.9989096370285314e-05,
+      "loss": 0.3029,
+      "step": 98
+    },
+    {
+      "epoch": 0.06336,
+      "grad_norm": 0.42408773456463106,
+      "learning_rate": 2.9987851370798936e-05,
+      "loss": 0.3302,
+      "step": 99
+    },
+    {
+      "epoch": 0.064,
+      "grad_norm": 0.36334101776279903,
+      "learning_rate": 2.998653911134421e-05,
+      "loss": 0.2749,
+      "step": 100
+    },
+    {
+      "epoch": 0.06464,
+      "grad_norm": 0.39914472077814156,
+      "learning_rate": 2.9985159597810067e-05,
+      "loss": 0.28,
+      "step": 101
+    },
+    {
+      "epoch": 0.06528,
+      "grad_norm": 0.4421338433902218,
+      "learning_rate": 2.9983712836387263e-05,
+      "loss": 0.3284,
+      "step": 102
+    },
+    {
+      "epoch": 0.06592,
+      "grad_norm": 0.4681114190701383,
+      "learning_rate": 2.9982198833568332e-05,
+      "loss": 0.3106,
+      "step": 103
+    },
+    {
+      "epoch": 0.06656,
+      "grad_norm": 0.43495699820223616,
+      "learning_rate": 2.998061759614756e-05,
+      "loss": 0.3034,
+      "step": 104
+    },
+    {
+      "epoch": 0.0672,
+      "grad_norm": 0.48523982047419434,
+      "learning_rate": 2.9978969131220956e-05,
+      "loss": 0.3199,
+      "step": 105
+    },
+    {
+      "epoch": 0.06784,
+      "grad_norm": 0.46970443054055944,
+      "learning_rate": 2.9977253446186236e-05,
+      "loss": 0.3145,
+      "step": 106
+    },
+    {
+      "epoch": 0.06848,
+      "grad_norm": 0.3762833085351204,
+      "learning_rate": 2.997547054874275e-05,
+      "loss": 0.2784,
+      "step": 107
+    },
+    {
+      "epoch": 0.06912,
+      "grad_norm": 0.4018479556059187,
+      "learning_rate": 2.9973620446891495e-05,
+      "loss": 0.29,
+      "step": 108
+    },
+    {
+      "epoch": 0.06976,
+      "grad_norm": 0.38557801552857196,
+      "learning_rate": 2.997170314893504e-05,
+      "loss": 0.2973,
+      "step": 109
+    },
+    {
+      "epoch": 0.0704,
+      "grad_norm": 0.42760504312755043,
+      "learning_rate": 2.9969718663477524e-05,
+      "loss": 0.3087,
+      "step": 110
+    },
+    {
+      "epoch": 0.07104,
+      "grad_norm": 0.4336880764497262,
+      "learning_rate": 2.9967666999424588e-05,
+      "loss": 0.3202,
+      "step": 111
+    },
+    {
+      "epoch": 0.07168,
+      "grad_norm": 0.4361380772072425,
+      "learning_rate": 2.9965548165983337e-05,
+      "loss": 0.2858,
+      "step": 112
+    },
+    {
+      "epoch": 0.07232,
+      "grad_norm": 0.36094868753422277,
+      "learning_rate": 2.9963362172662324e-05,
+      "loss": 0.2965,
+      "step": 113
+    },
+    {
+      "epoch": 0.07296,
+      "grad_norm": 0.5213407516270206,
+      "learning_rate": 2.9961109029271478e-05,
+      "loss": 0.2877,
+      "step": 114
+    },
+    {
+      "epoch": 0.0736,
+      "grad_norm": 0.39899966857979996,
+      "learning_rate": 2.9958788745922088e-05,
+      "loss": 0.2967,
+      "step": 115
+    },
+    {
+      "epoch": 0.07424,
+      "grad_norm": 0.4922285856357075,
+      "learning_rate": 2.995640133302672e-05,
+      "loss": 0.3135,
+      "step": 116
+    },
+    {
+      "epoch": 0.07488,
+      "grad_norm": 0.3840683744534582,
+      "learning_rate": 2.9953946801299213e-05,
+      "loss": 0.3021,
+      "step": 117
+    },
+    {
+      "epoch": 0.07552,
+      "grad_norm": 0.4183493778424685,
+      "learning_rate": 2.995142516175461e-05,
+      "loss": 0.307,
+      "step": 118
+    },
+    {
+      "epoch": 0.07616,
+      "grad_norm": 0.35220612158537656,
+      "learning_rate": 2.994883642570909e-05,
+      "loss": 0.2702,
+      "step": 119
+    },
+    {
+      "epoch": 0.0768,
+      "grad_norm": 0.3936594224655844,
+      "learning_rate": 2.9946180604779952e-05,
+      "loss": 0.2978,
+      "step": 120
+    },
+    {
+      "epoch": 0.07744,
+      "grad_norm": 0.4020931932900933,
+      "learning_rate": 2.9943457710885548e-05,
+      "loss": 0.3115,
+      "step": 121
+    },
+    {
+      "epoch": 0.07808,
+      "grad_norm": 0.413807873620709,
+      "learning_rate": 2.9940667756245218e-05,
+      "loss": 0.3084,
+      "step": 122
+    },
+    {
+      "epoch": 0.07872,
+      "grad_norm": 0.3861233473719098,
+      "learning_rate": 2.9937810753379266e-05,
+      "loss": 0.2844,
+      "step": 123
+    },
+    {
+      "epoch": 0.07936,
+      "grad_norm": 0.3386752443740303,
+      "learning_rate": 2.993488671510886e-05,
+      "loss": 0.3037,
+      "step": 124
+    },
+    {
+      "epoch": 0.08,
+      "grad_norm": 0.3972212865125969,
+      "learning_rate": 2.993189565455601e-05,
+      "loss": 0.3137,
+      "step": 125
+    },
+    {
+      "epoch": 0.08064,
+      "grad_norm": 0.38002895692953814,
+      "learning_rate": 2.9928837585143497e-05,
+      "loss": 0.2873,
+      "step": 126
+    },
+    {
+      "epoch": 0.08128,
+      "grad_norm": 0.44072168693518854,
+      "learning_rate": 2.992571252059482e-05,
+      "loss": 0.2782,
+      "step": 127
+    },
+    {
+      "epoch": 0.08192,
+      "grad_norm": 0.3479281011658149,
+      "learning_rate": 2.992252047493411e-05,
+      "loss": 0.2799,
+      "step": 128
+    },
+    {
+      "epoch": 0.08256,
+      "grad_norm": 0.4142266632731221,
+      "learning_rate": 2.9919261462486098e-05,
+      "loss": 0.3176,
+      "step": 129
+    },
+    {
+      "epoch": 0.0832,
+      "grad_norm": 0.35944919396457825,
+      "learning_rate": 2.991593549787604e-05,
+      "loss": 0.2963,
+      "step": 130
+    },
+    {
+      "epoch": 0.08384,
+      "grad_norm": 0.3837171188201694,
+      "learning_rate": 2.9912542596029635e-05,
+      "loss": 0.3057,
+      "step": 131
+    },
+    {
+      "epoch": 0.08448,
+      "grad_norm": 0.4763114305033498,
+      "learning_rate": 2.990908277217298e-05,
+      "loss": 0.3224,
+      "step": 132
+    },
+    {
+      "epoch": 0.08512,
+      "grad_norm": 0.3834181145690126,
+      "learning_rate": 2.9905556041832494e-05,
+      "loss": 0.2841,
+      "step": 133
+    },
+    {
+      "epoch": 0.08576,
+      "grad_norm": 0.40814627377992635,
+      "learning_rate": 2.990196242083485e-05,
+      "loss": 0.2871,
+      "step": 134
+    },
+    {
+      "epoch": 0.0864,
+      "grad_norm": 0.444873139895432,
+      "learning_rate": 2.98983019253069e-05,
+      "loss": 0.29,
+      "step": 135
+    },
+    {
+      "epoch": 0.08704,
+      "grad_norm": 0.4015570039172884,
+      "learning_rate": 2.9894574571675593e-05,
+      "loss": 0.29,
+      "step": 136
+    },
+    {
+      "epoch": 0.08768,
+      "grad_norm": 0.4192576277736737,
+      "learning_rate": 2.989078037666793e-05,
+      "loss": 0.3187,
+      "step": 137
+    },
+    {
+      "epoch": 0.08832,
+      "grad_norm": 0.4701185999968217,
+      "learning_rate": 2.988691935731086e-05,
+      "loss": 0.3039,
+      "step": 138
+    },
+    {
+      "epoch": 0.08896,
+      "grad_norm": 0.331862366916375,
+      "learning_rate": 2.988299153093122e-05,
+      "loss": 0.2814,
+      "step": 139
+    },
+    {
+      "epoch": 0.0896,
+      "grad_norm": 0.4503832184907915,
+      "learning_rate": 2.987899691515565e-05,
+      "loss": 0.3084,
+      "step": 140
+    },
+    {
+      "epoch": 0.09024,
+      "grad_norm": 0.3734705387771823,
+      "learning_rate": 2.987493552791052e-05,
+      "loss": 0.2837,
+      "step": 141
+    },
+    {
+      "epoch": 0.09088,
+      "grad_norm": 0.3978970762252604,
+      "learning_rate": 2.9870807387421843e-05,
+      "loss": 0.3032,
+      "step": 142
+    },
+    {
+      "epoch": 0.09152,
+      "grad_norm": 0.39348398869038875,
+      "learning_rate": 2.986661251221519e-05,
+      "loss": 0.3062,
+      "step": 143
+    },
+    {
+      "epoch": 0.09216,
+      "grad_norm": 0.4493064184937299,
+      "learning_rate": 2.986235092111563e-05,
+      "loss": 0.3071,
+      "step": 144
+    },
+    {
+      "epoch": 0.0928,
+      "grad_norm": 0.44946841788948644,
+      "learning_rate": 2.985802263324761e-05,
+      "loss": 0.3025,
+      "step": 145
+    },
+    {
+      "epoch": 0.09344,
+      "grad_norm": 0.5016226955727763,
+      "learning_rate": 2.9853627668034898e-05,
+      "loss": 0.3246,
+      "step": 146
+    },
+    {
+      "epoch": 0.09408,
+      "grad_norm": 0.399973023082225,
+      "learning_rate": 2.9849166045200476e-05,
+      "loss": 0.2811,
+      "step": 147
+    },
+    {
+      "epoch": 0.09472,
+      "grad_norm": 0.3739130134478663,
+      "learning_rate": 2.9844637784766478e-05,
+      "loss": 0.2855,
+      "step": 148
+    },
+    {
+      "epoch": 0.09536,
+      "grad_norm": 0.43805848638708655,
+      "learning_rate": 2.9840042907054068e-05,
+      "loss": 0.2798,
+      "step": 149
+    },
+    {
+      "epoch": 0.096,
+      "grad_norm": 0.4402314818129774,
+      "learning_rate": 2.9835381432683363e-05,
+      "loss": 0.2847,
+      "step": 150
+    },
+    {
+      "epoch": 0.09664,
+      "grad_norm": 0.4047049034331425,
+      "learning_rate": 2.9830653382573358e-05,
+      "loss": 0.2861,
+      "step": 151
+    },
+    {
+      "epoch": 0.09728,
+      "grad_norm": 0.43978737589580585,
+      "learning_rate": 2.982585877794179e-05,
+      "loss": 0.3054,
+      "step": 152
+    },
+    {
+      "epoch": 0.09792,
+      "grad_norm": 0.4730334615845102,
+      "learning_rate": 2.9820997640305097e-05,
+      "loss": 0.2817,
+      "step": 153
+    },
+    {
+      "epoch": 0.09856,
+      "grad_norm": 0.3740570479644525,
+      "learning_rate": 2.981606999147827e-05,
+      "loss": 0.2608,
+      "step": 154
+    },
+    {
+      "epoch": 0.0992,
+      "grad_norm": 0.481760965756092,
+      "learning_rate": 2.9811075853574788e-05,
+      "loss": 0.3113,
+      "step": 155
+    },
+    {
+      "epoch": 0.09984,
+      "grad_norm": 0.3961997084965111,
+      "learning_rate": 2.98060152490065e-05,
+      "loss": 0.3088,
+      "step": 156
+    },
+    {
+      "epoch": 0.10048,
+      "grad_norm": 0.428730401496227,
+      "learning_rate": 2.9800888200483552e-05,
+      "loss": 0.3129,
+      "step": 157
+    },
+    {
+      "epoch": 0.10112,
+      "grad_norm": 0.4178705790625643,
+      "learning_rate": 2.979569473101424e-05,
+      "loss": 0.3284,
+      "step": 158
+    },
+    {
+      "epoch": 0.10176,
+      "grad_norm": 0.4463567693300283,
+      "learning_rate": 2.9790434863904957e-05,
+      "loss": 0.3044,
+      "step": 159
+    },
+    {
+      "epoch": 0.1024,
+      "grad_norm": 0.3939775657775284,
+      "learning_rate": 2.9785108622760045e-05,
+      "loss": 0.2894,
+      "step": 160
+    },
+    {
+      "epoch": 0.10304,
+      "grad_norm": 0.38174144831499374,
+      "learning_rate": 2.9779716031481717e-05,
+      "loss": 0.2779,
+      "step": 161
+    },
+    {
+      "epoch": 0.10368,
+      "grad_norm": 0.38589717775088483,
+      "learning_rate": 2.9774257114269955e-05,
+      "loss": 0.3123,
+      "step": 162
+    },
+    {
+      "epoch": 0.10432,
+      "grad_norm": 0.3673496875585954,
+      "learning_rate": 2.9768731895622355e-05,
+      "loss": 0.2771,
+      "step": 163
+    },
+    {
+      "epoch": 0.10496,
+      "grad_norm": 0.3571608186420396,
+      "learning_rate": 2.9763140400334072e-05,
+      "loss": 0.2897,
+      "step": 164
+    },
+    {
+      "epoch": 0.1056,
+      "grad_norm": 0.3939651393603113,
+      "learning_rate": 2.975748265349769e-05,
+      "loss": 0.2966,
+      "step": 165
+    },
+    {
+      "epoch": 0.10624,
+      "grad_norm": 0.3805663760796052,
+      "learning_rate": 2.975175868050309e-05,
+      "loss": 0.3156,
+      "step": 166
+    },
+    {
+      "epoch": 0.10688,
+      "grad_norm": 0.3972270203399388,
+      "learning_rate": 2.9745968507037356e-05,
+      "loss": 0.2963,
+      "step": 167
+    },
+    {
+      "epoch": 0.10752,
+      "grad_norm": 0.39226032551563944,
+      "learning_rate": 2.974011215908467e-05,
+      "loss": 0.2933,
+      "step": 168
+    },
+    {
+      "epoch": 0.10816,
+      "grad_norm": 0.35875066299615677,
+      "learning_rate": 2.9734189662926163e-05,
+      "loss": 0.2758,
+      "step": 169
+    },
+    {
+      "epoch": 0.1088,
+      "grad_norm": 0.3709380144753209,
+      "learning_rate": 2.9728201045139813e-05,
+      "loss": 0.2927,
+      "step": 170
+    },
+    {
+      "epoch": 0.10944,
+      "grad_norm": 0.370201710217779,
+      "learning_rate": 2.972214633260035e-05,
+      "loss": 0.2866,
+      "step": 171
+    },
+    {
+      "epoch": 0.11008,
+      "grad_norm": 0.3758060771260273,
+      "learning_rate": 2.9716025552479093e-05,
+      "loss": 0.2994,
+      "step": 172
+    },
+    {
+      "epoch": 0.11072,
+      "grad_norm": 0.32580329325564505,
+      "learning_rate": 2.9709838732243844e-05,
+      "loss": 0.2802,
+      "step": 173
+    },
+    {
+      "epoch": 0.11136,
+      "grad_norm": 0.38248606018972797,
+      "learning_rate": 2.970358589965879e-05,
+      "loss": 0.2739,
+      "step": 174
+    },
+    {
+      "epoch": 0.112,
+      "grad_norm": 0.36493445057406637,
+      "learning_rate": 2.9697267082784342e-05,
+      "loss": 0.2874,
+      "step": 175
+    },
+    {
+      "epoch": 0.11264,
+      "grad_norm": 0.6384082314576618,
+      "learning_rate": 2.969088230997703e-05,
+      "loss": 0.2835,
+      "step": 176
+    },
+    {
+      "epoch": 0.11328,
+      "grad_norm": 0.4046365021981396,
+      "learning_rate": 2.9684431609889365e-05,
+      "loss": 0.2964,
+      "step": 177
+    },
+    {
+      "epoch": 0.11392,
+      "grad_norm": 0.3777231337256365,
+      "learning_rate": 2.9677915011469717e-05,
+      "loss": 0.302,
+      "step": 178
+    },
+    {
+      "epoch": 0.11456,
+      "grad_norm": 0.3550577304426545,
+      "learning_rate": 2.9671332543962183e-05,
+      "loss": 0.2809,
+      "step": 179
+    },
+    {
+      "epoch": 0.1152,
+      "grad_norm": 0.34027303690053873,
+      "learning_rate": 2.9664684236906466e-05,
+      "loss": 0.2998,
+      "step": 180
+    },
+    {
+      "epoch": 0.11584,
+      "grad_norm": 0.3705269161905217,
+      "learning_rate": 2.965797012013772e-05,
+      "loss": 0.3187,
+      "step": 181
+    },
+    {
+      "epoch": 0.11648,
+      "grad_norm": 0.36000193066901365,
+      "learning_rate": 2.9651190223786427e-05,
+      "loss": 0.2838,
+      "step": 182
+    },
+    {
+      "epoch": 0.11712,
+      "grad_norm": 0.32994641779773276,
+      "learning_rate": 2.9644344578278284e-05,
+      "loss": 0.2791,
+      "step": 183
+    },
+    {
+      "epoch": 0.11776,
+      "grad_norm": 0.35570577433544603,
+      "learning_rate": 2.963743321433402e-05,
+      "loss": 0.2773,
+      "step": 184
+    },
+    {
+      "epoch": 0.1184,
+      "grad_norm": 0.3596282933620285,
+      "learning_rate": 2.9630456162969298e-05,
+      "loss": 0.2797,
+      "step": 185
+    },
+    {
+      "epoch": 0.11904,
+      "grad_norm": 0.3262807239848876,
+      "learning_rate": 2.9623413455494563e-05,
+      "loss": 0.288,
+      "step": 186
+    },
+    {
+      "epoch": 0.11968,
+      "grad_norm": 0.3565865179949324,
+      "learning_rate": 2.9616305123514897e-05,
+      "loss": 0.2887,
+      "step": 187
+    },
+    {
+      "epoch": 0.12032,
+      "grad_norm": 0.39532927846862465,
+      "learning_rate": 2.9609131198929884e-05,
+      "loss": 0.2782,
+      "step": 188
+    },
+    {
+      "epoch": 0.12096,
+      "grad_norm": 0.40888574656849996,
+      "learning_rate": 2.9601891713933457e-05,
+      "loss": 0.2718,
+      "step": 189
+    },
+    {
+      "epoch": 0.1216,
+      "grad_norm": 0.3494105157933286,
+      "learning_rate": 2.9594586701013765e-05,
+      "loss": 0.2925,
+      "step": 190
+    },
+    {
+      "epoch": 0.12224,
+      "grad_norm": 0.38169153601979233,
+      "learning_rate": 2.958721619295302e-05,
+      "loss": 0.2783,
+      "step": 191
+    },
+    {
+      "epoch": 0.12288,
+      "grad_norm": 0.3650159986813621,
+      "learning_rate": 2.9579780222827354e-05,
+      "loss": 0.2839,
+      "step": 192
+    },
+    {
+      "epoch": 0.12352,
+      "grad_norm": 0.32047234144052256,
+      "learning_rate": 2.957227882400667e-05,
+      "loss": 0.2781,
+      "step": 193
+    },
+    {
+      "epoch": 0.12416,
+      "grad_norm": 0.39912184054626987,
+      "learning_rate": 2.9564712030154486e-05,
+      "loss": 0.2992,
+      "step": 194
+    },
+    {
+      "epoch": 0.1248,
+      "grad_norm": 0.3589964401769852,
+      "learning_rate": 2.9557079875227795e-05,
+      "loss": 0.2991,
+      "step": 195
+    },
+    {
+      "epoch": 0.12544,
+      "grad_norm": 0.3095249869798657,
+      "learning_rate": 2.95493823934769e-05,
+      "loss": 0.2886,
+      "step": 196
+    },
+    {
+      "epoch": 0.12608,
+      "grad_norm": 0.33996550944422993,
+      "learning_rate": 2.954161961944527e-05,
+      "loss": 0.2705,
+      "step": 197
+    },
+    {
+      "epoch": 0.12672,
+      "grad_norm": 0.3423071479543653,
+      "learning_rate": 2.953379158796938e-05,
+      "loss": 0.2795,
+      "step": 198
+    },
+    {
+      "epoch": 0.12736,
+      "grad_norm": 0.36082529918938056,
+      "learning_rate": 2.9525898334178566e-05,
+      "loss": 0.2887,
+      "step": 199
+    },
+    {
+      "epoch": 0.128,
+      "grad_norm": 0.37640661693173727,
+      "learning_rate": 2.951793989349484e-05,
+      "loss": 0.282,
+      "step": 200
+    },
+    {
+      "epoch": 0.12864,
+      "grad_norm": 0.3023070350545028,
+      "learning_rate": 2.950991630163277e-05,
+      "loss": 0.254,
+      "step": 201
+    },
+    {
+      "epoch": 0.12928,
+      "grad_norm": 0.3746495017193779,
+      "learning_rate": 2.950182759459928e-05,
+      "loss": 0.2913,
+      "step": 202
+    },
+    {
+      "epoch": 0.12992,
+      "grad_norm": 0.38483323478343573,
+      "learning_rate": 2.949367380869351e-05,
+      "loss": 0.2913,
+      "step": 203
+    },
+    {
+      "epoch": 0.13056,
+      "grad_norm": 0.4213879707012684,
+      "learning_rate": 2.9485454980506663e-05,
+      "loss": 0.2841,
+      "step": 204
+    },
+    {
+      "epoch": 0.1312,
+      "grad_norm": 0.4131757338780886,
+      "learning_rate": 2.9477171146921816e-05,
+      "loss": 0.2981,
+      "step": 205
+    },
+    {
+      "epoch": 0.13184,
+      "grad_norm": 0.4415959043983817,
+      "learning_rate": 2.946882234511377e-05,
+      "loss": 0.2975,
+      "step": 206
+    },
+    {
+      "epoch": 0.13248,
+      "grad_norm": 0.37499387107669097,
+      "learning_rate": 2.9460408612548876e-05,
+      "loss": 0.2844,
+      "step": 207
+    },
+    {
+      "epoch": 0.13312,
+      "grad_norm": 0.41827519988058354,
+      "learning_rate": 2.9451929986984875e-05,
+      "loss": 0.2958,
+      "step": 208
+    },
+    {
+      "epoch": 0.13376,
+      "grad_norm": 0.33683714204034787,
+      "learning_rate": 2.9443386506470725e-05,
+      "loss": 0.2686,
+      "step": 209
+    },
+    {
+      "epoch": 0.1344,
+      "grad_norm": 0.38867413035222104,
+      "learning_rate": 2.9434778209346427e-05,
+      "loss": 0.2876,
+      "step": 210
+    },
+    {
+      "epoch": 0.13504,
+      "grad_norm": 0.39089889654238963,
+      "learning_rate": 2.942610513424285e-05,
+      "loss": 0.2832,
+      "step": 211
+    },
+    {
+      "epoch": 0.13568,
+      "grad_norm": 0.3384476658736123,
+      "learning_rate": 2.9417367320081567e-05,
+      "loss": 0.2604,
+      "step": 212
+    },
+    {
+      "epoch": 0.13632,
+      "grad_norm": 0.38210814875894833,
+      "learning_rate": 2.940856480607468e-05,
+      "loss": 0.2986,
+      "step": 213
+    },
+    {
+      "epoch": 0.13696,
+      "grad_norm": 0.38075475653821916,
+      "learning_rate": 2.9399697631724637e-05,
+      "loss": 0.2761,
+      "step": 214
+    },
+    {
+      "epoch": 0.1376,
+      "grad_norm": 0.36451096593797544,
+      "learning_rate": 2.9390765836824053e-05,
+      "loss": 0.3053,
+      "step": 215
+    },
+    {
+      "epoch": 0.13824,
+      "grad_norm": 0.36903700040818893,
+      "learning_rate": 2.938176946145555e-05,
+      "loss": 0.2776,
+      "step": 216
+    },
+    {
+      "epoch": 0.13888,
+      "grad_norm": 0.3858847022411217,
+      "learning_rate": 2.9372708545991542e-05,
+      "loss": 0.2682,
+      "step": 217
+    },
+    {
+      "epoch": 0.13952,
+      "grad_norm": 0.3664388884748071,
+      "learning_rate": 2.936358313109409e-05,
+      "loss": 0.263,
+      "step": 218
+    },
+    {
+      "epoch": 0.14016,
+      "grad_norm": 0.3833106914044559,
+      "learning_rate": 2.935439325771471e-05,
+      "loss": 0.284,
+      "step": 219
+    },
+    {
+      "epoch": 0.1408,
+      "grad_norm": 0.4366332113795893,
+      "learning_rate": 2.9345138967094174e-05,
+      "loss": 0.2827,
+      "step": 220
+    },
+    {
+      "epoch": 0.14144,
+      "grad_norm": 0.35644258927978634,
+      "learning_rate": 2.9335820300762334e-05,
+      "loss": 0.2763,
+      "step": 221
+    },
+    {
+      "epoch": 0.14208,
+      "grad_norm": 0.4086383954714273,
+      "learning_rate": 2.9326437300537937e-05,
+      "loss": 0.2731,
+      "step": 222
+    },
+    {
+      "epoch": 0.14272,
+      "grad_norm": 0.37566830211795793,
+      "learning_rate": 2.9316990008528446e-05,
+      "loss": 0.3035,
+      "step": 223
+    },
+    {
+      "epoch": 0.14336,
+      "grad_norm": 0.37673883341537595,
+      "learning_rate": 2.9307478467129827e-05,
+      "loss": 0.2807,
+      "step": 224
+    },
+    {
+      "epoch": 0.144,
+      "grad_norm": 0.3658815413836672,
+      "learning_rate": 2.9297902719026392e-05,
+      "loss": 0.2872,
+      "step": 225
+    },
+    {
+      "epoch": 0.14464,
+      "grad_norm": 0.3723900174142704,
+      "learning_rate": 2.928826280719057e-05,
+      "loss": 0.2628,
+      "step": 226
+    },
+    {
+      "epoch": 0.14528,
+      "grad_norm": 0.3583483332892453,
+      "learning_rate": 2.9278558774882748e-05,
+      "loss": 0.2757,
+      "step": 227
+    },
+    {
+      "epoch": 0.14592,
+      "grad_norm": 0.36256098670512743,
+      "learning_rate": 2.9268790665651053e-05,
+      "loss": 0.2795,
+      "step": 228
+    },
+    {
+      "epoch": 0.14656,
+      "grad_norm": 0.4447205295517405,
+      "learning_rate": 2.925895852333117e-05,
+      "loss": 0.2917,
+      "step": 229
+    },
+    {
+      "epoch": 0.1472,
+      "grad_norm": 0.33653314086590325,
+      "learning_rate": 2.924906239204614e-05,
+      "loss": 0.2856,
+      "step": 230
+    },
+    {
+      "epoch": 0.14784,
+      "grad_norm": 0.3875121699387206,
+      "learning_rate": 2.9239102316206166e-05,
+      "loss": 0.2756,
+      "step": 231
+    },
+    {
+      "epoch": 0.14848,
+      "grad_norm": 0.4191113393584937,
+      "learning_rate": 2.9229078340508404e-05,
+      "loss": 0.2947,
+      "step": 232
+    },
+    {
+      "epoch": 0.14912,
+      "grad_norm": 0.3866679704383513,
+      "learning_rate": 2.9218990509936774e-05,
+      "loss": 0.2948,
+      "step": 233
+    },
+    {
+      "epoch": 0.14976,
+      "grad_norm": 0.44210860261964957,
+      "learning_rate": 2.9208838869761756e-05,
+      "loss": 0.2823,
+      "step": 234
+    },
+    {
+      "epoch": 0.1504,
+      "grad_norm": 0.3569394913477398,
+      "learning_rate": 2.9198623465540172e-05,
+      "loss": 0.2896,
+      "step": 235
+    },
+    {
+      "epoch": 0.15104,
+      "grad_norm": 0.3582653155321637,
+      "learning_rate": 2.9188344343115005e-05,
+      "loss": 0.2639,
+      "step": 236
+    },
+    {
+      "epoch": 0.15168,
+      "grad_norm": 0.3632299874629792,
+      "learning_rate": 2.9178001548615176e-05,
+      "loss": 0.2752,
+      "step": 237
+    },
+    {
+      "epoch": 0.15232,
+      "grad_norm": 0.3627690799272646,
+      "learning_rate": 2.9167595128455357e-05,
+      "loss": 0.2697,
+      "step": 238
+    },
+    {
+      "epoch": 0.15296,
+      "grad_norm": 0.31681761773997114,
+      "learning_rate": 2.915712512933572e-05,
+      "loss": 0.2886,
+      "step": 239
+    },
+    {
+      "epoch": 0.1536,
+      "grad_norm": 0.3652202557075009,
+      "learning_rate": 2.914659159824178e-05,
+      "loss": 0.291,
+      "step": 240
+    },
+    {
+      "epoch": 0.15424,
+      "grad_norm": 0.32355380076781154,
+      "learning_rate": 2.913599458244416e-05,
+      "loss": 0.288,
+      "step": 241
+    },
+    {
+      "epoch": 0.15488,
+      "grad_norm": 0.3263439505506079,
+      "learning_rate": 2.9125334129498368e-05,
+      "loss": 0.2669,
+      "step": 242
+    },
+    {
+      "epoch": 0.15552,
+      "grad_norm": 0.34698579986117173,
+      "learning_rate": 2.911461028724459e-05,
+      "loss": 0.2623,
+      "step": 243
+    },
+    {
+      "epoch": 0.15616,
+      "grad_norm": 0.3560347825174995,
+      "learning_rate": 2.9103823103807503e-05,
+      "loss": 0.2779,
+      "step": 244
+    },
+    {
+      "epoch": 0.1568,
+      "grad_norm": 0.3829349365942763,
+      "learning_rate": 2.9092972627596012e-05,
+      "loss": 0.285,
+      "step": 245
+    },
+    {
+      "epoch": 0.15744,
+      "grad_norm": 0.35634808472353574,
+      "learning_rate": 2.9082058907303064e-05,
+      "loss": 0.2787,
+      "step": 246
+    },
+    {
+      "epoch": 0.15808,
+      "grad_norm": 0.3973522847788536,
+      "learning_rate": 2.9071081991905436e-05,
+      "loss": 0.2686,
+      "step": 247
+    },
+    {
+      "epoch": 0.15872,
+      "grad_norm": 0.3503037505655835,
+      "learning_rate": 2.9060041930663477e-05,
+      "loss": 0.2601,
+      "step": 248
+    },
+    {
+      "epoch": 0.15936,
+      "grad_norm": 0.42641830259468066,
+      "learning_rate": 2.9048938773120933e-05,
+      "loss": 0.2571,
+      "step": 249
+    },
+    {
+      "epoch": 0.16,
+      "grad_norm": 0.33022262318082957,
+      "learning_rate": 2.903777256910469e-05,
+      "loss": 0.2539,
+      "step": 250
+    },
+    {
+      "epoch": 0.16064,
+      "grad_norm": 0.34806931009286257,
+      "learning_rate": 2.902654336872457e-05,
+      "loss": 0.2855,
+      "step": 251
+    },
+    {
+      "epoch": 0.16128,
+      "grad_norm": 0.3802291207724926,
+      "learning_rate": 2.9015251222373094e-05,
+      "loss": 0.2505,
+      "step": 252
+    },
+    {
+      "epoch": 0.16192,
+      "grad_norm": 0.4182370813509436,
+      "learning_rate": 2.9003896180725268e-05,
+      "loss": 0.2821,
+      "step": 253
+    },
+    {
+      "epoch": 0.16256,
+      "grad_norm": 0.45914515409569023,
+      "learning_rate": 2.8992478294738345e-05,
+      "loss": 0.2934,
+      "step": 254
+    },
+    {
+      "epoch": 0.1632,
+      "grad_norm": 0.3490846199737997,
+      "learning_rate": 2.8980997615651597e-05,
+      "loss": 0.2853,
+      "step": 255
+    },
+    {
+      "epoch": 0.16384,
+      "grad_norm": 0.3945607808525512,
+      "learning_rate": 2.8969454194986095e-05,
+      "loss": 0.2704,
+      "step": 256
+    },
+    {
+      "epoch": 0.16448,
+      "grad_norm": 0.4134287311468698,
+      "learning_rate": 2.8957848084544473e-05,
+      "loss": 0.2856,
+      "step": 257
+    },
+    {
+      "epoch": 0.16512,
+      "grad_norm": 0.388727953362913,
+      "learning_rate": 2.8946179336410682e-05,
+      "loss": 0.2707,
+      "step": 258
+    },
+    {
+      "epoch": 0.16576,
+      "grad_norm": 0.3707496935300037,
+      "learning_rate": 2.8934448002949775e-05,
+      "loss": 0.2736,
+      "step": 259
+    },
+    {
+      "epoch": 0.1664,
+      "grad_norm": 0.37509134347160566,
+      "learning_rate": 2.892265413680767e-05,
+      "loss": 0.2656,
+      "step": 260
+    },
+    {
+      "epoch": 0.16704,
+      "grad_norm": 0.37839278243532787,
+      "learning_rate": 2.8910797790910902e-05,
+      "loss": 0.2954,
+      "step": 261
+    },
+    {
+      "epoch": 0.16768,
+      "grad_norm": 0.3709934691909708,
+      "learning_rate": 2.889887901846639e-05,
+      "loss": 0.2771,
+      "step": 262
+    },
+    {
+      "epoch": 0.16832,
+      "grad_norm": 0.38441885239694634,
+      "learning_rate": 2.8886897872961203e-05,
+      "loss": 0.2872,
+      "step": 263
+    },
+    {
+      "epoch": 0.16896,
+      "grad_norm": 0.36199416231449505,
+      "learning_rate": 2.887485440816233e-05,
+      "loss": 0.2577,
+      "step": 264
+    },
+    {
+      "epoch": 0.1696,
+      "grad_norm": 0.35207892498047666,
+      "learning_rate": 2.8862748678116402e-05,
+      "loss": 0.2692,
+      "step": 265
+    },
+    {
+      "epoch": 0.17024,
+      "grad_norm": 0.3167681279515036,
+      "learning_rate": 2.885058073714949e-05,
+      "loss": 0.2517,
+      "step": 266
+    },
+    {
+      "epoch": 0.17088,
+      "grad_norm": 0.3836456596956298,
+      "learning_rate": 2.8838350639866843e-05,
+      "loss": 0.3038,
+      "step": 267
+    },
+    {
+      "epoch": 0.17152,
+      "grad_norm": 0.3604280243223731,
+      "learning_rate": 2.882605844115264e-05,
+      "loss": 0.2748,
+      "step": 268
+    },
+    {
+      "epoch": 0.17216,
+      "grad_norm": 0.33239517181664746,
+      "learning_rate": 2.8813704196169753e-05,
+      "loss": 0.2705,
+      "step": 269
+    },
+    {
+      "epoch": 0.1728,
+      "grad_norm": 0.4453367026441279,
+      "learning_rate": 2.8801287960359494e-05,
+      "loss": 0.2752,
+      "step": 270
+    },
+    {
+      "epoch": 0.17344,
+      "grad_norm": 0.3672385598637218,
+      "learning_rate": 2.8788809789441364e-05,
+      "loss": 0.2813,
+      "step": 271
+    },
+    {
+      "epoch": 0.17408,
+      "grad_norm": 0.42500271676682233,
+      "learning_rate": 2.8776269739412803e-05,
+      "loss": 0.2581,
+      "step": 272
+    },
+    {
+      "epoch": 0.17472,
+      "grad_norm": 0.3812194331681742,
+      "learning_rate": 2.8763667866548956e-05,
+      "loss": 0.2797,
+      "step": 273
+    },
+    {
+      "epoch": 0.17536,
+      "grad_norm": 0.398047671024292,
+      "learning_rate": 2.875100422740239e-05,
+      "loss": 0.2822,
+      "step": 274
+    },
+    {
+      "epoch": 0.176,
+      "grad_norm": 0.4382339198136674,
+      "learning_rate": 2.8738278878802865e-05,
+      "loss": 0.2951,
+      "step": 275
+    },
+    {
+      "epoch": 0.17664,
+      "grad_norm": 0.32914734501477555,
+      "learning_rate": 2.8725491877857073e-05,
+      "loss": 0.2505,
+      "step": 276
+    },
+    {
+      "epoch": 0.17728,
+      "grad_norm": 0.359221594407232,
+      "learning_rate": 2.8712643281948365e-05,
+      "loss": 0.2682,
+      "step": 277
+    },
+    {
+      "epoch": 0.17792,
+      "grad_norm": 0.3610201048162127,
+      "learning_rate": 2.8699733148736525e-05,
+      "loss": 0.2706,
+      "step": 278
+    },
+    {
+      "epoch": 0.17856,
+      "grad_norm": 0.3576575465005173,
+      "learning_rate": 2.868676153615748e-05,
+      "loss": 0.2725,
+      "step": 279
+    },
+    {
+      "epoch": 0.1792,
+      "grad_norm": 0.39648269977135403,
+      "learning_rate": 2.8673728502423067e-05,
+      "loss": 0.2638,
+      "step": 280
+    },
+    {
+      "epoch": 0.17984,
+      "grad_norm": 0.35937870544419476,
+      "learning_rate": 2.8660634106020747e-05,
+      "loss": 0.2753,
+      "step": 281
+    },
+    {
+      "epoch": 0.18048,
+      "grad_norm": 0.4065407489775043,
+      "learning_rate": 2.8647478405713355e-05,
+      "loss": 0.2814,
+      "step": 282
+    },
+    {
+      "epoch": 0.18112,
+      "grad_norm": 0.3401826548234812,
+      "learning_rate": 2.8634261460538845e-05,
+      "loss": 0.2734,
+      "step": 283
+    },
+    {
+      "epoch": 0.18176,
+      "grad_norm": 0.3871247999638719,
+      "learning_rate": 2.8620983329810004e-05,
+      "loss": 0.266,
+      "step": 284
+    },
+    {
+      "epoch": 0.1824,
+      "grad_norm": 0.40719492038320787,
+      "learning_rate": 2.860764407311421e-05,
+      "loss": 0.2727,
+      "step": 285
+    },
+    {
+      "epoch": 0.18304,
+      "grad_norm": 0.3399520864358679,
+      "learning_rate": 2.8594243750313137e-05,
+      "loss": 0.2689,
+      "step": 286
+    },
+    {
+      "epoch": 0.18368,
+      "grad_norm": 0.3953164307154916,
+      "learning_rate": 2.8580782421542514e-05,
+      "loss": 0.273,
+      "step": 287
+    },
+    {
+      "epoch": 0.18432,
+      "grad_norm": 0.4447124642066109,
+      "learning_rate": 2.8567260147211826e-05,
+      "loss": 0.282,
+      "step": 288
+    },
+    {
+      "epoch": 0.18496,
+      "grad_norm": 0.3479285885985358,
+      "learning_rate": 2.8553676988004083e-05,
+      "loss": 0.2846,
+      "step": 289
+    },
+    {
+      "epoch": 0.1856,
+      "grad_norm": 0.4427716710403484,
+      "learning_rate": 2.8540033004875506e-05,
+      "loss": 0.2731,
+      "step": 290
+    },
+    {
+      "epoch": 0.18624,
+      "grad_norm": 0.3389993255265989,
+      "learning_rate": 2.852632825905528e-05,
+      "loss": 0.2728,
+      "step": 291
+    },
+    {
+      "epoch": 0.18688,
+      "grad_norm": 0.36214801666992735,
+      "learning_rate": 2.851256281204526e-05,
+      "loss": 0.2963,
+      "step": 292
+    },
+    {
+      "epoch": 0.18752,
+      "grad_norm": 0.3225822277564517,
+      "learning_rate": 2.849873672561972e-05,
+      "loss": 0.256,
+      "step": 293
+    },
+    {
+      "epoch": 0.18816,
+      "grad_norm": 0.36090735754771,
+      "learning_rate": 2.8484850061825052e-05,
+      "loss": 0.274,
+      "step": 294
+    },
+    {
+      "epoch": 0.1888,
+      "grad_norm": 0.3619026443591473,
+      "learning_rate": 2.84709028829795e-05,
+      "loss": 0.2904,
+      "step": 295
+    },
+    {
+      "epoch": 0.18944,
+      "grad_norm": 0.3252751498399478,
+      "learning_rate": 2.8456895251672867e-05,
+      "loss": 0.2892,
+      "step": 296
+    },
+    {
+      "epoch": 0.19008,
+      "grad_norm": 0.3420708202146713,
+      "learning_rate": 2.8442827230766265e-05,
+      "loss": 0.2509,
+      "step": 297
+    },
+    {
+      "epoch": 0.19072,
+      "grad_norm": 0.3927376280930936,
+      "learning_rate": 2.8428698883391805e-05,
+      "loss": 0.2902,
+      "step": 298
+    },
+    {
+      "epoch": 0.19136,
+      "grad_norm": 0.33206871786295983,
+      "learning_rate": 2.8414510272952306e-05,
+      "loss": 0.2725,
+      "step": 299
+    },
+    {
+      "epoch": 0.192,
+      "grad_norm": 0.31541664358451077,
+      "learning_rate": 2.840026146312104e-05,
+      "loss": 0.2379,
+      "step": 300
+    },
+    {
+      "epoch": 0.19264,
+      "grad_norm": 0.3610931095304509,
+      "learning_rate": 2.8385952517841433e-05,
+      "loss": 0.2778,
+      "step": 301
+    },
+    {
+      "epoch": 0.19328,
+      "grad_norm": 0.32547764886975145,
+      "learning_rate": 2.837158350132677e-05,
+      "loss": 0.2786,
+      "step": 302
+    },
+    {
+      "epoch": 0.19392,
+      "grad_norm": 0.3428550126773416,
+      "learning_rate": 2.835715447805991e-05,
+      "loss": 0.2896,
+      "step": 303
+    },
+    {
+      "epoch": 0.19456,
+      "grad_norm": 0.34537169429710707,
+      "learning_rate": 2.8342665512793018e-05,
+      "loss": 0.2888,
+      "step": 304
+    },
+    {
+      "epoch": 0.1952,
+      "grad_norm": 0.3526668903699371,
+      "learning_rate": 2.8328116670547237e-05,
+      "loss": 0.2711,
+      "step": 305
+    },
+    {
+      "epoch": 0.19584,
+      "grad_norm": 0.3298129440685655,
+      "learning_rate": 2.8313508016612428e-05,
+      "loss": 0.2723,
+      "step": 306
+    },
+    {
+      "epoch": 0.19648,
+      "grad_norm": 0.3252403541009243,
+      "learning_rate": 2.8298839616546854e-05,
+      "loss": 0.2783,
+      "step": 307
+    },
+    {
+      "epoch": 0.19712,
+      "grad_norm": 0.31942049727047156,
+      "learning_rate": 2.8284111536176907e-05,
+      "loss": 0.3047,
+      "step": 308
+    },
+    {
+      "epoch": 0.19776,
+      "grad_norm": 0.28119673519843524,
+      "learning_rate": 2.8269323841596802e-05,
+      "loss": 0.2685,
+      "step": 309
+    },
+    {
+      "epoch": 0.1984,
+      "grad_norm": 0.3574212794376199,
+      "learning_rate": 2.825447659916827e-05,
+      "loss": 0.271,
+      "step": 310
+    },
+    {
+      "epoch": 0.19904,
+      "grad_norm": 0.30153919945084046,
+      "learning_rate": 2.823956987552028e-05,
+      "loss": 0.2529,
+      "step": 311
+    },
+    {
+      "epoch": 0.19968,
+      "grad_norm": 0.3494979054883366,
+      "learning_rate": 2.8224603737548737e-05,
+      "loss": 0.2886,
+      "step": 312
+    },
+    {
+      "epoch": 0.20032,
+      "grad_norm": 0.30740471706963013,
+      "learning_rate": 2.8209578252416162e-05,
+      "loss": 0.2876,
+      "step": 313
+    },
+    {
+      "epoch": 0.20096,
+      "grad_norm": 0.3279624895202358,
+      "learning_rate": 2.8194493487551402e-05,
+      "loss": 0.2946,
+      "step": 314
+    },
+    {
+      "epoch": 0.2016,
+      "grad_norm": 0.3430080164088275,
+      "learning_rate": 2.8179349510649354e-05,
+      "loss": 0.2891,
+      "step": 315
+    },
+    {
+      "epoch": 0.20224,
+      "grad_norm": 0.30496631367689814,
+      "learning_rate": 2.8164146389670605e-05,
+      "loss": 0.2842,
+      "step": 316
+    },
+    {
+      "epoch": 0.20288,
+      "grad_norm": 0.379409454373658,
+      "learning_rate": 2.8148884192841183e-05,
+      "loss": 0.294,
+      "step": 317
+    },
+    {
+      "epoch": 0.20352,
+      "grad_norm": 0.2948622980345085,
+      "learning_rate": 2.8133562988652218e-05,
+      "loss": 0.2584,
+      "step": 318
+    },
+    {
+      "epoch": 0.20416,
+      "grad_norm": 0.3443019691027711,
+      "learning_rate": 2.8118182845859636e-05,
+      "loss": 0.251,
+      "step": 319
+    },
+    {
+      "epoch": 0.2048,
+      "grad_norm": 0.36699154745018375,
+      "learning_rate": 2.810274383348387e-05,
+      "loss": 0.2952,
+      "step": 320
+    },
+    {
+      "epoch": 0.20544,
+      "grad_norm": 0.33967893078585726,
+      "learning_rate": 2.8087246020809536e-05,
+      "loss": 0.2796,
+      "step": 321
+    },
+    {
+      "epoch": 0.20608,
+      "grad_norm": 0.31170315997663467,
+      "learning_rate": 2.8071689477385117e-05,
+      "loss": 0.2721,
+      "step": 322
+    },
+    {
+      "epoch": 0.20672,
+      "grad_norm": 0.3242087780438831,
+      "learning_rate": 2.8056074273022666e-05,
+      "loss": 0.2774,
+      "step": 323
+    },
+    {
+      "epoch": 0.20736,
+      "grad_norm": 0.3555981636400892,
+      "learning_rate": 2.8040400477797476e-05,
+      "loss": 0.2797,
+      "step": 324
+    },
+    {
+      "epoch": 0.208,
+      "grad_norm": 0.309319481935765,
+      "learning_rate": 2.8024668162047783e-05,
+      "loss": 0.292,
+      "step": 325
+    },
+    {
+      "epoch": 0.20864,
+      "grad_norm": 0.38567954428634077,
+      "learning_rate": 2.8008877396374434e-05,
+      "loss": 0.2622,
+      "step": 326
+    },
+    {
+      "epoch": 0.20928,
+      "grad_norm": 0.3558299126479417,
+      "learning_rate": 2.799302825164058e-05,
+      "loss": 0.2741,
+      "step": 327
+    },
+    {
+      "epoch": 0.20992,
+      "grad_norm": 0.32961441916292467,
+      "learning_rate": 2.7977120798971374e-05,
+      "loss": 0.2669,
+      "step": 328
+    },
+    {
+      "epoch": 0.21056,
+      "grad_norm": 0.30783578854216875,
+      "learning_rate": 2.7961155109753596e-05,
+      "loss": 0.2645,
+      "step": 329
+    },
+    {
+      "epoch": 0.2112,
+      "grad_norm": 0.32966575581899893,
+      "learning_rate": 2.79451312556354e-05,
+      "loss": 0.2654,
+      "step": 330
+    },
+    {
+      "epoch": 0.21184,
+      "grad_norm": 0.31714659245735044,
+      "learning_rate": 2.7929049308525958e-05,
+      "loss": 0.2542,
+      "step": 331
+    },
+    {
+      "epoch": 0.21248,
+      "grad_norm": 0.3196236044042015,
+      "learning_rate": 2.7912909340595133e-05,
+      "loss": 0.2624,
+      "step": 332
+    },
+    {
+      "epoch": 0.21312,
+      "grad_norm": 0.3579856676787265,
+      "learning_rate": 2.7896711424273166e-05,
+      "loss": 0.272,
+      "step": 333
+    },
+    {
+      "epoch": 0.21376,
+      "grad_norm": 0.3327728411912822,
+      "learning_rate": 2.7880455632250358e-05,
+      "loss": 0.2588,
+      "step": 334
+    },
+    {
+      "epoch": 0.2144,
+      "grad_norm": 0.31027790499808855,
+      "learning_rate": 2.7864142037476727e-05,
+      "loss": 0.2728,
+      "step": 335
+    },
+    {
+      "epoch": 0.21504,
+      "grad_norm": 0.3202262124080976,
+      "learning_rate": 2.784777071316169e-05,
+      "loss": 0.2818,
+      "step": 336
+    },
+    {
+      "epoch": 0.21568,
+      "grad_norm": 0.29821110258387634,
+      "learning_rate": 2.7831341732773733e-05,
+      "loss": 0.2561,
+      "step": 337
+    },
+    {
+      "epoch": 0.21632,
+      "grad_norm": 0.33344963895852425,
+      "learning_rate": 2.7814855170040083e-05,
+      "loss": 0.2859,
+      "step": 338
+    },
+    {
+      "epoch": 0.21696,
+      "grad_norm": 0.3260713930863079,
+      "learning_rate": 2.7798311098946375e-05,
+      "loss": 0.2529,
+      "step": 339
+    },
+    {
+      "epoch": 0.2176,
+      "grad_norm": 0.3749841150106177,
+      "learning_rate": 2.7781709593736316e-05,
+      "loss": 0.2807,
+      "step": 340
+    },
+    {
+      "epoch": 0.21824,
+      "grad_norm": 0.34099250856014796,
+      "learning_rate": 2.7765050728911363e-05,
+      "loss": 0.2886,
+      "step": 341
+    },
+    {
+      "epoch": 0.21888,
+      "grad_norm": 0.33765459803209535,
+      "learning_rate": 2.7748334579230375e-05,
+      "loss": 0.2497,
+      "step": 342
+    },
+    {
+      "epoch": 0.21952,
+      "grad_norm": 0.2851983974846401,
+      "learning_rate": 2.773156121970929e-05,
+      "loss": 0.2401,
+      "step": 343
+    },
+    {
+      "epoch": 0.22016,
+      "grad_norm": 3.444718072231663,
+      "learning_rate": 2.7714730725620786e-05,
+      "loss": 0.2706,
+      "step": 344
+    },
+    {
+      "epoch": 0.2208,
+      "grad_norm": 0.41628559582587626,
+      "learning_rate": 2.7697843172493925e-05,
+      "loss": 0.2732,
+      "step": 345
+    },
+    {
+      "epoch": 0.22144,
+      "grad_norm": 0.33063598183417514,
+      "learning_rate": 2.7680898636113845e-05,
+      "loss": 0.2775,
+      "step": 346
+    },
+    {
+      "epoch": 0.22208,
+      "grad_norm": 0.3317549477105874,
+      "learning_rate": 2.7663897192521393e-05,
+      "loss": 0.2673,
+      "step": 347
+    },
+    {
+      "epoch": 0.22272,
+      "grad_norm": 0.39451049358358514,
+      "learning_rate": 2.764683891801281e-05,
+      "loss": 0.2853,
+      "step": 348
+    },
+    {
+      "epoch": 0.22336,
+      "grad_norm": 0.33825604933869896,
+      "learning_rate": 2.7629723889139345e-05,
+      "loss": 0.2836,
+      "step": 349
+    },
+    {
+      "epoch": 0.224,
+      "grad_norm": 0.3938596661138968,
+      "learning_rate": 2.7612552182706968e-05,
+      "loss": 0.2679,
+      "step": 350
+    },
+    {
+      "epoch": 0.22464,
+      "grad_norm": 0.3709101970004551,
+      "learning_rate": 2.759532387577599e-05,
+      "loss": 0.2819,
+      "step": 351
+    },
+    {
+      "epoch": 0.22528,
+      "grad_norm": 0.35417172495939264,
+      "learning_rate": 2.7578039045660713e-05,
+      "loss": 0.2639,
+      "step": 352
+    },
+    {
+      "epoch": 0.22592,
+      "grad_norm": 0.3327176351302671,
+      "learning_rate": 2.75606977699291e-05,
+      "loss": 0.2639,
+      "step": 353
+    },
+    {
+      "epoch": 0.22656,
+      "grad_norm": 0.3564326153038442,
+      "learning_rate": 2.754330012640243e-05,
+      "loss": 0.2556,
+      "step": 354
+    },
+    {
+      "epoch": 0.2272,
+      "grad_norm": 0.32512504523834834,
+      "learning_rate": 2.752584619315493e-05,
+      "loss": 0.2643,
+      "step": 355
+    },
+    {
+      "epoch": 0.22784,
+      "grad_norm": 0.3882343021011366,
+      "learning_rate": 2.7508336048513437e-05,
+      "loss": 0.2786,
+      "step": 356
+    },
+    {
+      "epoch": 0.22848,
+      "grad_norm": 0.3645455650133275,
+      "learning_rate": 2.7490769771057043e-05,
+      "loss": 0.2642,
+      "step": 357
+    },
+    {
+      "epoch": 0.22912,
+      "grad_norm": 0.36026260628073103,
+      "learning_rate": 2.747314743961675e-05,
+      "loss": 0.2639,
+      "step": 358
+    },
+    {
+      "epoch": 0.22976,
+      "grad_norm": 0.3515677194617887,
+      "learning_rate": 2.7455469133275095e-05,
+      "loss": 0.2814,
+      "step": 359
+    },
+    {
+      "epoch": 0.2304,
+      "grad_norm": 0.34391091885619035,
+      "learning_rate": 2.743773493136583e-05,
+      "loss": 0.2673,
+      "step": 360
+    },
+    {
+      "epoch": 0.23104,
+      "grad_norm": 0.3449879915659821,
+      "learning_rate": 2.7419944913473533e-05,
+      "loss": 0.2333,
+      "step": 361
+    },
+    {
+      "epoch": 0.23168,
+      "grad_norm": 0.31990363584739007,
+      "learning_rate": 2.7402099159433258e-05,
+      "loss": 0.2532,
+      "step": 362
+    },
+    {
+      "epoch": 0.23232,
+      "grad_norm": 0.3875818304841908,
+      "learning_rate": 2.73841977493302e-05,
+      "loss": 0.2811,
+      "step": 363
+    },
+    {
+      "epoch": 0.23296,
+      "grad_norm": 0.3566549850839257,
+      "learning_rate": 2.7366240763499302e-05,
+      "loss": 0.2507,
+      "step": 364
+    },
+    {
+      "epoch": 0.2336,
+      "grad_norm": 0.3080698897167787,
+      "learning_rate": 2.7348228282524916e-05,
+      "loss": 0.2603,
+      "step": 365
+    },
+    {
+      "epoch": 0.23424,
+      "grad_norm": 0.3357967412245159,
+      "learning_rate": 2.7330160387240443e-05,
+      "loss": 0.2451,
+      "step": 366
+    },
+    {
+      "epoch": 0.23488,
+      "grad_norm": 0.3527005868734707,
+      "learning_rate": 2.731203715872795e-05,
+      "loss": 0.275,
+      "step": 367
+    },
+    {
+      "epoch": 0.23552,
+      "grad_norm": 0.316985061769891,
+      "learning_rate": 2.729385867831783e-05,
+      "loss": 0.269,
+      "step": 368
+    },
+    {
+      "epoch": 0.23616,
+      "grad_norm": 0.33344202384537386,
+      "learning_rate": 2.7275625027588414e-05,
+      "loss": 0.2691,
+      "step": 369
+    },
+    {
+      "epoch": 0.2368,
+      "grad_norm": 0.34625417830177385,
+      "learning_rate": 2.7257336288365634e-05,
+      "loss": 0.2653,
+      "step": 370
+    },
+    {
+      "epoch": 0.23744,
+      "grad_norm": 0.3267922712231095,
+      "learning_rate": 2.7238992542722625e-05,
+      "loss": 0.2617,
+      "step": 371
+    },
+    {
+      "epoch": 0.23808,
+      "grad_norm": 0.330793248709517,
+      "learning_rate": 2.722059387297938e-05,
+      "loss": 0.2574,
+      "step": 372
+    },
+    {
+      "epoch": 0.23872,
+      "grad_norm": 0.3463369707262466,
+      "learning_rate": 2.720214036170236e-05,
+      "loss": 0.2627,
+      "step": 373
+    },
+    {
+      "epoch": 0.23936,
+      "grad_norm": 0.33043396116356993,
+      "learning_rate": 2.7183632091704143e-05,
+      "loss": 0.2708,
+      "step": 374
+    },
+    {
+      "epoch": 0.24,
+      "grad_norm": 0.3485722629331789,
+      "learning_rate": 2.716506914604305e-05,
+      "loss": 0.2612,
+      "step": 375
+    },
+    {
+      "epoch": 0.24064,
+      "grad_norm": 0.3412635617995565,
+      "learning_rate": 2.7146451608022748e-05,
+      "loss": 0.2607,
+      "step": 376
+    },
+    {
+      "epoch": 0.24128,
+      "grad_norm": 0.35685994147745786,
+      "learning_rate": 2.7127779561191905e-05,
+      "loss": 0.2577,
+      "step": 377
+    },
+    {
+      "epoch": 0.24192,
+      "grad_norm": 0.3039380817502753,
+      "learning_rate": 2.7109053089343815e-05,
+      "loss": 0.2611,
+      "step": 378
+    },
+    {
+      "epoch": 0.24256,
+      "grad_norm": 0.43279102042200523,
+      "learning_rate": 2.7090272276515993e-05,
+      "loss": 0.2756,
+      "step": 379
+    },
+    {
+      "epoch": 0.2432,
+      "grad_norm": 0.31054351672072755,
+      "learning_rate": 2.707143720698983e-05,
+      "loss": 0.2708,
+      "step": 380
+    },
+    {
+      "epoch": 0.24384,
+      "grad_norm": 0.36851176929762747,
+      "learning_rate": 2.7052547965290186e-05,
+      "loss": 0.276,
+      "step": 381
+    },
+    {
+      "epoch": 0.24448,
+      "grad_norm": 0.3627663453356185,
+      "learning_rate": 2.703360463618504e-05,
+      "loss": 0.2732,
+      "step": 382
+    },
+    {
+      "epoch": 0.24512,
+      "grad_norm": 0.3576702828639981,
+      "learning_rate": 2.7014607304685096e-05,
+      "loss": 0.2625,
+      "step": 383
+    },
+    {
+      "epoch": 0.24576,
+      "grad_norm": 0.3437585046911205,
+      "learning_rate": 2.699555605604339e-05,
+      "loss": 0.2717,
+      "step": 384
+    },
+    {
+      "epoch": 0.2464,
+      "grad_norm": 0.3563941457915153,
+      "learning_rate": 2.6976450975754923e-05,
+      "loss": 0.2599,
+      "step": 385
+    },
+    {
+      "epoch": 0.24704,
+      "grad_norm": 0.32499802120658194,
+      "learning_rate": 2.6957292149556276e-05,
+      "loss": 0.2677,
+      "step": 386
+    },
+    {
+      "epoch": 0.24768,
+      "grad_norm": 0.33344191861008443,
+      "learning_rate": 2.6938079663425218e-05,
+      "loss": 0.2882,
+      "step": 387
+    },
+    {
+      "epoch": 0.24832,
+      "grad_norm": 0.3517570106726394,
+      "learning_rate": 2.691881360358033e-05,
+      "loss": 0.2754,
+      "step": 388
+    },
+    {
+      "epoch": 0.24896,
+      "grad_norm": 0.3237247905400656,
+      "learning_rate": 2.6899494056480596e-05,
+      "loss": 0.2545,
+      "step": 389
+    },
+    {
+      "epoch": 0.2496,
+      "grad_norm": 0.3605247711441668,
+      "learning_rate": 2.6880121108825056e-05,
+      "loss": 0.2858,
+      "step": 390
+    },
+    {
+      "epoch": 0.25024,
+      "grad_norm": 0.3227764776694619,
+      "learning_rate": 2.6860694847552374e-05,
+      "loss": 0.2588,
+      "step": 391
+    },
+    {
+      "epoch": 0.25088,
+      "grad_norm": 0.30914534952022565,
+      "learning_rate": 2.6841215359840468e-05,
+      "loss": 0.265,
+      "step": 392
+    },
+    {
+      "epoch": 0.25152,
+      "grad_norm": 0.3040468595020032,
+      "learning_rate": 2.682168273310612e-05,
+      "loss": 0.2572,
+      "step": 393
+    },
+    {
+      "epoch": 0.25216,
+      "grad_norm": 0.3582628170978083,
+      "learning_rate": 2.680209705500458e-05,
+      "loss": 0.2692,
+      "step": 394
+    },
+    {
+      "epoch": 0.2528,
+      "grad_norm": 0.3288877765942096,
+      "learning_rate": 2.6782458413429177e-05,
+      "loss": 0.2556,
+      "step": 395
+    },
+    {
+      "epoch": 0.25344,
+      "grad_norm": 0.3483940416687742,
+      "learning_rate": 2.6762766896510914e-05,
+      "loss": 0.2509,
+      "step": 396
+    },
+    {
+      "epoch": 0.25408,
+      "grad_norm": 0.4148522716152593,
+      "learning_rate": 2.6743022592618075e-05,
+      "loss": 0.2855,
+      "step": 397
+    },
+    {
+      "epoch": 0.25472,
+      "grad_norm": 0.3153140469461542,
+      "learning_rate": 2.6723225590355852e-05,
+      "loss": 0.2422,
+      "step": 398
+    },
+    {
+      "epoch": 0.25536,
+      "grad_norm": 0.3414987747573693,
+      "learning_rate": 2.6703375978565906e-05,
+      "loss": 0.2566,
+      "step": 399
+    },
+    {
+      "epoch": 0.256,
+      "grad_norm": 0.3082723258062477,
+      "learning_rate": 2.6683473846326002e-05,
+      "loss": 0.2692,
+      "step": 400
+    },
+    {
+      "epoch": 0.25664,
+      "grad_norm": 0.3120921043255892,
+      "learning_rate": 2.666351928294959e-05,
+      "loss": 0.2484,
+      "step": 401
+    },
+    {
+      "epoch": 0.25728,
+      "grad_norm": 0.34491023806967935,
+      "learning_rate": 2.6643512377985423e-05,
+      "loss": 0.2634,
+      "step": 402
+    },
+    {
+      "epoch": 0.25792,
+      "grad_norm": 1.227578859205086,
+      "learning_rate": 2.6623453221217137e-05,
+      "loss": 0.2681,
+      "step": 403
+    },
+    {
+      "epoch": 0.25856,
+      "grad_norm": 0.36139208492186886,
+      "learning_rate": 2.660334190266285e-05,
+      "loss": 0.2781,
+      "step": 404
+    },
+    {
+      "epoch": 0.2592,
+      "grad_norm": 0.3146578283750008,
+      "learning_rate": 2.658317851257477e-05,
+      "loss": 0.2661,
+      "step": 405
+    },
+    {
+      "epoch": 0.25984,
+      "grad_norm": 0.3241167195918667,
+      "learning_rate": 2.6562963141438783e-05,
+      "loss": 0.2747,
+      "step": 406
+    },
+    {
+      "epoch": 0.26048,
+      "grad_norm": 0.32983966170598866,
+      "learning_rate": 2.6542695879974044e-05,
+      "loss": 0.2821,
+      "step": 407
+    },
+    {
+      "epoch": 0.26112,
+      "grad_norm": 0.3302761405872256,
+      "learning_rate": 2.652237681913257e-05,
+      "loss": 0.2677,
+      "step": 408
+    },
+    {
+      "epoch": 0.26176,
+      "grad_norm": 0.33848152131881326,
+      "learning_rate": 2.6502006050098842e-05,
+      "loss": 0.2542,
+      "step": 409
+    },
+    {
+      "epoch": 0.2624,
+      "grad_norm": 0.3413235227280745,
+      "learning_rate": 2.648158366428938e-05,
+      "loss": 0.2664,
+      "step": 410
+    },
+    {
+      "epoch": 0.26304,
+      "grad_norm": 0.32916338812223606,
+      "learning_rate": 2.6461109753352355e-05,
+      "loss": 0.2616,
+      "step": 411
+    },
+    {
+      "epoch": 0.26368,
+      "grad_norm": 0.3053892800899407,
+      "learning_rate": 2.6440584409167144e-05,
+      "loss": 0.2583,
+      "step": 412
+    },
+    {
+      "epoch": 0.26432,
+      "grad_norm": 0.3385379902087113,
+      "learning_rate": 2.6420007723843952e-05,
+      "loss": 0.256,
+      "step": 413
+    },
+    {
+      "epoch": 0.26496,
+      "grad_norm": 0.37571936271303413,
+      "learning_rate": 2.639937978972338e-05,
+      "loss": 0.2503,
+      "step": 414
+    },
+    {
+      "epoch": 0.2656,
+      "grad_norm": 0.2854100162554339,
+      "learning_rate": 2.6378700699376015e-05,
+      "loss": 0.2415,
+      "step": 415
+    },
+    {
+      "epoch": 0.26624,
+      "grad_norm": 0.31271921873205044,
+      "learning_rate": 2.6357970545602014e-05,
+      "loss": 0.2475,
+      "step": 416
+    },
+    {
+      "epoch": 0.26688,
+      "grad_norm": 0.34030180044647645,
+      "learning_rate": 2.6337189421430688e-05,
+      "loss": 0.2379,
+      "step": 417
+    },
+    {
+      "epoch": 0.26752,
+      "grad_norm": 0.4075617302343357,
+      "learning_rate": 2.631635742012009e-05,
+      "loss": 0.2742,
+      "step": 418
+    },
+    {
+      "epoch": 0.26816,
+      "grad_norm": 0.3607203044257088,
+      "learning_rate": 2.629547463515657e-05,
+      "loss": 0.2694,
+      "step": 419
+    },
+    {
+      "epoch": 0.2688,
+      "grad_norm": 0.3244858888191209,
+      "learning_rate": 2.6274541160254405e-05,
+      "loss": 0.2633,
+      "step": 420
+    },
+    {
+      "epoch": 0.26944,
+      "grad_norm": 0.3940598188912196,
+      "learning_rate": 2.6253557089355333e-05,
+      "loss": 0.2466,
+      "step": 421
+    },
+    {
+      "epoch": 0.27008,
+      "grad_norm": 0.3011262776053056,
+      "learning_rate": 2.6232522516628153e-05,
+      "loss": 0.2727,
+      "step": 422
+    },
+    {
+      "epoch": 0.27072,
+      "grad_norm": 0.3739061295982007,
+      "learning_rate": 2.6211437536468292e-05,
+      "loss": 0.2671,
+      "step": 423
+    },
+    {
+      "epoch": 0.27136,
+      "grad_norm": 0.32945664324477797,
+      "learning_rate": 2.6190302243497396e-05,
+      "loss": 0.2708,
+      "step": 424
+    },
+    {
+      "epoch": 0.272,
+      "grad_norm": 0.36534937038877574,
+      "learning_rate": 2.6169116732562892e-05,
+      "loss": 0.2298,
+      "step": 425
+    },
+    {
+      "epoch": 0.27264,
+      "grad_norm": 0.33670531669484066,
+      "learning_rate": 2.6147881098737562e-05,
+      "loss": 0.2681,
+      "step": 426
+    },
+    {
+      "epoch": 0.27328,
+      "grad_norm": 0.35965800175602797,
+      "learning_rate": 2.612659543731913e-05,
+      "loss": 0.2675,
+      "step": 427
+    },
+    {
+      "epoch": 0.27392,
+      "grad_norm": 0.3509912783523724,
+      "learning_rate": 2.610525984382982e-05,
+      "loss": 0.2752,
+      "step": 428
+    },
+    {
+      "epoch": 0.27456,
+      "grad_norm": 0.3341800947139574,
+      "learning_rate": 2.6083874414015934e-05,
+      "loss": 0.2651,
+      "step": 429
+    },
+    {
+      "epoch": 0.2752,
+      "grad_norm": 0.3512265177853723,
+      "learning_rate": 2.606243924384742e-05,
+      "loss": 0.2746,
+      "step": 430
+    },
+    {
+      "epoch": 0.27584,
+      "grad_norm": 0.3314007433299324,
+      "learning_rate": 2.6040954429517442e-05,
+      "loss": 0.2406,
+      "step": 431
+    },
+    {
+      "epoch": 0.27648,
+      "grad_norm": 0.31623472637208033,
+      "learning_rate": 2.6019420067441958e-05,
+      "loss": 0.2432,
+      "step": 432
+    },
+    {
+      "epoch": 0.27712,
+      "grad_norm": 0.33629626691329856,
+      "learning_rate": 2.599783625425926e-05,
+      "loss": 0.2642,
+      "step": 433
+    },
+    {
+      "epoch": 0.27776,
+      "grad_norm": 0.32932652640189397,
+      "learning_rate": 2.597620308682957e-05,
+      "loss": 0.2722,
+      "step": 434
+    },
+    {
+      "epoch": 0.2784,
+      "grad_norm": 0.3162553853731888,
+      "learning_rate": 2.5954520662234602e-05,
+      "loss": 0.253,
+      "step": 435
+    },
+    {
+      "epoch": 0.27904,
+      "grad_norm": 0.32294208690580367,
+      "learning_rate": 2.5932789077777103e-05,
+      "loss": 0.2453,
+      "step": 436
+    },
+    {
+      "epoch": 0.27968,
+      "grad_norm": 0.2860688977348506,
+      "learning_rate": 2.591100843098043e-05,
+      "loss": 0.2583,
+      "step": 437
+    },
+    {
+      "epoch": 0.28032,
+      "grad_norm": 0.3382649825825897,
+      "learning_rate": 2.5889178819588134e-05,
+      "loss": 0.243,
+      "step": 438
+    },
+    {
+      "epoch": 0.28096,
+      "grad_norm": 0.3426287563355019,
+      "learning_rate": 2.5867300341563477e-05,
+      "loss": 0.2832,
+      "step": 439
+    },
+    {
+      "epoch": 0.2816,
+      "grad_norm": 0.3616023005679438,
+      "learning_rate": 2.5845373095089028e-05,
+      "loss": 0.2586,
+      "step": 440
+    },
+    {
+      "epoch": 0.28224,
+      "grad_norm": 0.3680942120537055,
+      "learning_rate": 2.5823397178566217e-05,
+      "loss": 0.2449,
+      "step": 441
+    },
+    {
+      "epoch": 0.28288,
+      "grad_norm": 0.3266840109358284,
+      "learning_rate": 2.580137269061488e-05,
+      "loss": 0.2806,
+      "step": 442
+    },
+    {
+      "epoch": 0.28352,
+      "grad_norm": 0.3268151383316364,
+      "learning_rate": 2.5779299730072815e-05,
+      "loss": 0.264,
+      "step": 443
+    },
+    {
+      "epoch": 0.28416,
+      "grad_norm": 0.32651398276400556,
+      "learning_rate": 2.5757178395995358e-05,
+      "loss": 0.2672,
+      "step": 444
+    },
+    {
+      "epoch": 0.2848,
+      "grad_norm": 0.35504343797167337,
+      "learning_rate": 2.5735008787654933e-05,
+      "loss": 0.2563,
+      "step": 445
+    },
+    {
+      "epoch": 0.28544,
+      "grad_norm": 0.3332900931306675,
+      "learning_rate": 2.5712791004540592e-05,
+      "loss": 0.2679,
+      "step": 446
+    },
+    {
+      "epoch": 0.28608,
+      "grad_norm": 0.32153395285613084,
+      "learning_rate": 2.5690525146357575e-05,
+      "loss": 0.2572,
+      "step": 447
+    },
+    {
+      "epoch": 0.28672,
+      "grad_norm": 0.33090800287138983,
+      "learning_rate": 2.566821131302688e-05,
+      "loss": 0.2834,
+      "step": 448
+    },
+    {
+      "epoch": 0.28736,
+      "grad_norm": 0.336200932174992,
+      "learning_rate": 2.5645849604684775e-05,
+      "loss": 0.248,
+      "step": 449
+    },
+    {
+      "epoch": 0.288,
+      "grad_norm": 0.32237186462753925,
+      "learning_rate": 2.56234401216824e-05,
+      "loss": 0.2665,
+      "step": 450
+    },
+    {
+      "epoch": 0.28864,
+      "grad_norm": 0.3195380119597909,
+      "learning_rate": 2.5600982964585272e-05,
+      "loss": 0.2508,
+      "step": 451
+    },
+    {
+      "epoch": 0.28928,
+      "grad_norm": 0.3475060876563337,
+      "learning_rate": 2.5578478234172865e-05,
+      "loss": 0.2594,
+      "step": 452
+    },
+    {
+      "epoch": 0.28992,
+      "grad_norm": 0.627047630421656,
+      "learning_rate": 2.5555926031438134e-05,
+      "loss": 0.2485,
+      "step": 453
+    },
+    {
+      "epoch": 0.29056,
+      "grad_norm": 0.3005632921400129,
+      "learning_rate": 2.5533326457587072e-05,
+      "loss": 0.2363,
+      "step": 454
+    },
+    {
+      "epoch": 0.2912,
+      "grad_norm": 0.3907484843122131,
+      "learning_rate": 2.551067961403827e-05,
+      "loss": 0.2635,
+      "step": 455
+    },
+    {
+      "epoch": 0.29184,
+      "grad_norm": 0.30910241081220835,
+      "learning_rate": 2.5487985602422425e-05,
+      "loss": 0.2554,
+      "step": 456
+    },
+    {
+      "epoch": 0.29248,
+      "grad_norm": 0.3316657790671481,
+      "learning_rate": 2.546524452458193e-05,
+      "loss": 0.2465,
+      "step": 457
+    },
+    {
+      "epoch": 0.29312,
+      "grad_norm": 0.33514519152514977,
+      "learning_rate": 2.5442456482570383e-05,
+      "loss": 0.2509,
+      "step": 458
+    },
+    {
+      "epoch": 0.29376,
+      "grad_norm": 0.3135722919118478,
+      "learning_rate": 2.5419621578652137e-05,
+      "loss": 0.2584,
+      "step": 459
+    },
+    {
+      "epoch": 0.2944,
+      "grad_norm": 0.8767929120277721,
+      "learning_rate": 2.539673991530185e-05,
+      "loss": 0.2649,
+      "step": 460
+    },
+    {
+      "epoch": 0.29504,
+      "grad_norm": 0.32933106915081395,
+      "learning_rate": 2.5373811595204022e-05,
+      "loss": 0.2727,
+      "step": 461
+    },
+    {
+      "epoch": 0.29568,
+      "grad_norm": 0.4082581717544081,
+      "learning_rate": 2.5350836721252525e-05,
+      "loss": 0.2569,
+      "step": 462
+    },
+    {
+      "epoch": 0.29632,
+      "grad_norm": 0.3412102305943323,
+      "learning_rate": 2.5327815396550142e-05,
+      "loss": 0.2639,
+      "step": 463
+    },
+    {
+      "epoch": 0.29696,
+      "grad_norm": 0.31780550108168804,
+      "learning_rate": 2.530474772440812e-05,
+      "loss": 0.2418,
+      "step": 464
+    },
+    {
+      "epoch": 0.2976,
+      "grad_norm": 0.29529324755348607,
+      "learning_rate": 2.5281633808345702e-05,
+      "loss": 0.2471,
+      "step": 465
+    },
+    {
+      "epoch": 0.29824,
+      "grad_norm": 0.33728330149501135,
+      "learning_rate": 2.5258473752089636e-05,
+      "loss": 0.2801,
+      "step": 466
+    },
+    {
+      "epoch": 0.29888,
+      "grad_norm": 0.3267488434365671,
+      "learning_rate": 2.5235267659573746e-05,
+      "loss": 0.2414,
+      "step": 467
+    },
+    {
+      "epoch": 0.29952,
+      "grad_norm": 0.29150985108582134,
+      "learning_rate": 2.521201563493845e-05,
+      "loss": 0.265,
+      "step": 468
+    },
+    {
+      "epoch": 0.30016,
+      "grad_norm": 0.30336323784858,
+      "learning_rate": 2.5188717782530292e-05,
+      "loss": 0.2593,
+      "step": 469
+    },
+    {
+      "epoch": 0.3008,
+      "grad_norm": 0.29926798408682637,
+      "learning_rate": 2.516537420690146e-05,
+      "loss": 0.2604,
+      "step": 470
+    },
+    {
+      "epoch": 0.30144,
+      "grad_norm": 0.3065859162009593,
+      "learning_rate": 2.514198501280934e-05,
+      "loss": 0.2629,
+      "step": 471
+    },
+    {
+      "epoch": 0.30208,
+      "grad_norm": 0.3049740162502653,
+      "learning_rate": 2.5118550305216054e-05,
+      "loss": 0.2754,
+      "step": 472
+    },
+    {
+      "epoch": 0.30272,
+      "grad_norm": 0.30876372307121774,
+      "learning_rate": 2.5095070189287944e-05,
+      "loss": 0.2668,
+      "step": 473
+    },
+    {
+      "epoch": 0.30336,
+      "grad_norm": 0.29273269101927585,
+      "learning_rate": 2.5071544770395143e-05,
+      "loss": 0.2558,
+      "step": 474
+    },
+    {
+      "epoch": 0.304,
+      "grad_norm": 0.3076653994547958,
+      "learning_rate": 2.5047974154111092e-05,
+      "loss": 0.2489,
+      "step": 475
+    },
+    {
+      "epoch": 0.30464,
+      "grad_norm": 0.3333719011093153,
+      "learning_rate": 2.5024358446212046e-05,
+      "loss": 0.2435,
+      "step": 476
+    },
+    {
+      "epoch": 0.30528,
+      "grad_norm": 0.38541240875825966,
+      "learning_rate": 2.5000697752676622e-05,
+      "loss": 0.2791,
+      "step": 477
+    },
+    {
+      "epoch": 0.30592,
+      "grad_norm": 0.31076311872010254,
+      "learning_rate": 2.4976992179685317e-05,
+      "loss": 0.2693,
+      "step": 478
+    },
+    {
+      "epoch": 0.30656,
+      "grad_norm": 0.3140598892806967,
+      "learning_rate": 2.4953241833620034e-05,
+      "loss": 0.2453,
+      "step": 479
+    },
+    {
+      "epoch": 0.3072,
+      "grad_norm": 0.30866549699497225,
+      "learning_rate": 2.4929446821063596e-05,
+      "loss": 0.2494,
+      "step": 480
+    },
+    {
+      "epoch": 0.30784,
+      "grad_norm": 0.3057358078354188,
+      "learning_rate": 2.4905607248799265e-05,
+      "loss": 0.2583,
+      "step": 481
+    },
+    {
+      "epoch": 0.30848,
+      "grad_norm": 0.2832731794280749,
+      "learning_rate": 2.4881723223810295e-05,
+      "loss": 0.2687,
+      "step": 482
+    },
+    {
+      "epoch": 0.30912,
+      "grad_norm": 0.3051553661664977,
+      "learning_rate": 2.4857794853279396e-05,
+      "loss": 0.2599,
+      "step": 483
+    },
+    {
+      "epoch": 0.30976,
+      "grad_norm": 0.2998378381048481,
+      "learning_rate": 2.4833822244588312e-05,
+      "loss": 0.2451,
+      "step": 484
+    },
+    {
+      "epoch": 0.3104,
+      "grad_norm": 0.3174813787668605,
+      "learning_rate": 2.4809805505317296e-05,
+      "loss": 0.264,
+      "step": 485
+    },
+    {
+      "epoch": 0.31104,
+      "grad_norm": 0.28982862061791764,
+      "learning_rate": 2.4785744743244644e-05,
+      "loss": 0.2478,
+      "step": 486
+    },
+    {
+      "epoch": 0.31168,
+      "grad_norm": 0.3147239609211115,
+      "learning_rate": 2.4761640066346217e-05,
+      "loss": 0.2372,
+      "step": 487
+    },
+    {
+      "epoch": 0.31232,
+      "grad_norm": 0.2898871152482304,
+      "learning_rate": 2.4737491582794945e-05,
+      "loss": 0.2638,
+      "step": 488
+    },
+    {
+      "epoch": 0.31296,
+      "grad_norm": 0.33890483931828097,
+      "learning_rate": 2.4713299400960342e-05,
+      "loss": 0.2444,
+      "step": 489
+    },
+    {
+      "epoch": 0.3136,
+      "grad_norm": 0.28327565279880107,
+      "learning_rate": 2.4689063629408034e-05,
+      "loss": 0.272,
+      "step": 490
+    },
+    {
+      "epoch": 0.31424,
+      "grad_norm": 0.3212172431491676,
+      "learning_rate": 2.4664784376899257e-05,
+      "loss": 0.2545,
+      "step": 491
+    },
+    {
+      "epoch": 0.31488,
+      "grad_norm": 0.3367903918403918,
+      "learning_rate": 2.4640461752390367e-05,
+      "loss": 0.2632,
+      "step": 492
+    },
+    {
+      "epoch": 0.31552,
+      "grad_norm": 0.3230413061531243,
+      "learning_rate": 2.4616095865032366e-05,
+      "loss": 0.2825,
+      "step": 493
+    },
+    {
+      "epoch": 0.31616,
+      "grad_norm": 0.34034207513600895,
+      "learning_rate": 2.459168682417041e-05,
+      "loss": 0.271,
+      "step": 494
+    },
+    {
+      "epoch": 0.3168,
+      "grad_norm": 0.32141809222820283,
+      "learning_rate": 2.4567234739343283e-05,
+      "loss": 0.2396,
+      "step": 495
+    },
+    {
+      "epoch": 0.31744,
+      "grad_norm": 0.2753406394259948,
+      "learning_rate": 2.454273972028297e-05,
+      "loss": 0.2404,
+      "step": 496
+    },
+    {
+      "epoch": 0.31808,
+      "grad_norm": 0.3351114179983899,
+      "learning_rate": 2.451820187691411e-05,
+      "loss": 0.2763,
+      "step": 497
+    },
+    {
+      "epoch": 0.31872,
+      "grad_norm": 0.3513219675385064,
+      "learning_rate": 2.4493621319353525e-05,
+      "loss": 0.2609,
+      "step": 498
+    },
+    {
+      "epoch": 0.31936,
+      "grad_norm": 0.3110180234096309,
+      "learning_rate": 2.4468998157909723e-05,
+      "loss": 0.2402,
+      "step": 499
+    },
+    {
+      "epoch": 0.32,
+      "grad_norm": 0.2964890570904897,
+      "learning_rate": 2.444433250308241e-05,
+      "loss": 0.2659,
+      "step": 500
+    },
+    {
+      "epoch": 0.32064,
+      "grad_norm": 0.3579764939490037,
+      "learning_rate": 2.4419624465561964e-05,
+      "loss": 0.2514,
+      "step": 501
+    },
+    {
+      "epoch": 0.32128,
+      "grad_norm": 0.3144597113409246,
+      "learning_rate": 2.4394874156228988e-05,
+      "loss": 0.2608,
+      "step": 502
+    },
+    {
+      "epoch": 0.32192,
+      "grad_norm": 0.31609390625354067,
+      "learning_rate": 2.4370081686153767e-05,
+      "loss": 0.2443,
+      "step": 503
+    },
+    {
+      "epoch": 0.32256,
+      "grad_norm": 0.29746982935843747,
+      "learning_rate": 2.4345247166595803e-05,
+      "loss": 0.2377,
+      "step": 504
+    },
+    {
+      "epoch": 0.3232,
+      "grad_norm": 0.34126278199611504,
+      "learning_rate": 2.4320370709003284e-05,
+      "loss": 0.2554,
+      "step": 505
+    },
+    {
+      "epoch": 0.32384,
+      "grad_norm": 0.2990237684037871,
+      "learning_rate": 2.429545242501261e-05,
+      "loss": 0.2848,
+      "step": 506
+    },
+    {
+      "epoch": 0.32448,
+      "grad_norm": 0.32322206092320094,
+      "learning_rate": 2.4270492426447884e-05,
+      "loss": 0.2561,
+      "step": 507
+    },
+    {
+      "epoch": 0.32512,
+      "grad_norm": 0.3260943452277802,
+      "learning_rate": 2.424549082532041e-05,
+      "loss": 0.26,
+      "step": 508
+    },
+    {
+      "epoch": 0.32576,
+      "grad_norm": 0.4295758739365506,
+      "learning_rate": 2.422044773382817e-05,
+      "loss": 0.2513,
+      "step": 509
+    },
+    {
+      "epoch": 0.3264,
+      "grad_norm": 0.31835152340497586,
+      "learning_rate": 2.4195363264355365e-05,
+      "loss": 0.2505,
+      "step": 510
+    },
+    {
+      "epoch": 0.32704,
+      "grad_norm": 0.3531418180586244,
+      "learning_rate": 2.417023752947188e-05,
+      "loss": 0.2609,
+      "step": 511
+    },
+    {
+      "epoch": 0.32768,
+      "grad_norm": 0.28968278915332163,
+      "learning_rate": 2.4145070641932767e-05,
+      "loss": 0.2472,
+      "step": 512
+    },
+    {
+      "epoch": 0.32832,
+      "grad_norm": 0.3176579318606065,
+      "learning_rate": 2.4119862714677773e-05,
+      "loss": 0.2563,
+      "step": 513
+    },
+    {
+      "epoch": 0.32896,
+      "grad_norm": 0.3447755475779212,
+      "learning_rate": 2.4094613860830813e-05,
+      "loss": 0.299,
+      "step": 514
+    },
+    {
+      "epoch": 0.3296,
+      "grad_norm": 0.30246755944222914,
+      "learning_rate": 2.4069324193699453e-05,
+      "loss": 0.2285,
+      "step": 515
+    },
+    {
+      "epoch": 0.33024,
+      "grad_norm": 0.36003409699329864,
+      "learning_rate": 2.4043993826774433e-05,
+      "loss": 0.2473,
+      "step": 516
+    },
+    {
+      "epoch": 0.33088,
+      "grad_norm": 0.35159179010717023,
+      "learning_rate": 2.4018622873729136e-05,
+      "loss": 0.2654,
+      "step": 517
+    },
+    {
+      "epoch": 0.33152,
+      "grad_norm": 0.30073111767960253,
+      "learning_rate": 2.3993211448419055e-05,
+      "loss": 0.249,
+      "step": 518
+    },
+    {
+      "epoch": 0.33216,
+      "grad_norm": 0.3319898092071986,
+      "learning_rate": 2.3967759664881347e-05,
+      "loss": 0.2737,
+      "step": 519
+    },
+    {
+      "epoch": 0.3328,
+      "grad_norm": 0.4467760115064727,
+      "learning_rate": 2.394226763733425e-05,
+      "loss": 0.2585,
+      "step": 520
+    },
+    {
+      "epoch": 0.33344,
+      "grad_norm": 0.3041939438386619,
+      "learning_rate": 2.3916735480176618e-05,
+      "loss": 0.2266,
+      "step": 521
+    }
+  ],
+  "logging_steps": 1,
+  "max_steps": 1562,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 521,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 237774868512768.0,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c2aadc5b31371b9db6c8db28f809924e6b4be7369eb5e93afa0622aae716f332
+size 8017