Upload folder using huggingface_hub

ckpts_oct12/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.3_2e-1_connector-3.0_1.3_2e-1_ablation/added_tokens.json

@@ -0,0 +1,24 @@
+{
+  "</tool_call>": 151658,
+  "<tool_call>": 151657,
+  "<|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
+}

ckpts_oct12/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.3_2e-1_connector-3.0_1.3_2e-1_ablation/config.json

@@ -0,0 +1,88 @@
+{
+  "architectures": [
+    "TinyLlavaForConditionalGeneration"
+  ],
+  "backward_type_connector": "normal",
+  "cache_dir": null,
+  "connector_type": "mlp2x_gelu",
+  "hidden_size": 896,
+  "ignore_index": -100,
+  "image_aspect_ratio": "square",
+  "image_token_index": -200,
+  "llm_model_name_or_path": "Qwen/Qwen2.5-0.5B",
+  "mask_model": [
+    "llm",
+    "connector"
+  ],
+  "mask_type_connector": "soft",
+  "model_type": "tinyllava",
+  "num_queries": 128,
+  "num_resampler_layers": 3,
+  "pad_token": "<|endoftext|>",
+  "pad_token_id": 151643,
+  "resampler_hidden_size": 768,
+  "sparsity_connector": null,
+  "subnet_type_connector": "global",
+  "temperature_connector": 1.3,
+  "text_config": {
+    "_name_or_path": "Qwen/Qwen2.5-0.5B",
+    "architectures": [
+      "Qwen2ForCausalLM"
+    ],
+    "backward_type": "normal",
+    "bos_token_id": 151643,
+    "eos_token_id": 151643,
+    "hidden_size": 896,
+    "intermediate_size": 4864,
+    "mask_type": "soft",
+    "masked_layers": "all",
+    "max_position_embeddings": 32768,
+    "max_window_layers": 24,
+    "model_type": "qwen2",
+    "num_attention_heads": 14,
+    "num_hidden_layers": 24,
+    "num_key_value_heads": 2,
+    "rope_theta": 1000000.0,
+    "sliding_window": 32768,
+    "subnet_mode": "both",
+    "subnet_type": "None",
+    "temperature_attn": 1.3,
+    "temperature_mlp": 1.3,
+    "tie_word_embeddings": true,
+    "torch_dtype": "bfloat16",
+    "use_mrope": false,
+    "use_sliding_window": false,
+    "vocab_size": 151936
+  },
+  "threshold_connector": null,
+  "tokenizer_model_max_length": 2048,
+  "tokenizer_name_or_path": "Qwen/Qwen2.5-0.5B",
+  "tokenizer_padding_side": "right",
+  "tokenizer_use_fast": false,
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.40.1",
+  "tune_type_connector": "full",
+  "tune_type_llm": "full",
+  "tune_type_vision_tower": "frozen",
+  "tune_vision_tower_from_layer": 0,
+  "use_cache": true,
+  "vision_config": {
+    "hidden_act": "gelu_pytorch_tanh",
+    "hidden_size": 1152,
+    "image_size": 384,
+    "intermediate_size": 4304,
+    "layer_norm_eps": 1e-06,
+    "model_name_or_path": "google/siglip-so400m-patch14-384",
+    "model_name_or_path2": "",
+    "model_type": "siglip_vision_model",
+    "num_attention_heads": 16,
+    "num_hidden_layers": 27,
+    "patch_size": 14
+  },
+  "vision_feature_layer": -2,
+  "vision_feature_select_strategy": "patch",
+  "vision_hidden_size": 1152,
+  "vision_model_name_or_path": "google/siglip-so400m-patch14-384",
+  "vision_model_name_or_path2": "",
+  "vocab_size": 151936
+}

ckpts_oct12/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.3_2e-1_connector-3.0_1.3_2e-1_ablation/generation_config.json

@@ -0,0 +1,7 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 151643,
+  "eos_token_id": 151643,
+  "transformers_version": "4.40.1",
+  "use_cache": false
+}

ckpts_oct12/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.3_2e-1_connector-3.0_1.3_2e-1_ablation/log.txt

@@ -0,0 +1,2 @@
+2025-10-13 06:58:59,229 | INFO: Total Parameters: 1283756736, Total Trainable Parameters: 359661568
+2025-10-13 06:58:59,239 | INFO: Trainable Parameters:

ckpts_oct12/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.3_2e-1_connector-3.0_1.3_2e-1_ablation/mask_applied/added_tokens.json

@@ -0,0 +1,24 @@
+{
+  "</tool_call>": 151658,
+  "<tool_call>": 151657,
+  "<|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
+}

ckpts_oct12/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.3_2e-1_connector-3.0_1.3_2e-1_ablation/mask_applied/config.json

@@ -0,0 +1,70 @@
+{
+  "architectures": [
+    "TinyLlavaForConditionalGeneration"
+  ],
+  "cache_dir": null,
+  "connector_type": "mlp2x_gelu",
+  "hidden_size": 896,
+  "ignore_index": -100,
+  "image_aspect_ratio": "square",
+  "image_token_index": -200,
+  "llm_model_name_or_path": "Qwen/Qwen2.5-0.5B",
+  "model_type": "tinyllava",
+  "num_queries": 128,
+  "num_resampler_layers": 3,
+  "pad_token": "<|endoftext|>",
+  "resampler_hidden_size": 768,
+  "text_config": {
+    "_name_or_path": "Qwen/Qwen2.5-0.5B",
+    "architectures": [
+      "Qwen2ForCausalLM"
+    ],
+    "bos_token_id": 151643,
+    "eos_token_id": 151643,
+    "hidden_size": 896,
+    "intermediate_size": 4864,
+    "max_position_embeddings": 32768,
+    "max_window_layers": 24,
+    "model_type": "qwen2",
+    "num_attention_heads": 14,
+    "num_hidden_layers": 24,
+    "num_key_value_heads": 2,
+    "rope_theta": 1000000.0,
+    "sliding_window": 32768,
+    "tie_word_embeddings": true,
+    "torch_dtype": "bfloat16",
+    "use_mrope": false,
+    "use_sliding_window": false,
+    "vocab_size": 151936
+  },
+  "tokenizer_model_max_length": 2048,
+  "tokenizer_name_or_path": "Qwen/Qwen2.5-0.5B",
+  "tokenizer_padding_side": "right",
+  "tokenizer_use_fast": false,
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.40.1",
+  "tune_type_connector": "full",
+  "tune_type_llm": "full",
+  "tune_type_vision_tower": "frozen",
+  "tune_vision_tower_from_layer": 0,
+  "use_cache": true,
+  "vision_config": {
+    "hidden_act": "gelu_pytorch_tanh",
+    "hidden_size": 1152,
+    "image_size": 384,
+    "intermediate_size": 4304,
+    "layer_norm_eps": 1e-06,
+    "model_name_or_path": "google/siglip-so400m-patch14-384",
+    "model_name_or_path2": "",
+    "model_type": "siglip_vision_model",
+    "num_attention_heads": 16,
+    "num_hidden_layers": 27,
+    "patch_size": 14
+  },
+  "vision_feature_layer": -2,
+  "vision_feature_select_strategy": "patch",
+  "vision_hidden_size": 1152,
+  "vision_model_name_or_path": "google/siglip-so400m-patch14-384",
+  "vision_model_name_or_path2": "",
+  "vocab_size": 151936
+}

ckpts_oct12/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.3_2e-1_connector-3.0_1.3_2e-1_ablation/mask_applied/generation_config.json

@@ -0,0 +1,6 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 151643,
+  "eos_token_id": 151643,
+  "transformers_version": "4.40.1"
+}

ckpts_oct12/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.3_2e-1_connector-3.0_1.3_2e-1_ablation/mask_applied/masks_connector.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:15075161d71d8e73a1ba291b0489fe548cfae8c6b3f7a2712539f0ba919675b6
+size 3671560

ckpts_oct12/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.3_2e-1_connector-3.0_1.3_2e-1_ablation/mask_applied/masks_lm.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:907bffcaa6e2c2c81791f3df2a9c709b17b13ea5454566eee753c971f770d22c
+size 715707982

ckpts_oct12/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.3_2e-1_connector-3.0_1.3_2e-1_ablation/mask_applied/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:24e5d321084b81c045b6ac67c09b65b5bfae451e30041c6e5c0cd8d9645ac336
+size 1848295352

ckpts_oct12/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.3_2e-1_connector-3.0_1.3_2e-1_ablation/mask_applied/preprocessor_config.json

@@ -0,0 +1,37 @@
+{
+  "_valid_processor_keys": [
+    "images",
+    "do_resize",
+    "size",
+    "resample",
+    "do_rescale",
+    "rescale_factor",
+    "do_normalize",
+    "image_mean",
+    "image_std",
+    "return_tensors",
+    "data_format",
+    "input_data_format"
+  ],
+  "do_normalize": true,
+  "do_rescale": true,
+  "do_resize": true,
+  "image_mean": [
+    0.5,
+    0.5,
+    0.5
+  ],
+  "image_processor_type": "SiglipImageProcessor",
+  "image_std": [
+    0.5,
+    0.5,
+    0.5
+  ],
+  "processor_class": "SiglipProcessor",
+  "resample": 3,
+  "rescale_factor": 0.00392156862745098,
+  "size": {
+    "height": 384,
+    "width": 384
+  }
+}

ckpts_oct12/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.3_2e-1_connector-3.0_1.3_2e-1_ablation/mask_applied/special_tokens_map.json

@@ -0,0 +1,32 @@
+{
+  "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": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "unk_token": "<|endoftext|>"
+}

ckpts_oct12/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.3_2e-1_connector-3.0_1.3_2e-1_ablation/mask_applied/tokenizer_config.json

@@ -0,0 +1,208 @@
+{
+  "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
+    }
+  },
+  "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    {%- else %}\n        {{- 'You are a helpful assistant.' }}\n    {%- endif %}\n    {{- \"\\n\\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    {%- else %}\n        {{- '<|im_start|>system\\nYou are a helpful assistant.<|im_end|>\\n' }}\n    {%- endif %}\n{%- endif %}\n{%- for message in messages %}\n    {%- if (message.role == \"user\") or (message.role == \"system\" and not loop.first) or (message.role == \"assistant\" and not message.tool_calls) %}\n        {{- '<|im_start|>' + message.role + '\\n' + message.content + '<|im_end|>' + '\\n' }}\n    {%- elif message.role == \"assistant\" %}\n        {{- '<|im_start|>' + message.role }}\n        {%- if message.content %}\n            {{- '\\n' + message.content }}\n        {%- endif %}\n        {%- for tool_call in message.tool_calls %}\n            {%- if tool_call.function is defined %}\n                {%- set tool_call = tool_call.function %}\n            {%- endif %}\n            {{- '\\n<tool_call>\\n{\"name\": \"' }}\n            {{- tool_call.name }}\n            {{- '\", \"arguments\": ' }}\n            {{- tool_call.arguments | tojson }}\n            {{- '}\\n</tool_call>' }}\n        {%- endfor %}\n        {{- '<|im_end|>\\n' }}\n    {%- elif message.role == \"tool\" %}\n        {%- if (loop.index0 == 0) 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{%- endif %}\n",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "model_max_length": 2048,
+  "pad_token": "<|endoftext|>",
+  "padding_side": "right",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": "<|endoftext|>"
+}

ckpts_oct12/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.3_2e-1_connector-3.0_1.3_2e-1_ablation/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:46b9a746cdeefd3992ea32db3b4fa3d5179ed6688f7b069d712e59641ce7a229
+size 2567640568

ckpts_oct12/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.3_2e-1_connector-3.0_1.3_2e-1_ablation/runs/Oct13_06-57-53_ywang29-vrdb-test1-worker-0/events.out.tfevents.1760338744.ywang29-vrdb-test1-worker-0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:51bdd4d413e6123eca0168879623b4ee41f17424d547ec80785466670bf4198b
+size 116893

ckpts_oct12/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.3_2e-1_connector-3.0_1.3_2e-1_ablation/special_tokens_map.json

@@ -0,0 +1,32 @@
+{
+  "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": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "unk_token": "<|endoftext|>"
+}

ckpts_oct12/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.3_2e-1_connector-3.0_1.3_2e-1_ablation/tokenizer_config.json

@@ -0,0 +1,208 @@
+{
+  "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
+    }
+  },
+  "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    {%- else %}\n        {{- 'You are a helpful assistant.' }}\n    {%- endif %}\n    {{- \"\\n\\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    {%- else %}\n        {{- '<|im_start|>system\\nYou are a helpful assistant.<|im_end|>\\n' }}\n    {%- endif %}\n{%- endif %}\n{%- for message in messages %}\n    {%- if (message.role == \"user\") or (message.role == \"system\" and not loop.first) or (message.role == \"assistant\" and not message.tool_calls) %}\n        {{- '<|im_start|>' + message.role + '\\n' + message.content + '<|im_end|>' + '\\n' }}\n    {%- elif message.role == \"assistant\" %}\n        {{- '<|im_start|>' + message.role }}\n        {%- if message.content %}\n            {{- '\\n' + message.content }}\n        {%- endif %}\n        {%- for tool_call in message.tool_calls %}\n            {%- if tool_call.function is defined %}\n                {%- set tool_call = tool_call.function %}\n            {%- endif %}\n            {{- '\\n<tool_call>\\n{\"name\": \"' }}\n            {{- tool_call.name }}\n            {{- '\", \"arguments\": ' }}\n            {{- tool_call.arguments | tojson }}\n            {{- '}\\n</tool_call>' }}\n        {%- endfor %}\n        {{- '<|im_end|>\\n' }}\n    {%- elif message.role == \"tool\" %}\n        {%- if (loop.index0 == 0) 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{%- endif %}\n",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "model_max_length": 2048,
+  "pad_token": "<|endoftext|>",
+  "padding_side": "right",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": "<|endoftext|>"
+}

ckpts_oct12/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.3_2e-1_connector-3.0_1.3_2e-1_ablation/trainer_state.json

@@ -0,0 +1,3670 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.0,
+  "eval_steps": 500,
+  "global_step": 520,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0019230769230769232,
+      "grad_norm": 0.03814839847206242,
+      "learning_rate": 0.0125,
+      "loss": 2.1714,
+      "step": 1
+    },
+    {
+      "epoch": 0.0038461538461538464,
+      "grad_norm": 0.04067677085973634,
+      "learning_rate": 0.025,
+      "loss": 2.1502,
+      "step": 2
+    },
+    {
+      "epoch": 0.0057692307692307696,
+      "grad_norm": 0.015939706370972136,
+      "learning_rate": 0.037500000000000006,
+      "loss": 1.7078,
+      "step": 3
+    },
+    {
+      "epoch": 0.007692307692307693,
+      "grad_norm": 0.0068932964270646754,
+      "learning_rate": 0.05,
+      "loss": 1.5946,
+      "step": 4
+    },
+    {
+      "epoch": 0.009615384615384616,
+      "grad_norm": 0.01014886973296801,
+      "learning_rate": 0.0625,
+      "loss": 1.6227,
+      "step": 5
+    },
+    {
+      "epoch": 0.011538461538461539,
+      "grad_norm": 0.005890079859369931,
+      "learning_rate": 0.07500000000000001,
+      "loss": 1.4231,
+      "step": 6
+    },
+    {
+      "epoch": 0.013461538461538462,
+      "grad_norm": 0.005896559349666777,
+      "learning_rate": 0.08750000000000001,
+      "loss": 1.4266,
+      "step": 7
+    },
+    {
+      "epoch": 0.015384615384615385,
+      "grad_norm": 0.005227924105678162,
+      "learning_rate": 0.1,
+      "loss": 1.4832,
+      "step": 8
+    },
+    {
+      "epoch": 0.01730769230769231,
+      "grad_norm": 0.006563212671957713,
+      "learning_rate": 0.1125,
+      "loss": 1.5577,
+      "step": 9
+    },
+    {
+      "epoch": 0.019230769230769232,
+      "grad_norm": 0.004731250654941844,
+      "learning_rate": 0.125,
+      "loss": 1.3749,
+      "step": 10
+    },
+    {
+      "epoch": 0.021153846153846155,
+      "grad_norm": 0.0039738850997779445,
+      "learning_rate": 0.1375,
+      "loss": 1.4475,
+      "step": 11
+    },
+    {
+      "epoch": 0.023076923076923078,
+      "grad_norm": 0.00397221372500587,
+      "learning_rate": 0.15000000000000002,
+      "loss": 1.3878,
+      "step": 12
+    },
+    {
+      "epoch": 0.025,
+      "grad_norm": 0.003365932530092414,
+      "learning_rate": 0.1625,
+      "loss": 1.4063,
+      "step": 13
+    },
+    {
+      "epoch": 0.026923076923076925,
+      "grad_norm": 0.0037988365088866454,
+      "learning_rate": 0.17500000000000002,
+      "loss": 1.4577,
+      "step": 14
+    },
+    {
+      "epoch": 0.028846153846153848,
+      "grad_norm": 0.0032462572180738344,
+      "learning_rate": 0.1875,
+      "loss": 1.4456,
+      "step": 15
+    },
+    {
+      "epoch": 0.03076923076923077,
+      "grad_norm": 0.0032287861085325514,
+      "learning_rate": 0.2,
+      "loss": 1.414,
+      "step": 16
+    },
+    {
+      "epoch": 0.032692307692307694,
+      "grad_norm": 0.0030201015118433458,
+      "learning_rate": 0.1999980572931538,
+      "loss": 1.5109,
+      "step": 17
+    },
+    {
+      "epoch": 0.03461538461538462,
+      "grad_norm": 0.0032510411334095665,
+      "learning_rate": 0.19999222924809748,
+      "loss": 1.385,
+      "step": 18
+    },
+    {
+      "epoch": 0.03653846153846154,
+      "grad_norm": 0.0031235910962774493,
+      "learning_rate": 0.19998251609127465,
+      "loss": 1.4241,
+      "step": 19
+    },
+    {
+      "epoch": 0.038461538461538464,
+      "grad_norm": 0.0033510654340645897,
+      "learning_rate": 0.19996891820008164,
+      "loss": 1.3793,
+      "step": 20
+    },
+    {
+      "epoch": 0.04038461538461539,
+      "grad_norm": 0.0031372161547092298,
+      "learning_rate": 0.19995143610285276,
+      "loss": 1.4093,
+      "step": 21
+    },
+    {
+      "epoch": 0.04230769230769231,
+      "grad_norm": 0.003248863936960784,
+      "learning_rate": 0.19993007047883987,
+      "loss": 1.5217,
+      "step": 22
+    },
+    {
+      "epoch": 0.04423076923076923,
+      "grad_norm": 0.002744630830615553,
+      "learning_rate": 0.1999048221581858,
+      "loss": 1.4566,
+      "step": 23
+    },
+    {
+      "epoch": 0.046153846153846156,
+      "grad_norm": 0.002813657376694583,
+      "learning_rate": 0.19987569212189224,
+      "loss": 1.4108,
+      "step": 24
+    },
+    {
+      "epoch": 0.04807692307692308,
+      "grad_norm": 0.003055599584845224,
+      "learning_rate": 0.19984268150178167,
+      "loss": 1.4567,
+      "step": 25
+    },
+    {
+      "epoch": 0.05,
+      "grad_norm": 0.0025067944188097184,
+      "learning_rate": 0.1998057915804532,
+      "loss": 1.4517,
+      "step": 26
+    },
+    {
+      "epoch": 0.051923076923076926,
+      "grad_norm": 0.0030723577887423565,
+      "learning_rate": 0.1997650237912329,
+      "loss": 1.3732,
+      "step": 27
+    },
+    {
+      "epoch": 0.05384615384615385,
+      "grad_norm": 0.0029055764762655916,
+      "learning_rate": 0.199720379718118,
+      "loss": 1.3779,
+      "step": 28
+    },
+    {
+      "epoch": 0.05576923076923077,
+      "grad_norm": 0.0028010695705970156,
+      "learning_rate": 0.19967186109571552,
+      "loss": 1.3927,
+      "step": 29
+    },
+    {
+      "epoch": 0.057692307692307696,
+      "grad_norm": 0.0028552235732370373,
+      "learning_rate": 0.19961946980917455,
+      "loss": 1.5006,
+      "step": 30
+    },
+    {
+      "epoch": 0.05961538461538462,
+      "grad_norm": 0.0025425699819593946,
+      "learning_rate": 0.1995632078941134,
+      "loss": 1.3721,
+      "step": 31
+    },
+    {
+      "epoch": 0.06153846153846154,
+      "grad_norm": 0.003301888676646615,
+      "learning_rate": 0.19950307753654017,
+      "loss": 1.3718,
+      "step": 32
+    },
+    {
+      "epoch": 0.06346153846153846,
+      "grad_norm": 0.003187700562540326,
+      "learning_rate": 0.19943908107276798,
+      "loss": 1.3956,
+      "step": 33
+    },
+    {
+      "epoch": 0.06538461538461539,
+      "grad_norm": 0.0030113010635178165,
+      "learning_rate": 0.19937122098932428,
+      "loss": 1.3757,
+      "step": 34
+    },
+    {
+      "epoch": 0.0673076923076923,
+      "grad_norm": 0.002800497869582189,
+      "learning_rate": 0.19929949992285395,
+      "loss": 1.384,
+      "step": 35
+    },
+    {
+      "epoch": 0.06923076923076923,
+      "grad_norm": 0.0027821416973315334,
+      "learning_rate": 0.19922392066001723,
+      "loss": 1.504,
+      "step": 36
+    },
+    {
+      "epoch": 0.07115384615384615,
+      "grad_norm": 0.002475955016841264,
+      "learning_rate": 0.19914448613738106,
+      "loss": 1.4831,
+      "step": 37
+    },
+    {
+      "epoch": 0.07307692307692308,
+      "grad_norm": 0.00271507020797863,
+      "learning_rate": 0.1990611994413053,
+      "loss": 1.5787,
+      "step": 38
+    },
+    {
+      "epoch": 0.075,
+      "grad_norm": 0.0030041993464000214,
+      "learning_rate": 0.19897406380782262,
+      "loss": 1.4151,
+      "step": 39
+    },
+    {
+      "epoch": 0.07692307692307693,
+      "grad_norm": 0.0024019148948412187,
+      "learning_rate": 0.19888308262251286,
+      "loss": 1.4538,
+      "step": 40
+    },
+    {
+      "epoch": 0.07884615384615384,
+      "grad_norm": 0.0028430463589571154,
+      "learning_rate": 0.19878825942037148,
+      "loss": 1.4223,
+      "step": 41
+    },
+    {
+      "epoch": 0.08076923076923077,
+      "grad_norm": 0.0033940644869797274,
+      "learning_rate": 0.19868959788567211,
+      "loss": 1.4476,
+      "step": 42
+    },
+    {
+      "epoch": 0.08269230769230769,
+      "grad_norm": 0.0035925563157347016,
+      "learning_rate": 0.1985871018518236,
+      "loss": 1.4016,
+      "step": 43
+    },
+    {
+      "epoch": 0.08461538461538462,
+      "grad_norm": 0.0026264382574149206,
+      "learning_rate": 0.19848077530122082,
+      "loss": 1.4914,
+      "step": 44
+    },
+    {
+      "epoch": 0.08653846153846154,
+      "grad_norm": 0.002683211319764426,
+      "learning_rate": 0.19837062236509015,
+      "loss": 1.4591,
+      "step": 45
+    },
+    {
+      "epoch": 0.08846153846153847,
+      "grad_norm": 0.0027241097926793143,
+      "learning_rate": 0.19825664732332884,
+      "loss": 1.565,
+      "step": 46
+    },
+    {
+      "epoch": 0.09038461538461538,
+      "grad_norm": 0.0025015885738279937,
+      "learning_rate": 0.19813885460433878,
+      "loss": 1.4428,
+      "step": 47
+    },
+    {
+      "epoch": 0.09230769230769231,
+      "grad_norm": 0.0029370966734930046,
+      "learning_rate": 0.19801724878485438,
+      "loss": 1.4093,
+      "step": 48
+    },
+    {
+      "epoch": 0.09423076923076923,
+      "grad_norm": 0.0025682075162119437,
+      "learning_rate": 0.19789183458976486,
+      "loss": 1.4613,
+      "step": 49
+    },
+    {
+      "epoch": 0.09615384615384616,
+      "grad_norm": 0.0024960752555607,
+      "learning_rate": 0.19776261689193048,
+      "loss": 1.4502,
+      "step": 50
+    },
+    {
+      "epoch": 0.09807692307692308,
+      "grad_norm": 0.002816854564837006,
+      "learning_rate": 0.19762960071199334,
+      "loss": 1.3748,
+      "step": 51
+    },
+    {
+      "epoch": 0.1,
+      "grad_norm": 0.0026101923107874625,
+      "learning_rate": 0.19749279121818236,
+      "loss": 1.507,
+      "step": 52
+    },
+    {
+      "epoch": 0.10192307692307692,
+      "grad_norm": 0.0027646598510270006,
+      "learning_rate": 0.19735219372611235,
+      "loss": 1.5028,
+      "step": 53
+    },
+    {
+      "epoch": 0.10384615384615385,
+      "grad_norm": 0.002375376838643182,
+      "learning_rate": 0.19720781369857746,
+      "loss": 1.3989,
+      "step": 54
+    },
+    {
+      "epoch": 0.10576923076923077,
+      "grad_norm": 0.002537493145379656,
+      "learning_rate": 0.1970596567453391,
+      "loss": 1.3871,
+      "step": 55
+    },
+    {
+      "epoch": 0.1076923076923077,
+      "grad_norm": 0.0025207634856254197,
+      "learning_rate": 0.1969077286229078,
+      "loss": 1.5141,
+      "step": 56
+    },
+    {
+      "epoch": 0.10961538461538461,
+      "grad_norm": 0.0031753337474815365,
+      "learning_rate": 0.19675203523431964,
+      "loss": 1.3719,
+      "step": 57
+    },
+    {
+      "epoch": 0.11153846153846154,
+      "grad_norm": 0.0019763232451755586,
+      "learning_rate": 0.19659258262890683,
+      "loss": 1.5335,
+      "step": 58
+    },
+    {
+      "epoch": 0.11346153846153846,
+      "grad_norm": 0.0025586344100365236,
+      "learning_rate": 0.19642937700206278,
+      "loss": 1.3657,
+      "step": 59
+    },
+    {
+      "epoch": 0.11538461538461539,
+      "grad_norm": 0.003219037365566865,
+      "learning_rate": 0.19626242469500121,
+      "loss": 1.4619,
+      "step": 60
+    },
+    {
+      "epoch": 0.11730769230769231,
+      "grad_norm": 0.0034946622258303865,
+      "learning_rate": 0.19609173219450998,
+      "loss": 1.4912,
+      "step": 61
+    },
+    {
+      "epoch": 0.11923076923076924,
+      "grad_norm": 0.0026305275318732127,
+      "learning_rate": 0.19591730613269878,
+      "loss": 1.4328,
+      "step": 62
+    },
+    {
+      "epoch": 0.12115384615384615,
+      "grad_norm": 0.002856674753413912,
+      "learning_rate": 0.19573915328674182,
+      "loss": 1.4107,
+      "step": 63
+    },
+    {
+      "epoch": 0.12307692307692308,
+      "grad_norm": 0.0024829672287541645,
+      "learning_rate": 0.1955572805786141,
+      "loss": 1.464,
+      "step": 64
+    },
+    {
+      "epoch": 0.125,
+      "grad_norm": 0.0025565192596521207,
+      "learning_rate": 0.1953716950748227,
+      "loss": 1.4494,
+      "step": 65
+    },
+    {
+      "epoch": 0.12692307692307692,
+      "grad_norm": 0.003001393583908569,
+      "learning_rate": 0.19518240398613226,
+      "loss": 1.417,
+      "step": 66
+    },
+    {
+      "epoch": 0.12884615384615383,
+      "grad_norm": 0.0021485996931497296,
+      "learning_rate": 0.1949894146672846,
+      "loss": 1.3051,
+      "step": 67
+    },
+    {
+      "epoch": 0.13076923076923078,
+      "grad_norm": 0.0022228401551197285,
+      "learning_rate": 0.1947927346167132,
+      "loss": 1.3659,
+      "step": 68
+    },
+    {
+      "epoch": 0.1326923076923077,
+      "grad_norm": 0.0030842978039204473,
+      "learning_rate": 0.1945923714762516,
+      "loss": 1.3425,
+      "step": 69
+    },
+    {
+      "epoch": 0.1346153846153846,
+      "grad_norm": 0.0026180165970355333,
+      "learning_rate": 0.19438833303083677,
+      "loss": 1.3919,
+      "step": 70
+    },
+    {
+      "epoch": 0.13653846153846153,
+      "grad_norm": 0.0022132994870580807,
+      "learning_rate": 0.19418062720820636,
+      "loss": 1.3125,
+      "step": 71
+    },
+    {
+      "epoch": 0.13846153846153847,
+      "grad_norm": 0.002712484977588332,
+      "learning_rate": 0.19396926207859086,
+      "loss": 1.4632,
+      "step": 72
+    },
+    {
+      "epoch": 0.14038461538461539,
+      "grad_norm": 0.0022787297804940662,
+      "learning_rate": 0.19375424585439993,
+      "loss": 1.2917,
+      "step": 73
+    },
+    {
+      "epoch": 0.1423076923076923,
+      "grad_norm": 0.0024647144646780545,
+      "learning_rate": 0.1935355868899034,
+      "loss": 1.4093,
+      "step": 74
+    },
+    {
+      "epoch": 0.14423076923076922,
+      "grad_norm": 0.002441223784435101,
+      "learning_rate": 0.19331329368090666,
+      "loss": 1.3163,
+      "step": 75
+    },
+    {
+      "epoch": 0.14615384615384616,
+      "grad_norm": 0.003294189930559694,
+      "learning_rate": 0.19308737486442043,
+      "loss": 1.539,
+      "step": 76
+    },
+    {
+      "epoch": 0.14807692307692308,
+      "grad_norm": 0.0024181096877143,
+      "learning_rate": 0.19285783921832536,
+      "loss": 1.232,
+      "step": 77
+    },
+    {
+      "epoch": 0.15,
+      "grad_norm": 0.002595047787449542,
+      "learning_rate": 0.19262469566103088,
+      "loss": 1.3677,
+      "step": 78
+    },
+    {
+      "epoch": 0.1519230769230769,
+      "grad_norm": 0.0025088093314904842,
+      "learning_rate": 0.19238795325112867,
+      "loss": 1.3477,
+      "step": 79
+    },
+    {
+      "epoch": 0.15384615384615385,
+      "grad_norm": 0.0027886773804302817,
+      "learning_rate": 0.19214762118704076,
+      "loss": 1.5211,
+      "step": 80
+    },
+    {
+      "epoch": 0.15576923076923077,
+      "grad_norm": 0.003481034335154321,
+      "learning_rate": 0.19190370880666208,
+      "loss": 1.5011,
+      "step": 81
+    },
+    {
+      "epoch": 0.1576923076923077,
+      "grad_norm": 0.0023810124948324676,
+      "learning_rate": 0.19165622558699763,
+      "loss": 1.4348,
+      "step": 82
+    },
+    {
+      "epoch": 0.1596153846153846,
+      "grad_norm": 0.0023352234015946924,
+      "learning_rate": 0.19140518114379435,
+      "loss": 1.4377,
+      "step": 83
+    },
+    {
+      "epoch": 0.16153846153846155,
+      "grad_norm": 0.002856987942573144,
+      "learning_rate": 0.19115058523116735,
+      "loss": 1.4623,
+      "step": 84
+    },
+    {
+      "epoch": 0.16346153846153846,
+      "grad_norm": 0.002282735502342565,
+      "learning_rate": 0.1908924477412211,
+      "loss": 1.4694,
+      "step": 85
+    },
+    {
+      "epoch": 0.16538461538461538,
+      "grad_norm": 0.002480337700696306,
+      "learning_rate": 0.19063077870366502,
+      "loss": 1.4893,
+      "step": 86
+    },
+    {
+      "epoch": 0.1673076923076923,
+      "grad_norm": 0.002258738262257446,
+      "learning_rate": 0.1903655882854237,
+      "loss": 1.4641,
+      "step": 87
+    },
+    {
+      "epoch": 0.16923076923076924,
+      "grad_norm": 0.0024309958331379324,
+      "learning_rate": 0.19009688679024192,
+      "loss": 1.4303,
+      "step": 88
+    },
+    {
+      "epoch": 0.17115384615384616,
+      "grad_norm": 0.0025354156691821833,
+      "learning_rate": 0.18982468465828442,
+      "loss": 1.4392,
+      "step": 89
+    },
+    {
+      "epoch": 0.17307692307692307,
+      "grad_norm": 0.0024793183045657414,
+      "learning_rate": 0.1895489924657301,
+      "loss": 1.3739,
+      "step": 90
+    },
+    {
+      "epoch": 0.175,
+      "grad_norm": 0.0022106734516138993,
+      "learning_rate": 0.18926982092436118,
+      "loss": 1.4408,
+      "step": 91
+    },
+    {
+      "epoch": 0.17692307692307693,
+      "grad_norm": 0.0022928097431455263,
+      "learning_rate": 0.18898718088114688,
+      "loss": 1.3771,
+      "step": 92
+    },
+    {
+      "epoch": 0.17884615384615385,
+      "grad_norm": 0.0023543048478954797,
+      "learning_rate": 0.18870108331782218,
+      "loss": 1.3894,
+      "step": 93
+    },
+    {
+      "epoch": 0.18076923076923077,
+      "grad_norm": 0.0028397974342021137,
+      "learning_rate": 0.18841153935046098,
+      "loss": 1.4816,
+      "step": 94
+    },
+    {
+      "epoch": 0.18269230769230768,
+      "grad_norm": 0.0029613049041681333,
+      "learning_rate": 0.18811856022904425,
+      "loss": 1.3679,
+      "step": 95
+    },
+    {
+      "epoch": 0.18461538461538463,
+      "grad_norm": 0.002775261714490049,
+      "learning_rate": 0.18782215733702287,
+      "loss": 1.382,
+      "step": 96
+    },
+    {
+      "epoch": 0.18653846153846154,
+      "grad_norm": 0.0029672950526583083,
+      "learning_rate": 0.18752234219087538,
+      "loss": 1.3583,
+      "step": 97
+    },
+    {
+      "epoch": 0.18846153846153846,
+      "grad_norm": 0.002085699902403141,
+      "learning_rate": 0.18721912643966054,
+      "loss": 1.345,
+      "step": 98
+    },
+    {
+      "epoch": 0.19038461538461537,
+      "grad_norm": 0.002569788801573435,
+      "learning_rate": 0.18691252186456464,
+      "loss": 1.371,
+      "step": 99
+    },
+    {
+      "epoch": 0.19230769230769232,
+      "grad_norm": 0.0026102749406244274,
+      "learning_rate": 0.1866025403784439,
+      "loss": 1.4402,
+      "step": 100
+    },
+    {
+      "epoch": 0.19423076923076923,
+      "grad_norm": 0.0025066439097236343,
+      "learning_rate": 0.18628919402536132,
+      "loss": 1.3741,
+      "step": 101
+    },
+    {
+      "epoch": 0.19615384615384615,
+      "grad_norm": 0.002431404843765263,
+      "learning_rate": 0.18597249498011903,
+      "loss": 1.3837,
+      "step": 102
+    },
+    {
+      "epoch": 0.19807692307692307,
+      "grad_norm": 0.0019086657247981253,
+      "learning_rate": 0.18565245554778517,
+      "loss": 1.3005,
+      "step": 103
+    },
+    {
+      "epoch": 0.2,
+      "grad_norm": 0.0023146895051099382,
+      "learning_rate": 0.18532908816321558,
+      "loss": 1.3798,
+      "step": 104
+    },
+    {
+      "epoch": 0.20192307692307693,
+      "grad_norm": 0.0022659614716657533,
+      "learning_rate": 0.18500240539057092,
+      "loss": 1.3722,
+      "step": 105
+    },
+    {
+      "epoch": 0.20384615384615384,
+      "grad_norm": 0.0026261946424078323,
+      "learning_rate": 0.18467241992282843,
+      "loss": 1.4227,
+      "step": 106
+    },
+    {
+      "epoch": 0.20576923076923076,
+      "grad_norm": 0.0022464927129622536,
+      "learning_rate": 0.18433914458128858,
+      "loss": 1.3884,
+      "step": 107
+    },
+    {
+      "epoch": 0.2076923076923077,
+      "grad_norm": 0.002358669624623717,
+      "learning_rate": 0.18400259231507718,
+      "loss": 1.3307,
+      "step": 108
+    },
+    {
+      "epoch": 0.20961538461538462,
+      "grad_norm": 0.002010251831957762,
+      "learning_rate": 0.18366277620064198,
+      "loss": 1.3806,
+      "step": 109
+    },
+    {
+      "epoch": 0.21153846153846154,
+      "grad_norm": 0.002164369898530802,
+      "learning_rate": 0.1833197094412449,
+      "loss": 1.524,
+      "step": 110
+    },
+    {
+      "epoch": 0.21346153846153845,
+      "grad_norm": 0.0025084756295901717,
+      "learning_rate": 0.18297340536644877,
+      "loss": 1.524,
+      "step": 111
+    },
+    {
+      "epoch": 0.2153846153846154,
+      "grad_norm": 0.0021140771605330554,
+      "learning_rate": 0.1826238774315995,
+      "loss": 1.4094,
+      "step": 112
+    },
+    {
+      "epoch": 0.2173076923076923,
+      "grad_norm": 0.0019059483677735533,
+      "learning_rate": 0.18227113921730334,
+      "loss": 1.278,
+      "step": 113
+    },
+    {
+      "epoch": 0.21923076923076923,
+      "grad_norm": 0.0020124683438425403,
+      "learning_rate": 0.1819152044288992,
+      "loss": 1.3771,
+      "step": 114
+    },
+    {
+      "epoch": 0.22115384615384615,
+      "grad_norm": 0.0020371296271276458,
+      "learning_rate": 0.18155608689592603,
+      "loss": 1.4975,
+      "step": 115
+    },
+    {
+      "epoch": 0.2230769230769231,
+      "grad_norm": 0.0018481811307831423,
+      "learning_rate": 0.18119380057158568,
+      "loss": 1.4942,
+      "step": 116
+    },
+    {
+      "epoch": 0.225,
+      "grad_norm": 0.0022826331788792197,
+      "learning_rate": 0.18082835953220056,
+      "loss": 1.4711,
+      "step": 117
+    },
+    {
+      "epoch": 0.22692307692307692,
+      "grad_norm": 0.0018853795073498472,
+      "learning_rate": 0.18045977797666685,
+      "loss": 1.3523,
+      "step": 118
+    },
+    {
+      "epoch": 0.22884615384615384,
+      "grad_norm": 0.001965799662573032,
+      "learning_rate": 0.1800880702259028,
+      "loss": 1.3103,
+      "step": 119
+    },
+    {
+      "epoch": 0.23076923076923078,
+      "grad_norm": 0.002469289656337362,
+      "learning_rate": 0.17971325072229227,
+      "loss": 1.3325,
+      "step": 120
+    },
+    {
+      "epoch": 0.2326923076923077,
+      "grad_norm": 0.0025749805182455248,
+      "learning_rate": 0.17933533402912352,
+      "loss": 1.3913,
+      "step": 121
+    },
+    {
+      "epoch": 0.23461538461538461,
+      "grad_norm": 0.0020427815705227545,
+      "learning_rate": 0.17895433483002354,
+      "loss": 1.2874,
+      "step": 122
+    },
+    {
+      "epoch": 0.23653846153846153,
+      "grad_norm": 0.0022426018746006748,
+      "learning_rate": 0.17857026792838737,
+      "loss": 1.4567,
+      "step": 123
+    },
+    {
+      "epoch": 0.23846153846153847,
+      "grad_norm": 0.0022167591861620988,
+      "learning_rate": 0.178183148246803,
+      "loss": 1.3661,
+      "step": 124
+    },
+    {
+      "epoch": 0.2403846153846154,
+      "grad_norm": 0.0021485964089188766,
+      "learning_rate": 0.1777929908264715,
+      "loss": 1.3574,
+      "step": 125
+    },
+    {
+      "epoch": 0.2423076923076923,
+      "grad_norm": 0.001724717899626798,
+      "learning_rate": 0.17739981082662276,
+      "loss": 1.3687,
+      "step": 126
+    },
+    {
+      "epoch": 0.24423076923076922,
+      "grad_norm": 0.0026441639482556113,
+      "learning_rate": 0.1770036235239263,
+      "loss": 1.3381,
+      "step": 127
+    },
+    {
+      "epoch": 0.24615384615384617,
+      "grad_norm": 0.002076596671449281,
+      "learning_rate": 0.1766044443118978,
+      "loss": 1.389,
+      "step": 128
+    },
+    {
+      "epoch": 0.24807692307692308,
+      "grad_norm": 0.0017424642339292301,
+      "learning_rate": 0.17620228870030108,
+      "loss": 1.3174,
+      "step": 129
+    },
+    {
+      "epoch": 0.25,
+      "grad_norm": 0.001896489732855636,
+      "learning_rate": 0.1757971723145453,
+      "loss": 1.3717,
+      "step": 130
+    },
+    {
+      "epoch": 0.2519230769230769,
+      "grad_norm": 0.0021138815852579807,
+      "learning_rate": 0.175389110895078,
+      "loss": 1.3565,
+      "step": 131
+    },
+    {
+      "epoch": 0.25384615384615383,
+      "grad_norm": 0.0020449464189773933,
+      "learning_rate": 0.17497812029677343,
+      "loss": 1.4208,
+      "step": 132
+    },
+    {
+      "epoch": 0.25576923076923075,
+      "grad_norm": 0.0020477907697934606,
+      "learning_rate": 0.17456421648831655,
+      "loss": 1.3198,
+      "step": 133
+    },
+    {
+      "epoch": 0.25769230769230766,
+      "grad_norm": 0.002232787644982114,
+      "learning_rate": 0.17414741555158267,
+      "loss": 1.4099,
+      "step": 134
+    },
+    {
+      "epoch": 0.25961538461538464,
+      "grad_norm": 0.0020388918272079234,
+      "learning_rate": 0.1737277336810124,
+      "loss": 1.4743,
+      "step": 135
+    },
+    {
+      "epoch": 0.26153846153846155,
+      "grad_norm": 0.00207227140209728,
+      "learning_rate": 0.17330518718298263,
+      "loss": 1.3888,
+      "step": 136
+    },
+    {
+      "epoch": 0.26346153846153847,
+      "grad_norm": 0.0023090476064216037,
+      "learning_rate": 0.17287979247517285,
+      "loss": 1.3208,
+      "step": 137
+    },
+    {
+      "epoch": 0.2653846153846154,
+      "grad_norm": 0.0019278403960009776,
+      "learning_rate": 0.17245156608592727,
+      "loss": 1.3306,
+      "step": 138
+    },
+    {
+      "epoch": 0.2673076923076923,
+      "grad_norm": 0.0020481385125986516,
+      "learning_rate": 0.17202052465361267,
+      "loss": 1.2435,
+      "step": 139
+    },
+    {
+      "epoch": 0.2692307692307692,
+      "grad_norm": 0.0018946566368485598,
+      "learning_rate": 0.17158668492597184,
+      "loss": 1.3899,
+      "step": 140
+    },
+    {
+      "epoch": 0.27115384615384613,
+      "grad_norm": 0.0019277182918459703,
+      "learning_rate": 0.17115006375947303,
+      "loss": 1.4409,
+      "step": 141
+    },
+    {
+      "epoch": 0.27307692307692305,
+      "grad_norm": 0.0019017884142710493,
+      "learning_rate": 0.17071067811865476,
+      "loss": 1.4166,
+      "step": 142
+    },
+    {
+      "epoch": 0.275,
+      "grad_norm": 0.002120985443841013,
+      "learning_rate": 0.17026854507546693,
+      "loss": 1.3569,
+      "step": 143
+    },
+    {
+      "epoch": 0.27692307692307694,
+      "grad_norm": 0.001956971219571965,
+      "learning_rate": 0.1698236818086073,
+      "loss": 1.3051,
+      "step": 144
+    },
+    {
+      "epoch": 0.27884615384615385,
+      "grad_norm": 0.0017593598479267534,
+      "learning_rate": 0.16937610560285418,
+      "loss": 1.2509,
+      "step": 145
+    },
+    {
+      "epoch": 0.28076923076923077,
+      "grad_norm": 0.0019982131772829084,
+      "learning_rate": 0.1689258338483947,
+      "loss": 1.4789,
+      "step": 146
+    },
+    {
+      "epoch": 0.2826923076923077,
+      "grad_norm": 0.001966958226699274,
+      "learning_rate": 0.16847288404014937,
+      "loss": 1.294,
+      "step": 147
+    },
+    {
+      "epoch": 0.2846153846153846,
+      "grad_norm": 0.0019863794039199347,
+      "learning_rate": 0.16801727377709194,
+      "loss": 1.3299,
+      "step": 148
+    },
+    {
+      "epoch": 0.2865384615384615,
+      "grad_norm": 0.0019232143932359598,
+      "learning_rate": 0.16755902076156604,
+      "loss": 1.2694,
+      "step": 149
+    },
+    {
+      "epoch": 0.28846153846153844,
+      "grad_norm": 0.0020896835779997077,
+      "learning_rate": 0.16709814279859703,
+      "loss": 1.5086,
+      "step": 150
+    },
+    {
+      "epoch": 0.2903846153846154,
+      "grad_norm": 0.00188911127554728,
+      "learning_rate": 0.1666346577952004,
+      "loss": 1.3159,
+      "step": 151
+    },
+    {
+      "epoch": 0.2923076923076923,
+      "grad_norm": 0.001988614395997743,
+      "learning_rate": 0.16616858375968596,
+      "loss": 1.2821,
+      "step": 152
+    },
+    {
+      "epoch": 0.29423076923076924,
+      "grad_norm": 0.0020568405804363204,
+      "learning_rate": 0.16569993880095807,
+      "loss": 1.315,
+      "step": 153
+    },
+    {
+      "epoch": 0.29615384615384616,
+      "grad_norm": 0.0019417670188790713,
+      "learning_rate": 0.16522874112781213,
+      "loss": 1.4143,
+      "step": 154
+    },
+    {
+      "epoch": 0.2980769230769231,
+      "grad_norm": 0.0019316039996410928,
+      "learning_rate": 0.16475500904822704,
+      "loss": 1.3135,
+      "step": 155
+    },
+    {
+      "epoch": 0.3,
+      "grad_norm": 0.0020493230052184472,
+      "learning_rate": 0.16427876096865393,
+      "loss": 1.3389,
+      "step": 156
+    },
+    {
+      "epoch": 0.3019230769230769,
+      "grad_norm": 0.001997364374050798,
+      "learning_rate": 0.16380001539330089,
+      "loss": 1.4571,
+      "step": 157
+    },
+    {
+      "epoch": 0.3038461538461538,
+      "grad_norm": 0.002098158230862549,
+      "learning_rate": 0.163318790923414,
+      "loss": 1.3209,
+      "step": 158
+    },
+    {
+      "epoch": 0.3057692307692308,
+      "grad_norm": 0.0018761964068357178,
+      "learning_rate": 0.16283510625655473,
+      "loss": 1.3535,
+      "step": 159
+    },
+    {
+      "epoch": 0.3076923076923077,
+      "grad_norm": 0.0019642159677749076,
+      "learning_rate": 0.16234898018587338,
+      "loss": 1.3791,
+      "step": 160
+    },
+    {
+      "epoch": 0.3096153846153846,
+      "grad_norm": 0.001971229671621249,
+      "learning_rate": 0.16186043159937882,
+      "loss": 1.3607,
+      "step": 161
+    },
+    {
+      "epoch": 0.31153846153846154,
+      "grad_norm": 0.0020482509280399072,
+      "learning_rate": 0.16136947947920477,
+      "loss": 1.396,
+      "step": 162
+    },
+    {
+      "epoch": 0.31346153846153846,
+      "grad_norm": 0.002344137792459382,
+      "learning_rate": 0.16087614290087207,
+      "loss": 1.2363,
+      "step": 163
+    },
+    {
+      "epoch": 0.3153846153846154,
+      "grad_norm": 0.0018503862107493794,
+      "learning_rate": 0.16038044103254775,
+      "loss": 1.1982,
+      "step": 164
+    },
+    {
+      "epoch": 0.3173076923076923,
+      "grad_norm": 0.0017700001496928869,
+      "learning_rate": 0.15988239313430005,
+      "loss": 1.3454,
+      "step": 165
+    },
+    {
+      "epoch": 0.3192307692307692,
+      "grad_norm": 0.0019965981060615237,
+      "learning_rate": 0.15938201855735015,
+      "loss": 1.3406,
+      "step": 166
+    },
+    {
+      "epoch": 0.3211538461538462,
+      "grad_norm": 0.0019608943711386544,
+      "learning_rate": 0.15887933674332047,
+      "loss": 1.3374,
+      "step": 167
+    },
+    {
+      "epoch": 0.3230769230769231,
+      "grad_norm": 0.0018613130130918617,
+      "learning_rate": 0.158374367223479,
+      "loss": 1.2559,
+      "step": 168
+    },
+    {
+      "epoch": 0.325,
+      "grad_norm": 0.0018711989681940548,
+      "learning_rate": 0.1578671296179806,
+      "loss": 1.3507,
+      "step": 169
+    },
+    {
+      "epoch": 0.3269230769230769,
+      "grad_norm": 0.0017450569759756994,
+      "learning_rate": 0.15735764363510463,
+      "loss": 1.3244,
+      "step": 170
+    },
+    {
+      "epoch": 0.32884615384615384,
+      "grad_norm": 0.0020556058413007355,
+      "learning_rate": 0.15684592907048925,
+      "loss": 1.2801,
+      "step": 171
+    },
+    {
+      "epoch": 0.33076923076923076,
+      "grad_norm": 0.0019475944762560476,
+      "learning_rate": 0.1563320058063622,
+      "loss": 1.3479,
+      "step": 172
+    },
+    {
+      "epoch": 0.3326923076923077,
+      "grad_norm": 0.0024380131503897837,
+      "learning_rate": 0.15581589381076844,
+      "loss": 1.2875,
+      "step": 173
+    },
+    {
+      "epoch": 0.3346153846153846,
+      "grad_norm": 0.002096823845449059,
+      "learning_rate": 0.15529761313679394,
+      "loss": 1.3502,
+      "step": 174
+    },
+    {
+      "epoch": 0.33653846153846156,
+      "grad_norm": 0.001770358270063253,
+      "learning_rate": 0.15477718392178716,
+      "loss": 1.2519,
+      "step": 175
+    },
+    {
+      "epoch": 0.3384615384615385,
+      "grad_norm": 0.0018688420850506665,
+      "learning_rate": 0.15425462638657594,
+      "loss": 1.4133,
+      "step": 176
+    },
+    {
+      "epoch": 0.3403846153846154,
+      "grad_norm": 0.002075044927116544,
+      "learning_rate": 0.1537299608346824,
+      "loss": 1.2855,
+      "step": 177
+    },
+    {
+      "epoch": 0.3423076923076923,
+      "grad_norm": 0.0020736520416219976,
+      "learning_rate": 0.15320320765153367,
+      "loss": 1.324,
+      "step": 178
+    },
+    {
+      "epoch": 0.34423076923076923,
+      "grad_norm": 0.00179561484858704,
+      "learning_rate": 0.15267438730367008,
+      "loss": 1.41,
+      "step": 179
+    },
+    {
+      "epoch": 0.34615384615384615,
+      "grad_norm": 0.0020270382624602134,
+      "learning_rate": 0.1521435203379498,
+      "loss": 1.328,
+      "step": 180
+    },
+    {
+      "epoch": 0.34807692307692306,
+      "grad_norm": 0.0016492934799124752,
+      "learning_rate": 0.15161062738075068,
+      "loss": 1.2951,
+      "step": 181
+    },
+    {
+      "epoch": 0.35,
+      "grad_norm": 0.0018657480143007136,
+      "learning_rate": 0.1510757291371686,
+      "loss": 1.3038,
+      "step": 182
+    },
+    {
+      "epoch": 0.35192307692307695,
+      "grad_norm": 0.001850508700804724,
+      "learning_rate": 0.1505388463902131,
+      "loss": 1.3436,
+      "step": 183
+    },
+    {
+      "epoch": 0.35384615384615387,
+      "grad_norm": 0.001873209314519758,
+      "learning_rate": 0.15000000000000002,
+      "loss": 1.25,
+      "step": 184
+    },
+    {
+      "epoch": 0.3557692307692308,
+      "grad_norm": 0.0018402385735532016,
+      "learning_rate": 0.14945921090294076,
+      "loss": 1.4249,
+      "step": 185
+    },
+    {
+      "epoch": 0.3576923076923077,
+      "grad_norm": 0.0018025564834811538,
+      "learning_rate": 0.14891650011092894,
+      "loss": 1.2739,
+      "step": 186
+    },
+    {
+      "epoch": 0.3596153846153846,
+      "grad_norm": 0.0021184254140723573,
+      "learning_rate": 0.14837188871052398,
+      "loss": 1.2902,
+      "step": 187
+    },
+    {
+      "epoch": 0.36153846153846153,
+      "grad_norm": 0.0019266626675808075,
+      "learning_rate": 0.14782539786213184,
+      "loss": 1.3595,
+      "step": 188
+    },
+    {
+      "epoch": 0.36346153846153845,
+      "grad_norm": 0.0016985817669600309,
+      "learning_rate": 0.1472770487991827,
+      "loss": 1.372,
+      "step": 189
+    },
+    {
+      "epoch": 0.36538461538461536,
+      "grad_norm": 0.002093297423013253,
+      "learning_rate": 0.1467268628273062,
+      "loss": 1.2915,
+      "step": 190
+    },
+    {
+      "epoch": 0.36730769230769234,
+      "grad_norm": 0.0016841672241996834,
+      "learning_rate": 0.1461748613235034,
+      "loss": 1.2395,
+      "step": 191
+    },
+    {
+      "epoch": 0.36923076923076925,
+      "grad_norm": 0.0017978746011286784,
+      "learning_rate": 0.1456210657353163,
+      "loss": 1.336,
+      "step": 192
+    },
+    {
+      "epoch": 0.37115384615384617,
+      "grad_norm": 0.0021625504296969565,
+      "learning_rate": 0.14506549757999454,
+      "loss": 1.3379,
+      "step": 193
+    },
+    {
+      "epoch": 0.3730769230769231,
+      "grad_norm": 0.001687739983569612,
+      "learning_rate": 0.14450817844365924,
+      "loss": 1.2137,
+      "step": 194
+    },
+    {
+      "epoch": 0.375,
+      "grad_norm": 0.0017726156713259885,
+      "learning_rate": 0.1439491299804645,
+      "loss": 1.3468,
+      "step": 195
+    },
+    {
+      "epoch": 0.3769230769230769,
+      "grad_norm": 0.001959095374210495,
+      "learning_rate": 0.14338837391175582,
+      "loss": 1.3161,
+      "step": 196
+    },
+    {
+      "epoch": 0.37884615384615383,
+      "grad_norm": 0.0019099696278269624,
+      "learning_rate": 0.14282593202522628,
+      "loss": 1.2721,
+      "step": 197
+    },
+    {
+      "epoch": 0.38076923076923075,
+      "grad_norm": 0.0020633796120534047,
+      "learning_rate": 0.14226182617406996,
+      "loss": 1.3467,
+      "step": 198
+    },
+    {
+      "epoch": 0.38269230769230766,
+      "grad_norm": 0.0019063633624862503,
+      "learning_rate": 0.14169607827613281,
+      "loss": 1.2642,
+      "step": 199
+    },
+    {
+      "epoch": 0.38461538461538464,
+      "grad_norm": 0.0019048788385057887,
+      "learning_rate": 0.14112871031306118,
+      "loss": 1.2675,
+      "step": 200
+    },
+    {
+      "epoch": 0.38653846153846155,
+      "grad_norm": 0.001672545284898376,
+      "learning_rate": 0.1405597443294475,
+      "loss": 1.2963,
+      "step": 201
+    },
+    {
+      "epoch": 0.38846153846153847,
+      "grad_norm": 0.0017660722745018423,
+      "learning_rate": 0.13998920243197407,
+      "loss": 1.2614,
+      "step": 202
+    },
+    {
+      "epoch": 0.3903846153846154,
+      "grad_norm": 0.0019051487254877109,
+      "learning_rate": 0.13941710678855396,
+      "loss": 1.3164,
+      "step": 203
+    },
+    {
+      "epoch": 0.3923076923076923,
+      "grad_norm": 0.0018327160543885647,
+      "learning_rate": 0.13884347962746948,
+      "loss": 1.329,
+      "step": 204
+    },
+    {
+      "epoch": 0.3942307692307692,
+      "grad_norm": 0.001816651409761731,
+      "learning_rate": 0.138268343236509,
+      "loss": 1.2981,
+      "step": 205
+    },
+    {
+      "epoch": 0.39615384615384613,
+      "grad_norm": 0.001794748048178438,
+      "learning_rate": 0.13769171996210053,
+      "loss": 1.3789,
+      "step": 206
+    },
+    {
+      "epoch": 0.39807692307692305,
+      "grad_norm": 0.0016860920368524516,
+      "learning_rate": 0.1371136322084438,
+      "loss": 1.2807,
+      "step": 207
+    },
+    {
+      "epoch": 0.4,
+      "grad_norm": 0.002035843426413067,
+      "learning_rate": 0.13653410243663952,
+      "loss": 1.3526,
+      "step": 208
+    },
+    {
+      "epoch": 0.40192307692307694,
+      "grad_norm": 0.001750440164579015,
+      "learning_rate": 0.13595315316381676,
+      "loss": 1.2591,
+      "step": 209
+    },
+    {
+      "epoch": 0.40384615384615385,
+      "grad_norm": 0.0019471730758848227,
+      "learning_rate": 0.13537080696225814,
+      "loss": 1.3438,
+      "step": 210
+    },
+    {
+      "epoch": 0.40576923076923077,
+      "grad_norm": 0.0017097969309516508,
+      "learning_rate": 0.13478708645852272,
+      "loss": 1.3506,
+      "step": 211
+    },
+    {
+      "epoch": 0.4076923076923077,
+      "grad_norm": 0.0017832715273476403,
+      "learning_rate": 0.1342020143325669,
+      "loss": 1.3249,
+      "step": 212
+    },
+    {
+      "epoch": 0.4096153846153846,
+      "grad_norm": 0.002068562166710403,
+      "learning_rate": 0.13361561331686309,
+      "loss": 1.282,
+      "step": 213
+    },
+    {
+      "epoch": 0.4115384615384615,
+      "grad_norm": 0.001959421176206219,
+      "learning_rate": 0.13302790619551672,
+      "loss": 1.2747,
+      "step": 214
+    },
+    {
+      "epoch": 0.41346153846153844,
+      "grad_norm": 0.00181326678152697,
+      "learning_rate": 0.1324389158033807,
+      "loss": 1.2229,
+      "step": 215
+    },
+    {
+      "epoch": 0.4153846153846154,
+      "grad_norm": 0.0017835108874445192,
+      "learning_rate": 0.13184866502516845,
+      "loss": 1.1964,
+      "step": 216
+    },
+    {
+      "epoch": 0.4173076923076923,
+      "grad_norm": 0.001841257734392548,
+      "learning_rate": 0.13125717679456447,
+      "loss": 1.3202,
+      "step": 217
+    },
+    {
+      "epoch": 0.41923076923076924,
+      "grad_norm": 0.0018603971144532867,
+      "learning_rate": 0.13066447409333345,
+      "loss": 1.312,
+      "step": 218
+    },
+    {
+      "epoch": 0.42115384615384616,
+      "grad_norm": 0.0016330914282380706,
+      "learning_rate": 0.1300705799504273,
+      "loss": 1.2945,
+      "step": 219
+    },
+    {
+      "epoch": 0.4230769230769231,
+      "grad_norm": 0.0017839322810478326,
+      "learning_rate": 0.12947551744109043,
+      "loss": 1.2755,
+      "step": 220
+    },
+    {
+      "epoch": 0.425,
+      "grad_norm": 0.0017738349276177033,
+      "learning_rate": 0.128879309685963,
+      "loss": 1.3129,
+      "step": 221
+    },
+    {
+      "epoch": 0.4269230769230769,
+      "grad_norm": 0.001837147201948082,
+      "learning_rate": 0.12828197985018275,
+      "loss": 1.2397,
+      "step": 222
+    },
+    {
+      "epoch": 0.4288461538461538,
+      "grad_norm": 0.0016134831727139733,
+      "learning_rate": 0.12768355114248495,
+      "loss": 1.2252,
+      "step": 223
+    },
+    {
+      "epoch": 0.4307692307692308,
+      "grad_norm": 0.001968820923667599,
+      "learning_rate": 0.12708404681430052,
+      "loss": 1.3926,
+      "step": 224
+    },
+    {
+      "epoch": 0.4326923076923077,
+      "grad_norm": 0.0017554969794205308,
+      "learning_rate": 0.1264834901588527,
+      "loss": 1.2481,
+      "step": 225
+    },
+    {
+      "epoch": 0.4346153846153846,
+      "grad_norm": 0.0016918360385788994,
+      "learning_rate": 0.12588190451025208,
+      "loss": 1.3418,
+      "step": 226
+    },
+    {
+      "epoch": 0.43653846153846154,
+      "grad_norm": 0.001669865003083041,
+      "learning_rate": 0.12527931324258976,
+      "loss": 1.3316,
+      "step": 227
+    },
+    {
+      "epoch": 0.43846153846153846,
+      "grad_norm": 0.0018033112780445597,
+      "learning_rate": 0.12467573976902935,
+      "loss": 1.395,
+      "step": 228
+    },
+    {
+      "epoch": 0.4403846153846154,
+      "grad_norm": 0.0015651674574883592,
+      "learning_rate": 0.12407120754089732,
+      "loss": 1.2996,
+      "step": 229
+    },
+    {
+      "epoch": 0.4423076923076923,
+      "grad_norm": 0.0017180584671877414,
+      "learning_rate": 0.12346574004677154,
+      "loss": 1.1901,
+      "step": 230
+    },
+    {
+      "epoch": 0.4442307692307692,
+      "grad_norm": 0.0015557396691584788,
+      "learning_rate": 0.12285936081156897,
+      "loss": 1.2464,
+      "step": 231
+    },
+    {
+      "epoch": 0.4461538461538462,
+      "grad_norm": 0.002064527696538363,
+      "learning_rate": 0.12225209339563144,
+      "loss": 1.4332,
+      "step": 232
+    },
+    {
+      "epoch": 0.4480769230769231,
+      "grad_norm": 0.0020206026450596813,
+      "learning_rate": 0.12164396139381028,
+      "loss": 1.3153,
+      "step": 233
+    },
+    {
+      "epoch": 0.45,
+      "grad_norm": 0.0018051836751774427,
+      "learning_rate": 0.12103498843454959,
+      "loss": 1.1905,
+      "step": 234
+    },
+    {
+      "epoch": 0.4519230769230769,
+      "grad_norm": 0.0018952798191553262,
+      "learning_rate": 0.12042519817896805,
+      "loss": 1.2548,
+      "step": 235
+    },
+    {
+      "epoch": 0.45384615384615384,
+      "grad_norm": 0.0016753521715495048,
+      "learning_rate": 0.11981461431993977,
+      "loss": 1.353,
+      "step": 236
+    },
+    {
+      "epoch": 0.45576923076923076,
+      "grad_norm": 0.0018448488280946597,
+      "learning_rate": 0.11920326058117364,
+      "loss": 1.3323,
+      "step": 237
+    },
+    {
+      "epoch": 0.4576923076923077,
+      "grad_norm": 0.0018526687818276315,
+      "learning_rate": 0.11859116071629149,
+      "loss": 1.269,
+      "step": 238
+    },
+    {
+      "epoch": 0.4596153846153846,
+      "grad_norm": 0.0018481219115950194,
+      "learning_rate": 0.11797833850790528,
+      "loss": 1.3498,
+      "step": 239
+    },
+    {
+      "epoch": 0.46153846153846156,
+      "grad_norm": 0.0016415293224444757,
+      "learning_rate": 0.11736481776669305,
+      "loss": 1.1378,
+      "step": 240
+    },
+    {
+      "epoch": 0.4634615384615385,
+      "grad_norm": 0.0016600981096761287,
+      "learning_rate": 0.11675062233047365,
+      "loss": 1.2259,
+      "step": 241
+    },
+    {
+      "epoch": 0.4653846153846154,
+      "grad_norm": 0.0015767470137499273,
+      "learning_rate": 0.11613577606328068,
+      "loss": 1.2457,
+      "step": 242
+    },
+    {
+      "epoch": 0.4673076923076923,
+      "grad_norm": 0.0017167445582022098,
+      "learning_rate": 0.11552030285443515,
+      "loss": 1.2364,
+      "step": 243
+    },
+    {
+      "epoch": 0.46923076923076923,
+      "grad_norm": 0.0018475209496954582,
+      "learning_rate": 0.11490422661761744,
+      "loss": 1.3691,
+      "step": 244
+    },
+    {
+      "epoch": 0.47115384615384615,
+      "grad_norm": 0.0018642107213386438,
+      "learning_rate": 0.11428757128993801,
+      "loss": 1.2183,
+      "step": 245
+    },
+    {
+      "epoch": 0.47307692307692306,
+      "grad_norm": 0.001844896343827296,
+      "learning_rate": 0.11367036083100734,
+      "loss": 1.4068,
+      "step": 246
+    },
+    {
+      "epoch": 0.475,
+      "grad_norm": 0.001786916976884729,
+      "learning_rate": 0.11305261922200519,
+      "loss": 1.3991,
+      "step": 247
+    },
+    {
+      "epoch": 0.47692307692307695,
+      "grad_norm": 0.0017523724416176484,
+      "learning_rate": 0.11243437046474854,
+      "loss": 1.2308,
+      "step": 248
+    },
+    {
+      "epoch": 0.47884615384615387,
+      "grad_norm": 0.0017350899007910362,
+      "learning_rate": 0.1118156385807593,
+      "loss": 1.3218,
+      "step": 249
+    },
+    {
+      "epoch": 0.4807692307692308,
+      "grad_norm": 0.0018332702549989093,
+      "learning_rate": 0.11119644761033079,
+      "loss": 1.2592,
+      "step": 250
+    },
+    {
+      "epoch": 0.4826923076923077,
+      "grad_norm": 0.0016292282563774897,
+      "learning_rate": 0.1105768216115938,
+      "loss": 1.3284,
+      "step": 251
+    },
+    {
+      "epoch": 0.4846153846153846,
+      "grad_norm": 0.0017900743145921117,
+      "learning_rate": 0.10995678465958168,
+      "loss": 1.3052,
+      "step": 252
+    },
+    {
+      "epoch": 0.48653846153846153,
+      "grad_norm": 0.0019029681751188976,
+      "learning_rate": 0.10933636084529506,
+      "loss": 1.3285,
+      "step": 253
+    },
+    {
+      "epoch": 0.48846153846153845,
+      "grad_norm": 0.001592719417305566,
+      "learning_rate": 0.10871557427476584,
+      "loss": 1.2471,
+      "step": 254
+    },
+    {
+      "epoch": 0.49038461538461536,
+      "grad_norm": 0.0017927880288196637,
+      "learning_rate": 0.10809444906812034,
+      "loss": 1.2565,
+      "step": 255
+    },
+    {
+      "epoch": 0.49230769230769234,
+      "grad_norm": 0.0018198057167208592,
+      "learning_rate": 0.10747300935864244,
+      "loss": 1.2985,
+      "step": 256
+    },
+    {
+      "epoch": 0.49423076923076925,
+      "grad_norm": 0.001806520188792879,
+      "learning_rate": 0.10685127929183567,
+      "loss": 1.2919,
+      "step": 257
+    },
+    {
+      "epoch": 0.49615384615384617,
+      "grad_norm": 0.0015364637296544426,
+      "learning_rate": 0.10622928302448523,
+      "loss": 1.3005,
+      "step": 258
+    },
+    {
+      "epoch": 0.4980769230769231,
+      "grad_norm": 0.0019483285491073768,
+      "learning_rate": 0.10560704472371918,
+      "loss": 1.3734,
+      "step": 259
+    },
+    {
+      "epoch": 0.5,
+      "grad_norm": 0.0017550916605764977,
+      "learning_rate": 0.10498458856606972,
+      "loss": 1.3543,
+      "step": 260
+    },
+    {
+      "epoch": 0.5019230769230769,
+      "grad_norm": 0.001750484762691223,
+      "learning_rate": 0.10436193873653361,
+      "loss": 1.2926,
+      "step": 261
+    },
+    {
+      "epoch": 0.5038461538461538,
+      "grad_norm": 0.0017592665754914275,
+      "learning_rate": 0.10373911942763259,
+      "loss": 1.2162,
+      "step": 262
+    },
+    {
+      "epoch": 0.5057692307692307,
+      "grad_norm": 0.0018195715618310208,
+      "learning_rate": 0.10311615483847332,
+      "loss": 1.3139,
+      "step": 263
+    },
+    {
+      "epoch": 0.5076923076923077,
+      "grad_norm": 0.001713885117353041,
+      "learning_rate": 0.1024930691738073,
+      "loss": 1.3269,
+      "step": 264
+    },
+    {
+      "epoch": 0.5096153846153846,
+      "grad_norm": 0.0019161208943548611,
+      "learning_rate": 0.10186988664309023,
+      "loss": 1.2303,
+      "step": 265
+    },
+    {
+      "epoch": 0.5115384615384615,
+      "grad_norm": 0.0015728499243739412,
+      "learning_rate": 0.10124663145954152,
+      "loss": 1.0905,
+      "step": 266
+    },
+    {
+      "epoch": 0.5134615384615384,
+      "grad_norm": 0.0016484487711629719,
+      "learning_rate": 0.10062332783920336,
+      "loss": 1.2333,
+      "step": 267
+    },
+    {
+      "epoch": 0.5153846153846153,
+      "grad_norm": 0.0020301881604690274,
+      "learning_rate": 0.1,
+      "loss": 1.4231,
+      "step": 268
+    },
+    {
+      "epoch": 0.5173076923076924,
+      "grad_norm": 0.0017752185606448596,
+      "learning_rate": 0.09937667216079665,
+      "loss": 1.3362,
+      "step": 269
+    },
+    {
+      "epoch": 0.5192307692307693,
+      "grad_norm": 0.0016307556008379452,
+      "learning_rate": 0.0987533685404585,
+      "loss": 1.2303,
+      "step": 270
+    },
+    {
+      "epoch": 0.5211538461538462,
+      "grad_norm": 0.0017393087850451601,
+      "learning_rate": 0.0981301133569098,
+      "loss": 1.3259,
+      "step": 271
+    },
+    {
+      "epoch": 0.5230769230769231,
+      "grad_norm": 0.001790639755553169,
+      "learning_rate": 0.09750693082619273,
+      "loss": 1.2539,
+      "step": 272
+    },
+    {
+      "epoch": 0.525,
+      "grad_norm": 0.001799852633891133,
+      "learning_rate": 0.0968838451615267,
+      "loss": 1.3776,
+      "step": 273
+    },
+    {
+      "epoch": 0.5269230769230769,
+      "grad_norm": 0.00190901035307817,
+      "learning_rate": 0.09626088057236745,
+      "loss": 1.2952,
+      "step": 274
+    },
+    {
+      "epoch": 0.5288461538461539,
+      "grad_norm": 0.0018116798578691417,
+      "learning_rate": 0.09563806126346641,
+      "loss": 1.2344,
+      "step": 275
+    },
+    {
+      "epoch": 0.5307692307692308,
+      "grad_norm": 0.001980131447279998,
+      "learning_rate": 0.09501541143393027,
+      "loss": 1.3117,
+      "step": 276
+    },
+    {
+      "epoch": 0.5326923076923077,
+      "grad_norm": 0.0017213110400957346,
+      "learning_rate": 0.09439295527628082,
+      "loss": 1.3727,
+      "step": 277
+    },
+    {
+      "epoch": 0.5346153846153846,
+      "grad_norm": 0.0016146353429896993,
+      "learning_rate": 0.0937707169755148,
+      "loss": 1.1837,
+      "step": 278
+    },
+    {
+      "epoch": 0.5365384615384615,
+      "grad_norm": 0.0019286440565172923,
+      "learning_rate": 0.09314872070816434,
+      "loss": 1.251,
+      "step": 279
+    },
+    {
+      "epoch": 0.5384615384615384,
+      "grad_norm": 0.002157199870739311,
+      "learning_rate": 0.09252699064135758,
+      "loss": 1.2455,
+      "step": 280
+    },
+    {
+      "epoch": 0.5403846153846154,
+      "grad_norm": 0.0018122843862512104,
+      "learning_rate": 0.09190555093187967,
+      "loss": 1.3365,
+      "step": 281
+    },
+    {
+      "epoch": 0.5423076923076923,
+      "grad_norm": 0.0015953798979221347,
+      "learning_rate": 0.09128442572523418,
+      "loss": 1.1965,
+      "step": 282
+    },
+    {
+      "epoch": 0.5442307692307692,
+      "grad_norm": 0.0018737585884568049,
+      "learning_rate": 0.09066363915470495,
+      "loss": 1.3594,
+      "step": 283
+    },
+    {
+      "epoch": 0.5461538461538461,
+      "grad_norm": 0.0018675136120856334,
+      "learning_rate": 0.09004321534041836,
+      "loss": 1.2419,
+      "step": 284
+    },
+    {
+      "epoch": 0.5480769230769231,
+      "grad_norm": 0.0017652525295399034,
+      "learning_rate": 0.08942317838840624,
+      "loss": 1.2237,
+      "step": 285
+    },
+    {
+      "epoch": 0.55,
+      "grad_norm": 0.0018262194032546205,
+      "learning_rate": 0.08880355238966922,
+      "loss": 1.101,
+      "step": 286
+    },
+    {
+      "epoch": 0.551923076923077,
+      "grad_norm": 0.0017311697155055492,
+      "learning_rate": 0.08818436141924073,
+      "loss": 1.3366,
+      "step": 287
+    },
+    {
+      "epoch": 0.5538461538461539,
+      "grad_norm": 0.0016296851549918727,
+      "learning_rate": 0.08756562953525152,
+      "loss": 1.3793,
+      "step": 288
+    },
+    {
+      "epoch": 0.5557692307692308,
+      "grad_norm": 0.0016008767999666864,
+      "learning_rate": 0.08694738077799487,
+      "loss": 1.2386,
+      "step": 289
+    },
+    {
+      "epoch": 0.5576923076923077,
+      "grad_norm": 0.0015501945317164321,
+      "learning_rate": 0.08632963916899268,
+      "loss": 1.1574,
+      "step": 290
+    },
+    {
+      "epoch": 0.5596153846153846,
+      "grad_norm": 0.0018115808316203207,
+      "learning_rate": 0.08571242871006202,
+      "loss": 1.2135,
+      "step": 291
+    },
+    {
+      "epoch": 0.5615384615384615,
+      "grad_norm": 0.0016975500892055441,
+      "learning_rate": 0.08509577338238256,
+      "loss": 1.2679,
+      "step": 292
+    },
+    {
+      "epoch": 0.5634615384615385,
+      "grad_norm": 0.0017895528600354743,
+      "learning_rate": 0.08447969714556484,
+      "loss": 1.208,
+      "step": 293
+    },
+    {
+      "epoch": 0.5653846153846154,
+      "grad_norm": 0.001859297331732047,
+      "learning_rate": 0.08386422393671933,
+      "loss": 1.2313,
+      "step": 294
+    },
+    {
+      "epoch": 0.5673076923076923,
+      "grad_norm": 0.0017698584228706888,
+      "learning_rate": 0.08324937766952638,
+      "loss": 1.3027,
+      "step": 295
+    },
+    {
+      "epoch": 0.5692307692307692,
+      "grad_norm": 0.0018321714896463047,
+      "learning_rate": 0.08263518223330697,
+      "loss": 1.1769,
+      "step": 296
+    },
+    {
+      "epoch": 0.5711538461538461,
+      "grad_norm": 0.0018288038900690848,
+      "learning_rate": 0.08202166149209474,
+      "loss": 1.3001,
+      "step": 297
+    },
+    {
+      "epoch": 0.573076923076923,
+      "grad_norm": 0.0015500138873051717,
+      "learning_rate": 0.08140883928370855,
+      "loss": 1.2747,
+      "step": 298
+    },
+    {
+      "epoch": 0.575,
+      "grad_norm": 0.001590479045421572,
+      "learning_rate": 0.0807967394188264,
+      "loss": 1.3203,
+      "step": 299
+    },
+    {
+      "epoch": 0.5769230769230769,
+      "grad_norm": 0.001713005667970645,
+      "learning_rate": 0.08018538568006027,
+      "loss": 1.3249,
+      "step": 300
+    },
+    {
+      "epoch": 0.5788461538461539,
+      "grad_norm": 0.0016605344133729198,
+      "learning_rate": 0.07957480182103199,
+      "loss": 1.2999,
+      "step": 301
+    },
+    {
+      "epoch": 0.5807692307692308,
+      "grad_norm": 0.0018267333979554744,
+      "learning_rate": 0.07896501156545044,
+      "loss": 1.3425,
+      "step": 302
+    },
+    {
+      "epoch": 0.5826923076923077,
+      "grad_norm": 0.0019426873386282708,
+      "learning_rate": 0.07835603860618973,
+      "loss": 1.2357,
+      "step": 303
+    },
+    {
+      "epoch": 0.5846153846153846,
+      "grad_norm": 0.0017910488630856964,
+      "learning_rate": 0.07774790660436857,
+      "loss": 1.2252,
+      "step": 304
+    },
+    {
+      "epoch": 0.5865384615384616,
+      "grad_norm": 0.0019776779464906005,
+      "learning_rate": 0.07714063918843106,
+      "loss": 1.3405,
+      "step": 305
+    },
+    {
+      "epoch": 0.5884615384615385,
+      "grad_norm": 0.0017611848420070822,
+      "learning_rate": 0.0765342599532285,
+      "loss": 1.2864,
+      "step": 306
+    },
+    {
+      "epoch": 0.5903846153846154,
+      "grad_norm": 0.0016805610755363102,
+      "learning_rate": 0.07592879245910272,
+      "loss": 1.2197,
+      "step": 307
+    },
+    {
+      "epoch": 0.5923076923076923,
+      "grad_norm": 0.0017651097985566008,
+      "learning_rate": 0.07532426023097064,
+      "loss": 1.3392,
+      "step": 308
+    },
+    {
+      "epoch": 0.5942307692307692,
+      "grad_norm": 0.0015930148971587596,
+      "learning_rate": 0.07472068675741024,
+      "loss": 1.2204,
+      "step": 309
+    },
+    {
+      "epoch": 0.5961538461538461,
+      "grad_norm": 0.0017315003090512273,
+      "learning_rate": 0.07411809548974792,
+      "loss": 1.1999,
+      "step": 310
+    },
+    {
+      "epoch": 0.5980769230769231,
+      "grad_norm": 0.0016499104639604693,
+      "learning_rate": 0.07351650984114727,
+      "loss": 1.1664,
+      "step": 311
+    },
+    {
+      "epoch": 0.6,
+      "grad_norm": 0.0017614732162135958,
+      "learning_rate": 0.0729159531856995,
+      "loss": 1.1582,
+      "step": 312
+    },
+    {
+      "epoch": 0.6019230769230769,
+      "grad_norm": 0.001515163591939963,
+      "learning_rate": 0.07231644885751508,
+      "loss": 1.1466,
+      "step": 313
+    },
+    {
+      "epoch": 0.6038461538461538,
+      "grad_norm": 0.0015768704193946847,
+      "learning_rate": 0.07171802014981725,
+      "loss": 1.1852,
+      "step": 314
+    },
+    {
+      "epoch": 0.6057692307692307,
+      "grad_norm": 0.0020164187516767606,
+      "learning_rate": 0.07112069031403703,
+      "loss": 1.2826,
+      "step": 315
+    },
+    {
+      "epoch": 0.6076923076923076,
+      "grad_norm": 0.002017079986376746,
+      "learning_rate": 0.07052448255890957,
+      "loss": 1.1593,
+      "step": 316
+    },
+    {
+      "epoch": 0.6096153846153847,
+      "grad_norm": 0.001542834631307312,
+      "learning_rate": 0.0699294200495727,
+      "loss": 1.1801,
+      "step": 317
+    },
+    {
+      "epoch": 0.6115384615384616,
+      "grad_norm": 0.0018388905909277999,
+      "learning_rate": 0.06933552590666658,
+      "loss": 1.3054,
+      "step": 318
+    },
+    {
+      "epoch": 0.6134615384615385,
+      "grad_norm": 0.001581904684071986,
+      "learning_rate": 0.06874282320543557,
+      "loss": 1.1656,
+      "step": 319
+    },
+    {
+      "epoch": 0.6153846153846154,
+      "grad_norm": 0.0017320338559354727,
+      "learning_rate": 0.06815133497483157,
+      "loss": 1.1104,
+      "step": 320
+    },
+    {
+      "epoch": 0.6173076923076923,
+      "grad_norm": 0.0018007424755859475,
+      "learning_rate": 0.06756108419661931,
+      "loss": 1.3105,
+      "step": 321
+    },
+    {
+      "epoch": 0.6192307692307693,
+      "grad_norm": 0.00165199993060855,
+      "learning_rate": 0.06697209380448332,
+      "loss": 1.1726,
+      "step": 322
+    },
+    {
+      "epoch": 0.6211538461538462,
+      "grad_norm": 0.0016520592434930972,
+      "learning_rate": 0.06638438668313694,
+      "loss": 1.2338,
+      "step": 323
+    },
+    {
+      "epoch": 0.6230769230769231,
+      "grad_norm": 0.0016836681774878056,
+      "learning_rate": 0.06579798566743314,
+      "loss": 1.2449,
+      "step": 324
+    },
+    {
+      "epoch": 0.625,
+      "grad_norm": 0.0018289885394656377,
+      "learning_rate": 0.06521291354147728,
+      "loss": 1.2505,
+      "step": 325
+    },
+    {
+      "epoch": 0.6269230769230769,
+      "grad_norm": 0.0017275551936515178,
+      "learning_rate": 0.06462919303774187,
+      "loss": 1.2404,
+      "step": 326
+    },
+    {
+      "epoch": 0.6288461538461538,
+      "grad_norm": 0.0019022590379733683,
+      "learning_rate": 0.06404684683618325,
+      "loss": 1.3016,
+      "step": 327
+    },
+    {
+      "epoch": 0.6307692307692307,
+      "grad_norm": 0.001751797757856421,
+      "learning_rate": 0.0634658975633605,
+      "loss": 1.3027,
+      "step": 328
+    },
+    {
+      "epoch": 0.6326923076923077,
+      "grad_norm": 0.0014640516017159404,
+      "learning_rate": 0.06288636779155621,
+      "loss": 1.1636,
+      "step": 329
+    },
+    {
+      "epoch": 0.6346153846153846,
+      "grad_norm": 0.0015941913686773497,
+      "learning_rate": 0.06230828003789948,
+      "loss": 1.241,
+      "step": 330
+    },
+    {
+      "epoch": 0.6365384615384615,
+      "grad_norm": 0.001661445549851726,
+      "learning_rate": 0.06173165676349103,
+      "loss": 1.2004,
+      "step": 331
+    },
+    {
+      "epoch": 0.6384615384615384,
+      "grad_norm": 0.001621576083831744,
+      "learning_rate": 0.06115652037253053,
+      "loss": 1.3217,
+      "step": 332
+    },
+    {
+      "epoch": 0.6403846153846153,
+      "grad_norm": 0.001753465012168132,
+      "learning_rate": 0.06058289321144608,
+      "loss": 1.3503,
+      "step": 333
+    },
+    {
+      "epoch": 0.6423076923076924,
+      "grad_norm": 0.0018494240120512392,
+      "learning_rate": 0.06001079756802592,
+      "loss": 1.2442,
+      "step": 334
+    },
+    {
+      "epoch": 0.6442307692307693,
+      "grad_norm": 0.0015456255249585522,
+      "learning_rate": 0.059440255670552514,
+      "loss": 1.2442,
+      "step": 335
+    },
+    {
+      "epoch": 0.6461538461538462,
+      "grad_norm": 0.001802049207051288,
+      "learning_rate": 0.05887128968693887,
+      "loss": 1.1309,
+      "step": 336
+    },
+    {
+      "epoch": 0.6480769230769231,
+      "grad_norm": 0.0016552974451498018,
+      "learning_rate": 0.058303921723867225,
+      "loss": 1.1296,
+      "step": 337
+    },
+    {
+      "epoch": 0.65,
+      "grad_norm": 0.0017290346259557715,
+      "learning_rate": 0.05773817382593008,
+      "loss": 1.2552,
+      "step": 338
+    },
+    {
+      "epoch": 0.6519230769230769,
+      "grad_norm": 0.001644650146402024,
+      "learning_rate": 0.057174067974773715,
+      "loss": 1.193,
+      "step": 339
+    },
+    {
+      "epoch": 0.6538461538461539,
+      "grad_norm": 0.0016172433567801553,
+      "learning_rate": 0.056611626088244195,
+      "loss": 1.1869,
+      "step": 340
+    },
+    {
+      "epoch": 0.6557692307692308,
+      "grad_norm": 0.0017518694274739905,
+      "learning_rate": 0.056050870019535494,
+      "loss": 1.2051,
+      "step": 341
+    },
+    {
+      "epoch": 0.6576923076923077,
+      "grad_norm": 0.0020440904148119853,
+      "learning_rate": 0.05549182155634076,
+      "loss": 1.2834,
+      "step": 342
+    },
+    {
+      "epoch": 0.6596153846153846,
+      "grad_norm": 0.0016149436359766125,
+      "learning_rate": 0.054934502420005464,
+      "loss": 1.2413,
+      "step": 343
+    },
+    {
+      "epoch": 0.6615384615384615,
+      "grad_norm": 0.001647911037940789,
+      "learning_rate": 0.0543789342646837,
+      "loss": 1.1605,
+      "step": 344
+    },
+    {
+      "epoch": 0.6634615384615384,
+      "grad_norm": 0.001831931725729835,
+      "learning_rate": 0.05382513867649663,
+      "loss": 1.2812,
+      "step": 345
+    },
+    {
+      "epoch": 0.6653846153846154,
+      "grad_norm": 0.0015907277199899053,
+      "learning_rate": 0.0532731371726938,
+      "loss": 1.2486,
+      "step": 346
+    },
+    {
+      "epoch": 0.6673076923076923,
+      "grad_norm": 0.001594020940934309,
+      "learning_rate": 0.05272295120081732,
+      "loss": 1.1709,
+      "step": 347
+    },
+    {
+      "epoch": 0.6692307692307692,
+      "grad_norm": 0.0018678413757741573,
+      "learning_rate": 0.05217460213786821,
+      "loss": 1.1309,
+      "step": 348
+    },
+    {
+      "epoch": 0.6711538461538461,
+      "grad_norm": 0.001844082967677432,
+      "learning_rate": 0.051628111289476024,
+      "loss": 1.1775,
+      "step": 349
+    },
+    {
+      "epoch": 0.6730769230769231,
+      "grad_norm": 0.0017534600663280627,
+      "learning_rate": 0.051083499889071106,
+      "loss": 1.2202,
+      "step": 350
+    },
+    {
+      "epoch": 0.675,
+      "grad_norm": 0.0015441602077000001,
+      "learning_rate": 0.05054078909705926,
+      "loss": 1.1313,
+      "step": 351
+    },
+    {
+      "epoch": 0.676923076923077,
+      "grad_norm": 0.0016897907006289895,
+      "learning_rate": 0.050000000000000024,
+      "loss": 1.2477,
+      "step": 352
+    },
+    {
+      "epoch": 0.6788461538461539,
+      "grad_norm": 0.001516358497876371,
+      "learning_rate": 0.04946115360978696,
+      "loss": 1.198,
+      "step": 353
+    },
+    {
+      "epoch": 0.6807692307692308,
+      "grad_norm": 0.0016172885329030842,
+      "learning_rate": 0.048924270862831465,
+      "loss": 1.3268,
+      "step": 354
+    },
+    {
+      "epoch": 0.6826923076923077,
+      "grad_norm": 0.0016374556159027486,
+      "learning_rate": 0.04838937261924933,
+      "loss": 1.179,
+      "step": 355
+    },
+    {
+      "epoch": 0.6846153846153846,
+      "grad_norm": 0.0017162624639748866,
+      "learning_rate": 0.0478564796620502,
+      "loss": 1.1862,
+      "step": 356
+    },
+    {
+      "epoch": 0.6865384615384615,
+      "grad_norm": 0.0015722251949911268,
+      "learning_rate": 0.04732561269632993,
+      "loss": 1.2145,
+      "step": 357
+    },
+    {
+      "epoch": 0.6884615384615385,
+      "grad_norm": 0.0016063603466338328,
+      "learning_rate": 0.04679679234846636,
+      "loss": 1.1437,
+      "step": 358
+    },
+    {
+      "epoch": 0.6903846153846154,
+      "grad_norm": 0.001704312197385143,
+      "learning_rate": 0.046270039165317606,
+      "loss": 1.2534,
+      "step": 359
+    },
+    {
+      "epoch": 0.6923076923076923,
+      "grad_norm": 0.0017001188091410786,
+      "learning_rate": 0.04574537361342407,
+      "loss": 1.2665,
+      "step": 360
+    },
+    {
+      "epoch": 0.6942307692307692,
+      "grad_norm": 0.0015321589488624045,
+      "learning_rate": 0.04522281607821288,
+      "loss": 1.2692,
+      "step": 361
+    },
+    {
+      "epoch": 0.6961538461538461,
+      "grad_norm": 0.001812165996429982,
+      "learning_rate": 0.04470238686320606,
+      "loss": 1.2077,
+      "step": 362
+    },
+    {
+      "epoch": 0.698076923076923,
+      "grad_norm": 0.0016449144663692804,
+      "learning_rate": 0.044184106189231624,
+      "loss": 1.2294,
+      "step": 363
+    },
+    {
+      "epoch": 0.7,
+      "grad_norm": 0.0016794059108207475,
+      "learning_rate": 0.043667994193637795,
+      "loss": 1.2901,
+      "step": 364
+    },
+    {
+      "epoch": 0.7019230769230769,
+      "grad_norm": 0.0017164439408872277,
+      "learning_rate": 0.043154070929510784,
+      "loss": 1.2878,
+      "step": 365
+    },
+    {
+      "epoch": 0.7038461538461539,
+      "grad_norm": 0.0016383034087162991,
+      "learning_rate": 0.04264235636489542,
+      "loss": 1.2462,
+      "step": 366
+    },
+    {
+      "epoch": 0.7057692307692308,
+      "grad_norm": 0.0016394757652949774,
+      "learning_rate": 0.04213287038201943,
+      "loss": 1.2423,
+      "step": 367
+    },
+    {
+      "epoch": 0.7076923076923077,
+      "grad_norm": 0.0016629229483610495,
+      "learning_rate": 0.04162563277652104,
+      "loss": 1.0931,
+      "step": 368
+    },
+    {
+      "epoch": 0.7096153846153846,
+      "grad_norm": 0.0015294334877492964,
+      "learning_rate": 0.04112066325667954,
+      "loss": 1.2406,
+      "step": 369
+    },
+    {
+      "epoch": 0.7115384615384616,
+      "grad_norm": 0.0015962095851231736,
+      "learning_rate": 0.04061798144264986,
+      "loss": 1.1558,
+      "step": 370
+    },
+    {
+      "epoch": 0.7134615384615385,
+      "grad_norm": 0.001723744728855877,
+      "learning_rate": 0.04011760686569998,
+      "loss": 1.1545,
+      "step": 371
+    },
+    {
+      "epoch": 0.7153846153846154,
+      "grad_norm": 0.0015513246524669522,
+      "learning_rate": 0.03961955896745224,
+      "loss": 1.3296,
+      "step": 372
+    },
+    {
+      "epoch": 0.7173076923076923,
+      "grad_norm": 0.0017869968874958048,
+      "learning_rate": 0.03912385709912794,
+      "loss": 1.209,
+      "step": 373
+    },
+    {
+      "epoch": 0.7192307692307692,
+      "grad_norm": 0.0016272679975676922,
+      "learning_rate": 0.038630520520795276,
+      "loss": 1.2295,
+      "step": 374
+    },
+    {
+      "epoch": 0.7211538461538461,
+      "grad_norm": 0.0016608103842975003,
+      "learning_rate": 0.03813956840062119,
+      "loss": 1.1527,
+      "step": 375
+    },
+    {
+      "epoch": 0.7230769230769231,
+      "grad_norm": 0.0015462543614357164,
+      "learning_rate": 0.037651019814126656,
+      "loss": 1.2703,
+      "step": 376
+    },
+    {
+      "epoch": 0.725,
+      "grad_norm": 0.0018758798684679748,
+      "learning_rate": 0.037164893743445275,
+      "loss": 1.2093,
+      "step": 377
+    },
+    {
+      "epoch": 0.7269230769230769,
+      "grad_norm": 0.0016043481729127727,
+      "learning_rate": 0.03668120907658603,
+      "loss": 1.2571,
+      "step": 378
+    },
+    {
+      "epoch": 0.7288461538461538,
+      "grad_norm": 0.00156946570935217,
+      "learning_rate": 0.036199984606699154,
+      "loss": 1.2422,
+      "step": 379
+    },
+    {
+      "epoch": 0.7307692307692307,
+      "grad_norm": 0.001741096752598036,
+      "learning_rate": 0.035721239031346066,
+      "loss": 1.293,
+      "step": 380
+    },
+    {
+      "epoch": 0.7326923076923076,
+      "grad_norm": 0.001614209550126007,
+      "learning_rate": 0.03524499095177297,
+      "loss": 1.2424,
+      "step": 381
+    },
+    {
+      "epoch": 0.7346153846153847,
+      "grad_norm": 0.001601368228041657,
+      "learning_rate": 0.03477125887218792,
+      "loss": 1.2578,
+      "step": 382
+    },
+    {
+      "epoch": 0.7365384615384616,
+      "grad_norm": 0.0017196910047109712,
+      "learning_rate": 0.03430006119904196,
+      "loss": 1.0804,
+      "step": 383
+    },
+    {
+      "epoch": 0.7384615384615385,
+      "grad_norm": 0.0015493770337082129,
+      "learning_rate": 0.033831416240314084,
+      "loss": 1.324,
+      "step": 384
+    },
+    {
+      "epoch": 0.7403846153846154,
+      "grad_norm": 0.0015257429215887219,
+      "learning_rate": 0.03336534220479961,
+      "loss": 1.2161,
+      "step": 385
+    },
+    {
+      "epoch": 0.7423076923076923,
+      "grad_norm": 0.0013935855872166333,
+      "learning_rate": 0.032901857201403005,
+      "loss": 1.1693,
+      "step": 386
+    },
+    {
+      "epoch": 0.7442307692307693,
+      "grad_norm": 0.0016414242980987974,
+      "learning_rate": 0.032440979238433976,
+      "loss": 1.3286,
+      "step": 387
+    },
+    {
+      "epoch": 0.7461538461538462,
+      "grad_norm": 0.0014855608329890119,
+      "learning_rate": 0.03198272622290804,
+      "loss": 1.1144,
+      "step": 388
+    },
+    {
+      "epoch": 0.7480769230769231,
+      "grad_norm": 0.0019077390159117466,
+      "learning_rate": 0.03152711595985065,
+      "loss": 1.168,
+      "step": 389
+    },
+    {
+      "epoch": 0.75,
+      "grad_norm": 0.0015201996964961456,
+      "learning_rate": 0.031074166151605298,
+      "loss": 1.2361,
+      "step": 390
+    },
+    {
+      "epoch": 0.7519230769230769,
+      "grad_norm": 0.0016870248950836167,
+      "learning_rate": 0.030623894397145836,
+      "loss": 1.3165,
+      "step": 391
+    },
+    {
+      "epoch": 0.7538461538461538,
+      "grad_norm": 0.0015398875392807598,
+      "learning_rate": 0.03017631819139273,
+      "loss": 1.1252,
+      "step": 392
+    },
+    {
+      "epoch": 0.7557692307692307,
+      "grad_norm": 0.0014500829934791816,
+      "learning_rate": 0.029731454924533086,
+      "loss": 1.1465,
+      "step": 393
+    },
+    {
+      "epoch": 0.7576923076923077,
+      "grad_norm": 0.0017259217609586605,
+      "learning_rate": 0.029289321881345254,
+      "loss": 1.1893,
+      "step": 394
+    },
+    {
+      "epoch": 0.7596153846153846,
+      "grad_norm": 0.001739948607121838,
+      "learning_rate": 0.028849936240527008,
+      "loss": 1.1531,
+      "step": 395
+    },
+    {
+      "epoch": 0.7615384615384615,
+      "grad_norm": 0.0017589569680167866,
+      "learning_rate": 0.028413315074028157,
+      "loss": 1.2393,
+      "step": 396
+    },
+    {
+      "epoch": 0.7634615384615384,
+      "grad_norm": 0.0015913447586018193,
+      "learning_rate": 0.027979475346387363,
+      "loss": 1.2217,
+      "step": 397
+    },
+    {
+      "epoch": 0.7653846153846153,
+      "grad_norm": 0.001688917944087096,
+      "learning_rate": 0.027548433914072735,
+      "loss": 1.2195,
+      "step": 398
+    },
+    {
+      "epoch": 0.7673076923076924,
+      "grad_norm": 0.0015865185373342783,
+      "learning_rate": 0.027120207524827168,
+      "loss": 1.1989,
+      "step": 399
+    },
+    {
+      "epoch": 0.7692307692307693,
+      "grad_norm": 0.001551124245723901,
+      "learning_rate": 0.02669481281701739,
+      "loss": 1.2291,
+      "step": 400
+    },
+    {
+      "epoch": 0.7711538461538462,
+      "grad_norm": 0.001700020274306294,
+      "learning_rate": 0.026272266318987603,
+      "loss": 1.0415,
+      "step": 401
+    },
+    {
+      "epoch": 0.7730769230769231,
+      "grad_norm": 0.0016776202575816232,
+      "learning_rate": 0.02585258444841733,
+      "loss": 1.1653,
+      "step": 402
+    },
+    {
+      "epoch": 0.775,
+      "grad_norm": 0.0017695988956180737,
+      "learning_rate": 0.025435783511683442,
+      "loss": 1.197,
+      "step": 403
+    },
+    {
+      "epoch": 0.7769230769230769,
+      "grad_norm": 0.001930096886627576,
+      "learning_rate": 0.02502187970322657,
+      "loss": 1.097,
+      "step": 404
+    },
+    {
+      "epoch": 0.7788461538461539,
+      "grad_norm": 0.001614400345176137,
+      "learning_rate": 0.02461088910492202,
+      "loss": 1.199,
+      "step": 405
+    },
+    {
+      "epoch": 0.7807692307692308,
+      "grad_norm": 0.0018246939749030613,
+      "learning_rate": 0.02420282768545469,
+      "loss": 1.1258,
+      "step": 406
+    },
+    {
+      "epoch": 0.7826923076923077,
+      "grad_norm": 0.0016825872998105095,
+      "learning_rate": 0.02379771129969892,
+      "loss": 1.2858,
+      "step": 407
+    },
+    {
+      "epoch": 0.7846153846153846,
+      "grad_norm": 0.001768053290934491,
+      "learning_rate": 0.023395555688102213,
+      "loss": 1.1772,
+      "step": 408
+    },
+    {
+      "epoch": 0.7865384615384615,
+      "grad_norm": 0.001730895360984343,
+      "learning_rate": 0.02299637647607372,
+      "loss": 1.3043,
+      "step": 409
+    },
+    {
+      "epoch": 0.7884615384615384,
+      "grad_norm": 0.0015877283972124915,
+      "learning_rate": 0.022600189173377264,
+      "loss": 1.0303,
+      "step": 410
+    },
+    {
+      "epoch": 0.7903846153846154,
+      "grad_norm": 0.001915178932797099,
+      "learning_rate": 0.022207009173528525,
+      "loss": 1.2854,
+      "step": 411
+    },
+    {
+      "epoch": 0.7923076923076923,
+      "grad_norm": 0.001629821876090772,
+      "learning_rate": 0.02181685175319702,
+      "loss": 1.1925,
+      "step": 412
+    },
+    {
+      "epoch": 0.7942307692307692,
+      "grad_norm": 0.0016105290752491067,
+      "learning_rate": 0.021429732071612653,
+      "loss": 1.2242,
+      "step": 413
+    },
+    {
+      "epoch": 0.7961538461538461,
+      "grad_norm": 0.0014073409280485577,
+      "learning_rate": 0.02104566516997647,
+      "loss": 1.0226,
+      "step": 414
+    },
+    {
+      "epoch": 0.7980769230769231,
+      "grad_norm": 0.0015372816868932568,
+      "learning_rate": 0.020664665970876496,
+      "loss": 1.1666,
+      "step": 415
+    },
+    {
+      "epoch": 0.8,
+      "grad_norm": 0.001810765382480029,
+      "learning_rate": 0.020286749277707784,
+      "loss": 1.0893,
+      "step": 416
+    },
+    {
+      "epoch": 0.801923076923077,
+      "grad_norm": 0.0017573327941261495,
+      "learning_rate": 0.019911929774097215,
+      "loss": 1.2486,
+      "step": 417
+    },
+    {
+      "epoch": 0.8038461538461539,
+      "grad_norm": 0.0016140054937369458,
+      "learning_rate": 0.019540222023333165,
+      "loss": 1.2386,
+      "step": 418
+    },
+    {
+      "epoch": 0.8057692307692308,
+      "grad_norm": 0.0017506304643368441,
+      "learning_rate": 0.01917164046779948,
+      "loss": 1.224,
+      "step": 419
+    },
+    {
+      "epoch": 0.8076923076923077,
+      "grad_norm": 0.001704177911408827,
+      "learning_rate": 0.018806199428414352,
+      "loss": 1.1123,
+      "step": 420
+    },
+    {
+      "epoch": 0.8096153846153846,
+      "grad_norm": 0.0020188450998816923,
+      "learning_rate": 0.018443913104073985,
+      "loss": 1.0501,
+      "step": 421
+    },
+    {
+      "epoch": 0.8115384615384615,
+      "grad_norm": 0.0016441754781975168,
+      "learning_rate": 0.01808479557110081,
+      "loss": 1.1661,
+      "step": 422
+    },
+    {
+      "epoch": 0.8134615384615385,
+      "grad_norm": 0.0018663613830162133,
+      "learning_rate": 0.017728860782696667,
+      "loss": 1.1549,
+      "step": 423
+    },
+    {
+      "epoch": 0.8153846153846154,
+      "grad_norm": 0.0017212621395811104,
+      "learning_rate": 0.017376122568400532,
+      "loss": 1.3077,
+      "step": 424
+    },
+    {
+      "epoch": 0.8173076923076923,
+      "grad_norm": 0.001626299809599149,
+      "learning_rate": 0.017026594633551252,
+      "loss": 1.165,
+      "step": 425
+    },
+    {
+      "epoch": 0.8192307692307692,
+      "grad_norm": 0.0020879312135012747,
+      "learning_rate": 0.01668029055875512,
+      "loss": 1.1892,
+      "step": 426
+    },
+    {
+      "epoch": 0.8211538461538461,
+      "grad_norm": 0.0015421519245730874,
+      "learning_rate": 0.016337223799358026,
+      "loss": 1.1003,
+      "step": 427
+    },
+    {
+      "epoch": 0.823076923076923,
+      "grad_norm": 0.001669199031678073,
+      "learning_rate": 0.01599740768492286,
+      "loss": 1.0732,
+      "step": 428
+    },
+    {
+      "epoch": 0.825,
+      "grad_norm": 0.001614218609903845,
+      "learning_rate": 0.015660855418711452,
+      "loss": 1.173,
+      "step": 429
+    },
+    {
+      "epoch": 0.8269230769230769,
+      "grad_norm": 0.0014995162318101217,
+      "learning_rate": 0.015327580077171589,
+      "loss": 1.1733,
+      "step": 430
+    },
+    {
+      "epoch": 0.8288461538461539,
+      "grad_norm": 0.0018791305214014937,
+      "learning_rate": 0.014997594609429088,
+      "loss": 1.1947,
+      "step": 431
+    },
+    {
+      "epoch": 0.8307692307692308,
+      "grad_norm": 0.0016769752152827795,
+      "learning_rate": 0.01467091183678444,
+      "loss": 1.0874,
+      "step": 432
+    },
+    {
+      "epoch": 0.8326923076923077,
+      "grad_norm": 0.001634809824793844,
+      "learning_rate": 0.014347544452214867,
+      "loss": 1.2192,
+      "step": 433
+    },
+    {
+      "epoch": 0.8346153846153846,
+      "grad_norm": 0.0015957600097326484,
+      "learning_rate": 0.014027505019880971,
+      "loss": 0.9601,
+      "step": 434
+    },
+    {
+      "epoch": 0.8365384615384616,
+      "grad_norm": 0.0019307180912603791,
+      "learning_rate": 0.013710805974638696,
+      "loss": 1.2605,
+      "step": 435
+    },
+    {
+      "epoch": 0.8384615384615385,
+      "grad_norm": 0.00166425188658341,
+      "learning_rate": 0.01339745962155613,
+      "loss": 1.049,
+      "step": 436
+    },
+    {
+      "epoch": 0.8403846153846154,
+      "grad_norm": 0.0016713103596011339,
+      "learning_rate": 0.01308747813543536,
+      "loss": 1.2824,
+      "step": 437
+    },
+    {
+      "epoch": 0.8423076923076923,
+      "grad_norm": 0.0016712132054713617,
+      "learning_rate": 0.012780873560339467,
+      "loss": 1.0898,
+      "step": 438
+    },
+    {
+      "epoch": 0.8442307692307692,
+      "grad_norm": 0.0014277903836737555,
+      "learning_rate": 0.012477657809124632,
+      "loss": 1.1674,
+      "step": 439
+    },
+    {
+      "epoch": 0.8461538461538461,
+      "grad_norm": 0.001967941349281495,
+      "learning_rate": 0.012177842662977134,
+      "loss": 1.1375,
+      "step": 440
+    },
+    {
+      "epoch": 0.8480769230769231,
+      "grad_norm": 0.0016382119787190187,
+      "learning_rate": 0.01188143977095576,
+      "loss": 1.1829,
+      "step": 441
+    },
+    {
+      "epoch": 0.85,
+      "grad_norm": 0.0017736326917507721,
+      "learning_rate": 0.011588460649539035,
+      "loss": 1.1952,
+      "step": 442
+    },
+    {
+      "epoch": 0.8519230769230769,
+      "grad_norm": 0.0016699407309267577,
+      "learning_rate": 0.011298916682177829,
+      "loss": 1.2122,
+      "step": 443
+    },
+    {
+      "epoch": 0.8538461538461538,
+      "grad_norm": 0.0014724609926438752,
+      "learning_rate": 0.011012819118853146,
+      "loss": 1.1699,
+      "step": 444
+    },
+    {
+      "epoch": 0.8557692307692307,
+      "grad_norm": 0.0015976732146527326,
+      "learning_rate": 0.01073017907563887,
+      "loss": 1.1011,
+      "step": 445
+    },
+    {
+      "epoch": 0.8576923076923076,
+      "grad_norm": 0.001546317280518979,
+      "learning_rate": 0.010451007534269908,
+      "loss": 1.2669,
+      "step": 446
+    },
+    {
+      "epoch": 0.8596153846153847,
+      "grad_norm": 0.0018295788061838045,
+      "learning_rate": 0.010175315341715598,
+      "loss": 1.1902,
+      "step": 447
+    },
+    {
+      "epoch": 0.8615384615384616,
+      "grad_norm": 0.0017447038963714274,
+      "learning_rate": 0.009903113209758098,
+      "loss": 1.1675,
+      "step": 448
+    },
+    {
+      "epoch": 0.8634615384615385,
+      "grad_norm": 0.0017174765142991726,
+      "learning_rate": 0.009634411714576352,
+      "loss": 1.2203,
+      "step": 449
+    },
+    {
+      "epoch": 0.8653846153846154,
+      "grad_norm": 0.0016804127932740051,
+      "learning_rate": 0.009369221296335007,
+      "loss": 1.2053,
+      "step": 450
+    },
+    {
+      "epoch": 0.8673076923076923,
+      "grad_norm": 0.0016768764088767503,
+      "learning_rate": 0.009107552258778906,
+      "loss": 1.1945,
+      "step": 451
+    },
+    {
+      "epoch": 0.8692307692307693,
+      "grad_norm": 0.0015439029886146123,
+      "learning_rate": 0.008849414768832687,
+      "loss": 1.2605,
+      "step": 452
+    },
+    {
+      "epoch": 0.8711538461538462,
+      "grad_norm": 0.0015674341261523155,
+      "learning_rate": 0.008594818856205699,
+      "loss": 1.2326,
+      "step": 453
+    },
+    {
+      "epoch": 0.8730769230769231,
+      "grad_norm": 0.0017119494135829799,
+      "learning_rate": 0.00834377441300238,
+      "loss": 1.1077,
+      "step": 454
+    },
+    {
+      "epoch": 0.875,
+      "grad_norm": 0.0016239152425970632,
+      "learning_rate": 0.008096291193337934,
+      "loss": 1.2442,
+      "step": 455
+    },
+    {
+      "epoch": 0.8769230769230769,
+      "grad_norm": 0.0017084673499823916,
+      "learning_rate": 0.007852378812959226,
+      "loss": 1.1675,
+      "step": 456
+    },
+    {
+      "epoch": 0.8788461538461538,
+      "grad_norm": 0.001499046190700168,
+      "learning_rate": 0.007612046748871327,
+      "loss": 1.1658,
+      "step": 457
+    },
+    {
+      "epoch": 0.8807692307692307,
+      "grad_norm": 0.0018018573511216775,
+      "learning_rate": 0.007375304338969136,
+      "loss": 1.31,
+      "step": 458
+    },
+    {
+      "epoch": 0.8826923076923077,
+      "grad_norm": 0.0017770913534819715,
+      "learning_rate": 0.007142160781674645,
+      "loss": 1.2345,
+      "step": 459
+    },
+    {
+      "epoch": 0.8846153846153846,
+      "grad_norm": 0.0016345894166337977,
+      "learning_rate": 0.006912625135579587,
+      "loss": 1.113,
+      "step": 460
+    },
+    {
+      "epoch": 0.8865384615384615,
+      "grad_norm": 0.001337813628924635,
+      "learning_rate": 0.0066867063190933496,
+      "loss": 1.2498,
+      "step": 461
+    },
+    {
+      "epoch": 0.8884615384615384,
+      "grad_norm": 0.0015964313614033373,
+      "learning_rate": 0.006464413110096601,
+      "loss": 1.309,
+      "step": 462
+    },
+    {
+      "epoch": 0.8903846153846153,
+      "grad_norm": 0.0017293203522650673,
+      "learning_rate": 0.006245754145600091,
+      "loss": 1.0777,
+      "step": 463
+    },
+    {
+      "epoch": 0.8923076923076924,
+      "grad_norm": 0.0016810133478936666,
+      "learning_rate": 0.006030737921409169,
+      "loss": 1.2195,
+      "step": 464
+    },
+    {
+      "epoch": 0.8942307692307693,
+      "grad_norm": 0.0017782755002260582,
+      "learning_rate": 0.005819372791793654,
+      "loss": 1.325,
+      "step": 465
+    },
+    {
+      "epoch": 0.8961538461538462,
+      "grad_norm": 0.0014656046160945433,
+      "learning_rate": 0.005611666969163243,
+      "loss": 1.2014,
+      "step": 466
+    },
+    {
+      "epoch": 0.8980769230769231,
+      "grad_norm": 0.0015551682786066371,
+      "learning_rate": 0.005407628523748398,
+      "loss": 1.1975,
+      "step": 467
+    },
+    {
+      "epoch": 0.9,
+      "grad_norm": 0.0018930389065242943,
+      "learning_rate": 0.00520726538328683,
+      "loss": 1.1809,
+      "step": 468
+    },
+    {
+      "epoch": 0.9019230769230769,
+      "grad_norm": 0.001799383641692955,
+      "learning_rate": 0.005010585332715401,
+      "loss": 1.2407,
+      "step": 469
+    },
+    {
+      "epoch": 0.9038461538461539,
+      "grad_norm": 0.0014818708169123075,
+      "learning_rate": 0.004817596013867765,
+      "loss": 1.1198,
+      "step": 470
+    },
+    {
+      "epoch": 0.9057692307692308,
+      "grad_norm": 0.0017488959968762542,
+      "learning_rate": 0.004628304925177318,
+      "loss": 1.1442,
+      "step": 471
+    },
+    {
+      "epoch": 0.9076923076923077,
+      "grad_norm": 0.0016578903771245401,
+      "learning_rate": 0.004442719421385921,
+      "loss": 1.1136,
+      "step": 472
+    },
+    {
+      "epoch": 0.9096153846153846,
+      "grad_norm": 0.0016425978224233317,
+      "learning_rate": 0.004260846713258193,
+      "loss": 1.1699,
+      "step": 473
+    },
+    {
+      "epoch": 0.9115384615384615,
+      "grad_norm": 0.001555097603817431,
+      "learning_rate": 0.004082693867301224,
+      "loss": 1.2306,
+      "step": 474
+    },
+    {
+      "epoch": 0.9134615384615384,
+      "grad_norm": 0.0015852008857769441,
+      "learning_rate": 0.003908267805490051,
+      "loss": 1.1508,
+      "step": 475
+    },
+    {
+      "epoch": 0.9153846153846154,
+      "grad_norm": 0.0016671923260056787,
+      "learning_rate": 0.003737575304998797,
+      "loss": 1.1623,
+      "step": 476
+    },
+    {
+      "epoch": 0.9173076923076923,
+      "grad_norm": 0.0018068005829736569,
+      "learning_rate": 0.003570622997937234,
+      "loss": 1.1488,
+      "step": 477
+    },
+    {
+      "epoch": 0.9192307692307692,
+      "grad_norm": 0.0017134554043945194,
+      "learning_rate": 0.00340741737109318,
+      "loss": 1.1091,
+      "step": 478
+    },
+    {
+      "epoch": 0.9211538461538461,
+      "grad_norm": 0.0017932220525582424,
+      "learning_rate": 0.003247964765680389,
+      "loss": 1.1953,
+      "step": 479
+    },
+    {
+      "epoch": 0.9230769230769231,
+      "grad_norm": 0.0015977242672534431,
+      "learning_rate": 0.0030922713770922153,
+      "loss": 1.2233,
+      "step": 480
+    },
+    {
+      "epoch": 0.925,
+      "grad_norm": 0.0014925576866687904,
+      "learning_rate": 0.0029403432546609046,
+      "loss": 1.2153,
+      "step": 481
+    },
+    {
+      "epoch": 0.926923076923077,
+      "grad_norm": 0.0018649493938581562,
+      "learning_rate": 0.0027921863014225504,
+      "loss": 1.2297,
+      "step": 482
+    },
+    {
+      "epoch": 0.9288461538461539,
+      "grad_norm": 0.001742278718347871,
+      "learning_rate": 0.002647806273887665,
+      "loss": 1.1799,
+      "step": 483
+    },
+    {
+      "epoch": 0.9307692307692308,
+      "grad_norm": 0.001633192055176589,
+      "learning_rate": 0.0025072087818176383,
+      "loss": 1.1772,
+      "step": 484
+    },
+    {
+      "epoch": 0.9326923076923077,
+      "grad_norm": 0.001555706834688869,
+      "learning_rate": 0.002370399288006664,
+      "loss": 1.1344,
+      "step": 485
+    },
+    {
+      "epoch": 0.9346153846153846,
+      "grad_norm": 0.0016553174691554132,
+      "learning_rate": 0.0022373831080695463,
+      "loss": 1.2558,
+      "step": 486
+    },
+    {
+      "epoch": 0.9365384615384615,
+      "grad_norm": 0.001589086619041877,
+      "learning_rate": 0.0021081654102351635,
+      "loss": 1.1084,
+      "step": 487
+    },
+    {
+      "epoch": 0.9384615384615385,
+      "grad_norm": 0.0016357470683410066,
+      "learning_rate": 0.0019827512151456175,
+      "loss": 1.0539,
+      "step": 488
+    },
+    {
+      "epoch": 0.9403846153846154,
+      "grad_norm": 0.001472515571649712,
+      "learning_rate": 0.0018611453956612345,
+      "loss": 1.2245,
+      "step": 489
+    },
+    {
+      "epoch": 0.9423076923076923,
+      "grad_norm": 0.0017012417329146815,
+      "learning_rate": 0.0017433526766711727,
+      "loss": 1.177,
+      "step": 490
+    },
+    {
+      "epoch": 0.9442307692307692,
+      "grad_norm": 0.001709490220557341,
+      "learning_rate": 0.0016293776349098677,
+      "loss": 1.1369,
+      "step": 491
+    },
+    {
+      "epoch": 0.9461538461538461,
+      "grad_norm": 0.0017173343584238612,
+      "learning_rate": 0.0015192246987791981,
+      "loss": 1.2519,
+      "step": 492
+    },
+    {
+      "epoch": 0.948076923076923,
+      "grad_norm": 0.0016784976274621599,
+      "learning_rate": 0.0014128981481764114,
+      "loss": 1.2424,
+      "step": 493
+    },
+    {
+      "epoch": 0.95,
+      "grad_norm": 0.0015358453843140464,
+      "learning_rate": 0.0013104021143278911,
+      "loss": 1.194,
+      "step": 494
+    },
+    {
+      "epoch": 0.9519230769230769,
+      "grad_norm": 0.0016029869129127565,
+      "learning_rate": 0.0012117405796285285,
+      "loss": 1.1514,
+      "step": 495
+    },
+    {
+      "epoch": 0.9538461538461539,
+      "grad_norm": 0.0017094384739518747,
+      "learning_rate": 0.0011169173774871477,
+      "loss": 1.0714,
+      "step": 496
+    },
+    {
+      "epoch": 0.9557692307692308,
+      "grad_norm": 0.0014188185591408546,
+      "learning_rate": 0.0010259361921774012,
+      "loss": 1.1564,
+      "step": 497
+    },
+    {
+      "epoch": 0.9576923076923077,
+      "grad_norm": 0.0017447960034220513,
+      "learning_rate": 0.000938800558694719,
+      "loss": 1.1574,
+      "step": 498
+    },
+    {
+      "epoch": 0.9596153846153846,
+      "grad_norm": 0.0016768910790561108,
+      "learning_rate": 0.0008555138626189618,
+      "loss": 1.2977,
+      "step": 499
+    },
+    {
+      "epoch": 0.9615384615384616,
+      "grad_norm": 0.002133980950454975,
+      "learning_rate": 0.0007760793399827937,
+      "loss": 1.2709,
+      "step": 500
+    },
+    {
+      "epoch": 0.9634615384615385,
+      "grad_norm": 0.0017491335947608145,
+      "learning_rate": 0.000700500077146038,
+      "loss": 1.2042,
+      "step": 501
+    },
+    {
+      "epoch": 0.9653846153846154,
+      "grad_norm": 0.0015682406715703178,
+      "learning_rate": 0.0006287790106757397,
+      "loss": 1.1924,
+      "step": 502
+    },
+    {
+      "epoch": 0.9673076923076923,
+      "grad_norm": 0.0016392211231149461,
+      "learning_rate": 0.0005609189272320237,
+      "loss": 1.1881,
+      "step": 503
+    },
+    {
+      "epoch": 0.9692307692307692,
+      "grad_norm": 0.0018223272584551888,
+      "learning_rate": 0.000496922463459859,
+      "loss": 1.1896,
+      "step": 504
+    },
+    {
+      "epoch": 0.9711538461538461,
+      "grad_norm": 0.0016864792861672022,
+      "learning_rate": 0.0004367921058866187,
+      "loss": 1.2212,
+      "step": 505
+    },
+    {
+      "epoch": 0.9730769230769231,
+      "grad_norm": 0.0017109346531580904,
+      "learning_rate": 0.0003805301908254455,
+      "loss": 1.1433,
+      "step": 506
+    },
+    {
+      "epoch": 0.975,
+      "grad_norm": 0.001571419308734898,
+      "learning_rate": 0.0003281389042844918,
+      "loss": 1.3437,
+      "step": 507
+    },
+    {
+      "epoch": 0.9769230769230769,
+      "grad_norm": 0.0016677426691566874,
+      "learning_rate": 0.00027962028188198705,
+      "loss": 1.2639,
+      "step": 508
+    },
+    {
+      "epoch": 0.9788461538461538,
+      "grad_norm": 0.0015598831506800459,
+      "learning_rate": 0.00023497620876711256,
+      "loss": 1.2239,
+      "step": 509
+    },
+    {
+      "epoch": 0.9807692307692307,
+      "grad_norm": 0.001596772143606804,
+      "learning_rate": 0.00019420841954681523,
+      "loss": 1.1832,
+      "step": 510
+    },
+    {
+      "epoch": 0.9826923076923076,
+      "grad_norm": 0.001569582738961551,
+      "learning_rate": 0.00015731849821833956,
+      "loss": 1.157,
+      "step": 511
+    },
+    {
+      "epoch": 0.9846153846153847,
+      "grad_norm": 0.0016692872279000974,
+      "learning_rate": 0.00012430787810776555,
+      "loss": 1.0414,
+      "step": 512
+    },
+    {
+      "epoch": 0.9865384615384616,
+      "grad_norm": 0.0018327444459812839,
+      "learning_rate": 9.517784181422018e-05,
+      "loss": 1.2453,
+      "step": 513
+    },
+    {
+      "epoch": 0.9884615384615385,
+      "grad_norm": 0.001528422716491307,
+      "learning_rate": 6.992952116013917e-05,
+      "loss": 1.2085,
+      "step": 514
+    },
+    {
+      "epoch": 0.9903846153846154,
+      "grad_norm": 0.001859632754553769,
+      "learning_rate": 4.856389714723575e-05,
+      "loss": 1.2629,
+      "step": 515
+    },
+    {
+      "epoch": 0.9923076923076923,
+      "grad_norm": 0.0016081373824005788,
+      "learning_rate": 3.108179991837545e-05,
+      "loss": 1.1592,
+      "step": 516
+    },
+    {
+      "epoch": 0.9942307692307693,
+      "grad_norm": 0.0015865132222837675,
+      "learning_rate": 1.7483908725357544e-05,
+      "loss": 1.2356,
+      "step": 517
+    },
+    {
+      "epoch": 0.9961538461538462,
+      "grad_norm": 0.0017046381961448138,
+      "learning_rate": 7.770751902513862e-06,
+      "loss": 1.1792,
+      "step": 518
+    },
+    {
+      "epoch": 0.9980769230769231,
+      "grad_norm": 0.001622184294838896,
+      "learning_rate": 1.9427068461808086e-06,
+      "loss": 1.199,
+      "step": 519
+    },
+    {
+      "epoch": 1.0,
+      "grad_norm": 0.0015790765764115932,
+      "learning_rate": 0.0,
+      "loss": 1.2183,
+      "step": 520
+    },
+    {
+      "epoch": 1.0,
+      "step": 520,
+      "total_flos": 419724754747392.0,
+      "train_loss": 1.293173605432877,
+      "train_runtime": 1955.9457,
+      "train_samples_per_second": 34.014,
+      "train_steps_per_second": 0.266
+    }
+  ],
+  "logging_steps": 1.0,
+  "max_steps": 520,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 50000,
+  "total_flos": 419724754747392.0,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}

ckpts_oct12/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.3_2e-1_connector-3.0_1.3_2e-1_ablation/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:40f4f86367fee6828ad7e8b7fc771ac61fc81cad1a8bd74189494079e8ad4b25
+size 7608

ckpts_oct12/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.5_2e-1_connector-3.0_1.5_2e-1_ablation/added_tokens.json

@@ -0,0 +1,24 @@
+{
+  "</tool_call>": 151658,
+  "<tool_call>": 151657,
+  "<|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
+}

ckpts_oct12/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.5_2e-1_connector-3.0_1.5_2e-1_ablation/config.json

@@ -0,0 +1,88 @@
+{
+  "architectures": [
+    "TinyLlavaForConditionalGeneration"
+  ],
+  "backward_type_connector": "normal",
+  "cache_dir": null,
+  "connector_type": "mlp2x_gelu",
+  "hidden_size": 896,
+  "ignore_index": -100,
+  "image_aspect_ratio": "square",
+  "image_token_index": -200,
+  "llm_model_name_or_path": "Qwen/Qwen2.5-0.5B",
+  "mask_model": [
+    "llm",
+    "connector"
+  ],
+  "mask_type_connector": "soft",
+  "model_type": "tinyllava",
+  "num_queries": 128,
+  "num_resampler_layers": 3,
+  "pad_token": "<|endoftext|>",
+  "pad_token_id": 151643,
+  "resampler_hidden_size": 768,
+  "sparsity_connector": null,
+  "subnet_type_connector": "global",
+  "temperature_connector": 1.5,
+  "text_config": {
+    "_name_or_path": "Qwen/Qwen2.5-0.5B",
+    "architectures": [
+      "Qwen2ForCausalLM"
+    ],
+    "backward_type": "normal",
+    "bos_token_id": 151643,
+    "eos_token_id": 151643,
+    "hidden_size": 896,
+    "intermediate_size": 4864,
+    "mask_type": "soft",
+    "masked_layers": "all",
+    "max_position_embeddings": 32768,
+    "max_window_layers": 24,
+    "model_type": "qwen2",
+    "num_attention_heads": 14,
+    "num_hidden_layers": 24,
+    "num_key_value_heads": 2,
+    "rope_theta": 1000000.0,
+    "sliding_window": 32768,
+    "subnet_mode": "both",
+    "subnet_type": "None",
+    "temperature_attn": 1.5,
+    "temperature_mlp": 1.5,
+    "tie_word_embeddings": true,
+    "torch_dtype": "bfloat16",
+    "use_mrope": false,
+    "use_sliding_window": false,
+    "vocab_size": 151936
+  },
+  "threshold_connector": null,
+  "tokenizer_model_max_length": 2048,
+  "tokenizer_name_or_path": "Qwen/Qwen2.5-0.5B",
+  "tokenizer_padding_side": "right",
+  "tokenizer_use_fast": false,
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.40.1",
+  "tune_type_connector": "full",
+  "tune_type_llm": "full",
+  "tune_type_vision_tower": "frozen",
+  "tune_vision_tower_from_layer": 0,
+  "use_cache": true,
+  "vision_config": {
+    "hidden_act": "gelu_pytorch_tanh",
+    "hidden_size": 1152,
+    "image_size": 384,
+    "intermediate_size": 4304,
+    "layer_norm_eps": 1e-06,
+    "model_name_or_path": "google/siglip-so400m-patch14-384",
+    "model_name_or_path2": "",
+    "model_type": "siglip_vision_model",
+    "num_attention_heads": 16,
+    "num_hidden_layers": 27,
+    "patch_size": 14
+  },
+  "vision_feature_layer": -2,
+  "vision_feature_select_strategy": "patch",
+  "vision_hidden_size": 1152,
+  "vision_model_name_or_path": "google/siglip-so400m-patch14-384",
+  "vision_model_name_or_path2": "",
+  "vocab_size": 151936
+}

ckpts_oct12/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.5_2e-1_connector-3.0_1.5_2e-1_ablation/generation_config.json

@@ -0,0 +1,7 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 151643,
+  "eos_token_id": 151643,
+  "transformers_version": "4.40.1",
+  "use_cache": false
+}

ckpts_oct12/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.5_2e-1_connector-3.0_1.5_2e-1_ablation/log.txt

@@ -0,0 +1,2 @@
+2025-10-13 07:33:07,317 | INFO: Total Parameters: 1283756736, Total Trainable Parameters: 359661568
+2025-10-13 07:33:07,322 | INFO: Trainable Parameters:

ckpts_oct12/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.5_2e-1_connector-3.0_1.5_2e-1_ablation/mask_applied/added_tokens.json

@@ -0,0 +1,24 @@
+{
+  "</tool_call>": 151658,
+  "<tool_call>": 151657,
+  "<|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
+}

ckpts_oct12/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.5_2e-1_connector-3.0_1.5_2e-1_ablation/mask_applied/config.json

@@ -0,0 +1,70 @@
+{
+  "architectures": [
+    "TinyLlavaForConditionalGeneration"
+  ],
+  "cache_dir": null,
+  "connector_type": "mlp2x_gelu",
+  "hidden_size": 896,
+  "ignore_index": -100,
+  "image_aspect_ratio": "square",
+  "image_token_index": -200,
+  "llm_model_name_or_path": "Qwen/Qwen2.5-0.5B",
+  "model_type": "tinyllava",
+  "num_queries": 128,
+  "num_resampler_layers": 3,
+  "pad_token": "<|endoftext|>",
+  "resampler_hidden_size": 768,
+  "text_config": {
+    "_name_or_path": "Qwen/Qwen2.5-0.5B",
+    "architectures": [
+      "Qwen2ForCausalLM"
+    ],
+    "bos_token_id": 151643,
+    "eos_token_id": 151643,
+    "hidden_size": 896,
+    "intermediate_size": 4864,
+    "max_position_embeddings": 32768,
+    "max_window_layers": 24,
+    "model_type": "qwen2",
+    "num_attention_heads": 14,
+    "num_hidden_layers": 24,
+    "num_key_value_heads": 2,
+    "rope_theta": 1000000.0,
+    "sliding_window": 32768,
+    "tie_word_embeddings": true,
+    "torch_dtype": "bfloat16",
+    "use_mrope": false,
+    "use_sliding_window": false,
+    "vocab_size": 151936
+  },
+  "tokenizer_model_max_length": 2048,
+  "tokenizer_name_or_path": "Qwen/Qwen2.5-0.5B",
+  "tokenizer_padding_side": "right",
+  "tokenizer_use_fast": false,
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.40.1",
+  "tune_type_connector": "full",
+  "tune_type_llm": "full",
+  "tune_type_vision_tower": "frozen",
+  "tune_vision_tower_from_layer": 0,
+  "use_cache": true,
+  "vision_config": {
+    "hidden_act": "gelu_pytorch_tanh",
+    "hidden_size": 1152,
+    "image_size": 384,
+    "intermediate_size": 4304,
+    "layer_norm_eps": 1e-06,
+    "model_name_or_path": "google/siglip-so400m-patch14-384",
+    "model_name_or_path2": "",
+    "model_type": "siglip_vision_model",
+    "num_attention_heads": 16,
+    "num_hidden_layers": 27,
+    "patch_size": 14
+  },
+  "vision_feature_layer": -2,
+  "vision_feature_select_strategy": "patch",
+  "vision_hidden_size": 1152,
+  "vision_model_name_or_path": "google/siglip-so400m-patch14-384",
+  "vision_model_name_or_path2": "",
+  "vocab_size": 151936
+}

ckpts_oct12/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.5_2e-1_connector-3.0_1.5_2e-1_ablation/mask_applied/generation_config.json

@@ -0,0 +1,6 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 151643,
+  "eos_token_id": 151643,
+  "transformers_version": "4.40.1"
+}

ckpts_oct12/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.5_2e-1_connector-3.0_1.5_2e-1_ablation/mask_applied/masks_connector.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d61cdb1e040c8cc106ab673c632bb991aae9bd17bd46104961054fdf6bf265a5
+size 3671560

ckpts_oct12/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.5_2e-1_connector-3.0_1.5_2e-1_ablation/mask_applied/masks_lm.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:060676862e568e14a6e4156d6ffb6d24465b10194fb60ea008c3e9277f123706
+size 715707982

ckpts_oct12/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.5_2e-1_connector-3.0_1.5_2e-1_ablation/mask_applied/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:beff418f18c4995372551778fa3b25d2080f04c4a5a8eb093e04c52310b20ebb
+size 1848295352

ckpts_oct12/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.5_2e-1_connector-3.0_1.5_2e-1_ablation/mask_applied/preprocessor_config.json

@@ -0,0 +1,37 @@
+{
+  "_valid_processor_keys": [
+    "images",
+    "do_resize",
+    "size",
+    "resample",
+    "do_rescale",
+    "rescale_factor",
+    "do_normalize",
+    "image_mean",
+    "image_std",
+    "return_tensors",
+    "data_format",
+    "input_data_format"
+  ],
+  "do_normalize": true,
+  "do_rescale": true,
+  "do_resize": true,
+  "image_mean": [
+    0.5,
+    0.5,
+    0.5
+  ],
+  "image_processor_type": "SiglipImageProcessor",
+  "image_std": [
+    0.5,
+    0.5,
+    0.5
+  ],
+  "processor_class": "SiglipProcessor",
+  "resample": 3,
+  "rescale_factor": 0.00392156862745098,
+  "size": {
+    "height": 384,
+    "width": 384
+  }
+}

ckpts_oct12/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.5_2e-1_connector-3.0_1.5_2e-1_ablation/mask_applied/special_tokens_map.json

@@ -0,0 +1,32 @@
+{
+  "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": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "unk_token": "<|endoftext|>"
+}

ckpts_oct12/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.5_2e-1_connector-3.0_1.5_2e-1_ablation/mask_applied/tokenizer_config.json

@@ -0,0 +1,208 @@
+{
+  "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
+    }
+  },
+  "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    {%- else %}\n        {{- 'You are a helpful assistant.' }}\n    {%- endif %}\n    {{- \"\\n\\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    {%- else %}\n        {{- '<|im_start|>system\\nYou are a helpful assistant.<|im_end|>\\n' }}\n    {%- endif %}\n{%- endif %}\n{%- for message in messages %}\n    {%- if (message.role == \"user\") or (message.role == \"system\" and not loop.first) or (message.role == \"assistant\" and not message.tool_calls) %}\n        {{- '<|im_start|>' + message.role + '\\n' + message.content + '<|im_end|>' + '\\n' }}\n    {%- elif message.role == \"assistant\" %}\n        {{- '<|im_start|>' + message.role }}\n        {%- if message.content %}\n            {{- '\\n' + message.content }}\n        {%- endif %}\n        {%- for tool_call in message.tool_calls %}\n            {%- if tool_call.function is defined %}\n                {%- set tool_call = tool_call.function %}\n            {%- endif %}\n            {{- '\\n<tool_call>\\n{\"name\": \"' }}\n            {{- tool_call.name }}\n            {{- '\", \"arguments\": ' }}\n            {{- tool_call.arguments | tojson }}\n            {{- '}\\n</tool_call>' }}\n        {%- endfor %}\n        {{- '<|im_end|>\\n' }}\n    {%- elif message.role == \"tool\" %}\n        {%- if (loop.index0 == 0) 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{%- endif %}\n",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "model_max_length": 2048,
+  "pad_token": "<|endoftext|>",
+  "padding_side": "right",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": "<|endoftext|>"
+}

ckpts_oct12/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.5_2e-1_connector-3.0_1.5_2e-1_ablation/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4575e06e8c5320b664a36806fa163cc3c7a7d384467c2d246d149d982fc3205e
+size 2567640568

ckpts_oct12/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.5_2e-1_connector-3.0_1.5_2e-1_ablation/runs/Oct13_07-32-09_ywang29-vrdb-test1-worker-0/events.out.tfevents.1760340792.ywang29-vrdb-test1-worker-0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:092b9ae727818c61b502336e17333d299f62eac95926f2b547b8201d6305f891
+size 116893

ckpts_oct12/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.5_2e-1_connector-3.0_1.5_2e-1_ablation/special_tokens_map.json

@@ -0,0 +1,32 @@
+{
+  "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": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "unk_token": "<|endoftext|>"
+}

ckpts_oct12/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.5_2e-1_connector-3.0_1.5_2e-1_ablation/tokenizer_config.json

@@ -0,0 +1,208 @@
+{
+  "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
+    }
+  },
+  "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    {%- else %}\n        {{- 'You are a helpful assistant.' }}\n    {%- endif %}\n    {{- \"\\n\\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    {%- else %}\n        {{- '<|im_start|>system\\nYou are a helpful assistant.<|im_end|>\\n' }}\n    {%- endif %}\n{%- endif %}\n{%- for message in messages %}\n    {%- if (message.role == \"user\") or (message.role == \"system\" and not loop.first) or (message.role == \"assistant\" and not message.tool_calls) %}\n        {{- '<|im_start|>' + message.role + '\\n' + message.content + '<|im_end|>' + '\\n' }}\n    {%- elif message.role == \"assistant\" %}\n        {{- '<|im_start|>' + message.role }}\n        {%- if message.content %}\n            {{- '\\n' + message.content }}\n        {%- endif %}\n        {%- for tool_call in message.tool_calls %}\n            {%- if tool_call.function is defined %}\n                {%- set tool_call = tool_call.function %}\n            {%- endif %}\n            {{- '\\n<tool_call>\\n{\"name\": \"' }}\n            {{- tool_call.name }}\n            {{- '\", \"arguments\": ' }}\n            {{- tool_call.arguments | tojson }}\n            {{- '}\\n</tool_call>' }}\n        {%- endfor %}\n        {{- '<|im_end|>\\n' }}\n    {%- elif message.role == \"tool\" %}\n        {%- if (loop.index0 == 0) 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{%- endif %}\n",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "model_max_length": 2048,
+  "pad_token": "<|endoftext|>",
+  "padding_side": "right",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": "<|endoftext|>"
+}

ckpts_oct12/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.5_2e-1_connector-3.0_1.5_2e-1_ablation/trainer_state.json

@@ -0,0 +1,3670 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.0,
+  "eval_steps": 500,
+  "global_step": 520,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0019230769230769232,
+      "grad_norm": 0.045378101277949424,
+      "learning_rate": 0.0125,
+      "loss": 2.2867,
+      "step": 1
+    },
+    {
+      "epoch": 0.0038461538461538464,
+      "grad_norm": 0.04697132189897136,
+      "learning_rate": 0.025,
+      "loss": 2.2393,
+      "step": 2
+    },
+    {
+      "epoch": 0.0057692307692307696,
+      "grad_norm": 0.020763107716589704,
+      "learning_rate": 0.037500000000000006,
+      "loss": 1.792,
+      "step": 3
+    },
+    {
+      "epoch": 0.007692307692307693,
+      "grad_norm": 0.008500646175004328,
+      "learning_rate": 0.05,
+      "loss": 1.638,
+      "step": 4
+    },
+    {
+      "epoch": 0.009615384615384616,
+      "grad_norm": 0.01625219017795076,
+      "learning_rate": 0.0625,
+      "loss": 1.6977,
+      "step": 5
+    },
+    {
+      "epoch": 0.011538461538461539,
+      "grad_norm": 0.00789375345086673,
+      "learning_rate": 0.07500000000000001,
+      "loss": 1.4743,
+      "step": 6
+    },
+    {
+      "epoch": 0.013461538461538462,
+      "grad_norm": 0.007719990188295674,
+      "learning_rate": 0.08750000000000001,
+      "loss": 1.4541,
+      "step": 7
+    },
+    {
+      "epoch": 0.015384615384615385,
+      "grad_norm": 0.006743651929448017,
+      "learning_rate": 0.1,
+      "loss": 1.5047,
+      "step": 8
+    },
+    {
+      "epoch": 0.01730769230769231,
+      "grad_norm": 0.007201721039721101,
+      "learning_rate": 0.1125,
+      "loss": 1.5781,
+      "step": 9
+    },
+    {
+      "epoch": 0.019230769230769232,
+      "grad_norm": 0.005125762356592713,
+      "learning_rate": 0.125,
+      "loss": 1.3965,
+      "step": 10
+    },
+    {
+      "epoch": 0.021153846153846155,
+      "grad_norm": 0.00473650269619234,
+      "learning_rate": 0.1375,
+      "loss": 1.4714,
+      "step": 11
+    },
+    {
+      "epoch": 0.023076923076923078,
+      "grad_norm": 0.0043375869168087245,
+      "learning_rate": 0.15000000000000002,
+      "loss": 1.4038,
+      "step": 12
+    },
+    {
+      "epoch": 0.025,
+      "grad_norm": 0.003505359812509816,
+      "learning_rate": 0.1625,
+      "loss": 1.4248,
+      "step": 13
+    },
+    {
+      "epoch": 0.026923076923076925,
+      "grad_norm": 0.004145359787524275,
+      "learning_rate": 0.17500000000000002,
+      "loss": 1.4752,
+      "step": 14
+    },
+    {
+      "epoch": 0.028846153846153848,
+      "grad_norm": 0.0033095655054281186,
+      "learning_rate": 0.1875,
+      "loss": 1.4679,
+      "step": 15
+    },
+    {
+      "epoch": 0.03076923076923077,
+      "grad_norm": 0.003374028704451827,
+      "learning_rate": 0.2,
+      "loss": 1.4244,
+      "step": 16
+    },
+    {
+      "epoch": 0.032692307692307694,
+      "grad_norm": 0.0034781073571818837,
+      "learning_rate": 0.1999980572931538,
+      "loss": 1.5255,
+      "step": 17
+    },
+    {
+      "epoch": 0.03461538461538462,
+      "grad_norm": 0.003184551976515693,
+      "learning_rate": 0.19999222924809748,
+      "loss": 1.3938,
+      "step": 18
+    },
+    {
+      "epoch": 0.03653846153846154,
+      "grad_norm": 0.0037451447541712183,
+      "learning_rate": 0.19998251609127465,
+      "loss": 1.4534,
+      "step": 19
+    },
+    {
+      "epoch": 0.038461538461538464,
+      "grad_norm": 0.0035540055004907784,
+      "learning_rate": 0.19996891820008164,
+      "loss": 1.3951,
+      "step": 20
+    },
+    {
+      "epoch": 0.04038461538461539,
+      "grad_norm": 0.004390206215225588,
+      "learning_rate": 0.19995143610285276,
+      "loss": 1.4462,
+      "step": 21
+    },
+    {
+      "epoch": 0.04230769230769231,
+      "grad_norm": 0.004115756254952855,
+      "learning_rate": 0.19993007047883987,
+      "loss": 1.5501,
+      "step": 22
+    },
+    {
+      "epoch": 0.04423076923076923,
+      "grad_norm": 0.0033785619716500768,
+      "learning_rate": 0.1999048221581858,
+      "loss": 1.4852,
+      "step": 23
+    },
+    {
+      "epoch": 0.046153846153846156,
+      "grad_norm": 0.003358917938104122,
+      "learning_rate": 0.19987569212189224,
+      "loss": 1.4389,
+      "step": 24
+    },
+    {
+      "epoch": 0.04807692307692308,
+      "grad_norm": 0.0034162162364184535,
+      "learning_rate": 0.19984268150178167,
+      "loss": 1.484,
+      "step": 25
+    },
+    {
+      "epoch": 0.05,
+      "grad_norm": 0.002939719670816293,
+      "learning_rate": 0.1998057915804532,
+      "loss": 1.4829,
+      "step": 26
+    },
+    {
+      "epoch": 0.051923076923076926,
+      "grad_norm": 0.0034395539918342065,
+      "learning_rate": 0.1997650237912329,
+      "loss": 1.4009,
+      "step": 27
+    },
+    {
+      "epoch": 0.05384615384615385,
+      "grad_norm": 0.0029404742489057023,
+      "learning_rate": 0.199720379718118,
+      "loss": 1.3928,
+      "step": 28
+    },
+    {
+      "epoch": 0.05576923076923077,
+      "grad_norm": 0.0033700481915348657,
+      "learning_rate": 0.19967186109571552,
+      "loss": 1.4079,
+      "step": 29
+    },
+    {
+      "epoch": 0.057692307692307696,
+      "grad_norm": 0.0030480303975775357,
+      "learning_rate": 0.19961946980917455,
+      "loss": 1.5244,
+      "step": 30
+    },
+    {
+      "epoch": 0.05961538461538462,
+      "grad_norm": 0.00260924567952519,
+      "learning_rate": 0.1995632078941134,
+      "loss": 1.3934,
+      "step": 31
+    },
+    {
+      "epoch": 0.06153846153846154,
+      "grad_norm": 0.0037806545010756294,
+      "learning_rate": 0.19950307753654017,
+      "loss": 1.4034,
+      "step": 32
+    },
+    {
+      "epoch": 0.06346153846153846,
+      "grad_norm": 0.003101530285778325,
+      "learning_rate": 0.19943908107276798,
+      "loss": 1.4089,
+      "step": 33
+    },
+    {
+      "epoch": 0.06538461538461539,
+      "grad_norm": 0.00306040442230196,
+      "learning_rate": 0.19937122098932428,
+      "loss": 1.3878,
+      "step": 34
+    },
+    {
+      "epoch": 0.0673076923076923,
+      "grad_norm": 0.003137229844780806,
+      "learning_rate": 0.19929949992285395,
+      "loss": 1.4026,
+      "step": 35
+    },
+    {
+      "epoch": 0.06923076923076923,
+      "grad_norm": 0.002863825745893692,
+      "learning_rate": 0.19922392066001723,
+      "loss": 1.5189,
+      "step": 36
+    },
+    {
+      "epoch": 0.07115384615384615,
+      "grad_norm": 0.0028411164670203016,
+      "learning_rate": 0.19914448613738106,
+      "loss": 1.5049,
+      "step": 37
+    },
+    {
+      "epoch": 0.07307692307692308,
+      "grad_norm": 0.003268263828925368,
+      "learning_rate": 0.1990611994413053,
+      "loss": 1.5974,
+      "step": 38
+    },
+    {
+      "epoch": 0.075,
+      "grad_norm": 0.003302249662789975,
+      "learning_rate": 0.19897406380782262,
+      "loss": 1.4244,
+      "step": 39
+    },
+    {
+      "epoch": 0.07692307692307693,
+      "grad_norm": 0.0026978723301392775,
+      "learning_rate": 0.19888308262251286,
+      "loss": 1.4678,
+      "step": 40
+    },
+    {
+      "epoch": 0.07884615384615384,
+      "grad_norm": 0.003384878839686392,
+      "learning_rate": 0.19878825942037148,
+      "loss": 1.4412,
+      "step": 41
+    },
+    {
+      "epoch": 0.08076923076923077,
+      "grad_norm": 0.0035427553168805945,
+      "learning_rate": 0.19868959788567211,
+      "loss": 1.4644,
+      "step": 42
+    },
+    {
+      "epoch": 0.08269230769230769,
+      "grad_norm": 0.003132232419872934,
+      "learning_rate": 0.1985871018518236,
+      "loss": 1.4225,
+      "step": 43
+    },
+    {
+      "epoch": 0.08461538461538462,
+      "grad_norm": 0.0028758318994219602,
+      "learning_rate": 0.19848077530122082,
+      "loss": 1.5162,
+      "step": 44
+    },
+    {
+      "epoch": 0.08653846153846154,
+      "grad_norm": 0.0028430318455470076,
+      "learning_rate": 0.19837062236509015,
+      "loss": 1.4757,
+      "step": 45
+    },
+    {
+      "epoch": 0.08846153846153847,
+      "grad_norm": 0.003068377365940242,
+      "learning_rate": 0.19825664732332884,
+      "loss": 1.5933,
+      "step": 46
+    },
+    {
+      "epoch": 0.09038461538461538,
+      "grad_norm": 0.002763992671823621,
+      "learning_rate": 0.19813885460433878,
+      "loss": 1.4581,
+      "step": 47
+    },
+    {
+      "epoch": 0.09230769230769231,
+      "grad_norm": 0.0030834677912234343,
+      "learning_rate": 0.19801724878485438,
+      "loss": 1.4316,
+      "step": 48
+    },
+    {
+      "epoch": 0.09423076923076923,
+      "grad_norm": 0.00292219417278853,
+      "learning_rate": 0.19789183458976486,
+      "loss": 1.4749,
+      "step": 49
+    },
+    {
+      "epoch": 0.09615384615384616,
+      "grad_norm": 0.002763802166460018,
+      "learning_rate": 0.19776261689193048,
+      "loss": 1.4603,
+      "step": 50
+    },
+    {
+      "epoch": 0.09807692307692308,
+      "grad_norm": 0.0030584841908596976,
+      "learning_rate": 0.19762960071199334,
+      "loss": 1.3835,
+      "step": 51
+    },
+    {
+      "epoch": 0.1,
+      "grad_norm": 0.002735861398303265,
+      "learning_rate": 0.19749279121818236,
+      "loss": 1.5208,
+      "step": 52
+    },
+    {
+      "epoch": 0.10192307692307692,
+      "grad_norm": 0.0030020663538372293,
+      "learning_rate": 0.19735219372611235,
+      "loss": 1.5189,
+      "step": 53
+    },
+    {
+      "epoch": 0.10384615384615385,
+      "grad_norm": 0.0025971637321121925,
+      "learning_rate": 0.19720781369857746,
+      "loss": 1.4084,
+      "step": 54
+    },
+    {
+      "epoch": 0.10576923076923077,
+      "grad_norm": 0.002922053596768616,
+      "learning_rate": 0.1970596567453391,
+      "loss": 1.3979,
+      "step": 55
+    },
+    {
+      "epoch": 0.1076923076923077,
+      "grad_norm": 0.0027464188325000945,
+      "learning_rate": 0.1969077286229078,
+      "loss": 1.5279,
+      "step": 56
+    },
+    {
+      "epoch": 0.10961538461538461,
+      "grad_norm": 0.003717926700986435,
+      "learning_rate": 0.19675203523431964,
+      "loss": 1.3843,
+      "step": 57
+    },
+    {
+      "epoch": 0.11153846153846154,
+      "grad_norm": 0.0021841779873059426,
+      "learning_rate": 0.19659258262890683,
+      "loss": 1.5419,
+      "step": 58
+    },
+    {
+      "epoch": 0.11346153846153846,
+      "grad_norm": 0.002666049472470836,
+      "learning_rate": 0.19642937700206278,
+      "loss": 1.3759,
+      "step": 59
+    },
+    {
+      "epoch": 0.11538461538461539,
+      "grad_norm": 0.003620741595115023,
+      "learning_rate": 0.19626242469500121,
+      "loss": 1.4799,
+      "step": 60
+    },
+    {
+      "epoch": 0.11730769230769231,
+      "grad_norm": 0.0036635069435368237,
+      "learning_rate": 0.19609173219450998,
+      "loss": 1.4946,
+      "step": 61
+    },
+    {
+      "epoch": 0.11923076923076924,
+      "grad_norm": 0.002731448857001095,
+      "learning_rate": 0.19591730613269878,
+      "loss": 1.4466,
+      "step": 62
+    },
+    {
+      "epoch": 0.12115384615384615,
+      "grad_norm": 0.003215021015176617,
+      "learning_rate": 0.19573915328674182,
+      "loss": 1.4274,
+      "step": 63
+    },
+    {
+      "epoch": 0.12307692307692308,
+      "grad_norm": 0.002596186365224578,
+      "learning_rate": 0.1955572805786141,
+      "loss": 1.4667,
+      "step": 64
+    },
+    {
+      "epoch": 0.125,
+      "grad_norm": 0.0026441834790114746,
+      "learning_rate": 0.1953716950748227,
+      "loss": 1.4666,
+      "step": 65
+    },
+    {
+      "epoch": 0.12692307692307692,
+      "grad_norm": 0.0034551049795108445,
+      "learning_rate": 0.19518240398613226,
+      "loss": 1.4339,
+      "step": 66
+    },
+    {
+      "epoch": 0.12884615384615383,
+      "grad_norm": 0.0023628445433301897,
+      "learning_rate": 0.1949894146672846,
+      "loss": 1.3108,
+      "step": 67
+    },
+    {
+      "epoch": 0.13076923076923078,
+      "grad_norm": 0.002366510258992572,
+      "learning_rate": 0.1947927346167132,
+      "loss": 1.3753,
+      "step": 68
+    },
+    {
+      "epoch": 0.1326923076923077,
+      "grad_norm": 0.003336940459396765,
+      "learning_rate": 0.1945923714762516,
+      "loss": 1.3567,
+      "step": 69
+    },
+    {
+      "epoch": 0.1346153846153846,
+      "grad_norm": 0.002787767681937568,
+      "learning_rate": 0.19438833303083677,
+      "loss": 1.4043,
+      "step": 70
+    },
+    {
+      "epoch": 0.13653846153846153,
+      "grad_norm": 0.002411556251554849,
+      "learning_rate": 0.19418062720820636,
+      "loss": 1.3332,
+      "step": 71
+    },
+    {
+      "epoch": 0.13846153846153847,
+      "grad_norm": 0.0029167099283612004,
+      "learning_rate": 0.19396926207859086,
+      "loss": 1.4845,
+      "step": 72
+    },
+    {
+      "epoch": 0.14038461538461539,
+      "grad_norm": 0.0023452950885406902,
+      "learning_rate": 0.19375424585439993,
+      "loss": 1.3072,
+      "step": 73
+    },
+    {
+      "epoch": 0.1423076923076923,
+      "grad_norm": 0.0025689970389686265,
+      "learning_rate": 0.1935355868899034,
+      "loss": 1.4244,
+      "step": 74
+    },
+    {
+      "epoch": 0.14423076923076922,
+      "grad_norm": 0.0025928080128953023,
+      "learning_rate": 0.19331329368090666,
+      "loss": 1.3281,
+      "step": 75
+    },
+    {
+      "epoch": 0.14615384615384616,
+      "grad_norm": 0.003771637327681654,
+      "learning_rate": 0.19308737486442043,
+      "loss": 1.5576,
+      "step": 76
+    },
+    {
+      "epoch": 0.14807692307692308,
+      "grad_norm": 0.002589642234271887,
+      "learning_rate": 0.19285783921832536,
+      "loss": 1.2515,
+      "step": 77
+    },
+    {
+      "epoch": 0.15,
+      "grad_norm": 0.0026342452258640136,
+      "learning_rate": 0.19262469566103088,
+      "loss": 1.3838,
+      "step": 78
+    },
+    {
+      "epoch": 0.1519230769230769,
+      "grad_norm": 0.002479993743398484,
+      "learning_rate": 0.19238795325112867,
+      "loss": 1.3613,
+      "step": 79
+    },
+    {
+      "epoch": 0.15384615384615385,
+      "grad_norm": 0.0029740089112267627,
+      "learning_rate": 0.19214762118704076,
+      "loss": 1.5477,
+      "step": 80
+    },
+    {
+      "epoch": 0.15576923076923077,
+      "grad_norm": 0.0034527084165741607,
+      "learning_rate": 0.19190370880666208,
+      "loss": 1.5217,
+      "step": 81
+    },
+    {
+      "epoch": 0.1576923076923077,
+      "grad_norm": 0.002518555927049685,
+      "learning_rate": 0.19165622558699763,
+      "loss": 1.4461,
+      "step": 82
+    },
+    {
+      "epoch": 0.1596153846153846,
+      "grad_norm": 0.002667893377921186,
+      "learning_rate": 0.19140518114379435,
+      "loss": 1.4658,
+      "step": 83
+    },
+    {
+      "epoch": 0.16153846153846155,
+      "grad_norm": 0.0029081098398688137,
+      "learning_rate": 0.19115058523116735,
+      "loss": 1.4672,
+      "step": 84
+    },
+    {
+      "epoch": 0.16346153846153846,
+      "grad_norm": 0.0024129963207500564,
+      "learning_rate": 0.1908924477412211,
+      "loss": 1.4907,
+      "step": 85
+    },
+    {
+      "epoch": 0.16538461538461538,
+      "grad_norm": 0.0026433806796283014,
+      "learning_rate": 0.19063077870366502,
+      "loss": 1.5015,
+      "step": 86
+    },
+    {
+      "epoch": 0.1673076923076923,
+      "grad_norm": 0.002783006731820975,
+      "learning_rate": 0.1903655882854237,
+      "loss": 1.4975,
+      "step": 87
+    },
+    {
+      "epoch": 0.16923076923076924,
+      "grad_norm": 0.003201156597198366,
+      "learning_rate": 0.19009688679024192,
+      "loss": 1.4621,
+      "step": 88
+    },
+    {
+      "epoch": 0.17115384615384616,
+      "grad_norm": 0.002650689621398623,
+      "learning_rate": 0.18982468465828442,
+      "loss": 1.4493,
+      "step": 89
+    },
+    {
+      "epoch": 0.17307692307692307,
+      "grad_norm": 0.0024341159054140564,
+      "learning_rate": 0.1895489924657301,
+      "loss": 1.3823,
+      "step": 90
+    },
+    {
+      "epoch": 0.175,
+      "grad_norm": 0.0022918144374339863,
+      "learning_rate": 0.18926982092436118,
+      "loss": 1.457,
+      "step": 91
+    },
+    {
+      "epoch": 0.17692307692307693,
+      "grad_norm": 0.0024236677560793003,
+      "learning_rate": 0.18898718088114688,
+      "loss": 1.3895,
+      "step": 92
+    },
+    {
+      "epoch": 0.17884615384615385,
+      "grad_norm": 0.002717863921759231,
+      "learning_rate": 0.18870108331782218,
+      "loss": 1.4091,
+      "step": 93
+    },
+    {
+      "epoch": 0.18076923076923077,
+      "grad_norm": 0.002837018436174238,
+      "learning_rate": 0.18841153935046098,
+      "loss": 1.4939,
+      "step": 94
+    },
+    {
+      "epoch": 0.18269230769230768,
+      "grad_norm": 0.002965685513275199,
+      "learning_rate": 0.18811856022904425,
+      "loss": 1.3752,
+      "step": 95
+    },
+    {
+      "epoch": 0.18461538461538463,
+      "grad_norm": 0.0022003626251208114,
+      "learning_rate": 0.18782215733702287,
+      "loss": 1.4006,
+      "step": 96
+    },
+    {
+      "epoch": 0.18653846153846154,
+      "grad_norm": 0.0029564910663042827,
+      "learning_rate": 0.18752234219087538,
+      "loss": 1.3584,
+      "step": 97
+    },
+    {
+      "epoch": 0.18846153846153846,
+      "grad_norm": 0.0021779557853213887,
+      "learning_rate": 0.18721912643966054,
+      "loss": 1.3593,
+      "step": 98
+    },
+    {
+      "epoch": 0.19038461538461537,
+      "grad_norm": 0.0025229569122321677,
+      "learning_rate": 0.18691252186456464,
+      "loss": 1.3788,
+      "step": 99
+    },
+    {
+      "epoch": 0.19230769230769232,
+      "grad_norm": 0.00392883494631525,
+      "learning_rate": 0.1866025403784439,
+      "loss": 1.4736,
+      "step": 100
+    },
+    {
+      "epoch": 0.19423076923076923,
+      "grad_norm": 0.002741295008206236,
+      "learning_rate": 0.18628919402536132,
+      "loss": 1.3789,
+      "step": 101
+    },
+    {
+      "epoch": 0.19615384615384615,
+      "grad_norm": 0.0026177825656032304,
+      "learning_rate": 0.18597249498011903,
+      "loss": 1.3867,
+      "step": 102
+    },
+    {
+      "epoch": 0.19807692307692307,
+      "grad_norm": 0.001999427890969773,
+      "learning_rate": 0.18565245554778517,
+      "loss": 1.3074,
+      "step": 103
+    },
+    {
+      "epoch": 0.2,
+      "grad_norm": 0.0023163598005425915,
+      "learning_rate": 0.18532908816321558,
+      "loss": 1.386,
+      "step": 104
+    },
+    {
+      "epoch": 0.20192307692307693,
+      "grad_norm": 0.0022961803674992156,
+      "learning_rate": 0.18500240539057092,
+      "loss": 1.3856,
+      "step": 105
+    },
+    {
+      "epoch": 0.20384615384615384,
+      "grad_norm": 0.0029919067771560785,
+      "learning_rate": 0.18467241992282843,
+      "loss": 1.4702,
+      "step": 106
+    },
+    {
+      "epoch": 0.20576923076923076,
+      "grad_norm": 0.0029555048850722168,
+      "learning_rate": 0.18433914458128858,
+      "loss": 1.4473,
+      "step": 107
+    },
+    {
+      "epoch": 0.2076923076923077,
+      "grad_norm": 0.002581090139025784,
+      "learning_rate": 0.18400259231507718,
+      "loss": 1.345,
+      "step": 108
+    },
+    {
+      "epoch": 0.20961538461538462,
+      "grad_norm": 0.0022135417569099525,
+      "learning_rate": 0.18366277620064198,
+      "loss": 1.4452,
+      "step": 109
+    },
+    {
+      "epoch": 0.21153846153846154,
+      "grad_norm": 0.002232341924742974,
+      "learning_rate": 0.1833197094412449,
+      "loss": 1.5354,
+      "step": 110
+    },
+    {
+      "epoch": 0.21346153846153845,
+      "grad_norm": 0.0025435349947964363,
+      "learning_rate": 0.18297340536644877,
+      "loss": 1.5364,
+      "step": 111
+    },
+    {
+      "epoch": 0.2153846153846154,
+      "grad_norm": 0.0021602396249646963,
+      "learning_rate": 0.1826238774315995,
+      "loss": 1.4139,
+      "step": 112
+    },
+    {
+      "epoch": 0.2173076923076923,
+      "grad_norm": 0.0020438454690316957,
+      "learning_rate": 0.18227113921730334,
+      "loss": 1.2895,
+      "step": 113
+    },
+    {
+      "epoch": 0.21923076923076923,
+      "grad_norm": 0.0022672981669240207,
+      "learning_rate": 0.1819152044288992,
+      "loss": 1.3883,
+      "step": 114
+    },
+    {
+      "epoch": 0.22115384615384615,
+      "grad_norm": 0.0021800436941815694,
+      "learning_rate": 0.18155608689592603,
+      "loss": 1.5118,
+      "step": 115
+    },
+    {
+      "epoch": 0.2230769230769231,
+      "grad_norm": 0.001958585635038842,
+      "learning_rate": 0.18119380057158568,
+      "loss": 1.5063,
+      "step": 116
+    },
+    {
+      "epoch": 0.225,
+      "grad_norm": 0.0025922212099791964,
+      "learning_rate": 0.18082835953220056,
+      "loss": 1.4833,
+      "step": 117
+    },
+    {
+      "epoch": 0.22692307692307692,
+      "grad_norm": 0.0019067914970774418,
+      "learning_rate": 0.18045977797666685,
+      "loss": 1.3633,
+      "step": 118
+    },
+    {
+      "epoch": 0.22884615384615384,
+      "grad_norm": 0.002210756759966818,
+      "learning_rate": 0.1800880702259028,
+      "loss": 1.3204,
+      "step": 119
+    },
+    {
+      "epoch": 0.23076923076923078,
+      "grad_norm": 0.0024441051302525317,
+      "learning_rate": 0.17971325072229227,
+      "loss": 1.3399,
+      "step": 120
+    },
+    {
+      "epoch": 0.2326923076923077,
+      "grad_norm": 0.0026702946419293197,
+      "learning_rate": 0.17933533402912352,
+      "loss": 1.4027,
+      "step": 121
+    },
+    {
+      "epoch": 0.23461538461538461,
+      "grad_norm": 0.0021227072041655262,
+      "learning_rate": 0.17895433483002354,
+      "loss": 1.294,
+      "step": 122
+    },
+    {
+      "epoch": 0.23653846153846153,
+      "grad_norm": 0.01099242489750551,
+      "learning_rate": 0.17857026792838737,
+      "loss": 1.518,
+      "step": 123
+    },
+    {
+      "epoch": 0.23846153846153847,
+      "grad_norm": 0.002570384105282688,
+      "learning_rate": 0.178183148246803,
+      "loss": 1.3745,
+      "step": 124
+    },
+    {
+      "epoch": 0.2403846153846154,
+      "grad_norm": 0.0022559646803342478,
+      "learning_rate": 0.1777929908264715,
+      "loss": 1.3649,
+      "step": 125
+    },
+    {
+      "epoch": 0.2423076923076923,
+      "grad_norm": 0.0021514019497935253,
+      "learning_rate": 0.17739981082662276,
+      "loss": 1.4242,
+      "step": 126
+    },
+    {
+      "epoch": 0.24423076923076922,
+      "grad_norm": 0.0029204119611455845,
+      "learning_rate": 0.1770036235239263,
+      "loss": 1.3461,
+      "step": 127
+    },
+    {
+      "epoch": 0.24615384615384617,
+      "grad_norm": 0.01130354756304454,
+      "learning_rate": 0.1766044443118978,
+      "loss": 1.42,
+      "step": 128
+    },
+    {
+      "epoch": 0.24807692307692308,
+      "grad_norm": 0.0018837725987765582,
+      "learning_rate": 0.17620228870030108,
+      "loss": 1.3233,
+      "step": 129
+    },
+    {
+      "epoch": 0.25,
+      "grad_norm": 0.0021762108331150917,
+      "learning_rate": 0.1757971723145453,
+      "loss": 1.3868,
+      "step": 130
+    },
+    {
+      "epoch": 0.2519230769230769,
+      "grad_norm": 0.0026453439304667055,
+      "learning_rate": 0.175389110895078,
+      "loss": 1.3784,
+      "step": 131
+    },
+    {
+      "epoch": 0.25384615384615383,
+      "grad_norm": 0.002234163138236021,
+      "learning_rate": 0.17497812029677343,
+      "loss": 1.4302,
+      "step": 132
+    },
+    {
+      "epoch": 0.25576923076923075,
+      "grad_norm": 0.002178573841758172,
+      "learning_rate": 0.17456421648831655,
+      "loss": 1.3336,
+      "step": 133
+    },
+    {
+      "epoch": 0.25769230769230766,
+      "grad_norm": 0.002383369795406837,
+      "learning_rate": 0.17414741555158267,
+      "loss": 1.4187,
+      "step": 134
+    },
+    {
+      "epoch": 0.25961538461538464,
+      "grad_norm": 0.0022816283836005627,
+      "learning_rate": 0.1737277336810124,
+      "loss": 1.4953,
+      "step": 135
+    },
+    {
+      "epoch": 0.26153846153846155,
+      "grad_norm": 0.002278872820064295,
+      "learning_rate": 0.17330518718298263,
+      "loss": 1.4059,
+      "step": 136
+    },
+    {
+      "epoch": 0.26346153846153847,
+      "grad_norm": 0.0025536620229928315,
+      "learning_rate": 0.17287979247517285,
+      "loss": 1.3305,
+      "step": 137
+    },
+    {
+      "epoch": 0.2653846153846154,
+      "grad_norm": 0.0021553638415712185,
+      "learning_rate": 0.17245156608592727,
+      "loss": 1.3425,
+      "step": 138
+    },
+    {
+      "epoch": 0.2673076923076923,
+      "grad_norm": 0.0022912980666559865,
+      "learning_rate": 0.17202052465361267,
+      "loss": 1.2828,
+      "step": 139
+    },
+    {
+      "epoch": 0.2692307692307692,
+      "grad_norm": 0.0026727329448286933,
+      "learning_rate": 0.17158668492597184,
+      "loss": 1.4367,
+      "step": 140
+    },
+    {
+      "epoch": 0.27115384615384613,
+      "grad_norm": 0.0021725421057655872,
+      "learning_rate": 0.17115006375947303,
+      "loss": 1.4507,
+      "step": 141
+    },
+    {
+      "epoch": 0.27307692307692305,
+      "grad_norm": 0.002220464720316986,
+      "learning_rate": 0.17071067811865476,
+      "loss": 1.4708,
+      "step": 142
+    },
+    {
+      "epoch": 0.275,
+      "grad_norm": 0.002302438550568782,
+      "learning_rate": 0.17026854507546693,
+      "loss": 1.3827,
+      "step": 143
+    },
+    {
+      "epoch": 0.27692307692307694,
+      "grad_norm": 0.002175140906883818,
+      "learning_rate": 0.1698236818086073,
+      "loss": 1.322,
+      "step": 144
+    },
+    {
+      "epoch": 0.27884615384615385,
+      "grad_norm": 0.0018987756492918578,
+      "learning_rate": 0.16937610560285418,
+      "loss": 1.2636,
+      "step": 145
+    },
+    {
+      "epoch": 0.28076923076923077,
+      "grad_norm": 0.0021189507793586445,
+      "learning_rate": 0.1689258338483947,
+      "loss": 1.5021,
+      "step": 146
+    },
+    {
+      "epoch": 0.2826923076923077,
+      "grad_norm": 0.0020449372642768796,
+      "learning_rate": 0.16847288404014937,
+      "loss": 1.2972,
+      "step": 147
+    },
+    {
+      "epoch": 0.2846153846153846,
+      "grad_norm": 0.002006911683886497,
+      "learning_rate": 0.16801727377709194,
+      "loss": 1.3408,
+      "step": 148
+    },
+    {
+      "epoch": 0.2865384615384615,
+      "grad_norm": 0.002172586126186971,
+      "learning_rate": 0.16755902076156604,
+      "loss": 1.2854,
+      "step": 149
+    },
+    {
+      "epoch": 0.28846153846153844,
+      "grad_norm": 0.0022664449329225685,
+      "learning_rate": 0.16709814279859703,
+      "loss": 1.5219,
+      "step": 150
+    },
+    {
+      "epoch": 0.2903846153846154,
+      "grad_norm": 0.0020449734157795708,
+      "learning_rate": 0.1666346577952004,
+      "loss": 1.3266,
+      "step": 151
+    },
+    {
+      "epoch": 0.2923076923076923,
+      "grad_norm": 0.002083151216649477,
+      "learning_rate": 0.16616858375968596,
+      "loss": 1.2932,
+      "step": 152
+    },
+    {
+      "epoch": 0.29423076923076924,
+      "grad_norm": 0.0021882247950676162,
+      "learning_rate": 0.16569993880095807,
+      "loss": 1.3309,
+      "step": 153
+    },
+    {
+      "epoch": 0.29615384615384616,
+      "grad_norm": 0.0020397993982194293,
+      "learning_rate": 0.16522874112781213,
+      "loss": 1.4226,
+      "step": 154
+    },
+    {
+      "epoch": 0.2980769230769231,
+      "grad_norm": 0.0020714866945855414,
+      "learning_rate": 0.16475500904822704,
+      "loss": 1.3252,
+      "step": 155
+    },
+    {
+      "epoch": 0.3,
+      "grad_norm": 0.0021787655910149822,
+      "learning_rate": 0.16427876096865393,
+      "loss": 1.356,
+      "step": 156
+    },
+    {
+      "epoch": 0.3019230769230769,
+      "grad_norm": 0.003143499121369811,
+      "learning_rate": 0.16380001539330089,
+      "loss": 1.5059,
+      "step": 157
+    },
+    {
+      "epoch": 0.3038461538461538,
+      "grad_norm": 0.002390082104938572,
+      "learning_rate": 0.163318790923414,
+      "loss": 1.3349,
+      "step": 158
+    },
+    {
+      "epoch": 0.3057692307692308,
+      "grad_norm": 0.0019617906995362516,
+      "learning_rate": 0.16283510625655473,
+      "loss": 1.36,
+      "step": 159
+    },
+    {
+      "epoch": 0.3076923076923077,
+      "grad_norm": 0.002082132238910029,
+      "learning_rate": 0.16234898018587338,
+      "loss": 1.3878,
+      "step": 160
+    },
+    {
+      "epoch": 0.3096153846153846,
+      "grad_norm": 0.0020552419026430365,
+      "learning_rate": 0.16186043159937882,
+      "loss": 1.369,
+      "step": 161
+    },
+    {
+      "epoch": 0.31153846153846154,
+      "grad_norm": 0.0028246321389691466,
+      "learning_rate": 0.16136947947920477,
+      "loss": 1.437,
+      "step": 162
+    },
+    {
+      "epoch": 0.31346153846153846,
+      "grad_norm": 0.002540646543107811,
+      "learning_rate": 0.16087614290087207,
+      "loss": 1.2468,
+      "step": 163
+    },
+    {
+      "epoch": 0.3153846153846154,
+      "grad_norm": 0.0020870811304053542,
+      "learning_rate": 0.16038044103254775,
+      "loss": 1.212,
+      "step": 164
+    },
+    {
+      "epoch": 0.3173076923076923,
+      "grad_norm": 0.001833116006669458,
+      "learning_rate": 0.15988239313430005,
+      "loss": 1.3597,
+      "step": 165
+    },
+    {
+      "epoch": 0.3192307692307692,
+      "grad_norm": 0.002164479291956821,
+      "learning_rate": 0.15938201855735015,
+      "loss": 1.3529,
+      "step": 166
+    },
+    {
+      "epoch": 0.3211538461538462,
+      "grad_norm": 0.002173886458131699,
+      "learning_rate": 0.15887933674332047,
+      "loss": 1.3481,
+      "step": 167
+    },
+    {
+      "epoch": 0.3230769230769231,
+      "grad_norm": 0.002028160110290444,
+      "learning_rate": 0.158374367223479,
+      "loss": 1.2659,
+      "step": 168
+    },
+    {
+      "epoch": 0.325,
+      "grad_norm": 0.0019551265695328703,
+      "learning_rate": 0.1578671296179806,
+      "loss": 1.3602,
+      "step": 169
+    },
+    {
+      "epoch": 0.3269230769230769,
+      "grad_norm": 0.0019881218097369236,
+      "learning_rate": 0.15735764363510463,
+      "loss": 1.3513,
+      "step": 170
+    },
+    {
+      "epoch": 0.32884615384615384,
+      "grad_norm": 0.0021184038673805034,
+      "learning_rate": 0.15684592907048925,
+      "loss": 1.2885,
+      "step": 171
+    },
+    {
+      "epoch": 0.33076923076923076,
+      "grad_norm": 0.0020627568599592425,
+      "learning_rate": 0.1563320058063622,
+      "loss": 1.3586,
+      "step": 172
+    },
+    {
+      "epoch": 0.3326923076923077,
+      "grad_norm": 0.0019292451804717719,
+      "learning_rate": 0.15581589381076844,
+      "loss": 1.2894,
+      "step": 173
+    },
+    {
+      "epoch": 0.3346153846153846,
+      "grad_norm": 0.002226644389508938,
+      "learning_rate": 0.15529761313679394,
+      "loss": 1.3599,
+      "step": 174
+    },
+    {
+      "epoch": 0.33653846153846156,
+      "grad_norm": 0.001972743148985222,
+      "learning_rate": 0.15477718392178716,
+      "loss": 1.2642,
+      "step": 175
+    },
+    {
+      "epoch": 0.3384615384615385,
+      "grad_norm": 0.0020916329619356847,
+      "learning_rate": 0.15425462638657594,
+      "loss": 1.4455,
+      "step": 176
+    },
+    {
+      "epoch": 0.3403846153846154,
+      "grad_norm": 0.002170192236744153,
+      "learning_rate": 0.1537299608346824,
+      "loss": 1.3062,
+      "step": 177
+    },
+    {
+      "epoch": 0.3423076923076923,
+      "grad_norm": 0.002076299641803403,
+      "learning_rate": 0.15320320765153367,
+      "loss": 1.337,
+      "step": 178
+    },
+    {
+      "epoch": 0.34423076923076923,
+      "grad_norm": 0.0018925374992674942,
+      "learning_rate": 0.15267438730367008,
+      "loss": 1.424,
+      "step": 179
+    },
+    {
+      "epoch": 0.34615384615384615,
+      "grad_norm": 0.0020846601453478578,
+      "learning_rate": 0.1521435203379498,
+      "loss": 1.3383,
+      "step": 180
+    },
+    {
+      "epoch": 0.34807692307692306,
+      "grad_norm": 0.0018318477637558762,
+      "learning_rate": 0.15161062738075068,
+      "loss": 1.3029,
+      "step": 181
+    },
+    {
+      "epoch": 0.35,
+      "grad_norm": 0.0020194527393682157,
+      "learning_rate": 0.1510757291371686,
+      "loss": 1.3096,
+      "step": 182
+    },
+    {
+      "epoch": 0.35192307692307695,
+      "grad_norm": 0.0020336421606418006,
+      "learning_rate": 0.1505388463902131,
+      "loss": 1.352,
+      "step": 183
+    },
+    {
+      "epoch": 0.35384615384615387,
+      "grad_norm": 0.0020077241198259814,
+      "learning_rate": 0.15000000000000002,
+      "loss": 1.2597,
+      "step": 184
+    },
+    {
+      "epoch": 0.3557692307692308,
+      "grad_norm": 0.002010322702844064,
+      "learning_rate": 0.14945921090294076,
+      "loss": 1.4327,
+      "step": 185
+    },
+    {
+      "epoch": 0.3576923076923077,
+      "grad_norm": 0.0019301224418169396,
+      "learning_rate": 0.14891650011092894,
+      "loss": 1.2833,
+      "step": 186
+    },
+    {
+      "epoch": 0.3596153846153846,
+      "grad_norm": 0.0022870518155233884,
+      "learning_rate": 0.14837188871052398,
+      "loss": 1.3011,
+      "step": 187
+    },
+    {
+      "epoch": 0.36153846153846153,
+      "grad_norm": 0.002116689573891423,
+      "learning_rate": 0.14782539786213184,
+      "loss": 1.3717,
+      "step": 188
+    },
+    {
+      "epoch": 0.36346153846153845,
+      "grad_norm": 0.0017870271201307379,
+      "learning_rate": 0.1472770487991827,
+      "loss": 1.3824,
+      "step": 189
+    },
+    {
+      "epoch": 0.36538461538461536,
+      "grad_norm": 0.0021076267459302447,
+      "learning_rate": 0.1467268628273062,
+      "loss": 1.2986,
+      "step": 190
+    },
+    {
+      "epoch": 0.36730769230769234,
+      "grad_norm": 0.0017721197412291676,
+      "learning_rate": 0.1461748613235034,
+      "loss": 1.2499,
+      "step": 191
+    },
+    {
+      "epoch": 0.36923076923076925,
+      "grad_norm": 0.0017824778304023152,
+      "learning_rate": 0.1456210657353163,
+      "loss": 1.3403,
+      "step": 192
+    },
+    {
+      "epoch": 0.37115384615384617,
+      "grad_norm": 0.0030397245265361803,
+      "learning_rate": 0.14506549757999454,
+      "loss": 1.3932,
+      "step": 193
+    },
+    {
+      "epoch": 0.3730769230769231,
+      "grad_norm": 0.0019625232689332263,
+      "learning_rate": 0.14450817844365924,
+      "loss": 1.2458,
+      "step": 194
+    },
+    {
+      "epoch": 0.375,
+      "grad_norm": 0.001924912511123542,
+      "learning_rate": 0.1439491299804645,
+      "loss": 1.3568,
+      "step": 195
+    },
+    {
+      "epoch": 0.3769230769230769,
+      "grad_norm": 0.0021667906308159538,
+      "learning_rate": 0.14338837391175582,
+      "loss": 1.3255,
+      "step": 196
+    },
+    {
+      "epoch": 0.37884615384615383,
+      "grad_norm": 0.0018935444703060223,
+      "learning_rate": 0.14282593202522628,
+      "loss": 1.277,
+      "step": 197
+    },
+    {
+      "epoch": 0.38076923076923075,
+      "grad_norm": 0.00196592343673127,
+      "learning_rate": 0.14226182617406996,
+      "loss": 1.3525,
+      "step": 198
+    },
+    {
+      "epoch": 0.38269230769230766,
+      "grad_norm": 0.0019949426054931803,
+      "learning_rate": 0.14169607827613281,
+      "loss": 1.2704,
+      "step": 199
+    },
+    {
+      "epoch": 0.38461538461538464,
+      "grad_norm": 0.00202208635270333,
+      "learning_rate": 0.14112871031306118,
+      "loss": 1.2898,
+      "step": 200
+    },
+    {
+      "epoch": 0.38653846153846155,
+      "grad_norm": 0.0019407538657155684,
+      "learning_rate": 0.1405597443294475,
+      "loss": 1.3183,
+      "step": 201
+    },
+    {
+      "epoch": 0.38846153846153847,
+      "grad_norm": 0.001854817555970354,
+      "learning_rate": 0.13998920243197407,
+      "loss": 1.2672,
+      "step": 202
+    },
+    {
+      "epoch": 0.3903846153846154,
+      "grad_norm": 0.0020415975983161393,
+      "learning_rate": 0.13941710678855396,
+      "loss": 1.3196,
+      "step": 203
+    },
+    {
+      "epoch": 0.3923076923076923,
+      "grad_norm": 0.0020382897871836538,
+      "learning_rate": 0.13884347962746948,
+      "loss": 1.3431,
+      "step": 204
+    },
+    {
+      "epoch": 0.3942307692307692,
+      "grad_norm": 0.0021174252687392783,
+      "learning_rate": 0.138268343236509,
+      "loss": 1.3221,
+      "step": 205
+    },
+    {
+      "epoch": 0.39615384615384613,
+      "grad_norm": 0.0019493065698673942,
+      "learning_rate": 0.13769171996210053,
+      "loss": 1.3867,
+      "step": 206
+    },
+    {
+      "epoch": 0.39807692307692305,
+      "grad_norm": 0.0019555393677294737,
+      "learning_rate": 0.1371136322084438,
+      "loss": 1.3032,
+      "step": 207
+    },
+    {
+      "epoch": 0.4,
+      "grad_norm": 0.0021942780825240423,
+      "learning_rate": 0.13653410243663952,
+      "loss": 1.3579,
+      "step": 208
+    },
+    {
+      "epoch": 0.40192307692307694,
+      "grad_norm": 0.001968724371615986,
+      "learning_rate": 0.13595315316381676,
+      "loss": 1.2747,
+      "step": 209
+    },
+    {
+      "epoch": 0.40384615384615385,
+      "grad_norm": 0.002124548387444516,
+      "learning_rate": 0.13537080696225814,
+      "loss": 1.3607,
+      "step": 210
+    },
+    {
+      "epoch": 0.40576923076923077,
+      "grad_norm": 0.001811570759324109,
+      "learning_rate": 0.13478708645852272,
+      "loss": 1.3605,
+      "step": 211
+    },
+    {
+      "epoch": 0.4076923076923077,
+      "grad_norm": 0.002007761615095212,
+      "learning_rate": 0.1342020143325669,
+      "loss": 1.3377,
+      "step": 212
+    },
+    {
+      "epoch": 0.4096153846153846,
+      "grad_norm": 0.0023782017771847015,
+      "learning_rate": 0.13361561331686309,
+      "loss": 1.2967,
+      "step": 213
+    },
+    {
+      "epoch": 0.4115384615384615,
+      "grad_norm": 0.002044665683479174,
+      "learning_rate": 0.13302790619551672,
+      "loss": 1.2871,
+      "step": 214
+    },
+    {
+      "epoch": 0.41346153846153844,
+      "grad_norm": 0.0019970543792148174,
+      "learning_rate": 0.1324389158033807,
+      "loss": 1.2387,
+      "step": 215
+    },
+    {
+      "epoch": 0.4153846153846154,
+      "grad_norm": 0.0018097422252602335,
+      "learning_rate": 0.13184866502516845,
+      "loss": 1.2078,
+      "step": 216
+    },
+    {
+      "epoch": 0.4173076923076923,
+      "grad_norm": 0.0019365842577466989,
+      "learning_rate": 0.13125717679456447,
+      "loss": 1.3299,
+      "step": 217
+    },
+    {
+      "epoch": 0.41923076923076924,
+      "grad_norm": 0.0019770892511836798,
+      "learning_rate": 0.13066447409333345,
+      "loss": 1.3239,
+      "step": 218
+    },
+    {
+      "epoch": 0.42115384615384616,
+      "grad_norm": 0.0017115239537278369,
+      "learning_rate": 0.1300705799504273,
+      "loss": 1.3004,
+      "step": 219
+    },
+    {
+      "epoch": 0.4230769230769231,
+      "grad_norm": 0.002242868024498637,
+      "learning_rate": 0.12947551744109043,
+      "loss": 1.3145,
+      "step": 220
+    },
+    {
+      "epoch": 0.425,
+      "grad_norm": 0.002062274192443457,
+      "learning_rate": 0.128879309685963,
+      "loss": 1.3291,
+      "step": 221
+    },
+    {
+      "epoch": 0.4269230769230769,
+      "grad_norm": 0.0019821879476358225,
+      "learning_rate": 0.12828197985018275,
+      "loss": 1.2416,
+      "step": 222
+    },
+    {
+      "epoch": 0.4288461538461538,
+      "grad_norm": 0.0017885876368598837,
+      "learning_rate": 0.12768355114248495,
+      "loss": 1.2332,
+      "step": 223
+    },
+    {
+      "epoch": 0.4307692307692308,
+      "grad_norm": 0.0026879675448863833,
+      "learning_rate": 0.12708404681430052,
+      "loss": 1.4249,
+      "step": 224
+    },
+    {
+      "epoch": 0.4326923076923077,
+      "grad_norm": 0.0019819195452595016,
+      "learning_rate": 0.1264834901588527,
+      "loss": 1.2577,
+      "step": 225
+    },
+    {
+      "epoch": 0.4346153846153846,
+      "grad_norm": 0.0018114893240739826,
+      "learning_rate": 0.12588190451025208,
+      "loss": 1.3548,
+      "step": 226
+    },
+    {
+      "epoch": 0.43653846153846154,
+      "grad_norm": 0.0017891431602696701,
+      "learning_rate": 0.12527931324258976,
+      "loss": 1.3367,
+      "step": 227
+    },
+    {
+      "epoch": 0.43846153846153846,
+      "grad_norm": 0.002042131355005374,
+      "learning_rate": 0.12467573976902935,
+      "loss": 1.4347,
+      "step": 228
+    },
+    {
+      "epoch": 0.4403846153846154,
+      "grad_norm": 0.001700904982707256,
+      "learning_rate": 0.12407120754089732,
+      "loss": 1.3084,
+      "step": 229
+    },
+    {
+      "epoch": 0.4423076923076923,
+      "grad_norm": 0.0019798809418815027,
+      "learning_rate": 0.12346574004677154,
+      "loss": 1.1943,
+      "step": 230
+    },
+    {
+      "epoch": 0.4442307692307692,
+      "grad_norm": 0.0017034791809830677,
+      "learning_rate": 0.12285936081156897,
+      "loss": 1.2532,
+      "step": 231
+    },
+    {
+      "epoch": 0.4461538461538462,
+      "grad_norm": 0.0026204869306797555,
+      "learning_rate": 0.12225209339563144,
+      "loss": 1.4729,
+      "step": 232
+    },
+    {
+      "epoch": 0.4480769230769231,
+      "grad_norm": 0.0020802131499292865,
+      "learning_rate": 0.12164396139381028,
+      "loss": 1.3349,
+      "step": 233
+    },
+    {
+      "epoch": 0.45,
+      "grad_norm": 0.0020398917445271377,
+      "learning_rate": 0.12103498843454959,
+      "loss": 1.2051,
+      "step": 234
+    },
+    {
+      "epoch": 0.4519230769230769,
+      "grad_norm": 0.002075270596536038,
+      "learning_rate": 0.12042519817896805,
+      "loss": 1.2631,
+      "step": 235
+    },
+    {
+      "epoch": 0.45384615384615384,
+      "grad_norm": 0.0018984799986342287,
+      "learning_rate": 0.11981461431993977,
+      "loss": 1.3706,
+      "step": 236
+    },
+    {
+      "epoch": 0.45576923076923076,
+      "grad_norm": 0.0019950561491145374,
+      "learning_rate": 0.11920326058117364,
+      "loss": 1.3391,
+      "step": 237
+    },
+    {
+      "epoch": 0.4576923076923077,
+      "grad_norm": 0.001832845006276492,
+      "learning_rate": 0.11859116071629149,
+      "loss": 1.2694,
+      "step": 238
+    },
+    {
+      "epoch": 0.4596153846153846,
+      "grad_norm": 0.0019356056305728363,
+      "learning_rate": 0.11797833850790528,
+      "loss": 1.3652,
+      "step": 239
+    },
+    {
+      "epoch": 0.46153846153846156,
+      "grad_norm": 0.001986801597050041,
+      "learning_rate": 0.11736481776669305,
+      "loss": 1.1444,
+      "step": 240
+    },
+    {
+      "epoch": 0.4634615384615385,
+      "grad_norm": 0.0017367332779446953,
+      "learning_rate": 0.11675062233047365,
+      "loss": 1.2317,
+      "step": 241
+    },
+    {
+      "epoch": 0.4653846153846154,
+      "grad_norm": 0.001782361005921778,
+      "learning_rate": 0.11613577606328068,
+      "loss": 1.255,
+      "step": 242
+    },
+    {
+      "epoch": 0.4673076923076923,
+      "grad_norm": 0.001887926247667154,
+      "learning_rate": 0.11552030285443515,
+      "loss": 1.2431,
+      "step": 243
+    },
+    {
+      "epoch": 0.46923076923076923,
+      "grad_norm": 0.0020043540156241104,
+      "learning_rate": 0.11490422661761744,
+      "loss": 1.3709,
+      "step": 244
+    },
+    {
+      "epoch": 0.47115384615384615,
+      "grad_norm": 0.0018474367364264008,
+      "learning_rate": 0.11428757128993801,
+      "loss": 1.2265,
+      "step": 245
+    },
+    {
+      "epoch": 0.47307692307692306,
+      "grad_norm": 0.0019847989053645705,
+      "learning_rate": 0.11367036083100734,
+      "loss": 1.4203,
+      "step": 246
+    },
+    {
+      "epoch": 0.475,
+      "grad_norm": 0.001917041350317627,
+      "learning_rate": 0.11305261922200519,
+      "loss": 1.4098,
+      "step": 247
+    },
+    {
+      "epoch": 0.47692307692307695,
+      "grad_norm": 0.002031993050399593,
+      "learning_rate": 0.11243437046474854,
+      "loss": 1.2373,
+      "step": 248
+    },
+    {
+      "epoch": 0.47884615384615387,
+      "grad_norm": 0.0018373684686496883,
+      "learning_rate": 0.1118156385807593,
+      "loss": 1.3307,
+      "step": 249
+    },
+    {
+      "epoch": 0.4807692307692308,
+      "grad_norm": 0.0020380999477915087,
+      "learning_rate": 0.11119644761033079,
+      "loss": 1.2626,
+      "step": 250
+    },
+    {
+      "epoch": 0.4826923076923077,
+      "grad_norm": 0.0017373833081173985,
+      "learning_rate": 0.1105768216115938,
+      "loss": 1.3365,
+      "step": 251
+    },
+    {
+      "epoch": 0.4846153846153846,
+      "grad_norm": 0.001843186023374389,
+      "learning_rate": 0.10995678465958168,
+      "loss": 1.3193,
+      "step": 252
+    },
+    {
+      "epoch": 0.48653846153846153,
+      "grad_norm": 0.0020830516220618645,
+      "learning_rate": 0.10933636084529506,
+      "loss": 1.3337,
+      "step": 253
+    },
+    {
+      "epoch": 0.48846153846153845,
+      "grad_norm": 0.0018240368446772008,
+      "learning_rate": 0.10871557427476584,
+      "loss": 1.2577,
+      "step": 254
+    },
+    {
+      "epoch": 0.49038461538461536,
+      "grad_norm": 0.002022581711459243,
+      "learning_rate": 0.10809444906812034,
+      "loss": 1.2622,
+      "step": 255
+    },
+    {
+      "epoch": 0.49230769230769234,
+      "grad_norm": 0.001969894477577831,
+      "learning_rate": 0.10747300935864244,
+      "loss": 1.3101,
+      "step": 256
+    },
+    {
+      "epoch": 0.49423076923076925,
+      "grad_norm": 0.0018750896086650649,
+      "learning_rate": 0.10685127929183567,
+      "loss": 1.3008,
+      "step": 257
+    },
+    {
+      "epoch": 0.49615384615384617,
+      "grad_norm": 0.0016732286278647928,
+      "learning_rate": 0.10622928302448523,
+      "loss": 1.3104,
+      "step": 258
+    },
+    {
+      "epoch": 0.4980769230769231,
+      "grad_norm": 0.002146324978856792,
+      "learning_rate": 0.10560704472371918,
+      "loss": 1.3859,
+      "step": 259
+    },
+    {
+      "epoch": 0.5,
+      "grad_norm": 0.0020686063644514605,
+      "learning_rate": 0.10498458856606972,
+      "loss": 1.3834,
+      "step": 260
+    },
+    {
+      "epoch": 0.5019230769230769,
+      "grad_norm": 0.002049713922500538,
+      "learning_rate": 0.10436193873653361,
+      "loss": 1.3214,
+      "step": 261
+    },
+    {
+      "epoch": 0.5038461538461538,
+      "grad_norm": 0.0018761131092795476,
+      "learning_rate": 0.10373911942763259,
+      "loss": 1.2234,
+      "step": 262
+    },
+    {
+      "epoch": 0.5057692307692307,
+      "grad_norm": 0.0020275555944458848,
+      "learning_rate": 0.10311615483847332,
+      "loss": 1.3327,
+      "step": 263
+    },
+    {
+      "epoch": 0.5076923076923077,
+      "grad_norm": 0.001870032964042851,
+      "learning_rate": 0.1024930691738073,
+      "loss": 1.3398,
+      "step": 264
+    },
+    {
+      "epoch": 0.5096153846153846,
+      "grad_norm": 0.002034819469943561,
+      "learning_rate": 0.10186988664309023,
+      "loss": 1.2382,
+      "step": 265
+    },
+    {
+      "epoch": 0.5115384615384615,
+      "grad_norm": 0.0016625115279263545,
+      "learning_rate": 0.10124663145954152,
+      "loss": 1.097,
+      "step": 266
+    },
+    {
+      "epoch": 0.5134615384615384,
+      "grad_norm": 0.0018541577537325982,
+      "learning_rate": 0.10062332783920336,
+      "loss": 1.2395,
+      "step": 267
+    },
+    {
+      "epoch": 0.5153846153846153,
+      "grad_norm": 0.00233354169750256,
+      "learning_rate": 0.1,
+      "loss": 1.4447,
+      "step": 268
+    },
+    {
+      "epoch": 0.5173076923076924,
+      "grad_norm": 0.0018803124575082293,
+      "learning_rate": 0.09937667216079665,
+      "loss": 1.3468,
+      "step": 269
+    },
+    {
+      "epoch": 0.5192307692307693,
+      "grad_norm": 0.0017690274393655188,
+      "learning_rate": 0.0987533685404585,
+      "loss": 1.2466,
+      "step": 270
+    },
+    {
+      "epoch": 0.5211538461538462,
+      "grad_norm": 0.0019006796288003466,
+      "learning_rate": 0.0981301133569098,
+      "loss": 1.3386,
+      "step": 271
+    },
+    {
+      "epoch": 0.5230769230769231,
+      "grad_norm": 0.0019081005551189927,
+      "learning_rate": 0.09750693082619273,
+      "loss": 1.2631,
+      "step": 272
+    },
+    {
+      "epoch": 0.525,
+      "grad_norm": 0.00204379120753758,
+      "learning_rate": 0.0968838451615267,
+      "loss": 1.3989,
+      "step": 273
+    },
+    {
+      "epoch": 0.5269230769230769,
+      "grad_norm": 0.00206475290425934,
+      "learning_rate": 0.09626088057236745,
+      "loss": 1.293,
+      "step": 274
+    },
+    {
+      "epoch": 0.5288461538461539,
+      "grad_norm": 0.002010575368667895,
+      "learning_rate": 0.09563806126346641,
+      "loss": 1.2392,
+      "step": 275
+    },
+    {
+      "epoch": 0.5307692307692308,
+      "grad_norm": 0.0022035961007159915,
+      "learning_rate": 0.09501541143393027,
+      "loss": 1.3192,
+      "step": 276
+    },
+    {
+      "epoch": 0.5326923076923077,
+      "grad_norm": 0.0018470706531543482,
+      "learning_rate": 0.09439295527628082,
+      "loss": 1.3833,
+      "step": 277
+    },
+    {
+      "epoch": 0.5346153846153846,
+      "grad_norm": 0.0017463701086020463,
+      "learning_rate": 0.0937707169755148,
+      "loss": 1.1917,
+      "step": 278
+    },
+    {
+      "epoch": 0.5365384615384615,
+      "grad_norm": 0.002193375330313697,
+      "learning_rate": 0.09314872070816434,
+      "loss": 1.27,
+      "step": 279
+    },
+    {
+      "epoch": 0.5384615384615384,
+      "grad_norm": 0.002243701084039769,
+      "learning_rate": 0.09252699064135758,
+      "loss": 1.2442,
+      "step": 280
+    },
+    {
+      "epoch": 0.5403846153846154,
+      "grad_norm": 0.0019816609753914834,
+      "learning_rate": 0.09190555093187967,
+      "loss": 1.3468,
+      "step": 281
+    },
+    {
+      "epoch": 0.5423076923076923,
+      "grad_norm": 0.0017114626375470438,
+      "learning_rate": 0.09128442572523418,
+      "loss": 1.2072,
+      "step": 282
+    },
+    {
+      "epoch": 0.5442307692307692,
+      "grad_norm": 0.002064462344791014,
+      "learning_rate": 0.09066363915470495,
+      "loss": 1.3685,
+      "step": 283
+    },
+    {
+      "epoch": 0.5461538461538461,
+      "grad_norm": 0.0020428421215791734,
+      "learning_rate": 0.09004321534041836,
+      "loss": 1.2541,
+      "step": 284
+    },
+    {
+      "epoch": 0.5480769230769231,
+      "grad_norm": 0.0018946634070589614,
+      "learning_rate": 0.08942317838840624,
+      "loss": 1.2306,
+      "step": 285
+    },
+    {
+      "epoch": 0.55,
+      "grad_norm": 0.0020107173701820226,
+      "learning_rate": 0.08880355238966922,
+      "loss": 1.1027,
+      "step": 286
+    },
+    {
+      "epoch": 0.551923076923077,
+      "grad_norm": 0.0019791652399223088,
+      "learning_rate": 0.08818436141924073,
+      "loss": 1.345,
+      "step": 287
+    },
+    {
+      "epoch": 0.5538461538461539,
+      "grad_norm": 0.0018969296426998177,
+      "learning_rate": 0.08756562953525152,
+      "loss": 1.3871,
+      "step": 288
+    },
+    {
+      "epoch": 0.5557692307692308,
+      "grad_norm": 0.001711957049336324,
+      "learning_rate": 0.08694738077799487,
+      "loss": 1.2434,
+      "step": 289
+    },
+    {
+      "epoch": 0.5576923076923077,
+      "grad_norm": 0.0017311077865144707,
+      "learning_rate": 0.08632963916899268,
+      "loss": 1.165,
+      "step": 290
+    },
+    {
+      "epoch": 0.5596153846153846,
+      "grad_norm": 0.0019541079833854027,
+      "learning_rate": 0.08571242871006202,
+      "loss": 1.2237,
+      "step": 291
+    },
+    {
+      "epoch": 0.5615384615384615,
+      "grad_norm": 0.0018472317804026361,
+      "learning_rate": 0.08509577338238256,
+      "loss": 1.274,
+      "step": 292
+    },
+    {
+      "epoch": 0.5634615384615385,
+      "grad_norm": 0.0019215945725959064,
+      "learning_rate": 0.08447969714556484,
+      "loss": 1.2096,
+      "step": 293
+    },
+    {
+      "epoch": 0.5653846153846154,
+      "grad_norm": 0.002038965919537297,
+      "learning_rate": 0.08386422393671933,
+      "loss": 1.2394,
+      "step": 294
+    },
+    {
+      "epoch": 0.5673076923076923,
+      "grad_norm": 0.0018890268334391081,
+      "learning_rate": 0.08324937766952638,
+      "loss": 1.3154,
+      "step": 295
+    },
+    {
+      "epoch": 0.5692307692307692,
+      "grad_norm": 0.0019458661970966815,
+      "learning_rate": 0.08263518223330697,
+      "loss": 1.1855,
+      "step": 296
+    },
+    {
+      "epoch": 0.5711538461538461,
+      "grad_norm": 0.0019824805485196125,
+      "learning_rate": 0.08202166149209474,
+      "loss": 1.3105,
+      "step": 297
+    },
+    {
+      "epoch": 0.573076923076923,
+      "grad_norm": 0.0016619098723183203,
+      "learning_rate": 0.08140883928370855,
+      "loss": 1.2827,
+      "step": 298
+    },
+    {
+      "epoch": 0.575,
+      "grad_norm": 0.001794526653649084,
+      "learning_rate": 0.0807967394188264,
+      "loss": 1.332,
+      "step": 299
+    },
+    {
+      "epoch": 0.5769230769230769,
+      "grad_norm": 0.0018242602716706784,
+      "learning_rate": 0.08018538568006027,
+      "loss": 1.3348,
+      "step": 300
+    },
+    {
+      "epoch": 0.5788461538461539,
+      "grad_norm": 0.0018228642732679404,
+      "learning_rate": 0.07957480182103199,
+      "loss": 1.31,
+      "step": 301
+    },
+    {
+      "epoch": 0.5807692307692308,
+      "grad_norm": 0.0019802700350512368,
+      "learning_rate": 0.07896501156545044,
+      "loss": 1.3527,
+      "step": 302
+    },
+    {
+      "epoch": 0.5826923076923077,
+      "grad_norm": 0.002059672141024724,
+      "learning_rate": 0.07835603860618973,
+      "loss": 1.2482,
+      "step": 303
+    },
+    {
+      "epoch": 0.5846153846153846,
+      "grad_norm": 0.0019886353455798316,
+      "learning_rate": 0.07774790660436857,
+      "loss": 1.2401,
+      "step": 304
+    },
+    {
+      "epoch": 0.5865384615384616,
+      "grad_norm": 0.002078511817195078,
+      "learning_rate": 0.07714063918843106,
+      "loss": 1.3512,
+      "step": 305
+    },
+    {
+      "epoch": 0.5884615384615385,
+      "grad_norm": 0.0018290551809598812,
+      "learning_rate": 0.0765342599532285,
+      "loss": 1.2901,
+      "step": 306
+    },
+    {
+      "epoch": 0.5903846153846154,
+      "grad_norm": 0.0017447486041320863,
+      "learning_rate": 0.07592879245910272,
+      "loss": 1.2223,
+      "step": 307
+    },
+    {
+      "epoch": 0.5923076923076923,
+      "grad_norm": 0.002005421013224939,
+      "learning_rate": 0.07532426023097064,
+      "loss": 1.3429,
+      "step": 308
+    },
+    {
+      "epoch": 0.5942307692307692,
+      "grad_norm": 0.0016860171364833738,
+      "learning_rate": 0.07472068675741024,
+      "loss": 1.2219,
+      "step": 309
+    },
+    {
+      "epoch": 0.5961538461538461,
+      "grad_norm": 0.001780858683274138,
+      "learning_rate": 0.07411809548974792,
+      "loss": 1.2031,
+      "step": 310
+    },
+    {
+      "epoch": 0.5980769230769231,
+      "grad_norm": 0.001805653983836877,
+      "learning_rate": 0.07351650984114727,
+      "loss": 1.1765,
+      "step": 311
+    },
+    {
+      "epoch": 0.6,
+      "grad_norm": 0.0019263627769412846,
+      "learning_rate": 0.0729159531856995,
+      "loss": 1.1676,
+      "step": 312
+    },
+    {
+      "epoch": 0.6019230769230769,
+      "grad_norm": 0.001613950954976428,
+      "learning_rate": 0.07231644885751508,
+      "loss": 1.1571,
+      "step": 313
+    },
+    {
+      "epoch": 0.6038461538461538,
+      "grad_norm": 0.0016638551507284869,
+      "learning_rate": 0.07171802014981725,
+      "loss": 1.1916,
+      "step": 314
+    },
+    {
+      "epoch": 0.6057692307692307,
+      "grad_norm": 0.0021720887839922517,
+      "learning_rate": 0.07112069031403703,
+      "loss": 1.3055,
+      "step": 315
+    },
+    {
+      "epoch": 0.6076923076923076,
+      "grad_norm": 0.0021838519441203354,
+      "learning_rate": 0.07052448255890957,
+      "loss": 1.1638,
+      "step": 316
+    },
+    {
+      "epoch": 0.6096153846153847,
+      "grad_norm": 0.001616551293575509,
+      "learning_rate": 0.0699294200495727,
+      "loss": 1.1897,
+      "step": 317
+    },
+    {
+      "epoch": 0.6115384615384616,
+      "grad_norm": 0.0019850405682130617,
+      "learning_rate": 0.06933552590666658,
+      "loss": 1.3088,
+      "step": 318
+    },
+    {
+      "epoch": 0.6134615384615385,
+      "grad_norm": 0.0018246971951640582,
+      "learning_rate": 0.06874282320543557,
+      "loss": 1.1754,
+      "step": 319
+    },
+    {
+      "epoch": 0.6153846153846154,
+      "grad_norm": 0.0019097801002015357,
+      "learning_rate": 0.06815133497483157,
+      "loss": 1.1139,
+      "step": 320
+    },
+    {
+      "epoch": 0.6173076923076923,
+      "grad_norm": 0.0019592265259293835,
+      "learning_rate": 0.06756108419661931,
+      "loss": 1.3208,
+      "step": 321
+    },
+    {
+      "epoch": 0.6192307692307693,
+      "grad_norm": 0.0018436900726176199,
+      "learning_rate": 0.06697209380448332,
+      "loss": 1.1823,
+      "step": 322
+    },
+    {
+      "epoch": 0.6211538461538462,
+      "grad_norm": 0.001966987155336665,
+      "learning_rate": 0.06638438668313694,
+      "loss": 1.2547,
+      "step": 323
+    },
+    {
+      "epoch": 0.6230769230769231,
+      "grad_norm": 0.0019281825692393753,
+      "learning_rate": 0.06579798566743314,
+      "loss": 1.2522,
+      "step": 324
+    },
+    {
+      "epoch": 0.625,
+      "grad_norm": 0.0019281891678048927,
+      "learning_rate": 0.06521291354147728,
+      "loss": 1.2601,
+      "step": 325
+    },
+    {
+      "epoch": 0.6269230769230769,
+      "grad_norm": 0.001819737721032606,
+      "learning_rate": 0.06462919303774187,
+      "loss": 1.2513,
+      "step": 326
+    },
+    {
+      "epoch": 0.6288461538461538,
+      "grad_norm": 0.0020296586746981074,
+      "learning_rate": 0.06404684683618325,
+      "loss": 1.3175,
+      "step": 327
+    },
+    {
+      "epoch": 0.6307692307692307,
+      "grad_norm": 0.0019210794581622944,
+      "learning_rate": 0.0634658975633605,
+      "loss": 1.3086,
+      "step": 328
+    },
+    {
+      "epoch": 0.6326923076923077,
+      "grad_norm": 0.001630870184872911,
+      "learning_rate": 0.06288636779155621,
+      "loss": 1.1678,
+      "step": 329
+    },
+    {
+      "epoch": 0.6346153846153846,
+      "grad_norm": 0.0016676095451111728,
+      "learning_rate": 0.06230828003789948,
+      "loss": 1.2459,
+      "step": 330
+    },
+    {
+      "epoch": 0.6365384615384615,
+      "grad_norm": 0.0017345291235235463,
+      "learning_rate": 0.06173165676349103,
+      "loss": 1.2016,
+      "step": 331
+    },
+    {
+      "epoch": 0.6384615384615384,
+      "grad_norm": 0.0017118487353533444,
+      "learning_rate": 0.06115652037253053,
+      "loss": 1.3306,
+      "step": 332
+    },
+    {
+      "epoch": 0.6403846153846153,
+      "grad_norm": 0.001953386046224116,
+      "learning_rate": 0.06058289321144608,
+      "loss": 1.3562,
+      "step": 333
+    },
+    {
+      "epoch": 0.6423076923076924,
+      "grad_norm": 0.002009867700523358,
+      "learning_rate": 0.06001079756802592,
+      "loss": 1.2531,
+      "step": 334
+    },
+    {
+      "epoch": 0.6442307692307693,
+      "grad_norm": 0.0016635702832438184,
+      "learning_rate": 0.059440255670552514,
+      "loss": 1.2463,
+      "step": 335
+    },
+    {
+      "epoch": 0.6461538461538462,
+      "grad_norm": 0.0019918449403549194,
+      "learning_rate": 0.05887128968693887,
+      "loss": 1.1359,
+      "step": 336
+    },
+    {
+      "epoch": 0.6480769230769231,
+      "grad_norm": 0.0018098607239414595,
+      "learning_rate": 0.058303921723867225,
+      "loss": 1.1363,
+      "step": 337
+    },
+    {
+      "epoch": 0.65,
+      "grad_norm": 0.0018005672587287607,
+      "learning_rate": 0.05773817382593008,
+      "loss": 1.2581,
+      "step": 338
+    },
+    {
+      "epoch": 0.6519230769230769,
+      "grad_norm": 0.001760807841509949,
+      "learning_rate": 0.057174067974773715,
+      "loss": 1.2011,
+      "step": 339
+    },
+    {
+      "epoch": 0.6538461538461539,
+      "grad_norm": 0.0018042743597355261,
+      "learning_rate": 0.056611626088244195,
+      "loss": 1.1965,
+      "step": 340
+    },
+    {
+      "epoch": 0.6557692307692308,
+      "grad_norm": 0.0018609654932741188,
+      "learning_rate": 0.056050870019535494,
+      "loss": 1.2115,
+      "step": 341
+    },
+    {
+      "epoch": 0.6576923076923077,
+      "grad_norm": 0.002283227636184851,
+      "learning_rate": 0.05549182155634076,
+      "loss": 1.2981,
+      "step": 342
+    },
+    {
+      "epoch": 0.6596153846153846,
+      "grad_norm": 0.0017874883084867455,
+      "learning_rate": 0.054934502420005464,
+      "loss": 1.2522,
+      "step": 343
+    },
+    {
+      "epoch": 0.6615384615384615,
+      "grad_norm": 0.0017884437432287657,
+      "learning_rate": 0.0543789342646837,
+      "loss": 1.1655,
+      "step": 344
+    },
+    {
+      "epoch": 0.6634615384615384,
+      "grad_norm": 0.002112235705692318,
+      "learning_rate": 0.05382513867649663,
+      "loss": 1.2856,
+      "step": 345
+    },
+    {
+      "epoch": 0.6653846153846154,
+      "grad_norm": 0.0017299830338927645,
+      "learning_rate": 0.0532731371726938,
+      "loss": 1.2587,
+      "step": 346
+    },
+    {
+      "epoch": 0.6673076923076923,
+      "grad_norm": 0.0017301923300043838,
+      "learning_rate": 0.05272295120081732,
+      "loss": 1.1801,
+      "step": 347
+    },
+    {
+      "epoch": 0.6692307692307692,
+      "grad_norm": 0.002078244321857179,
+      "learning_rate": 0.05217460213786821,
+      "loss": 1.1399,
+      "step": 348
+    },
+    {
+      "epoch": 0.6711538461538461,
+      "grad_norm": 0.001819076302669517,
+      "learning_rate": 0.051628111289476024,
+      "loss": 1.1803,
+      "step": 349
+    },
+    {
+      "epoch": 0.6730769230769231,
+      "grad_norm": 0.0018474822017365646,
+      "learning_rate": 0.051083499889071106,
+      "loss": 1.223,
+      "step": 350
+    },
+    {
+      "epoch": 0.675,
+      "grad_norm": 0.001635472962910989,
+      "learning_rate": 0.05054078909705926,
+      "loss": 1.1317,
+      "step": 351
+    },
+    {
+      "epoch": 0.676923076923077,
+      "grad_norm": 0.0018090953966598264,
+      "learning_rate": 0.050000000000000024,
+      "loss": 1.259,
+      "step": 352
+    },
+    {
+      "epoch": 0.6788461538461539,
+      "grad_norm": 0.0016349635485458152,
+      "learning_rate": 0.04946115360978696,
+      "loss": 1.2064,
+      "step": 353
+    },
+    {
+      "epoch": 0.6807692307692308,
+      "grad_norm": 0.0017183816004449682,
+      "learning_rate": 0.048924270862831465,
+      "loss": 1.3352,
+      "step": 354
+    },
+    {
+      "epoch": 0.6826923076923077,
+      "grad_norm": 0.001798620398606143,
+      "learning_rate": 0.04838937261924933,
+      "loss": 1.1908,
+      "step": 355
+    },
+    {
+      "epoch": 0.6846153846153846,
+      "grad_norm": 0.0018477199168563687,
+      "learning_rate": 0.0478564796620502,
+      "loss": 1.1878,
+      "step": 356
+    },
+    {
+      "epoch": 0.6865384615384615,
+      "grad_norm": 0.0016855145400003111,
+      "learning_rate": 0.04732561269632993,
+      "loss": 1.215,
+      "step": 357
+    },
+    {
+      "epoch": 0.6884615384615385,
+      "grad_norm": 0.0017207631857037331,
+      "learning_rate": 0.04679679234846636,
+      "loss": 1.1509,
+      "step": 358
+    },
+    {
+      "epoch": 0.6903846153846154,
+      "grad_norm": 0.0018764231778032876,
+      "learning_rate": 0.046270039165317606,
+      "loss": 1.2652,
+      "step": 359
+    },
+    {
+      "epoch": 0.6923076923076923,
+      "grad_norm": 0.0018778996937584573,
+      "learning_rate": 0.04574537361342407,
+      "loss": 1.2769,
+      "step": 360
+    },
+    {
+      "epoch": 0.6942307692307692,
+      "grad_norm": 0.0016570901921947258,
+      "learning_rate": 0.04522281607821288,
+      "loss": 1.2777,
+      "step": 361
+    },
+    {
+      "epoch": 0.6961538461538461,
+      "grad_norm": 0.001935004899464571,
+      "learning_rate": 0.04470238686320606,
+      "loss": 1.2109,
+      "step": 362
+    },
+    {
+      "epoch": 0.698076923076923,
+      "grad_norm": 0.0017224941698611633,
+      "learning_rate": 0.044184106189231624,
+      "loss": 1.2279,
+      "step": 363
+    },
+    {
+      "epoch": 0.7,
+      "grad_norm": 0.0018366090049641755,
+      "learning_rate": 0.043667994193637795,
+      "loss": 1.3007,
+      "step": 364
+    },
+    {
+      "epoch": 0.7019230769230769,
+      "grad_norm": 0.0018618381727355056,
+      "learning_rate": 0.043154070929510784,
+      "loss": 1.2978,
+      "step": 365
+    },
+    {
+      "epoch": 0.7038461538461539,
+      "grad_norm": 0.001704666457247974,
+      "learning_rate": 0.04264235636489542,
+      "loss": 1.2502,
+      "step": 366
+    },
+    {
+      "epoch": 0.7057692307692308,
+      "grad_norm": 0.0017691309472415667,
+      "learning_rate": 0.04213287038201943,
+      "loss": 1.2473,
+      "step": 367
+    },
+    {
+      "epoch": 0.7076923076923077,
+      "grad_norm": 0.0020169783477486263,
+      "learning_rate": 0.04162563277652104,
+      "loss": 1.1,
+      "step": 368
+    },
+    {
+      "epoch": 0.7096153846153846,
+      "grad_norm": 0.0016465782686624223,
+      "learning_rate": 0.04112066325667954,
+      "loss": 1.2494,
+      "step": 369
+    },
+    {
+      "epoch": 0.7115384615384616,
+      "grad_norm": 0.0016466753588309927,
+      "learning_rate": 0.04061798144264986,
+      "loss": 1.159,
+      "step": 370
+    },
+    {
+      "epoch": 0.7134615384615385,
+      "grad_norm": 0.0018442639237441385,
+      "learning_rate": 0.04011760686569998,
+      "loss": 1.1582,
+      "step": 371
+    },
+    {
+      "epoch": 0.7153846153846154,
+      "grad_norm": 0.0016184167045702805,
+      "learning_rate": 0.03961955896745224,
+      "loss": 1.3398,
+      "step": 372
+    },
+    {
+      "epoch": 0.7173076923076923,
+      "grad_norm": 0.001866565537744226,
+      "learning_rate": 0.03912385709912794,
+      "loss": 1.2166,
+      "step": 373
+    },
+    {
+      "epoch": 0.7192307692307692,
+      "grad_norm": 0.0017138544993580834,
+      "learning_rate": 0.038630520520795276,
+      "loss": 1.2397,
+      "step": 374
+    },
+    {
+      "epoch": 0.7211538461538461,
+      "grad_norm": 0.0018877387906157657,
+      "learning_rate": 0.03813956840062119,
+      "loss": 1.1556,
+      "step": 375
+    },
+    {
+      "epoch": 0.7230769230769231,
+      "grad_norm": 0.0016807149671783103,
+      "learning_rate": 0.037651019814126656,
+      "loss": 1.2765,
+      "step": 376
+    },
+    {
+      "epoch": 0.725,
+      "grad_norm": 0.001789690612811793,
+      "learning_rate": 0.037164893743445275,
+      "loss": 1.2109,
+      "step": 377
+    },
+    {
+      "epoch": 0.7269230769230769,
+      "grad_norm": 0.0016959501472444875,
+      "learning_rate": 0.03668120907658603,
+      "loss": 1.2622,
+      "step": 378
+    },
+    {
+      "epoch": 0.7288461538461538,
+      "grad_norm": 0.0016537034459015076,
+      "learning_rate": 0.036199984606699154,
+      "loss": 1.2495,
+      "step": 379
+    },
+    {
+      "epoch": 0.7307692307692307,
+      "grad_norm": 0.0022217377353371355,
+      "learning_rate": 0.035721239031346066,
+      "loss": 1.3138,
+      "step": 380
+    },
+    {
+      "epoch": 0.7326923076923076,
+      "grad_norm": 0.0017349340573131204,
+      "learning_rate": 0.03524499095177297,
+      "loss": 1.244,
+      "step": 381
+    },
+    {
+      "epoch": 0.7346153846153847,
+      "grad_norm": 0.0017617940318522952,
+      "learning_rate": 0.03477125887218792,
+      "loss": 1.2669,
+      "step": 382
+    },
+    {
+      "epoch": 0.7365384615384616,
+      "grad_norm": 0.001829467828681564,
+      "learning_rate": 0.03430006119904196,
+      "loss": 1.0813,
+      "step": 383
+    },
+    {
+      "epoch": 0.7384615384615385,
+      "grad_norm": 0.0016549794339755684,
+      "learning_rate": 0.033831416240314084,
+      "loss": 1.336,
+      "step": 384
+    },
+    {
+      "epoch": 0.7403846153846154,
+      "grad_norm": 0.0016428396098067798,
+      "learning_rate": 0.03336534220479961,
+      "loss": 1.2222,
+      "step": 385
+    },
+    {
+      "epoch": 0.7423076923076923,
+      "grad_norm": 0.001505151130014924,
+      "learning_rate": 0.032901857201403005,
+      "loss": 1.1703,
+      "step": 386
+    },
+    {
+      "epoch": 0.7442307692307693,
+      "grad_norm": 0.0017606806783378218,
+      "learning_rate": 0.032440979238433976,
+      "loss": 1.3362,
+      "step": 387
+    },
+    {
+      "epoch": 0.7461538461538462,
+      "grad_norm": 0.001689001789541186,
+      "learning_rate": 0.03198272622290804,
+      "loss": 1.1182,
+      "step": 388
+    },
+    {
+      "epoch": 0.7480769230769231,
+      "grad_norm": 0.0022641376726381564,
+      "learning_rate": 0.03152711595985065,
+      "loss": 1.1759,
+      "step": 389
+    },
+    {
+      "epoch": 0.75,
+      "grad_norm": 0.0016879082512593958,
+      "learning_rate": 0.031074166151605298,
+      "loss": 1.2427,
+      "step": 390
+    },
+    {
+      "epoch": 0.7519230769230769,
+      "grad_norm": 0.001835907655455613,
+      "learning_rate": 0.030623894397145836,
+      "loss": 1.3253,
+      "step": 391
+    },
+    {
+      "epoch": 0.7538461538461538,
+      "grad_norm": 0.0016608930131525894,
+      "learning_rate": 0.03017631819139273,
+      "loss": 1.1268,
+      "step": 392
+    },
+    {
+      "epoch": 0.7557692307692307,
+      "grad_norm": 0.0015328183570992634,
+      "learning_rate": 0.029731454924533086,
+      "loss": 1.1542,
+      "step": 393
+    },
+    {
+      "epoch": 0.7576923076923077,
+      "grad_norm": 0.0019159334362267488,
+      "learning_rate": 0.029289321881345254,
+      "loss": 1.1942,
+      "step": 394
+    },
+    {
+      "epoch": 0.7596153846153846,
+      "grad_norm": 0.0019618949795048636,
+      "learning_rate": 0.028849936240527008,
+      "loss": 1.1599,
+      "step": 395
+    },
+    {
+      "epoch": 0.7615384615384615,
+      "grad_norm": 0.0018705845534838588,
+      "learning_rate": 0.028413315074028157,
+      "loss": 1.2409,
+      "step": 396
+    },
+    {
+      "epoch": 0.7634615384615384,
+      "grad_norm": 0.0016627265978662253,
+      "learning_rate": 0.027979475346387363,
+      "loss": 1.2278,
+      "step": 397
+    },
+    {
+      "epoch": 0.7653846153846153,
+      "grad_norm": 0.001786315940263252,
+      "learning_rate": 0.027548433914072735,
+      "loss": 1.2264,
+      "step": 398
+    },
+    {
+      "epoch": 0.7673076923076924,
+      "grad_norm": 0.0016971904439370942,
+      "learning_rate": 0.027120207524827168,
+      "loss": 1.2044,
+      "step": 399
+    },
+    {
+      "epoch": 0.7692307692307693,
+      "grad_norm": 0.0016412591454362708,
+      "learning_rate": 0.02669481281701739,
+      "loss": 1.2374,
+      "step": 400
+    },
+    {
+      "epoch": 0.7711538461538462,
+      "grad_norm": 0.001998912056969298,
+      "learning_rate": 0.026272266318987603,
+      "loss": 1.0517,
+      "step": 401
+    },
+    {
+      "epoch": 0.7730769230769231,
+      "grad_norm": 0.0018708737561761153,
+      "learning_rate": 0.02585258444841733,
+      "loss": 1.1721,
+      "step": 402
+    },
+    {
+      "epoch": 0.775,
+      "grad_norm": 0.001967393594936681,
+      "learning_rate": 0.025435783511683442,
+      "loss": 1.1993,
+      "step": 403
+    },
+    {
+      "epoch": 0.7769230769230769,
+      "grad_norm": 0.0021466488393323854,
+      "learning_rate": 0.02502187970322657,
+      "loss": 1.0999,
+      "step": 404
+    },
+    {
+      "epoch": 0.7788461538461539,
+      "grad_norm": 0.0017458705551823743,
+      "learning_rate": 0.02461088910492202,
+      "loss": 1.208,
+      "step": 405
+    },
+    {
+      "epoch": 0.7807692307692308,
+      "grad_norm": 0.0020247254158974764,
+      "learning_rate": 0.02420282768545469,
+      "loss": 1.1349,
+      "step": 406
+    },
+    {
+      "epoch": 0.7826923076923077,
+      "grad_norm": 0.001764718465751009,
+      "learning_rate": 0.02379771129969892,
+      "loss": 1.2917,
+      "step": 407
+    },
+    {
+      "epoch": 0.7846153846153846,
+      "grad_norm": 0.0018592788882771515,
+      "learning_rate": 0.023395555688102213,
+      "loss": 1.1823,
+      "step": 408
+    },
+    {
+      "epoch": 0.7865384615384615,
+      "grad_norm": 0.001948383781083682,
+      "learning_rate": 0.02299637647607372,
+      "loss": 1.3108,
+      "step": 409
+    },
+    {
+      "epoch": 0.7884615384615384,
+      "grad_norm": 0.001740883243036254,
+      "learning_rate": 0.022600189173377264,
+      "loss": 1.0318,
+      "step": 410
+    },
+    {
+      "epoch": 0.7903846153846154,
+      "grad_norm": 0.00198065059794587,
+      "learning_rate": 0.022207009173528525,
+      "loss": 1.2828,
+      "step": 411
+    },
+    {
+      "epoch": 0.7923076923076923,
+      "grad_norm": 0.0017443778028226392,
+      "learning_rate": 0.02181685175319702,
+      "loss": 1.1937,
+      "step": 412
+    },
+    {
+      "epoch": 0.7942307692307692,
+      "grad_norm": 0.0017615300838455196,
+      "learning_rate": 0.021429732071612653,
+      "loss": 1.2324,
+      "step": 413
+    },
+    {
+      "epoch": 0.7961538461538461,
+      "grad_norm": 0.001475900933277607,
+      "learning_rate": 0.02104566516997647,
+      "loss": 1.0329,
+      "step": 414
+    },
+    {
+      "epoch": 0.7980769230769231,
+      "grad_norm": 0.0016929700527951055,
+      "learning_rate": 0.020664665970876496,
+      "loss": 1.1704,
+      "step": 415
+    },
+    {
+      "epoch": 0.8,
+      "grad_norm": 0.0019169580534935764,
+      "learning_rate": 0.020286749277707784,
+      "loss": 1.0929,
+      "step": 416
+    },
+    {
+      "epoch": 0.801923076923077,
+      "grad_norm": 0.0019255807933238338,
+      "learning_rate": 0.019911929774097215,
+      "loss": 1.2539,
+      "step": 417
+    },
+    {
+      "epoch": 0.8038461538461539,
+      "grad_norm": 0.001717306792967696,
+      "learning_rate": 0.019540222023333165,
+      "loss": 1.2418,
+      "step": 418
+    },
+    {
+      "epoch": 0.8057692307692308,
+      "grad_norm": 0.0018664904203607685,
+      "learning_rate": 0.01917164046779948,
+      "loss": 1.2294,
+      "step": 419
+    },
+    {
+      "epoch": 0.8076923076923077,
+      "grad_norm": 0.0019455344506997781,
+      "learning_rate": 0.018806199428414352,
+      "loss": 1.118,
+      "step": 420
+    },
+    {
+      "epoch": 0.8096153846153846,
+      "grad_norm": 0.00212775616235756,
+      "learning_rate": 0.018443913104073985,
+      "loss": 1.0491,
+      "step": 421
+    },
+    {
+      "epoch": 0.8115384615384615,
+      "grad_norm": 0.001795162751958178,
+      "learning_rate": 0.01808479557110081,
+      "loss": 1.1679,
+      "step": 422
+    },
+    {
+      "epoch": 0.8134615384615385,
+      "grad_norm": 0.001967167274699136,
+      "learning_rate": 0.017728860782696667,
+      "loss": 1.1605,
+      "step": 423
+    },
+    {
+      "epoch": 0.8153846153846154,
+      "grad_norm": 0.0018606482549505577,
+      "learning_rate": 0.017376122568400532,
+      "loss": 1.3145,
+      "step": 424
+    },
+    {
+      "epoch": 0.8173076923076923,
+      "grad_norm": 0.001712339863591348,
+      "learning_rate": 0.017026594633551252,
+      "loss": 1.1697,
+      "step": 425
+    },
+    {
+      "epoch": 0.8192307692307692,
+      "grad_norm": 0.0023220033480005485,
+      "learning_rate": 0.01668029055875512,
+      "loss": 1.1905,
+      "step": 426
+    },
+    {
+      "epoch": 0.8211538461538461,
+      "grad_norm": 0.0016683908997586607,
+      "learning_rate": 0.016337223799358026,
+      "loss": 1.1042,
+      "step": 427
+    },
+    {
+      "epoch": 0.823076923076923,
+      "grad_norm": 0.001801188929310371,
+      "learning_rate": 0.01599740768492286,
+      "loss": 1.0802,
+      "step": 428
+    },
+    {
+      "epoch": 0.825,
+      "grad_norm": 0.0017247396248211374,
+      "learning_rate": 0.015660855418711452,
+      "loss": 1.1782,
+      "step": 429
+    },
+    {
+      "epoch": 0.8269230769230769,
+      "grad_norm": 0.0016312624008005092,
+      "learning_rate": 0.015327580077171589,
+      "loss": 1.1757,
+      "step": 430
+    },
+    {
+      "epoch": 0.8288461538461539,
+      "grad_norm": 0.001799990595192518,
+      "learning_rate": 0.014997594609429088,
+      "loss": 1.1985,
+      "step": 431
+    },
+    {
+      "epoch": 0.8307692307692308,
+      "grad_norm": 0.0018870097215196863,
+      "learning_rate": 0.01467091183678444,
+      "loss": 1.088,
+      "step": 432
+    },
+    {
+      "epoch": 0.8326923076923077,
+      "grad_norm": 0.0016849592372375636,
+      "learning_rate": 0.014347544452214867,
+      "loss": 1.2225,
+      "step": 433
+    },
+    {
+      "epoch": 0.8346153846153846,
+      "grad_norm": 0.0017971972226316119,
+      "learning_rate": 0.014027505019880971,
+      "loss": 0.9647,
+      "step": 434
+    },
+    {
+      "epoch": 0.8365384615384616,
+      "grad_norm": 0.0019679457495197167,
+      "learning_rate": 0.013710805974638696,
+      "loss": 1.2663,
+      "step": 435
+    },
+    {
+      "epoch": 0.8384615384615385,
+      "grad_norm": 0.001768346873979825,
+      "learning_rate": 0.01339745962155613,
+      "loss": 1.0496,
+      "step": 436
+    },
+    {
+      "epoch": 0.8403846153846154,
+      "grad_norm": 0.0017408512288308725,
+      "learning_rate": 0.01308747813543536,
+      "loss": 1.2824,
+      "step": 437
+    },
+    {
+      "epoch": 0.8423076923076923,
+      "grad_norm": 0.001757522677179555,
+      "learning_rate": 0.012780873560339467,
+      "loss": 1.0906,
+      "step": 438
+    },
+    {
+      "epoch": 0.8442307692307692,
+      "grad_norm": 0.001515942611306163,
+      "learning_rate": 0.012477657809124632,
+      "loss": 1.1742,
+      "step": 439
+    },
+    {
+      "epoch": 0.8461538461538461,
+      "grad_norm": 0.001731896305786343,
+      "learning_rate": 0.012177842662977134,
+      "loss": 1.1424,
+      "step": 440
+    },
+    {
+      "epoch": 0.8480769230769231,
+      "grad_norm": 0.001882482079025467,
+      "learning_rate": 0.01188143977095576,
+      "loss": 1.2013,
+      "step": 441
+    },
+    {
+      "epoch": 0.85,
+      "grad_norm": 0.0019180494940922055,
+      "learning_rate": 0.011588460649539035,
+      "loss": 1.1984,
+      "step": 442
+    },
+    {
+      "epoch": 0.8519230769230769,
+      "grad_norm": 0.001765296275351436,
+      "learning_rate": 0.011298916682177829,
+      "loss": 1.2107,
+      "step": 443
+    },
+    {
+      "epoch": 0.8538461538461538,
+      "grad_norm": 0.0016177859402696205,
+      "learning_rate": 0.011012819118853146,
+      "loss": 1.1791,
+      "step": 444
+    },
+    {
+      "epoch": 0.8557692307692307,
+      "grad_norm": 0.0017408419897363358,
+      "learning_rate": 0.01073017907563887,
+      "loss": 1.1023,
+      "step": 445
+    },
+    {
+      "epoch": 0.8576923076923076,
+      "grad_norm": 0.0016577197965964652,
+      "learning_rate": 0.010451007534269908,
+      "loss": 1.2727,
+      "step": 446
+    },
+    {
+      "epoch": 0.8596153846153847,
+      "grad_norm": 0.0017355497740829744,
+      "learning_rate": 0.010175315341715598,
+      "loss": 1.1902,
+      "step": 447
+    },
+    {
+      "epoch": 0.8615384615384616,
+      "grad_norm": 0.0017627058787641511,
+      "learning_rate": 0.009903113209758098,
+      "loss": 1.1718,
+      "step": 448
+    },
+    {
+      "epoch": 0.8634615384615385,
+      "grad_norm": 0.0019513860897464456,
+      "learning_rate": 0.009634411714576352,
+      "loss": 1.2255,
+      "step": 449
+    },
+    {
+      "epoch": 0.8653846153846154,
+      "grad_norm": 0.00182481702521829,
+      "learning_rate": 0.009369221296335007,
+      "loss": 1.2085,
+      "step": 450
+    },
+    {
+      "epoch": 0.8673076923076923,
+      "grad_norm": 0.0017495315501774742,
+      "learning_rate": 0.009107552258778906,
+      "loss": 1.1972,
+      "step": 451
+    },
+    {
+      "epoch": 0.8692307692307693,
+      "grad_norm": 0.001647513771091212,
+      "learning_rate": 0.008849414768832687,
+      "loss": 1.268,
+      "step": 452
+    },
+    {
+      "epoch": 0.8711538461538462,
+      "grad_norm": 0.0017104854479968248,
+      "learning_rate": 0.008594818856205699,
+      "loss": 1.2416,
+      "step": 453
+    },
+    {
+      "epoch": 0.8730769230769231,
+      "grad_norm": 0.001957419495187883,
+      "learning_rate": 0.00834377441300238,
+      "loss": 1.1137,
+      "step": 454
+    },
+    {
+      "epoch": 0.875,
+      "grad_norm": 0.001739358641367216,
+      "learning_rate": 0.008096291193337934,
+      "loss": 1.2508,
+      "step": 455
+    },
+    {
+      "epoch": 0.8769230769230769,
+      "grad_norm": 0.0018005707016834295,
+      "learning_rate": 0.007852378812959226,
+      "loss": 1.1673,
+      "step": 456
+    },
+    {
+      "epoch": 0.8788461538461538,
+      "grad_norm": 0.001583138232470233,
+      "learning_rate": 0.007612046748871327,
+      "loss": 1.1762,
+      "step": 457
+    },
+    {
+      "epoch": 0.8807692307692307,
+      "grad_norm": 0.0018744662452583954,
+      "learning_rate": 0.007375304338969136,
+      "loss": 1.3097,
+      "step": 458
+    },
+    {
+      "epoch": 0.8826923076923077,
+      "grad_norm": 0.0017919705147192092,
+      "learning_rate": 0.007142160781674645,
+      "loss": 1.2419,
+      "step": 459
+    },
+    {
+      "epoch": 0.8846153846153846,
+      "grad_norm": 0.0017174421035306493,
+      "learning_rate": 0.006912625135579587,
+      "loss": 1.1175,
+      "step": 460
+    },
+    {
+      "epoch": 0.8865384615384615,
+      "grad_norm": 0.0014327536558487097,
+      "learning_rate": 0.0066867063190933496,
+      "loss": 1.2626,
+      "step": 461
+    },
+    {
+      "epoch": 0.8884615384615384,
+      "grad_norm": 0.0017051346513362966,
+      "learning_rate": 0.006464413110096601,
+      "loss": 1.3176,
+      "step": 462
+    },
+    {
+      "epoch": 0.8903846153846153,
+      "grad_norm": 0.0018923442926620394,
+      "learning_rate": 0.006245754145600091,
+      "loss": 1.0832,
+      "step": 463
+    },
+    {
+      "epoch": 0.8923076923076924,
+      "grad_norm": 0.0018384472554677986,
+      "learning_rate": 0.006030737921409169,
+      "loss": 1.2238,
+      "step": 464
+    },
+    {
+      "epoch": 0.8942307692307693,
+      "grad_norm": 0.0019190275613438377,
+      "learning_rate": 0.005819372791793654,
+      "loss": 1.332,
+      "step": 465
+    },
+    {
+      "epoch": 0.8961538461538462,
+      "grad_norm": 0.001611357356901353,
+      "learning_rate": 0.005611666969163243,
+      "loss": 1.2052,
+      "step": 466
+    },
+    {
+      "epoch": 0.8980769230769231,
+      "grad_norm": 0.0016195340665133688,
+      "learning_rate": 0.005407628523748398,
+      "loss": 1.2023,
+      "step": 467
+    },
+    {
+      "epoch": 0.9,
+      "grad_norm": 0.0020087591358493805,
+      "learning_rate": 0.00520726538328683,
+      "loss": 1.187,
+      "step": 468
+    },
+    {
+      "epoch": 0.9019230769230769,
+      "grad_norm": 0.0019331168314998862,
+      "learning_rate": 0.005010585332715401,
+      "loss": 1.2417,
+      "step": 469
+    },
+    {
+      "epoch": 0.9038461538461539,
+      "grad_norm": 0.001562529546586759,
+      "learning_rate": 0.004817596013867765,
+      "loss": 1.1179,
+      "step": 470
+    },
+    {
+      "epoch": 0.9057692307692308,
+      "grad_norm": 0.0018078560030878644,
+      "learning_rate": 0.004628304925177318,
+      "loss": 1.1437,
+      "step": 471
+    },
+    {
+      "epoch": 0.9076923076923077,
+      "grad_norm": 0.001769038830552537,
+      "learning_rate": 0.004442719421385921,
+      "loss": 1.1135,
+      "step": 472
+    },
+    {
+      "epoch": 0.9096153846153846,
+      "grad_norm": 0.001781591959468375,
+      "learning_rate": 0.004260846713258193,
+      "loss": 1.1721,
+      "step": 473
+    },
+    {
+      "epoch": 0.9115384615384615,
+      "grad_norm": 0.0016726117584839285,
+      "learning_rate": 0.004082693867301224,
+      "loss": 1.2359,
+      "step": 474
+    },
+    {
+      "epoch": 0.9134615384615384,
+      "grad_norm": 0.0016929753144215962,
+      "learning_rate": 0.003908267805490051,
+      "loss": 1.1567,
+      "step": 475
+    },
+    {
+      "epoch": 0.9153846153846154,
+      "grad_norm": 0.001782319493005969,
+      "learning_rate": 0.003737575304998797,
+      "loss": 1.1686,
+      "step": 476
+    },
+    {
+      "epoch": 0.9173076923076923,
+      "grad_norm": 0.001926333632861426,
+      "learning_rate": 0.003570622997937234,
+      "loss": 1.1528,
+      "step": 477
+    },
+    {
+      "epoch": 0.9192307692307692,
+      "grad_norm": 0.0018057780362503032,
+      "learning_rate": 0.00340741737109318,
+      "loss": 1.114,
+      "step": 478
+    },
+    {
+      "epoch": 0.9211538461538461,
+      "grad_norm": 0.0017963522162677652,
+      "learning_rate": 0.003247964765680389,
+      "loss": 1.2001,
+      "step": 479
+    },
+    {
+      "epoch": 0.9230769230769231,
+      "grad_norm": 0.001641454286539415,
+      "learning_rate": 0.0030922713770922153,
+      "loss": 1.2256,
+      "step": 480
+    },
+    {
+      "epoch": 0.925,
+      "grad_norm": 0.0016080278708384774,
+      "learning_rate": 0.0029403432546609046,
+      "loss": 1.2249,
+      "step": 481
+    },
+    {
+      "epoch": 0.926923076923077,
+      "grad_norm": 0.0021218314518878854,
+      "learning_rate": 0.0027921863014225504,
+      "loss": 1.2412,
+      "step": 482
+    },
+    {
+      "epoch": 0.9288461538461539,
+      "grad_norm": 0.0018600584214373175,
+      "learning_rate": 0.002647806273887665,
+      "loss": 1.1785,
+      "step": 483
+    },
+    {
+      "epoch": 0.9307692307692308,
+      "grad_norm": 0.0018368986637581934,
+      "learning_rate": 0.0025072087818176383,
+      "loss": 1.1841,
+      "step": 484
+    },
+    {
+      "epoch": 0.9326923076923077,
+      "grad_norm": 0.0016634703686010202,
+      "learning_rate": 0.002370399288006664,
+      "loss": 1.1317,
+      "step": 485
+    },
+    {
+      "epoch": 0.9346153846153846,
+      "grad_norm": 0.001809654185025008,
+      "learning_rate": 0.0022373831080695463,
+      "loss": 1.2576,
+      "step": 486
+    },
+    {
+      "epoch": 0.9365384615384615,
+      "grad_norm": 0.0016445584482093626,
+      "learning_rate": 0.0021081654102351635,
+      "loss": 1.1071,
+      "step": 487
+    },
+    {
+      "epoch": 0.9384615384615385,
+      "grad_norm": 0.0017502034377247955,
+      "learning_rate": 0.0019827512151456175,
+      "loss": 1.0504,
+      "step": 488
+    },
+    {
+      "epoch": 0.9403846153846154,
+      "grad_norm": 0.0015087597176024238,
+      "learning_rate": 0.0018611453956612345,
+      "loss": 1.2307,
+      "step": 489
+    },
+    {
+      "epoch": 0.9423076923076923,
+      "grad_norm": 0.0017733749387127936,
+      "learning_rate": 0.0017433526766711727,
+      "loss": 1.1757,
+      "step": 490
+    },
+    {
+      "epoch": 0.9442307692307692,
+      "grad_norm": 0.001827726825514985,
+      "learning_rate": 0.0016293776349098677,
+      "loss": 1.1349,
+      "step": 491
+    },
+    {
+      "epoch": 0.9461538461538461,
+      "grad_norm": 0.0018049379400307702,
+      "learning_rate": 0.0015192246987791981,
+      "loss": 1.2541,
+      "step": 492
+    },
+    {
+      "epoch": 0.948076923076923,
+      "grad_norm": 0.0018586073138955074,
+      "learning_rate": 0.0014128981481764114,
+      "loss": 1.2639,
+      "step": 493
+    },
+    {
+      "epoch": 0.95,
+      "grad_norm": 0.001639137383745666,
+      "learning_rate": 0.0013104021143278911,
+      "loss": 1.1987,
+      "step": 494
+    },
+    {
+      "epoch": 0.9519230769230769,
+      "grad_norm": 0.0017845044817163377,
+      "learning_rate": 0.0012117405796285285,
+      "loss": 1.1511,
+      "step": 495
+    },
+    {
+      "epoch": 0.9538461538461539,
+      "grad_norm": 0.0017943426783809326,
+      "learning_rate": 0.0011169173774871477,
+      "loss": 1.0759,
+      "step": 496
+    },
+    {
+      "epoch": 0.9557692307692308,
+      "grad_norm": 0.0015180353167847944,
+      "learning_rate": 0.0010259361921774012,
+      "loss": 1.1634,
+      "step": 497
+    },
+    {
+      "epoch": 0.9576923076923077,
+      "grad_norm": 0.0018004002552187213,
+      "learning_rate": 0.000938800558694719,
+      "loss": 1.1525,
+      "step": 498
+    },
+    {
+      "epoch": 0.9596153846153846,
+      "grad_norm": 0.0018394632194405148,
+      "learning_rate": 0.0008555138626189618,
+      "loss": 1.3065,
+      "step": 499
+    },
+    {
+      "epoch": 0.9615384615384616,
+      "grad_norm": 0.0020519473811755643,
+      "learning_rate": 0.0007760793399827937,
+      "loss": 1.2709,
+      "step": 500
+    },
+    {
+      "epoch": 0.9634615384615385,
+      "grad_norm": 0.0019103695401836494,
+      "learning_rate": 0.000700500077146038,
+      "loss": 1.2134,
+      "step": 501
+    },
+    {
+      "epoch": 0.9653846153846154,
+      "grad_norm": 0.0016337581403707033,
+      "learning_rate": 0.0006287790106757397,
+      "loss": 1.1953,
+      "step": 502
+    },
+    {
+      "epoch": 0.9673076923076923,
+      "grad_norm": 0.0017804862581599216,
+      "learning_rate": 0.0005609189272320237,
+      "loss": 1.1967,
+      "step": 503
+    },
+    {
+      "epoch": 0.9692307692307692,
+      "grad_norm": 0.0020830233555271006,
+      "learning_rate": 0.000496922463459859,
+      "loss": 1.1907,
+      "step": 504
+    },
+    {
+      "epoch": 0.9711538461538461,
+      "grad_norm": 0.001796505805656371,
+      "learning_rate": 0.0004367921058866187,
+      "loss": 1.2284,
+      "step": 505
+    },
+    {
+      "epoch": 0.9730769230769231,
+      "grad_norm": 0.0018563066367581421,
+      "learning_rate": 0.0003805301908254455,
+      "loss": 1.1445,
+      "step": 506
+    },
+    {
+      "epoch": 0.975,
+      "grad_norm": 0.0016888674996207236,
+      "learning_rate": 0.0003281389042844918,
+      "loss": 1.3471,
+      "step": 507
+    },
+    {
+      "epoch": 0.9769230769230769,
+      "grad_norm": 0.0017843455119250446,
+      "learning_rate": 0.00027962028188198705,
+      "loss": 1.2668,
+      "step": 508
+    },
+    {
+      "epoch": 0.9788461538461538,
+      "grad_norm": 0.0017126348951132893,
+      "learning_rate": 0.00023497620876711256,
+      "loss": 1.2309,
+      "step": 509
+    },
+    {
+      "epoch": 0.9807692307692307,
+      "grad_norm": 0.0017073019548309691,
+      "learning_rate": 0.00019420841954681523,
+      "loss": 1.1828,
+      "step": 510
+    },
+    {
+      "epoch": 0.9826923076923076,
+      "grad_norm": 0.0016811212449231337,
+      "learning_rate": 0.00015731849821833956,
+      "loss": 1.1611,
+      "step": 511
+    },
+    {
+      "epoch": 0.9846153846153847,
+      "grad_norm": 0.001820250260597972,
+      "learning_rate": 0.00012430787810776555,
+      "loss": 1.0436,
+      "step": 512
+    },
+    {
+      "epoch": 0.9865384615384616,
+      "grad_norm": 0.0020457912904106555,
+      "learning_rate": 9.517784181422018e-05,
+      "loss": 1.2473,
+      "step": 513
+    },
+    {
+      "epoch": 0.9884615384615385,
+      "grad_norm": 0.001663528411427663,
+      "learning_rate": 6.992952116013917e-05,
+      "loss": 1.2123,
+      "step": 514
+    },
+    {
+      "epoch": 0.9903846153846154,
+      "grad_norm": 0.0020152422521012947,
+      "learning_rate": 4.856389714723575e-05,
+      "loss": 1.2624,
+      "step": 515
+    },
+    {
+      "epoch": 0.9923076923076923,
+      "grad_norm": 0.001671439470027453,
+      "learning_rate": 3.108179991837545e-05,
+      "loss": 1.1565,
+      "step": 516
+    },
+    {
+      "epoch": 0.9942307692307693,
+      "grad_norm": 0.0017172775913499498,
+      "learning_rate": 1.7483908725357544e-05,
+      "loss": 1.2469,
+      "step": 517
+    },
+    {
+      "epoch": 0.9961538461538462,
+      "grad_norm": 0.0017825779963981006,
+      "learning_rate": 7.770751902513862e-06,
+      "loss": 1.1824,
+      "step": 518
+    },
+    {
+      "epoch": 0.9980769230769231,
+      "grad_norm": 0.001717098647369745,
+      "learning_rate": 1.9427068461808086e-06,
+      "loss": 1.2058,
+      "step": 519
+    },
+    {
+      "epoch": 1.0,
+      "grad_norm": 0.0016998398817481683,
+      "learning_rate": 0.0,
+      "loss": 1.2262,
+      "step": 520
+    },
+    {
+      "epoch": 1.0,
+      "step": 520,
+      "total_flos": 419724754747392.0,
+      "train_loss": 1.3048745036125182,
+      "train_runtime": 1954.7282,
+      "train_samples_per_second": 34.035,
+      "train_steps_per_second": 0.266
+    }
+  ],
+  "logging_steps": 1.0,
+  "max_steps": 520,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 50000,
+  "total_flos": 419724754747392.0,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}

ckpts_oct12/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.5_2e-1_connector-3.0_1.5_2e-1_ablation/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e38f6cbb5b75fa003dddd78d1786a36d1fd7b58c38c781bf1ffc2d204d38fee9
+size 7608

ckpts_oct12/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.7_2e-1_connector-3.0_1.7_2e-1_ablation/added_tokens.json

@@ -0,0 +1,24 @@
+{
+  "</tool_call>": 151658,
+  "<tool_call>": 151657,
+  "<|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
+}

ckpts_oct12/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.7_2e-1_connector-3.0_1.7_2e-1_ablation/config.json

@@ -0,0 +1,88 @@
+{
+  "architectures": [
+    "TinyLlavaForConditionalGeneration"
+  ],
+  "backward_type_connector": "normal",
+  "cache_dir": null,
+  "connector_type": "mlp2x_gelu",
+  "hidden_size": 896,
+  "ignore_index": -100,
+  "image_aspect_ratio": "square",
+  "image_token_index": -200,
+  "llm_model_name_or_path": "Qwen/Qwen2.5-0.5B",
+  "mask_model": [
+    "llm",
+    "connector"
+  ],
+  "mask_type_connector": "soft",
+  "model_type": "tinyllava",
+  "num_queries": 128,
+  "num_resampler_layers": 3,
+  "pad_token": "<|endoftext|>",
+  "pad_token_id": 151643,
+  "resampler_hidden_size": 768,
+  "sparsity_connector": null,
+  "subnet_type_connector": "global",
+  "temperature_connector": 1.7,
+  "text_config": {
+    "_name_or_path": "Qwen/Qwen2.5-0.5B",
+    "architectures": [
+      "Qwen2ForCausalLM"
+    ],
+    "backward_type": "normal",
+    "bos_token_id": 151643,
+    "eos_token_id": 151643,
+    "hidden_size": 896,
+    "intermediate_size": 4864,
+    "mask_type": "soft",
+    "masked_layers": "all",
+    "max_position_embeddings": 32768,
+    "max_window_layers": 24,
+    "model_type": "qwen2",
+    "num_attention_heads": 14,
+    "num_hidden_layers": 24,
+    "num_key_value_heads": 2,
+    "rope_theta": 1000000.0,
+    "sliding_window": 32768,
+    "subnet_mode": "both",
+    "subnet_type": "None",
+    "temperature_attn": 1.7,
+    "temperature_mlp": 1.7,
+    "tie_word_embeddings": true,
+    "torch_dtype": "bfloat16",
+    "use_mrope": false,
+    "use_sliding_window": false,
+    "vocab_size": 151936
+  },
+  "threshold_connector": null,
+  "tokenizer_model_max_length": 2048,
+  "tokenizer_name_or_path": "Qwen/Qwen2.5-0.5B",
+  "tokenizer_padding_side": "right",
+  "tokenizer_use_fast": false,
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.40.1",
+  "tune_type_connector": "full",
+  "tune_type_llm": "full",
+  "tune_type_vision_tower": "frozen",
+  "tune_vision_tower_from_layer": 0,
+  "use_cache": true,
+  "vision_config": {
+    "hidden_act": "gelu_pytorch_tanh",
+    "hidden_size": 1152,
+    "image_size": 384,
+    "intermediate_size": 4304,
+    "layer_norm_eps": 1e-06,
+    "model_name_or_path": "google/siglip-so400m-patch14-384",
+    "model_name_or_path2": "",
+    "model_type": "siglip_vision_model",
+    "num_attention_heads": 16,
+    "num_hidden_layers": 27,
+    "patch_size": 14
+  },
+  "vision_feature_layer": -2,
+  "vision_feature_select_strategy": "patch",
+  "vision_hidden_size": 1152,
+  "vision_model_name_or_path": "google/siglip-so400m-patch14-384",
+  "vision_model_name_or_path2": "",
+  "vocab_size": 151936
+}

ckpts_oct12/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.7_2e-1_connector-3.0_1.7_2e-1_ablation/generation_config.json

@@ -0,0 +1,7 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 151643,
+  "eos_token_id": 151643,
+  "transformers_version": "4.40.1",
+  "use_cache": false
+}

ckpts_oct12/qwen2.5-0_5b_base_masktune_42_llm-connector_text-3.0_1.7_2e-1_connector-3.0_1.7_2e-1_ablation/log.txt

@@ -0,0 +1,2 @@
+2025-10-13 09:17:00,685 | INFO: Total Parameters: 1283756736, Total Trainable Parameters: 359661568
+2025-10-13 09:17:00,689 | INFO: Trainable Parameters: