Upload continuous evaluation checkpoints, batch 12

.gitattributes

@@ -134,3 +134,6 @@ continuous/qwen3/mc1_g14_224_2.5b/finetune/tokenizer.json filter=lfs diff=lfs me
 continuous/qwen3/mc1_h14_224_2.5b/finetune/tokenizer.json filter=lfs diff=lfs merge=lfs -text
 continuous/qwen3/mc1_h14_224_v1.2/finetune/tokenizer.json filter=lfs diff=lfs merge=lfs -text
 continuous/qwen3/mc1_l14_224_2.5b/finetune/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+continuous/qwen3/mc1_l14_224_400m/finetune/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+continuous/qwen3/mc2_b16_224/finetune/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+continuous/qwen3/mc2_b16_384/finetune/tokenizer.json filter=lfs diff=lfs merge=lfs -text

continuous/qwen3/mc1_l14_224_400m/finetune/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:09149cf0190e785321454ee58637e76c19fc858e7b009b7c4503bbcc154fd5d3
+size 4696673872

continuous/qwen3/mc1_l14_224_400m/finetune/special_tokens_map.json

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

continuous/qwen3/mc1_l14_224_400m/finetune/tokenizer.json

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

continuous/qwen3/mc1_l14_224_400m/finetune/tokenizer_config.json

@@ -0,0 +1,241 @@
+{
+  "add_bos_token": false,
+  "add_prefix_space": false,
+  "added_tokens_decoder": {
+    "151643": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151644": {
+      "content": "<|im_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151645": {
+      "content": "<|im_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151646": {
+      "content": "<|object_ref_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151647": {
+      "content": "<|object_ref_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151648": {
+      "content": "<|box_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151649": {
+      "content": "<|box_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151650": {
+      "content": "<|quad_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151651": {
+      "content": "<|quad_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151652": {
+      "content": "<|vision_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151653": {
+      "content": "<|vision_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151654": {
+      "content": "<|vision_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151655": {
+      "content": "<|image_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151656": {
+      "content": "<|video_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151657": {
+      "content": "<tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151658": {
+      "content": "</tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151659": {
+      "content": "<|fim_prefix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151660": {
+      "content": "<|fim_middle|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151661": {
+      "content": "<|fim_suffix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151662": {
+      "content": "<|fim_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151663": {
+      "content": "<|repo_name|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151664": {
+      "content": "<|file_sep|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151665": {
+      "content": "<tool_response>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151666": {
+      "content": "</tool_response>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151667": {
+      "content": "<think>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151668": {
+      "content": "</think>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    }
+  },
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "bos_token": null,
+  "chat_template": "{%- if tools %}\n    {{- '<|im_start|>system\\n' }}\n    {%- if messages[0].role == 'system' %}\n        {{- messages[0].content + '\\n\\n' }}\n    {%- endif %}\n    {{- \"# Tools\\n\\nYou may call one or more functions to assist with the user query.\\n\\nYou are provided with function signatures within <tools></tools> XML tags:\\n<tools>\" }}\n    {%- for tool in tools %}\n        {{- \"\\n\" }}\n        {{- tool | tojson }}\n    {%- endfor %}\n    {{- \"\\n</tools>\\n\\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\\n<tool_call>\\n{\\\"name\\\": <function-name>, \\\"arguments\\\": <args-json-object>}\\n</tool_call><|im_end|>\\n\" }}\n{%- else %}\n    {%- if messages[0].role == 'system' %}\n        {{- '<|im_start|>system\\n' + messages[0].content + '<|im_end|>\\n' }}\n    {%- endif %}\n{%- endif %}\n{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}\n{%- for message in messages[::-1] %}\n    {%- set index = (messages|length - 1) - loop.index0 %}\n    {%- if ns.multi_step_tool and message.role == \"user\" and message.content is string and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}\n        {%- set ns.multi_step_tool = false %}\n        {%- set ns.last_query_index = index %}\n    {%- endif %}\n{%- endfor %}\n{%- for message in messages %}\n    {%- if message.content is string %}\n        {%- set content = message.content %}\n    {%- else %}\n        {%- set content = '' %}\n    {%- endif %}\n    {%- if (message.role == \"user\") or (message.role == \"system\" and not loop.first) %}\n        {{- '<|im_start|>' + message.role + '\\n' + content + '<|im_end|>' + '\\n' }}\n    {%- elif message.role == \"assistant\" %}\n        {%- set reasoning_content = '' %}\n        {%- if message.reasoning_content is string %}\n            {%- set reasoning_content = message.reasoning_content %}\n        {%- else %}\n            {%- if '</think>' in content %}\n                {%- set reasoning_content = content.split('</think>')[0].rstrip('\\n').split('<think>')[-1].lstrip('\\n') %}\n                {%- set content = content.split('</think>')[-1].lstrip('\\n') %}\n            {%- endif %}\n        {%- endif %}\n        {%- if loop.index0 > ns.last_query_index %}\n            {%- if loop.last or (not loop.last and reasoning_content) %}\n                {{- '<|im_start|>' + message.role + '\\n<think>\\n' + reasoning_content.strip('\\n') + '\\n</think>\\n\\n' + content.lstrip('\\n') }}\n            {%- else %}\n                {{- '<|im_start|>' + message.role + '\\n' + content }}\n            {%- endif %}\n        {%- else %}\n            {{- '<|im_start|>' + message.role + '\\n' + content }}\n        {%- endif %}\n        {%- if message.tool_calls %}\n            {%- for tool_call in message.tool_calls %}\n                {%- if (loop.first and content) or (not loop.first) %}\n                    {{- '\\n' }}\n                {%- endif %}\n                {%- if tool_call.function %}\n                    {%- set tool_call = tool_call.function %}\n                {%- endif %}\n                {{- '<tool_call>\\n{\"name\": \"' }}\n                {{- tool_call.name }}\n                {{- '\", \"arguments\": ' }}\n                {%- if tool_call.arguments is string %}\n                    {{- tool_call.arguments }}\n                {%- else %}\n                    {{- tool_call.arguments | tojson }}\n                {%- endif %}\n                {{- '}\\n</tool_call>' }}\n            {%- endfor %}\n        {%- endif %}\n        {{- '<|im_end|>\\n' }}\n    {%- elif message.role == \"tool\" %}\n        {%- if loop.first or (messages[loop.index0 - 1].role != \"tool\") %}\n            {{- '<|im_start|>user' }}\n        {%- endif %}\n        {{- '\\n<tool_response>\\n' }}\n        {{- content }}\n        {{- '\\n</tool_response>' }}\n        {%- if loop.last or (messages[loop.index0 + 1].role != \"tool\") %}\n            {{- '<|im_end|>\\n' }}\n        {%- endif %}\n    {%- endif %}\n{%- endfor %}\n{%- if add_generation_prompt %}\n    {{- '<|im_start|>assistant\\n' }}\n    {%- if enable_thinking is defined and enable_thinking is false %}\n        {{- '<think>\\n\\n</think>\\n\\n' }}\n    {%- endif %}\n{%- endif %}",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|im_end|>",
+  "errors": "replace",
+  "extra_special_tokens": {},
+  "model_max_length": 8192,
+  "pad_token": "<|endoftext|>",
+  "padding_side": "right",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}

continuous/qwen3/mc1_l14_224_400m/finetune/trainer_state.json

@@ -0,0 +1,1856 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.9999037443449803,
+  "eval_steps": 500,
+  "global_step": 2597,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.003850226200789296,
+      "grad_norm": 20.429611206054688,
+      "learning_rate": 1.153846153846154e-06,
+      "loss": 2.6651,
+      "step": 10
+    },
+    {
+      "epoch": 0.007700452401578592,
+      "grad_norm": 7.410536766052246,
+      "learning_rate": 2.435897435897436e-06,
+      "loss": 2.0274,
+      "step": 20
+    },
+    {
+      "epoch": 0.01155067860236789,
+      "grad_norm": 5.182516574859619,
+      "learning_rate": 3.7179487179487184e-06,
+      "loss": 1.5509,
+      "step": 30
+    },
+    {
+      "epoch": 0.015400904803157185,
+      "grad_norm": 1.9049382209777832,
+      "learning_rate": 5e-06,
+      "loss": 1.1955,
+      "step": 40
+    },
+    {
+      "epoch": 0.019251131003946483,
+      "grad_norm": 0.8399803042411804,
+      "learning_rate": 6.282051282051282e-06,
+      "loss": 1.063,
+      "step": 50
+    },
+    {
+      "epoch": 0.02310135720473578,
+      "grad_norm": 0.42456743121147156,
+      "learning_rate": 7.564102564102564e-06,
+      "loss": 0.9708,
+      "step": 60
+    },
+    {
+      "epoch": 0.026951583405525073,
+      "grad_norm": 0.31009483337402344,
+      "learning_rate": 8.846153846153847e-06,
+      "loss": 0.9202,
+      "step": 70
+    },
+    {
+      "epoch": 0.03080180960631437,
+      "grad_norm": 0.3100764751434326,
+      "learning_rate": 9.999996111488723e-06,
+      "loss": 0.9174,
+      "step": 80
+    },
+    {
+      "epoch": 0.034652035807103666,
+      "grad_norm": 0.29028603434562683,
+      "learning_rate": 9.999529497453782e-06,
+      "loss": 0.8836,
+      "step": 90
+    },
+    {
+      "epoch": 0.038502262007892966,
+      "grad_norm": 0.2627231478691101,
+      "learning_rate": 9.998285264324193e-06,
+      "loss": 0.8645,
+      "step": 100
+    },
+    {
+      "epoch": 0.04235248820868226,
+      "grad_norm": 0.2563619315624237,
+      "learning_rate": 9.996263605626052e-06,
+      "loss": 0.8461,
+      "step": 110
+    },
+    {
+      "epoch": 0.04620271440947156,
+      "grad_norm": 0.2703174948692322,
+      "learning_rate": 9.99346483580503e-06,
+      "loss": 0.8555,
+      "step": 120
+    },
+    {
+      "epoch": 0.05005294061026085,
+      "grad_norm": 0.26077234745025635,
+      "learning_rate": 9.989889390177467e-06,
+      "loss": 0.8485,
+      "step": 130
+    },
+    {
+      "epoch": 0.053903166811050146,
+      "grad_norm": 0.2980562746524811,
+      "learning_rate": 9.985537824862647e-06,
+      "loss": 0.8439,
+      "step": 140
+    },
+    {
+      "epoch": 0.057753393011839446,
+      "grad_norm": 0.30622270703315735,
+      "learning_rate": 9.980410816696317e-06,
+      "loss": 0.8275,
+      "step": 150
+    },
+    {
+      "epoch": 0.06160361921262874,
+      "grad_norm": 0.2755362391471863,
+      "learning_rate": 9.97450916312541e-06,
+      "loss": 0.8333,
+      "step": 160
+    },
+    {
+      "epoch": 0.06545384541341803,
+      "grad_norm": 0.23848822712898254,
+      "learning_rate": 9.967833782084004e-06,
+      "loss": 0.8338,
+      "step": 170
+    },
+    {
+      "epoch": 0.06930407161420733,
+      "grad_norm": 0.2696181535720825,
+      "learning_rate": 9.960385711850554e-06,
+      "loss": 0.8589,
+      "step": 180
+    },
+    {
+      "epoch": 0.07315429781499663,
+      "grad_norm": 0.2689056396484375,
+      "learning_rate": 9.952166110886394e-06,
+      "loss": 0.8497,
+      "step": 190
+    },
+    {
+      "epoch": 0.07700452401578593,
+      "grad_norm": 0.29626214504241943,
+      "learning_rate": 9.943176257655567e-06,
+      "loss": 0.8359,
+      "step": 200
+    },
+    {
+      "epoch": 0.08085475021657522,
+      "grad_norm": 0.2611771523952484,
+      "learning_rate": 9.933417550425952e-06,
+      "loss": 0.8333,
+      "step": 210
+    },
+    {
+      "epoch": 0.08470497641736452,
+      "grad_norm": 0.2807839810848236,
+      "learning_rate": 9.9228915070518e-06,
+      "loss": 0.8173,
+      "step": 220
+    },
+    {
+      "epoch": 0.08855520261815382,
+      "grad_norm": 0.2607056200504303,
+      "learning_rate": 9.911599764737632e-06,
+      "loss": 0.8184,
+      "step": 230
+    },
+    {
+      "epoch": 0.09240542881894312,
+      "grad_norm": 0.27480337023735046,
+      "learning_rate": 9.899544079783606e-06,
+      "loss": 0.8125,
+      "step": 240
+    },
+    {
+      "epoch": 0.0962556550197324,
+      "grad_norm": 0.29196396470069885,
+      "learning_rate": 9.886726327312338e-06,
+      "loss": 0.8308,
+      "step": 250
+    },
+    {
+      "epoch": 0.1001058812205217,
+      "grad_norm": 0.24465692043304443,
+      "learning_rate": 9.87314850097724e-06,
+      "loss": 0.8245,
+      "step": 260
+    },
+    {
+      "epoch": 0.103956107421311,
+      "grad_norm": 0.2484912872314453,
+      "learning_rate": 9.858812712652452e-06,
+      "loss": 0.8133,
+      "step": 270
+    },
+    {
+      "epoch": 0.10780633362210029,
+      "grad_norm": 0.2494082748889923,
+      "learning_rate": 9.843721192104341e-06,
+      "loss": 0.8113,
+      "step": 280
+    },
+    {
+      "epoch": 0.11165655982288959,
+      "grad_norm": 0.25708091259002686,
+      "learning_rate": 9.827876286644698e-06,
+      "loss": 0.83,
+      "step": 290
+    },
+    {
+      "epoch": 0.11550678602367889,
+      "grad_norm": 0.22752511501312256,
+      "learning_rate": 9.811280460765643e-06,
+      "loss": 0.7914,
+      "step": 300
+    },
+    {
+      "epoch": 0.11935701222446819,
+      "grad_norm": 0.24338088929653168,
+      "learning_rate": 9.793936295756292e-06,
+      "loss": 0.8032,
+      "step": 310
+    },
+    {
+      "epoch": 0.12320723842525748,
+      "grad_norm": 0.24953772127628326,
+      "learning_rate": 9.775846489301279e-06,
+      "loss": 0.8036,
+      "step": 320
+    },
+    {
+      "epoch": 0.12705746462604678,
+      "grad_norm": 0.2560637891292572,
+      "learning_rate": 9.757013855061157e-06,
+      "loss": 0.8022,
+      "step": 330
+    },
+    {
+      "epoch": 0.13090769082683607,
+      "grad_norm": 0.23933368921279907,
+      "learning_rate": 9.737441322234758e-06,
+      "loss": 0.8149,
+      "step": 340
+    },
+    {
+      "epoch": 0.13475791702762538,
+      "grad_norm": 0.2594987452030182,
+      "learning_rate": 9.717131935103605e-06,
+      "loss": 0.8155,
+      "step": 350
+    },
+    {
+      "epoch": 0.13860814322841467,
+      "grad_norm": 0.24666482210159302,
+      "learning_rate": 9.696088852558398e-06,
+      "loss": 0.8019,
+      "step": 360
+    },
+    {
+      "epoch": 0.14245836942920398,
+      "grad_norm": 0.25465869903564453,
+      "learning_rate": 9.674315347607686e-06,
+      "loss": 0.8283,
+      "step": 370
+    },
+    {
+      "epoch": 0.14630859562999327,
+      "grad_norm": 0.23826292157173157,
+      "learning_rate": 9.651814806868796e-06,
+      "loss": 0.7878,
+      "step": 380
+    },
+    {
+      "epoch": 0.15015882183078255,
+      "grad_norm": 0.248566672205925,
+      "learning_rate": 9.62859073004107e-06,
+      "loss": 0.793,
+      "step": 390
+    },
+    {
+      "epoch": 0.15400904803157187,
+      "grad_norm": 0.23663872480392456,
+      "learning_rate": 9.604646729361547e-06,
+      "loss": 0.7937,
+      "step": 400
+    },
+    {
+      "epoch": 0.15785927423236115,
+      "grad_norm": 0.23697665333747864,
+      "learning_rate": 9.579986529043099e-06,
+      "loss": 0.7821,
+      "step": 410
+    },
+    {
+      "epoch": 0.16170950043315044,
+      "grad_norm": 0.2467927187681198,
+      "learning_rate": 9.554613964695189e-06,
+      "loss": 0.8021,
+      "step": 420
+    },
+    {
+      "epoch": 0.16555972663393975,
+      "grad_norm": 0.2488505095243454,
+      "learning_rate": 9.528532982727281e-06,
+      "loss": 0.7926,
+      "step": 430
+    },
+    {
+      "epoch": 0.16940995283472904,
+      "grad_norm": 0.24139896035194397,
+      "learning_rate": 9.501747639735018e-06,
+      "loss": 0.7877,
+      "step": 440
+    },
+    {
+      "epoch": 0.17326017903551832,
+      "grad_norm": 0.25130051374435425,
+      "learning_rate": 9.474262101869272e-06,
+      "loss": 0.7782,
+      "step": 450
+    },
+    {
+      "epoch": 0.17711040523630764,
+      "grad_norm": 0.2491873800754547,
+      "learning_rate": 9.446080644188141e-06,
+      "loss": 0.7842,
+      "step": 460
+    },
+    {
+      "epoch": 0.18096063143709692,
+      "grad_norm": 0.2577859163284302,
+      "learning_rate": 9.417207649992017e-06,
+      "loss": 0.7894,
+      "step": 470
+    },
+    {
+      "epoch": 0.18481085763788624,
+      "grad_norm": 0.24168357253074646,
+      "learning_rate": 9.387647610141816e-06,
+      "loss": 0.7883,
+      "step": 480
+    },
+    {
+      "epoch": 0.18866108383867553,
+      "grad_norm": 0.23632767796516418,
+      "learning_rate": 9.35740512236047e-06,
+      "loss": 0.7761,
+      "step": 490
+    },
+    {
+      "epoch": 0.1925113100394648,
+      "grad_norm": 0.23387444019317627,
+      "learning_rate": 9.326484890517809e-06,
+      "loss": 0.7839,
+      "step": 500
+    },
+    {
+      "epoch": 0.19636153624025413,
+      "grad_norm": 0.23738186061382294,
+      "learning_rate": 9.294891723898926e-06,
+      "loss": 0.7774,
+      "step": 510
+    },
+    {
+      "epoch": 0.2002117624410434,
+      "grad_norm": 0.24920213222503662,
+      "learning_rate": 9.262630536456153e-06,
+      "loss": 0.8068,
+      "step": 520
+    },
+    {
+      "epoch": 0.2040619886418327,
+      "grad_norm": 0.2566337287425995,
+      "learning_rate": 9.229706346044749e-06,
+      "loss": 0.7777,
+      "step": 530
+    },
+    {
+      "epoch": 0.207912214842622,
+      "grad_norm": 0.2610238790512085,
+      "learning_rate": 9.196124273642422e-06,
+      "loss": 0.7923,
+      "step": 540
+    },
+    {
+      "epoch": 0.2117624410434113,
+      "grad_norm": 0.23474909365177155,
+      "learning_rate": 9.161889542552842e-06,
+      "loss": 0.7977,
+      "step": 550
+    },
+    {
+      "epoch": 0.21561266724420058,
+      "grad_norm": 0.2307462990283966,
+      "learning_rate": 9.127007477593193e-06,
+      "loss": 0.7977,
+      "step": 560
+    },
+    {
+      "epoch": 0.2194628934449899,
+      "grad_norm": 0.26065587997436523,
+      "learning_rate": 9.091483504265974e-06,
+      "loss": 0.8029,
+      "step": 570
+    },
+    {
+      "epoch": 0.22331311964577918,
+      "grad_norm": 0.2518492639064789,
+      "learning_rate": 9.055323147915116e-06,
+      "loss": 0.7871,
+      "step": 580
+    },
+    {
+      "epoch": 0.2271633458465685,
+      "grad_norm": 0.24906082451343536,
+      "learning_rate": 9.018532032866585e-06,
+      "loss": 0.7954,
+      "step": 590
+    },
+    {
+      "epoch": 0.23101357204735778,
+      "grad_norm": 0.22207732498645782,
+      "learning_rate": 8.981115881553579e-06,
+      "loss": 0.7861,
+      "step": 600
+    },
+    {
+      "epoch": 0.23486379824814707,
+      "grad_norm": 0.26233604550361633,
+      "learning_rate": 8.943080513626476e-06,
+      "loss": 0.8069,
+      "step": 610
+    },
+    {
+      "epoch": 0.23871402444893638,
+      "grad_norm": 0.23801080882549286,
+      "learning_rate": 8.904431845047641e-06,
+      "loss": 0.7917,
+      "step": 620
+    },
+    {
+      "epoch": 0.24256425064972567,
+      "grad_norm": 0.24589622020721436,
+      "learning_rate": 8.86517588717128e-06,
+      "loss": 0.8074,
+      "step": 630
+    },
+    {
+      "epoch": 0.24641447685051496,
+      "grad_norm": 0.2348746508359909,
+      "learning_rate": 8.82531874580844e-06,
+      "loss": 0.7999,
+      "step": 640
+    },
+    {
+      "epoch": 0.25026470305130427,
+      "grad_norm": 0.23529697954654694,
+      "learning_rate": 8.784866620277314e-06,
+      "loss": 0.7569,
+      "step": 650
+    },
+    {
+      "epoch": 0.25411492925209356,
+      "grad_norm": 0.24850475788116455,
+      "learning_rate": 8.743825802439018e-06,
+      "loss": 0.7887,
+      "step": 660
+    },
+    {
+      "epoch": 0.25796515545288284,
+      "grad_norm": 0.271946519613266,
+      "learning_rate": 8.702202675718956e-06,
+      "loss": 0.7994,
+      "step": 670
+    },
+    {
+      "epoch": 0.26181538165367213,
+      "grad_norm": 0.2309792935848236,
+      "learning_rate": 8.660003714113959e-06,
+      "loss": 0.7651,
+      "step": 680
+    },
+    {
+      "epoch": 0.26566560785446147,
+      "grad_norm": 0.23769214749336243,
+      "learning_rate": 8.617235481185323e-06,
+      "loss": 0.7887,
+      "step": 690
+    },
+    {
+      "epoch": 0.26951583405525076,
+      "grad_norm": 0.24355381727218628,
+      "learning_rate": 8.573904629037931e-06,
+      "loss": 0.7964,
+      "step": 700
+    },
+    {
+      "epoch": 0.27336606025604004,
+      "grad_norm": 0.23923350870609283,
+      "learning_rate": 8.530017897285585e-06,
+      "loss": 0.7776,
+      "step": 710
+    },
+    {
+      "epoch": 0.27721628645682933,
+      "grad_norm": 0.2191125452518463,
+      "learning_rate": 8.485582112002746e-06,
+      "loss": 0.7816,
+      "step": 720
+    },
+    {
+      "epoch": 0.2810665126576186,
+      "grad_norm": 0.2501249611377716,
+      "learning_rate": 8.440604184662811e-06,
+      "loss": 0.7848,
+      "step": 730
+    },
+    {
+      "epoch": 0.28491673885840796,
+      "grad_norm": 0.24289488792419434,
+      "learning_rate": 8.395091111063118e-06,
+      "loss": 0.7718,
+      "step": 740
+    },
+    {
+      "epoch": 0.28876696505919724,
+      "grad_norm": 0.2237260639667511,
+      "learning_rate": 8.349049970236822e-06,
+      "loss": 0.7758,
+      "step": 750
+    },
+    {
+      "epoch": 0.29261719125998653,
+      "grad_norm": 0.2678987979888916,
+      "learning_rate": 8.302487923351842e-06,
+      "loss": 0.766,
+      "step": 760
+    },
+    {
+      "epoch": 0.2964674174607758,
+      "grad_norm": 0.2348799705505371,
+      "learning_rate": 8.255412212597021e-06,
+      "loss": 0.7833,
+      "step": 770
+    },
+    {
+      "epoch": 0.3003176436615651,
+      "grad_norm": 0.2380896806716919,
+      "learning_rate": 8.207830160055688e-06,
+      "loss": 0.7807,
+      "step": 780
+    },
+    {
+      "epoch": 0.3041678698623544,
+      "grad_norm": 0.24251551926136017,
+      "learning_rate": 8.159749166566786e-06,
+      "loss": 0.7646,
+      "step": 790
+    },
+    {
+      "epoch": 0.30801809606314373,
+      "grad_norm": 0.2226421982049942,
+      "learning_rate": 8.11117671057377e-06,
+      "loss": 0.7917,
+      "step": 800
+    },
+    {
+      "epoch": 0.311868322263933,
+      "grad_norm": 0.24054287374019623,
+      "learning_rate": 8.062120346961404e-06,
+      "loss": 0.7835,
+      "step": 810
+    },
+    {
+      "epoch": 0.3157185484647223,
+      "grad_norm": 0.22916527092456818,
+      "learning_rate": 8.012587705880705e-06,
+      "loss": 0.7763,
+      "step": 820
+    },
+    {
+      "epoch": 0.3195687746655116,
+      "grad_norm": 0.24169644713401794,
+      "learning_rate": 7.962586491562146e-06,
+      "loss": 0.7627,
+      "step": 830
+    },
+    {
+      "epoch": 0.3234190008663009,
+      "grad_norm": 0.24502651393413544,
+      "learning_rate": 7.912124481117357e-06,
+      "loss": 0.7696,
+      "step": 840
+    },
+    {
+      "epoch": 0.3272692270670902,
+      "grad_norm": 0.23691031336784363,
+      "learning_rate": 7.861209523329494e-06,
+      "loss": 0.7825,
+      "step": 850
+    },
+    {
+      "epoch": 0.3311194532678795,
+      "grad_norm": 0.23765712976455688,
+      "learning_rate": 7.809849537432432e-06,
+      "loss": 0.7617,
+      "step": 860
+    },
+    {
+      "epoch": 0.3349696794686688,
+      "grad_norm": 0.2313634753227234,
+      "learning_rate": 7.758052511879035e-06,
+      "loss": 0.7746,
+      "step": 870
+    },
+    {
+      "epoch": 0.3388199056694581,
+      "grad_norm": 0.2180565744638443,
+      "learning_rate": 7.705826503098642e-06,
+      "loss": 0.7699,
+      "step": 880
+    },
+    {
+      "epoch": 0.34267013187024736,
+      "grad_norm": 0.23273739218711853,
+      "learning_rate": 7.653179634243979e-06,
+      "loss": 0.77,
+      "step": 890
+    },
+    {
+      "epoch": 0.34652035807103665,
+      "grad_norm": 0.21656735241413116,
+      "learning_rate": 7.600120093927696e-06,
+      "loss": 0.7719,
+      "step": 900
+    },
+    {
+      "epoch": 0.350370584271826,
+      "grad_norm": 0.2414737194776535,
+      "learning_rate": 7.546656134948729e-06,
+      "loss": 0.7617,
+      "step": 910
+    },
+    {
+      "epoch": 0.3542208104726153,
+      "grad_norm": 0.24607038497924805,
+      "learning_rate": 7.492796073008661e-06,
+      "loss": 0.7635,
+      "step": 920
+    },
+    {
+      "epoch": 0.35807103667340456,
+      "grad_norm": 0.25156834721565247,
+      "learning_rate": 7.438548285418326e-06,
+      "loss": 0.7485,
+      "step": 930
+    },
+    {
+      "epoch": 0.36192126287419385,
+      "grad_norm": 0.2642505466938019,
+      "learning_rate": 7.3839212097948e-06,
+      "loss": 0.7843,
+      "step": 940
+    },
+    {
+      "epoch": 0.36577148907498314,
+      "grad_norm": 0.2332247793674469,
+      "learning_rate": 7.328923342749039e-06,
+      "loss": 0.7987,
+      "step": 950
+    },
+    {
+      "epoch": 0.3696217152757725,
+      "grad_norm": 0.2522306442260742,
+      "learning_rate": 7.273563238564319e-06,
+      "loss": 0.7827,
+      "step": 960
+    },
+    {
+      "epoch": 0.37347194147656176,
+      "grad_norm": 0.23498684167861938,
+      "learning_rate": 7.217849507865724e-06,
+      "loss": 0.7624,
+      "step": 970
+    },
+    {
+      "epoch": 0.37732216767735105,
+      "grad_norm": 0.22976292669773102,
+      "learning_rate": 7.161790816280858e-06,
+      "loss": 0.77,
+      "step": 980
+    },
+    {
+      "epoch": 0.38117239387814034,
+      "grad_norm": 0.25127989053726196,
+      "learning_rate": 7.1053958830920125e-06,
+      "loss": 0.7512,
+      "step": 990
+    },
+    {
+      "epoch": 0.3850226200789296,
+      "grad_norm": 0.24100199341773987,
+      "learning_rate": 7.0486734798799715e-06,
+      "loss": 0.7726,
+      "step": 1000
+    },
+    {
+      "epoch": 0.3888728462797189,
+      "grad_norm": 0.23490232229232788,
+      "learning_rate": 6.991632429159708e-06,
+      "loss": 0.7947,
+      "step": 1010
+    },
+    {
+      "epoch": 0.39272307248050825,
+      "grad_norm": 0.26322200894355774,
+      "learning_rate": 6.934281603008132e-06,
+      "loss": 0.784,
+      "step": 1020
+    },
+    {
+      "epoch": 0.39657329868129754,
+      "grad_norm": 0.24126212298870087,
+      "learning_rate": 6.876629921684143e-06,
+      "loss": 0.7821,
+      "step": 1030
+    },
+    {
+      "epoch": 0.4004235248820868,
+      "grad_norm": 0.24527502059936523,
+      "learning_rate": 6.818686352241199e-06,
+      "loss": 0.7727,
+      "step": 1040
+    },
+    {
+      "epoch": 0.4042737510828761,
+      "grad_norm": 0.21318113803863525,
+      "learning_rate": 6.760459907132573e-06,
+      "loss": 0.768,
+      "step": 1050
+    },
+    {
+      "epoch": 0.4081239772836654,
+      "grad_norm": 0.22305262088775635,
+      "learning_rate": 6.701959642809589e-06,
+      "loss": 0.7727,
+      "step": 1060
+    },
+    {
+      "epoch": 0.41197420348445474,
+      "grad_norm": 0.23872095346450806,
+      "learning_rate": 6.643194658312987e-06,
+      "loss": 0.7659,
+      "step": 1070
+    },
+    {
+      "epoch": 0.415824429685244,
+      "grad_norm": 0.22792795300483704,
+      "learning_rate": 6.584174093857676e-06,
+      "loss": 0.7893,
+      "step": 1080
+    },
+    {
+      "epoch": 0.4196746558860333,
+      "grad_norm": 0.2658918797969818,
+      "learning_rate": 6.5249071294110735e-06,
+      "loss": 0.7721,
+      "step": 1090
+    },
+    {
+      "epoch": 0.4235248820868226,
+      "grad_norm": 0.24555674195289612,
+      "learning_rate": 6.465402983265274e-06,
+      "loss": 0.7853,
+      "step": 1100
+    },
+    {
+      "epoch": 0.4273751082876119,
+      "grad_norm": 0.2436600625514984,
+      "learning_rate": 6.405670910603243e-06,
+      "loss": 0.7802,
+      "step": 1110
+    },
+    {
+      "epoch": 0.43122533448840117,
+      "grad_norm": 0.2627989947795868,
+      "learning_rate": 6.345720202059291e-06,
+      "loss": 0.7767,
+      "step": 1120
+    },
+    {
+      "epoch": 0.4350755606891905,
+      "grad_norm": 0.24053417146205902,
+      "learning_rate": 6.285560182274008e-06,
+      "loss": 0.7555,
+      "step": 1130
+    },
+    {
+      "epoch": 0.4389257868899798,
+      "grad_norm": 0.23731058835983276,
+      "learning_rate": 6.225200208443936e-06,
+      "loss": 0.7786,
+      "step": 1140
+    },
+    {
+      "epoch": 0.4427760130907691,
+      "grad_norm": 0.22336068749427795,
+      "learning_rate": 6.16464966886616e-06,
+      "loss": 0.7665,
+      "step": 1150
+    },
+    {
+      "epoch": 0.44662623929155837,
+      "grad_norm": 0.2633199691772461,
+      "learning_rate": 6.103917981478065e-06,
+      "loss": 0.7908,
+      "step": 1160
+    },
+    {
+      "epoch": 0.45047646549234766,
+      "grad_norm": 0.22687651216983795,
+      "learning_rate": 6.043014592392489e-06,
+      "loss": 0.8015,
+      "step": 1170
+    },
+    {
+      "epoch": 0.454326691693137,
+      "grad_norm": 0.22555126249790192,
+      "learning_rate": 5.981948974428485e-06,
+      "loss": 0.7773,
+      "step": 1180
+    },
+    {
+      "epoch": 0.4581769178939263,
+      "grad_norm": 0.24943314492702484,
+      "learning_rate": 5.920730625637934e-06,
+      "loss": 0.7536,
+      "step": 1190
+    },
+    {
+      "epoch": 0.46202714409471557,
+      "grad_norm": 0.2478240579366684,
+      "learning_rate": 5.8593690678282355e-06,
+      "loss": 0.7669,
+      "step": 1200
+    },
+    {
+      "epoch": 0.46587737029550486,
+      "grad_norm": 0.25187209248542786,
+      "learning_rate": 5.797873845081301e-06,
+      "loss": 0.7716,
+      "step": 1210
+    },
+    {
+      "epoch": 0.46972759649629414,
+      "grad_norm": 0.23382095992565155,
+      "learning_rate": 5.736254522269075e-06,
+      "loss": 0.7874,
+      "step": 1220
+    },
+    {
+      "epoch": 0.47357782269708343,
+      "grad_norm": 0.23809374868869781,
+      "learning_rate": 5.6745206835658386e-06,
+      "loss": 0.7898,
+      "step": 1230
+    },
+    {
+      "epoch": 0.47742804889787277,
+      "grad_norm": 0.23236331343650818,
+      "learning_rate": 5.612681930957495e-06,
+      "loss": 0.7609,
+      "step": 1240
+    },
+    {
+      "epoch": 0.48127827509866206,
+      "grad_norm": 0.22258979082107544,
+      "learning_rate": 5.550747882748095e-06,
+      "loss": 0.7577,
+      "step": 1250
+    },
+    {
+      "epoch": 0.48512850129945134,
+      "grad_norm": 0.25479453802108765,
+      "learning_rate": 5.4887281720638195e-06,
+      "loss": 0.7653,
+      "step": 1260
+    },
+    {
+      "epoch": 0.48897872750024063,
+      "grad_norm": 0.26307424902915955,
+      "learning_rate": 5.42663244535466e-06,
+      "loss": 0.7487,
+      "step": 1270
+    },
+    {
+      "epoch": 0.4928289537010299,
+      "grad_norm": 0.25793325901031494,
+      "learning_rate": 5.3644703608940205e-06,
+      "loss": 0.7593,
+      "step": 1280
+    },
+    {
+      "epoch": 0.49667917990181926,
+      "grad_norm": 0.2353559136390686,
+      "learning_rate": 5.302251587276491e-06,
+      "loss": 0.7689,
+      "step": 1290
+    },
+    {
+      "epoch": 0.5005294061026085,
+      "grad_norm": 0.22789326310157776,
+      "learning_rate": 5.2399858019140005e-06,
+      "loss": 0.7693,
+      "step": 1300
+    },
+    {
+      "epoch": 0.5043796323033978,
+      "grad_norm": 0.23442833125591278,
+      "learning_rate": 5.177682689530615e-06,
+      "loss": 0.7853,
+      "step": 1310
+    },
+    {
+      "epoch": 0.5082298585041871,
+      "grad_norm": 0.24031957983970642,
+      "learning_rate": 5.11535194065619e-06,
+      "loss": 0.7512,
+      "step": 1320
+    },
+    {
+      "epoch": 0.5120800847049765,
+      "grad_norm": 0.25515758991241455,
+      "learning_rate": 5.053003250119113e-06,
+      "loss": 0.7612,
+      "step": 1330
+    },
+    {
+      "epoch": 0.5159303109057657,
+      "grad_norm": 0.24716448783874512,
+      "learning_rate": 4.9906463155383925e-06,
+      "loss": 0.7705,
+      "step": 1340
+    },
+    {
+      "epoch": 0.519780537106555,
+      "grad_norm": 0.2333715260028839,
+      "learning_rate": 4.9282908358153046e-06,
+      "loss": 0.7642,
+      "step": 1350
+    },
+    {
+      "epoch": 0.5236307633073443,
+      "grad_norm": 0.25636494159698486,
+      "learning_rate": 4.8659465096248345e-06,
+      "loss": 0.7784,
+      "step": 1360
+    },
+    {
+      "epoch": 0.5274809895081336,
+      "grad_norm": 0.24976707994937897,
+      "learning_rate": 4.803623033907169e-06,
+      "loss": 0.7765,
+      "step": 1370
+    },
+    {
+      "epoch": 0.5313312157089229,
+      "grad_norm": 0.22607001662254333,
+      "learning_rate": 4.741330102359441e-06,
+      "loss": 0.7838,
+      "step": 1380
+    },
+    {
+      "epoch": 0.5351814419097122,
+      "grad_norm": 0.2550511658191681,
+      "learning_rate": 4.679077403927996e-06,
+      "loss": 0.7629,
+      "step": 1390
+    },
+    {
+      "epoch": 0.5390316681105015,
+      "grad_norm": 0.23168377578258514,
+      "learning_rate": 4.616874621301373e-06,
+      "loss": 0.7643,
+      "step": 1400
+    },
+    {
+      "epoch": 0.5428818943112907,
+      "grad_norm": 0.24735374748706818,
+      "learning_rate": 4.554731429404293e-06,
+      "loss": 0.7776,
+      "step": 1410
+    },
+    {
+      "epoch": 0.5467321205120801,
+      "grad_norm": 0.24885909259319305,
+      "learning_rate": 4.492657493892822e-06,
+      "loss": 0.7615,
+      "step": 1420
+    },
+    {
+      "epoch": 0.5505823467128694,
+      "grad_norm": 0.2702668011188507,
+      "learning_rate": 4.430662469650999e-06,
+      "loss": 0.7772,
+      "step": 1430
+    },
+    {
+      "epoch": 0.5544325729136587,
+      "grad_norm": 0.27038612961769104,
+      "learning_rate": 4.368755999289118e-06,
+      "loss": 0.7814,
+      "step": 1440
+    },
+    {
+      "epoch": 0.558282799114448,
+      "grad_norm": 0.25525403022766113,
+      "learning_rate": 4.306947711643949e-06,
+      "loss": 0.7651,
+      "step": 1450
+    },
+    {
+      "epoch": 0.5621330253152372,
+      "grad_norm": 0.24610450863838196,
+      "learning_rate": 4.245247220281068e-06,
+      "loss": 0.7682,
+      "step": 1460
+    },
+    {
+      "epoch": 0.5659832515160266,
+      "grad_norm": 0.23824715614318848,
+      "learning_rate": 4.183664121999587e-06,
+      "loss": 0.7645,
+      "step": 1470
+    },
+    {
+      "epoch": 0.5698334777168159,
+      "grad_norm": 0.24462740123271942,
+      "learning_rate": 4.122207995339474e-06,
+      "loss": 0.7635,
+      "step": 1480
+    },
+    {
+      "epoch": 0.5736837039176051,
+      "grad_norm": 0.22978441417217255,
+      "learning_rate": 4.060888399091734e-06,
+      "loss": 0.7663,
+      "step": 1490
+    },
+    {
+      "epoch": 0.5775339301183945,
+      "grad_norm": 0.2306719720363617,
+      "learning_rate": 3.999714870811639e-06,
+      "loss": 0.7769,
+      "step": 1500
+    },
+    {
+      "epoch": 0.5813841563191837,
+      "grad_norm": 0.23302052915096283,
+      "learning_rate": 3.938696925335275e-06,
+      "loss": 0.7694,
+      "step": 1510
+    },
+    {
+      "epoch": 0.5852343825199731,
+      "grad_norm": 0.24033910036087036,
+      "learning_rate": 3.87784405329962e-06,
+      "loss": 0.7766,
+      "step": 1520
+    },
+    {
+      "epoch": 0.5890846087207623,
+      "grad_norm": 0.23780210316181183,
+      "learning_rate": 3.817165719666382e-06,
+      "loss": 0.7785,
+      "step": 1530
+    },
+    {
+      "epoch": 0.5929348349215516,
+      "grad_norm": 0.23227883875370026,
+      "learning_rate": 3.756671362249841e-06,
+      "loss": 0.7835,
+      "step": 1540
+    },
+    {
+      "epoch": 0.596785061122341,
+      "grad_norm": 0.23924824595451355,
+      "learning_rate": 3.6963703902488907e-06,
+      "loss": 0.7693,
+      "step": 1550
+    },
+    {
+      "epoch": 0.6006352873231302,
+      "grad_norm": 0.23137369751930237,
+      "learning_rate": 3.6362721827835655e-06,
+      "loss": 0.7845,
+      "step": 1560
+    },
+    {
+      "epoch": 0.6044855135239195,
+      "grad_norm": 0.24796825647354126,
+      "learning_rate": 3.576386087436212e-06,
+      "loss": 0.7815,
+      "step": 1570
+    },
+    {
+      "epoch": 0.6083357397247088,
+      "grad_norm": 0.2561451494693756,
+      "learning_rate": 3.5167214187975913e-06,
+      "loss": 0.7705,
+      "step": 1580
+    },
+    {
+      "epoch": 0.6121859659254981,
+      "grad_norm": 0.2110113799571991,
+      "learning_rate": 3.457287457018086e-06,
+      "loss": 0.7738,
+      "step": 1590
+    },
+    {
+      "epoch": 0.6160361921262875,
+      "grad_norm": 0.23232227563858032,
+      "learning_rate": 3.3980934463643003e-06,
+      "loss": 0.7616,
+      "step": 1600
+    },
+    {
+      "epoch": 0.6198864183270767,
+      "grad_norm": 0.22184711694717407,
+      "learning_rate": 3.3391485937812053e-06,
+      "loss": 0.7433,
+      "step": 1610
+    },
+    {
+      "epoch": 0.623736644527866,
+      "grad_norm": 0.2389945536851883,
+      "learning_rate": 3.280462067460115e-06,
+      "loss": 0.7422,
+      "step": 1620
+    },
+    {
+      "epoch": 0.6275868707286553,
+      "grad_norm": 0.2308284044265747,
+      "learning_rate": 3.222042995412669e-06,
+      "loss": 0.7734,
+      "step": 1630
+    },
+    {
+      "epoch": 0.6314370969294446,
+      "grad_norm": 0.21970754861831665,
+      "learning_rate": 3.163900464051089e-06,
+      "loss": 0.7643,
+      "step": 1640
+    },
+    {
+      "epoch": 0.635287323130234,
+      "grad_norm": 0.24234025180339813,
+      "learning_rate": 3.1060435167748826e-06,
+      "loss": 0.774,
+      "step": 1650
+    },
+    {
+      "epoch": 0.6391375493310232,
+      "grad_norm": 0.23716484010219574,
+      "learning_rate": 3.048481152564253e-06,
+      "loss": 0.7539,
+      "step": 1660
+    },
+    {
+      "epoch": 0.6429877755318125,
+      "grad_norm": 0.24449272453784943,
+      "learning_rate": 2.9912223245804017e-06,
+      "loss": 0.7776,
+      "step": 1670
+    },
+    {
+      "epoch": 0.6468380017326018,
+      "grad_norm": 0.24895402789115906,
+      "learning_rate": 2.934275938772981e-06,
+      "loss": 0.7704,
+      "step": 1680
+    },
+    {
+      "epoch": 0.6506882279333911,
+      "grad_norm": 0.24309934675693512,
+      "learning_rate": 2.8776508524948676e-06,
+      "loss": 0.7617,
+      "step": 1690
+    },
+    {
+      "epoch": 0.6545384541341804,
+      "grad_norm": 0.2123984396457672,
+      "learning_rate": 2.821355873124503e-06,
+      "loss": 0.759,
+      "step": 1700
+    },
+    {
+      "epoch": 0.6583886803349697,
+      "grad_norm": 0.2443847805261612,
+      "learning_rate": 2.7653997566960144e-06,
+      "loss": 0.7759,
+      "step": 1710
+    },
+    {
+      "epoch": 0.662238906535759,
+      "grad_norm": 0.24432224035263062,
+      "learning_rate": 2.709791206537309e-06,
+      "loss": 0.7646,
+      "step": 1720
+    },
+    {
+      "epoch": 0.6660891327365482,
+      "grad_norm": 0.24617986381053925,
+      "learning_rate": 2.6545388719163747e-06,
+      "loss": 0.768,
+      "step": 1730
+    },
+    {
+      "epoch": 0.6699393589373376,
+      "grad_norm": 0.2258128821849823,
+      "learning_rate": 2.599651346695979e-06,
+      "loss": 0.7674,
+      "step": 1740
+    },
+    {
+      "epoch": 0.6737895851381268,
+      "grad_norm": 0.24280114471912384,
+      "learning_rate": 2.545137167997007e-06,
+      "loss": 0.7645,
+      "step": 1750
+    },
+    {
+      "epoch": 0.6776398113389162,
+      "grad_norm": 0.22707784175872803,
+      "learning_rate": 2.4910048148705972e-06,
+      "loss": 0.7512,
+      "step": 1760
+    },
+    {
+      "epoch": 0.6814900375397055,
+      "grad_norm": 0.2536712884902954,
+      "learning_rate": 2.4372627069793327e-06,
+      "loss": 0.7984,
+      "step": 1770
+    },
+    {
+      "epoch": 0.6853402637404947,
+      "grad_norm": 0.23620164394378662,
+      "learning_rate": 2.3839192032876486e-06,
+      "loss": 0.7805,
+      "step": 1780
+    },
+    {
+      "epoch": 0.6891904899412841,
+      "grad_norm": 0.23412460088729858,
+      "learning_rate": 2.3309826007617047e-06,
+      "loss": 0.7924,
+      "step": 1790
+    },
+    {
+      "epoch": 0.6930407161420733,
+      "grad_norm": 0.23768353462219238,
+      "learning_rate": 2.278461133078881e-06,
+      "loss": 0.7573,
+      "step": 1800
+    },
+    {
+      "epoch": 0.6968909423428626,
+      "grad_norm": 0.24352246522903442,
+      "learning_rate": 2.2263629693471258e-06,
+      "loss": 0.7475,
+      "step": 1810
+    },
+    {
+      "epoch": 0.700741168543652,
+      "grad_norm": 0.23348534107208252,
+      "learning_rate": 2.1746962128343414e-06,
+      "loss": 0.7709,
+      "step": 1820
+    },
+    {
+      "epoch": 0.7045913947444412,
+      "grad_norm": 0.22494001686573029,
+      "learning_rate": 2.1234688997080183e-06,
+      "loss": 0.7601,
+      "step": 1830
+    },
+    {
+      "epoch": 0.7084416209452306,
+      "grad_norm": 0.2611626982688904,
+      "learning_rate": 2.0726889977853053e-06,
+      "loss": 0.7647,
+      "step": 1840
+    },
+    {
+      "epoch": 0.7122918471460198,
+      "grad_norm": 0.25041767954826355,
+      "learning_rate": 2.022364405293703e-06,
+      "loss": 0.7792,
+      "step": 1850
+    },
+    {
+      "epoch": 0.7161420733468091,
+      "grad_norm": 0.22937117516994476,
+      "learning_rate": 1.9725029496425814e-06,
+      "loss": 0.7709,
+      "step": 1860
+    },
+    {
+      "epoch": 0.7199922995475985,
+      "grad_norm": 0.24669761955738068,
+      "learning_rate": 1.923112386205727e-06,
+      "loss": 0.7675,
+      "step": 1870
+    },
+    {
+      "epoch": 0.7238425257483877,
+      "grad_norm": 0.22329692542552948,
+      "learning_rate": 1.874200397115079e-06,
+      "loss": 0.7748,
+      "step": 1880
+    },
+    {
+      "epoch": 0.727692751949177,
+      "grad_norm": 0.24617932736873627,
+      "learning_rate": 1.825774590065857e-06,
+      "loss": 0.7588,
+      "step": 1890
+    },
+    {
+      "epoch": 0.7315429781499663,
+      "grad_norm": 0.2504587471485138,
+      "learning_rate": 1.7778424971332813e-06,
+      "loss": 0.7668,
+      "step": 1900
+    },
+    {
+      "epoch": 0.7353932043507556,
+      "grad_norm": 0.23082894086837769,
+      "learning_rate": 1.7304115736010407e-06,
+      "loss": 0.7635,
+      "step": 1910
+    },
+    {
+      "epoch": 0.739243430551545,
+      "grad_norm": 0.2332499474287033,
+      "learning_rate": 1.6834891968017108e-06,
+      "loss": 0.7576,
+      "step": 1920
+    },
+    {
+      "epoch": 0.7430936567523342,
+      "grad_norm": 0.24186600744724274,
+      "learning_rate": 1.6370826649692884e-06,
+      "loss": 0.7744,
+      "step": 1930
+    },
+    {
+      "epoch": 0.7469438829531235,
+      "grad_norm": 0.23990845680236816,
+      "learning_rate": 1.5911991961040463e-06,
+      "loss": 0.784,
+      "step": 1940
+    },
+    {
+      "epoch": 0.7507941091539128,
+      "grad_norm": 0.2412080466747284,
+      "learning_rate": 1.545845926849851e-06,
+      "loss": 0.7716,
+      "step": 1950
+    },
+    {
+      "epoch": 0.7546443353547021,
+      "grad_norm": 0.2554948329925537,
+      "learning_rate": 1.5010299113841397e-06,
+      "loss": 0.7879,
+      "step": 1960
+    },
+    {
+      "epoch": 0.7584945615554913,
+      "grad_norm": 0.23692354559898376,
+      "learning_rate": 1.4567581203207238e-06,
+      "loss": 0.7782,
+      "step": 1970
+    },
+    {
+      "epoch": 0.7623447877562807,
+      "grad_norm": 0.2554529011249542,
+      "learning_rate": 1.413037439625593e-06,
+      "loss": 0.7744,
+      "step": 1980
+    },
+    {
+      "epoch": 0.76619501395707,
+      "grad_norm": 0.2627207636833191,
+      "learning_rate": 1.3698746695458847e-06,
+      "loss": 0.7711,
+      "step": 1990
+    },
+    {
+      "epoch": 0.7700452401578592,
+      "grad_norm": 0.2332954853773117,
+      "learning_rate": 1.3272765235521745e-06,
+      "loss": 0.762,
+      "step": 2000
+    },
+    {
+      "epoch": 0.7738954663586486,
+      "grad_norm": 0.2289002537727356,
+      "learning_rate": 1.2852496272942877e-06,
+      "loss": 0.771,
+      "step": 2010
+    },
+    {
+      "epoch": 0.7777456925594378,
+      "grad_norm": 0.22354260087013245,
+      "learning_rate": 1.2438005175707396e-06,
+      "loss": 0.7579,
+      "step": 2020
+    },
+    {
+      "epoch": 0.7815959187602272,
+      "grad_norm": 0.23141302168369293,
+      "learning_rate": 1.2029356413120297e-06,
+      "loss": 0.743,
+      "step": 2030
+    },
+    {
+      "epoch": 0.7854461449610165,
+      "grad_norm": 0.23089073598384857,
+      "learning_rate": 1.1626613545778786e-06,
+      "loss": 0.7596,
+      "step": 2040
+    },
+    {
+      "epoch": 0.7892963711618057,
+      "grad_norm": 0.2222256362438202,
+      "learning_rate": 1.1229839215686294e-06,
+      "loss": 0.7644,
+      "step": 2050
+    },
+    {
+      "epoch": 0.7931465973625951,
+      "grad_norm": 0.22386392951011658,
+      "learning_rate": 1.0839095136509188e-06,
+      "loss": 0.7483,
+      "step": 2060
+    },
+    {
+      "epoch": 0.7969968235633843,
+      "grad_norm": 0.2468927651643753,
+      "learning_rate": 1.045444208397791e-06,
+      "loss": 0.7698,
+      "step": 2070
+    },
+    {
+      "epoch": 0.8008470497641736,
+      "grad_norm": 0.2262856513261795,
+      "learning_rate": 1.007593988643402e-06,
+      "loss": 0.7514,
+      "step": 2080
+    },
+    {
+      "epoch": 0.804697275964963,
+      "grad_norm": 0.2303187996149063,
+      "learning_rate": 9.703647415524613e-07,
+      "loss": 0.7742,
+      "step": 2090
+    },
+    {
+      "epoch": 0.8085475021657522,
+      "grad_norm": 0.23217792809009552,
+      "learning_rate": 9.337622577045485e-07,
+      "loss": 0.7626,
+      "step": 2100
+    },
+    {
+      "epoch": 0.8123977283665416,
+      "grad_norm": 0.23152533173561096,
+      "learning_rate": 8.977922301934588e-07,
+      "loss": 0.7727,
+      "step": 2110
+    },
+    {
+      "epoch": 0.8162479545673308,
+      "grad_norm": 0.23491141200065613,
+      "learning_rate": 8.624602537417009e-07,
+      "loss": 0.7745,
+      "step": 2120
+    },
+    {
+      "epoch": 0.8200981807681201,
+      "grad_norm": 0.2488638460636139,
+      "learning_rate": 8.277718238303123e-07,
+      "loss": 0.7624,
+      "step": 2130
+    },
+    {
+      "epoch": 0.8239484069689095,
+      "grad_norm": 0.22829313576221466,
+      "learning_rate": 7.937323358440935e-07,
+      "loss": 0.7642,
+      "step": 2140
+    },
+    {
+      "epoch": 0.8277986331696987,
+      "grad_norm": 0.2461923360824585,
+      "learning_rate": 7.603470842324223e-07,
+      "loss": 0.7575,
+      "step": 2150
+    },
+    {
+      "epoch": 0.831648859370488,
+      "grad_norm": 0.22681856155395508,
+      "learning_rate": 7.276212616857575e-07,
+      "loss": 0.775,
+      "step": 2160
+    },
+    {
+      "epoch": 0.8354990855712773,
+      "grad_norm": 0.2307906150817871,
+      "learning_rate": 6.955599583279837e-07,
+      "loss": 0.7722,
+      "step": 2170
+    },
+    {
+      "epoch": 0.8393493117720666,
+      "grad_norm": 0.23845583200454712,
+      "learning_rate": 6.641681609246981e-07,
+      "loss": 0.7664,
+      "step": 2180
+    },
+    {
+      "epoch": 0.8431995379728559,
+      "grad_norm": 0.22303789854049683,
+      "learning_rate": 6.334507521075711e-07,
+      "loss": 0.7708,
+      "step": 2190
+    },
+    {
+      "epoch": 0.8470497641736452,
+      "grad_norm": 0.2522687017917633,
+      "learning_rate": 6.034125096149179e-07,
+      "loss": 0.766,
+      "step": 2200
+    },
+    {
+      "epoch": 0.8508999903744345,
+      "grad_norm": 0.245204359292984,
+      "learning_rate": 5.740581055485722e-07,
+      "loss": 0.7396,
+      "step": 2210
+    },
+    {
+      "epoch": 0.8547502165752238,
+      "grad_norm": 0.22333519160747528,
+      "learning_rate": 5.453921056471961e-07,
+      "loss": 0.772,
+      "step": 2220
+    },
+    {
+      "epoch": 0.8586004427760131,
+      "grad_norm": 0.24742162227630615,
+      "learning_rate": 5.174189685761288e-07,
+      "loss": 0.7655,
+      "step": 2230
+    },
+    {
+      "epoch": 0.8624506689768023,
+      "grad_norm": 0.22996990382671356,
+      "learning_rate": 4.901430452338979e-07,
+      "loss": 0.7785,
+      "step": 2240
+    },
+    {
+      "epoch": 0.8663008951775917,
+      "grad_norm": 0.23695597052574158,
+      "learning_rate": 4.6356857807548286e-07,
+      "loss": 0.7592,
+      "step": 2250
+    },
+    {
+      "epoch": 0.870151121378381,
+      "grad_norm": 0.24071873724460602,
+      "learning_rate": 4.376997004524541e-07,
+      "loss": 0.7502,
+      "step": 2260
+    },
+    {
+      "epoch": 0.8740013475791703,
+      "grad_norm": 0.24638807773590088,
+      "learning_rate": 4.12540435970073e-07,
+      "loss": 0.7586,
+      "step": 2270
+    },
+    {
+      "epoch": 0.8778515737799596,
+      "grad_norm": 0.23039661347866058,
+      "learning_rate": 3.880946978614703e-07,
+      "loss": 0.7722,
+      "step": 2280
+    },
+    {
+      "epoch": 0.8817017999807488,
+      "grad_norm": 0.2324901670217514,
+      "learning_rate": 3.643662883789878e-07,
+      "loss": 0.7737,
+      "step": 2290
+    },
+    {
+      "epoch": 0.8855520261815382,
+      "grad_norm": 0.22476065158843994,
+      "learning_rate": 3.41358898202781e-07,
+      "loss": 0.7522,
+      "step": 2300
+    },
+    {
+      "epoch": 0.8894022523823275,
+      "grad_norm": 0.22126555442810059,
+      "learning_rate": 3.190761058667741e-07,
+      "loss": 0.7624,
+      "step": 2310
+    },
+    {
+      "epoch": 0.8932524785831167,
+      "grad_norm": 0.2299666702747345,
+      "learning_rate": 2.9752137720206463e-07,
+      "loss": 0.7676,
+      "step": 2320
+    },
+    {
+      "epoch": 0.8971027047839061,
+      "grad_norm": 0.23159314692020416,
+      "learning_rate": 2.7669806479785067e-07,
+      "loss": 0.7738,
+      "step": 2330
+    },
+    {
+      "epoch": 0.9009529309846953,
+      "grad_norm": 0.2464313805103302,
+      "learning_rate": 2.566094074799763e-07,
+      "loss": 0.7681,
+      "step": 2340
+    },
+    {
+      "epoch": 0.9048031571854847,
+      "grad_norm": 0.2339472472667694,
+      "learning_rate": 2.3725852980716557e-07,
+      "loss": 0.7495,
+      "step": 2350
+    },
+    {
+      "epoch": 0.908653383386274,
+      "grad_norm": 0.2298385500907898,
+      "learning_rate": 2.1864844158503983e-07,
+      "loss": 0.7628,
+      "step": 2360
+    },
+    {
+      "epoch": 0.9125036095870632,
+      "grad_norm": 0.25483641028404236,
+      "learning_rate": 2.007820373979702e-07,
+      "loss": 0.7559,
+      "step": 2370
+    },
+    {
+      "epoch": 0.9163538357878526,
+      "grad_norm": 0.2433347851037979,
+      "learning_rate": 1.8366209615886532e-07,
+      "loss": 0.763,
+      "step": 2380
+    },
+    {
+      "epoch": 0.9202040619886418,
+      "grad_norm": 0.2319941371679306,
+      "learning_rate": 1.6729128067693613e-07,
+      "loss": 0.7769,
+      "step": 2390
+    },
+    {
+      "epoch": 0.9240542881894311,
+      "grad_norm": 0.25519171357154846,
+      "learning_rate": 1.5167213724353426e-07,
+      "loss": 0.7769,
+      "step": 2400
+    },
+    {
+      "epoch": 0.9279045143902205,
+      "grad_norm": 0.22719712555408478,
+      "learning_rate": 1.3680709523610058e-07,
+      "loss": 0.7596,
+      "step": 2410
+    },
+    {
+      "epoch": 0.9317547405910097,
+      "grad_norm": 0.22994057834148407,
+      "learning_rate": 1.2269846674030418e-07,
+      "loss": 0.7795,
+      "step": 2420
+    },
+    {
+      "epoch": 0.935604966791799,
+      "grad_norm": 0.2410043329000473,
+      "learning_rate": 1.0934844619042595e-07,
+      "loss": 0.7783,
+      "step": 2430
+    },
+    {
+      "epoch": 0.9394551929925883,
+      "grad_norm": 0.2494816929101944,
+      "learning_rate": 9.675911002803961e-08,
+      "loss": 0.7649,
+      "step": 2440
+    },
+    {
+      "epoch": 0.9433054191933776,
+      "grad_norm": 0.2485540509223938,
+      "learning_rate": 8.493241637904381e-08,
+      "loss": 0.7612,
+      "step": 2450
+    },
+    {
+      "epoch": 0.9471556453941669,
+      "grad_norm": 0.24908973276615143,
+      "learning_rate": 7.387020474909689e-08,
+      "loss": 0.7829,
+      "step": 2460
+    },
+    {
+      "epoch": 0.9510058715949562,
+      "grad_norm": 0.2574099898338318,
+      "learning_rate": 6.357419573750856e-08,
+      "loss": 0.7762,
+      "step": 2470
+    },
+    {
+      "epoch": 0.9548560977957455,
+      "grad_norm": 0.23951750993728638,
+      "learning_rate": 5.4045990769613365e-08,
+      "loss": 0.7603,
+      "step": 2480
+    },
+    {
+      "epoch": 0.9587063239965348,
+      "grad_norm": 0.2879584729671478,
+      "learning_rate": 4.5287071847694434e-08,
+      "loss": 0.7645,
+      "step": 2490
+    },
+    {
+      "epoch": 0.9625565501973241,
+      "grad_norm": 0.2307216078042984,
+      "learning_rate": 3.7298801320467817e-08,
+      "loss": 0.7601,
+      "step": 2500
+    },
+    {
+      "epoch": 0.9664067763981133,
+      "grad_norm": 0.23883534967899323,
+      "learning_rate": 3.0082421671192576e-08,
+      "loss": 0.7536,
+      "step": 2510
+    },
+    {
+      "epoch": 0.9702570025989027,
+      "grad_norm": 0.24071452021598816,
+      "learning_rate": 2.3639055324410355e-08,
+      "loss": 0.7766,
+      "step": 2520
+    },
+    {
+      "epoch": 0.974107228799692,
+      "grad_norm": 0.2307554930448532,
+      "learning_rate": 1.7969704471369477e-08,
+      "loss": 0.7587,
+      "step": 2530
+    },
+    {
+      "epoch": 0.9779574550004813,
+      "grad_norm": 0.2208103984594345,
+      "learning_rate": 1.3075250914142413e-08,
+      "loss": 0.7621,
+      "step": 2540
+    },
+    {
+      "epoch": 0.9818076812012706,
+      "grad_norm": 0.22277307510375977,
+      "learning_rate": 8.956455928476049e-09,
+      "loss": 0.77,
+      "step": 2550
+    },
+    {
+      "epoch": 0.9856579074020598,
+      "grad_norm": 0.23475295305252075,
+      "learning_rate": 5.6139601453786275e-09,
+      "loss": 0.7628,
+      "step": 2560
+    },
+    {
+      "epoch": 0.9895081336028492,
+      "grad_norm": 0.2369798868894577,
+      "learning_rate": 3.0482834514805602e-09,
+      "loss": 0.7523,
+      "step": 2570
+    },
+    {
+      "epoch": 0.9933583598036385,
+      "grad_norm": 0.2121664136648178,
+      "learning_rate": 1.2598249081713366e-09,
+      "loss": 0.746,
+      "step": 2580
+    },
+    {
+      "epoch": 0.9972085860044277,
+      "grad_norm": 0.2545640170574188,
+      "learning_rate": 2.488626895302826e-10,
+      "loss": 0.7784,
+      "step": 2590
+    },
+    {
+      "epoch": 0.9999037443449803,
+      "step": 2597,
+      "total_flos": 4.3217624287286395e+18,
+      "train_loss": 0.7982175262974095,
+      "train_runtime": 10867.2939,
+      "train_samples_per_second": 61.184,
+      "train_steps_per_second": 0.239
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 2597,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 4.3217624287286395e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}

continuous/qwen3/mc1_l14_224_400m/finetune/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a37e821c3a0043c96fede8bad4a0f6065e61117c5ce896fafaf79dd16585954c
+size 6136

continuous/qwen3/mc1_l14_224_400m/pretrain/config.json

@@ -0,0 +1,60 @@
+{
+  "_attn_implementation_autoset": true,
+  "add_faster_video": false,
+  "add_time_instruction": false,
+  "architectures": [
+    "Qwen3ForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "bos_token_id": 151643,
+  "eos_token_id": 151645,
+  "faster_token_stride": 10,
+  "force_sample": false,
+  "head_dim": 128,
+  "hidden_act": "silu",
+  "hidden_size": 2048,
+  "image_aspect_ratio": "square",
+  "image_crop_resolution": null,
+  "image_grid_pinpoints": null,
+  "image_split_resolution": null,
+  "initializer_range": 0.02,
+  "intermediate_size": 6144,
+  "max_position_embeddings": 40960,
+  "max_window_layers": 28,
+  "mm_hidden_size": 1024,
+  "mm_newline_position": "grid",
+  "mm_patch_merge_type": "flat",
+  "mm_projector_lr": null,
+  "mm_projector_type": "mlp2x_gelu",
+  "mm_resampler_type": null,
+  "mm_spatial_pool_mode": "bilinear",
+  "mm_spatial_pool_stride": null,
+  "mm_tunable_parts": "mm_mlp_adapter",
+  "mm_use_im_patch_token": false,
+  "mm_use_im_start_end": false,
+  "mm_vision_select_feature": "patch",
+  "mm_vision_select_layer": -2,
+  "mm_vision_tower": "open_clip_hub:ViT-L-14",
+  "mm_vision_tower_lr": null,
+  "model_type": "qwen3",
+  "num_attention_heads": 16,
+  "num_hidden_layers": 28,
+  "num_key_value_heads": 8,
+  "pos_skipping_range": 4096,
+  "rms_norm_eps": 1e-06,
+  "rope_scaling": null,
+  "rope_theta": 1000000,
+  "sliding_window": null,
+  "tie_word_embeddings": true,
+  "tokenizer_model_max_length": 8192,
+  "tokenizer_padding_side": "right",
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.51.3",
+  "use_cache": true,
+  "use_mm_proj": true,
+  "use_pos_skipping": false,
+  "use_sliding_window": false,
+  "vision_tower_pretrained": "/cache/ckpt/download/tokenizer/continuous/mc1/mc1_l14_224_400m.pt",
+  "vocab_size": 151936
+}

continuous/qwen3/mc1_l14_224_400m/pretrain/mm_projector.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8c7884497e51ad7a05aa99ae6d92c22d543e87552953d554a618b914aacf2c2a
+size 25184312

continuous/qwen3/mc1_l14_224_400m/pretrain/trainer_state.json

@@ -0,0 +1,421 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.0,
+  "eval_steps": 500,
+  "global_step": 545,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.01834862385321101,
+      "grad_norm": 16.137203216552734,
+      "learning_rate": 0.00032142857142857147,
+      "loss": 9.6844,
+      "step": 10
+    },
+    {
+      "epoch": 0.03669724770642202,
+      "grad_norm": 3.938124656677246,
+      "learning_rate": 0.0006785714285714287,
+      "loss": 6.5844,
+      "step": 20
+    },
+    {
+      "epoch": 0.05504587155963303,
+      "grad_norm": 0.43354448676109314,
+      "learning_rate": 0.0009999907688176173,
+      "loss": 5.1683,
+      "step": 30
+    },
+    {
+      "epoch": 0.07339449541284404,
+      "grad_norm": 0.08924254029989243,
+      "learning_rate": 0.0009988834393115766,
+      "loss": 4.7866,
+      "step": 40
+    },
+    {
+      "epoch": 0.09174311926605505,
+      "grad_norm": 0.05785101279616356,
+      "learning_rate": 0.0009959345572940772,
+      "loss": 4.7057,
+      "step": 50
+    },
+    {
+      "epoch": 0.11009174311926606,
+      "grad_norm": 0.10574612021446228,
+      "learning_rate": 0.0009911550081155983,
+      "loss": 4.6699,
+      "step": 60
+    },
+    {
+      "epoch": 0.12844036697247707,
+      "grad_norm": 0.07545497268438339,
+      "learning_rate": 0.0009845624347566433,
+      "loss": 4.6319,
+      "step": 70
+    },
+    {
+      "epoch": 0.14678899082568808,
+      "grad_norm": 0.09925920516252518,
+      "learning_rate": 0.0009761811727013547,
+      "loss": 4.6124,
+      "step": 80
+    },
+    {
+      "epoch": 0.1651376146788991,
+      "grad_norm": 0.08575291186571121,
+      "learning_rate": 0.0009660421601067666,
+      "loss": 4.5799,
+      "step": 90
+    },
+    {
+      "epoch": 0.1834862385321101,
+      "grad_norm": 0.10520287603139877,
+      "learning_rate": 0.0009541828235992926,
+      "loss": 4.5537,
+      "step": 100
+    },
+    {
+      "epoch": 0.2018348623853211,
+      "grad_norm": 0.11979979276657104,
+      "learning_rate": 0.0009406469401200151,
+      "loss": 4.5104,
+      "step": 110
+    },
+    {
+      "epoch": 0.22018348623853212,
+      "grad_norm": 0.14063483476638794,
+      "learning_rate": 0.0009254844753287492,
+      "loss": 4.4834,
+      "step": 120
+    },
+    {
+      "epoch": 0.23853211009174313,
+      "grad_norm": 0.08678723126649857,
+      "learning_rate": 0.0009087513991633924,
+      "loss": 4.44,
+      "step": 130
+    },
+    {
+      "epoch": 0.25688073394495414,
+      "grad_norm": 0.15796740353107452,
+      "learning_rate": 0.0008905094792353917,
+      "loss": 4.4111,
+      "step": 140
+    },
+    {
+      "epoch": 0.27522935779816515,
+      "grad_norm": 0.2151155322790146,
+      "learning_rate": 0.0008708260528239789,
+      "loss": 4.398,
+      "step": 150
+    },
+    {
+      "epoch": 0.29357798165137616,
+      "grad_norm": 0.12783434987068176,
+      "learning_rate": 0.0008497737783108237,
+      "loss": 4.3699,
+      "step": 160
+    },
+    {
+      "epoch": 0.3119266055045872,
+      "grad_norm": 0.10626783221960068,
+      "learning_rate": 0.0008274303669726426,
+      "loss": 4.3471,
+      "step": 170
+    },
+    {
+      "epoch": 0.3302752293577982,
+      "grad_norm": 0.12742184102535248,
+      "learning_rate": 0.0008038782961218137,
+      "loss": 4.3242,
+      "step": 180
+    },
+    {
+      "epoch": 0.3486238532110092,
+      "grad_norm": 0.14109306037425995,
+      "learning_rate": 0.000779204504653894,
+      "loss": 4.3143,
+      "step": 190
+    },
+    {
+      "epoch": 0.3669724770642202,
+      "grad_norm": 0.1309584230184555,
+      "learning_rate": 0.0007535000721258791,
+      "loss": 4.2902,
+      "step": 200
+    },
+    {
+      "epoch": 0.3853211009174312,
+      "grad_norm": 0.12982024252414703,
+      "learning_rate": 0.00072685988254984,
+      "loss": 4.2796,
+      "step": 210
+    },
+    {
+      "epoch": 0.4036697247706422,
+      "grad_norm": 0.12430711835622787,
+      "learning_rate": 0.0006993822741429906,
+      "loss": 4.2866,
+      "step": 220
+    },
+    {
+      "epoch": 0.42201834862385323,
+      "grad_norm": 0.13445371389389038,
+      "learning_rate": 0.0006711686763270818,
+      "loss": 4.2761,
+      "step": 230
+    },
+    {
+      "epoch": 0.44036697247706424,
+      "grad_norm": 0.10663062334060669,
+      "learning_rate": 0.0006423232353170831,
+      "loss": 4.2614,
+      "step": 240
+    },
+    {
+      "epoch": 0.45871559633027525,
+      "grad_norm": 0.11505359411239624,
+      "learning_rate": 0.0006129524296812335,
+      "loss": 4.24,
+      "step": 250
+    },
+    {
+      "epoch": 0.47706422018348627,
+      "grad_norm": 0.11271247267723083,
+      "learning_rate": 0.0005831646772915651,
+      "loss": 4.2482,
+      "step": 260
+    },
+    {
+      "epoch": 0.4954128440366973,
+      "grad_norm": 0.10678277909755707,
+      "learning_rate": 0.0005530699351157809,
+      "loss": 4.2243,
+      "step": 270
+    },
+    {
+      "epoch": 0.5137614678899083,
+      "grad_norm": 0.11792612820863724,
+      "learning_rate": 0.0005227792933277943,
+      "loss": 4.2159,
+      "step": 280
+    },
+    {
+      "epoch": 0.5321100917431193,
+      "grad_norm": 0.10429419577121735,
+      "learning_rate": 0.0004924045652352092,
+      "loss": 4.1995,
+      "step": 290
+    },
+    {
+      "epoch": 0.5504587155963303,
+      "grad_norm": 0.11379282176494598,
+      "learning_rate": 0.0004620578745374633,
+      "loss": 4.1949,
+      "step": 300
+    },
+    {
+      "epoch": 0.5688073394495413,
+      "grad_norm": 0.1280711144208908,
+      "learning_rate": 0.0004318512414382059,
+      "loss": 4.2041,
+      "step": 310
+    },
+    {
+      "epoch": 0.5871559633027523,
+      "grad_norm": 0.11540122330188751,
+      "learning_rate": 0.0004018961691397148,
+      "loss": 4.1935,
+      "step": 320
+    },
+    {
+      "epoch": 0.6055045871559633,
+      "grad_norm": 0.10324843227863312,
+      "learning_rate": 0.0003723032322457458,
+      "loss": 4.1686,
+      "step": 330
+    },
+    {
+      "epoch": 0.6238532110091743,
+      "grad_norm": 0.16367778182029724,
+      "learning_rate": 0.0003431816685921625,
+      "loss": 4.1653,
+      "step": 340
+    },
+    {
+      "epoch": 0.6422018348623854,
+      "grad_norm": 0.13228970766067505,
+      "learning_rate": 0.0003146389760120416,
+      "loss": 4.164,
+      "step": 350
+    },
+    {
+      "epoch": 0.6605504587155964,
+      "grad_norm": 0.11159172654151917,
+      "learning_rate": 0.00028678051552373485,
+      "loss": 4.1694,
+      "step": 360
+    },
+    {
+      "epoch": 0.6788990825688074,
+      "grad_norm": 0.10820302367210388,
+      "learning_rate": 0.0002597091224066581,
+      "loss": 4.145,
+      "step": 370
+    },
+    {
+      "epoch": 0.6972477064220184,
+      "grad_norm": 0.1050753965973854,
+      "learning_rate": 0.0002335247266004629,
+      "loss": 4.1576,
+      "step": 380
+    },
+    {
+      "epoch": 0.7155963302752294,
+      "grad_norm": 0.12846793234348297,
+      "learning_rate": 0.000208323983828828,
+      "loss": 4.1363,
+      "step": 390
+    },
+    {
+      "epoch": 0.7339449541284404,
+      "grad_norm": 0.1170160248875618,
+      "learning_rate": 0.0001841999188095224,
+      "loss": 4.1335,
+      "step": 400
+    },
+    {
+      "epoch": 0.7522935779816514,
+      "grad_norm": 0.10726332664489746,
+      "learning_rate": 0.00016124158186777676,
+      "loss": 4.1282,
+      "step": 410
+    },
+    {
+      "epoch": 0.7706422018348624,
+      "grad_norm": 0.10912300646305084,
+      "learning_rate": 0.0001395337202205211,
+      "loss": 4.1174,
+      "step": 420
+    },
+    {
+      "epoch": 0.7889908256880734,
+      "grad_norm": 0.10090328007936478,
+      "learning_rate": 0.00011915646514489292,
+      "loss": 4.1203,
+      "step": 430
+    },
+    {
+      "epoch": 0.8073394495412844,
+      "grad_norm": 0.1216532364487648,
+      "learning_rate": 0.00010018503618578818,
+      "loss": 4.1226,
+      "step": 440
+    },
+    {
+      "epoch": 0.8256880733944955,
+      "grad_norm": 0.11492498219013214,
+      "learning_rate": 8.268946349432583e-05,
+      "loss": 4.1057,
+      "step": 450
+    },
+    {
+      "epoch": 0.8440366972477065,
+      "grad_norm": 0.10929980129003525,
+      "learning_rate": 6.673432932217117e-05,
+      "loss": 4.0948,
+      "step": 460
+    },
+    {
+      "epoch": 0.8623853211009175,
+      "grad_norm": 0.1102873906493187,
+      "learning_rate": 5.2378529625954684e-05,
+      "loss": 4.0832,
+      "step": 470
+    },
+    {
+      "epoch": 0.8807339449541285,
+      "grad_norm": 0.1124844029545784,
+      "learning_rate": 3.9675056661785556e-05,
+      "loss": 4.0868,
+      "step": 480
+    },
+    {
+      "epoch": 0.8990825688073395,
+      "grad_norm": 0.1044391617178917,
+      "learning_rate": 2.8670803372379538e-05,
+      "loss": 4.0898,
+      "step": 490
+    },
+    {
+      "epoch": 0.9174311926605505,
+      "grad_norm": 0.10904878377914429,
+      "learning_rate": 1.9406390288876586e-05,
+      "loss": 4.0979,
+      "step": 500
+    },
+    {
+      "epoch": 0.9357798165137615,
+      "grad_norm": 0.11043853312730789,
+      "learning_rate": 1.191601558631461e-05,
+      "loss": 4.0854,
+      "step": 510
+    },
+    {
+      "epoch": 0.9541284403669725,
+      "grad_norm": 0.1087583675980568,
+      "learning_rate": 6.2273288462559305e-06,
+      "loss": 4.0782,
+      "step": 520
+    },
+    {
+      "epoch": 0.9724770642201835,
+      "grad_norm": 0.10724236071109772,
+      "learning_rate": 2.361328992552314e-06,
+      "loss": 4.0808,
+      "step": 530
+    },
+    {
+      "epoch": 0.9908256880733946,
+      "grad_norm": 0.10512278228998184,
+      "learning_rate": 3.322867770044202e-07,
+      "loss": 4.0825,
+      "step": 540
+    },
+    {
+      "epoch": 1.0,
+      "step": 545,
+      "total_flos": 3.4196958409549414e+17,
+      "train_loss": 4.4287514747829615,
+      "train_runtime": 5667.8399,
+      "train_samples_per_second": 98.473,
+      "train_steps_per_second": 0.096
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 545,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": false,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.4196958409549414e+17,
+  "train_batch_size": 64,
+  "trial_name": null,
+  "trial_params": null
+}

continuous/qwen3/mc2_b16_224/finetune/added_tokens.json

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

continuous/qwen3/mc2_b16_224/finetune/config.json

@@ -0,0 +1,59 @@
+{
+  "add_faster_video": false,
+  "add_time_instruction": false,
+  "architectures": [
+    "LlavaQwen3ForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "bos_token_id": 151643,
+  "eos_token_id": 151645,
+  "faster_token_stride": 10,
+  "force_sample": false,
+  "head_dim": 128,
+  "hidden_act": "silu",
+  "hidden_size": 2048,
+  "image_aspect_ratio": "square",
+  "image_crop_resolution": null,
+  "image_grid_pinpoints": null,
+  "image_split_resolution": null,
+  "initializer_range": 0.02,
+  "intermediate_size": 6144,
+  "max_position_embeddings": 40960,
+  "max_window_layers": 28,
+  "mm_hidden_size": 768,
+  "mm_newline_position": "grid",
+  "mm_patch_merge_type": "flat",
+  "mm_projector_lr": 0.0001,
+  "mm_projector_type": "mlp2x_gelu",
+  "mm_resampler_type": null,
+  "mm_spatial_pool_mode": "bilinear",
+  "mm_spatial_pool_stride": null,
+  "mm_tunable_parts": "mm_mlp_adapter,mm_language_model",
+  "mm_use_im_patch_token": false,
+  "mm_use_im_start_end": false,
+  "mm_vision_select_feature": "patch",
+  "mm_vision_select_layer": -2,
+  "mm_vision_tower": "open_clip_hub:ViT-B-16",
+  "mm_vision_tower_lr": null,
+  "model_type": "qwen3",
+  "num_attention_heads": 16,
+  "num_hidden_layers": 28,
+  "num_key_value_heads": 8,
+  "pos_skipping_range": 4096,
+  "rms_norm_eps": 1e-06,
+  "rope_scaling": null,
+  "rope_theta": 1000000,
+  "sliding_window": null,
+  "tie_word_embeddings": true,
+  "tokenizer_model_max_length": 8192,
+  "tokenizer_padding_side": "right",
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.51.3",
+  "use_cache": true,
+  "use_mm_proj": true,
+  "use_pos_skipping": false,
+  "use_sliding_window": false,
+  "vision_tower_pretrained": "/cache/ckpt/download/tokenizer/continuous/mc2/mc2_b16_224.pt",
+  "vocab_size": 151936
+}

continuous/qwen3/mc2_b16_224/finetune/generation_config.json

@@ -0,0 +1,13 @@
+{
+  "bos_token_id": 151643,
+  "do_sample": true,
+  "eos_token_id": [
+    151645,
+    151643
+  ],
+  "pad_token_id": 151643,
+  "temperature": 0.6,
+  "top_k": 20,
+  "top_p": 0.95,
+  "transformers_version": "4.51.3"
+}

continuous/qwen3/mc2_b16_224/finetune/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6fb981e0c78353f4c0e1c1b246fddad050b185ddbfb0bbc3ecbae801c97e6cae
+size 4259008248

continuous/qwen3/mc2_b16_224/finetune/special_tokens_map.json

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

continuous/qwen3/mc2_b16_224/finetune/tokenizer.json

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

continuous/qwen3/mc2_b16_224/finetune/tokenizer_config.json

@@ -0,0 +1,241 @@
+{
+  "add_bos_token": false,
+  "add_prefix_space": false,
+  "added_tokens_decoder": {
+    "151643": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151644": {
+      "content": "<|im_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151645": {
+      "content": "<|im_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151646": {
+      "content": "<|object_ref_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151647": {
+      "content": "<|object_ref_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151648": {
+      "content": "<|box_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151649": {
+      "content": "<|box_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151650": {
+      "content": "<|quad_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151651": {
+      "content": "<|quad_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151652": {
+      "content": "<|vision_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151653": {
+      "content": "<|vision_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151654": {
+      "content": "<|vision_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151655": {
+      "content": "<|image_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151656": {
+      "content": "<|video_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151657": {
+      "content": "<tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151658": {
+      "content": "</tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151659": {
+      "content": "<|fim_prefix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151660": {
+      "content": "<|fim_middle|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151661": {
+      "content": "<|fim_suffix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151662": {
+      "content": "<|fim_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151663": {
+      "content": "<|repo_name|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151664": {
+      "content": "<|file_sep|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151665": {
+      "content": "<tool_response>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151666": {
+      "content": "</tool_response>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151667": {
+      "content": "<think>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151668": {
+      "content": "</think>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    }
+  },
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "bos_token": null,
+  "chat_template": "{%- if tools %}\n    {{- '<|im_start|>system\\n' }}\n    {%- if messages[0].role == 'system' %}\n        {{- messages[0].content + '\\n\\n' }}\n    {%- endif %}\n    {{- \"# Tools\\n\\nYou may call one or more functions to assist with the user query.\\n\\nYou are provided with function signatures within <tools></tools> XML tags:\\n<tools>\" }}\n    {%- for tool in tools %}\n        {{- \"\\n\" }}\n        {{- tool | tojson }}\n    {%- endfor %}\n    {{- \"\\n</tools>\\n\\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\\n<tool_call>\\n{\\\"name\\\": <function-name>, \\\"arguments\\\": <args-json-object>}\\n</tool_call><|im_end|>\\n\" }}\n{%- else %}\n    {%- if messages[0].role == 'system' %}\n        {{- '<|im_start|>system\\n' + messages[0].content + '<|im_end|>\\n' }}\n    {%- endif %}\n{%- endif %}\n{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}\n{%- for message in messages[::-1] %}\n    {%- set index = (messages|length - 1) - loop.index0 %}\n    {%- if ns.multi_step_tool and message.role == \"user\" and message.content is string and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}\n        {%- set ns.multi_step_tool = false %}\n        {%- set ns.last_query_index = index %}\n    {%- endif %}\n{%- endfor %}\n{%- for message in messages %}\n    {%- if message.content is string %}\n        {%- set content = message.content %}\n    {%- else %}\n        {%- set content = '' %}\n    {%- endif %}\n    {%- if (message.role == \"user\") or (message.role == \"system\" and not loop.first) %}\n        {{- '<|im_start|>' + message.role + '\\n' + content + '<|im_end|>' + '\\n' }}\n    {%- elif message.role == \"assistant\" %}\n        {%- set reasoning_content = '' %}\n        {%- if message.reasoning_content is string %}\n            {%- set reasoning_content = message.reasoning_content %}\n        {%- else %}\n            {%- if '</think>' in content %}\n                {%- set reasoning_content = content.split('</think>')[0].rstrip('\\n').split('<think>')[-1].lstrip('\\n') %}\n                {%- set content = content.split('</think>')[-1].lstrip('\\n') %}\n            {%- endif %}\n        {%- endif %}\n        {%- if loop.index0 > ns.last_query_index %}\n            {%- if loop.last or (not loop.last and reasoning_content) %}\n                {{- '<|im_start|>' + message.role + '\\n<think>\\n' + reasoning_content.strip('\\n') + '\\n</think>\\n\\n' + content.lstrip('\\n') }}\n            {%- else %}\n                {{- '<|im_start|>' + message.role + '\\n' + content }}\n            {%- endif %}\n        {%- else %}\n            {{- '<|im_start|>' + message.role + '\\n' + content }}\n        {%- endif %}\n        {%- if message.tool_calls %}\n            {%- for tool_call in message.tool_calls %}\n                {%- if (loop.first and content) or (not loop.first) %}\n                    {{- '\\n' }}\n                {%- endif %}\n                {%- if tool_call.function %}\n                    {%- set tool_call = tool_call.function %}\n                {%- endif %}\n                {{- '<tool_call>\\n{\"name\": \"' }}\n                {{- tool_call.name }}\n                {{- '\", \"arguments\": ' }}\n                {%- if tool_call.arguments is string %}\n                    {{- tool_call.arguments }}\n                {%- else %}\n                    {{- tool_call.arguments | tojson }}\n                {%- endif %}\n                {{- '}\\n</tool_call>' }}\n            {%- endfor %}\n        {%- endif %}\n        {{- '<|im_end|>\\n' }}\n    {%- elif message.role == \"tool\" %}\n        {%- if loop.first or (messages[loop.index0 - 1].role != \"tool\") %}\n            {{- '<|im_start|>user' }}\n        {%- endif %}\n        {{- '\\n<tool_response>\\n' }}\n        {{- content }}\n        {{- '\\n</tool_response>' }}\n        {%- if loop.last or (messages[loop.index0 + 1].role != \"tool\") %}\n            {{- '<|im_end|>\\n' }}\n        {%- endif %}\n    {%- endif %}\n{%- endfor %}\n{%- if add_generation_prompt %}\n    {{- '<|im_start|>assistant\\n' }}\n    {%- if enable_thinking is defined and enable_thinking is false %}\n        {{- '<think>\\n\\n</think>\\n\\n' }}\n    {%- endif %}\n{%- endif %}",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|im_end|>",
+  "errors": "replace",
+  "extra_special_tokens": {},
+  "model_max_length": 8192,
+  "pad_token": "<|endoftext|>",
+  "padding_side": "right",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}

continuous/qwen3/mc2_b16_224/finetune/trainer_state.json

@@ -0,0 +1,1856 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.9999037443449803,
+  "eval_steps": 500,
+  "global_step": 2597,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.003850226200789296,
+      "grad_norm": 24.383756637573242,
+      "learning_rate": 1.153846153846154e-06,
+      "loss": 2.6764,
+      "step": 10
+    },
+    {
+      "epoch": 0.007700452401578592,
+      "grad_norm": 8.538256645202637,
+      "learning_rate": 2.435897435897436e-06,
+      "loss": 2.0943,
+      "step": 20
+    },
+    {
+      "epoch": 0.01155067860236789,
+      "grad_norm": 4.648759841918945,
+      "learning_rate": 3.7179487179487184e-06,
+      "loss": 1.5625,
+      "step": 30
+    },
+    {
+      "epoch": 0.015400904803157185,
+      "grad_norm": 1.5024904012680054,
+      "learning_rate": 5e-06,
+      "loss": 1.1778,
+      "step": 40
+    },
+    {
+      "epoch": 0.019251131003946483,
+      "grad_norm": 0.7831487655639648,
+      "learning_rate": 6.282051282051282e-06,
+      "loss": 1.047,
+      "step": 50
+    },
+    {
+      "epoch": 0.02310135720473578,
+      "grad_norm": 0.4106180965900421,
+      "learning_rate": 7.564102564102564e-06,
+      "loss": 0.9502,
+      "step": 60
+    },
+    {
+      "epoch": 0.026951583405525073,
+      "grad_norm": 0.3043743371963501,
+      "learning_rate": 8.846153846153847e-06,
+      "loss": 0.9016,
+      "step": 70
+    },
+    {
+      "epoch": 0.03080180960631437,
+      "grad_norm": 0.2947506606578827,
+      "learning_rate": 9.999996111488723e-06,
+      "loss": 0.8986,
+      "step": 80
+    },
+    {
+      "epoch": 0.034652035807103666,
+      "grad_norm": 0.28330516815185547,
+      "learning_rate": 9.999529497453782e-06,
+      "loss": 0.867,
+      "step": 90
+    },
+    {
+      "epoch": 0.038502262007892966,
+      "grad_norm": 0.24412478506565094,
+      "learning_rate": 9.998285264324193e-06,
+      "loss": 0.8464,
+      "step": 100
+    },
+    {
+      "epoch": 0.04235248820868226,
+      "grad_norm": 0.2534524202346802,
+      "learning_rate": 9.996263605626052e-06,
+      "loss": 0.8307,
+      "step": 110
+    },
+    {
+      "epoch": 0.04620271440947156,
+      "grad_norm": 0.25276076793670654,
+      "learning_rate": 9.99346483580503e-06,
+      "loss": 0.8401,
+      "step": 120
+    },
+    {
+      "epoch": 0.05005294061026085,
+      "grad_norm": 0.24338100850582123,
+      "learning_rate": 9.989889390177467e-06,
+      "loss": 0.8332,
+      "step": 130
+    },
+    {
+      "epoch": 0.053903166811050146,
+      "grad_norm": 0.2805628478527069,
+      "learning_rate": 9.985537824862647e-06,
+      "loss": 0.8308,
+      "step": 140
+    },
+    {
+      "epoch": 0.057753393011839446,
+      "grad_norm": 0.2955642342567444,
+      "learning_rate": 9.980410816696317e-06,
+      "loss": 0.8143,
+      "step": 150
+    },
+    {
+      "epoch": 0.06160361921262874,
+      "grad_norm": 0.25917601585388184,
+      "learning_rate": 9.97450916312541e-06,
+      "loss": 0.8223,
+      "step": 160
+    },
+    {
+      "epoch": 0.06545384541341803,
+      "grad_norm": 0.21919578313827515,
+      "learning_rate": 9.967833782084004e-06,
+      "loss": 0.8228,
+      "step": 170
+    },
+    {
+      "epoch": 0.06930407161420733,
+      "grad_norm": 0.2581384479999542,
+      "learning_rate": 9.960385711850554e-06,
+      "loss": 0.8477,
+      "step": 180
+    },
+    {
+      "epoch": 0.07315429781499663,
+      "grad_norm": 0.25717639923095703,
+      "learning_rate": 9.952166110886394e-06,
+      "loss": 0.8395,
+      "step": 190
+    },
+    {
+      "epoch": 0.07700452401578593,
+      "grad_norm": 0.2712290585041046,
+      "learning_rate": 9.943176257655567e-06,
+      "loss": 0.8263,
+      "step": 200
+    },
+    {
+      "epoch": 0.08085475021657522,
+      "grad_norm": 0.2560865879058838,
+      "learning_rate": 9.933417550425952e-06,
+      "loss": 0.8257,
+      "step": 210
+    },
+    {
+      "epoch": 0.08470497641736452,
+      "grad_norm": 0.24958431720733643,
+      "learning_rate": 9.9228915070518e-06,
+      "loss": 0.8085,
+      "step": 220
+    },
+    {
+      "epoch": 0.08855520261815382,
+      "grad_norm": 0.24898356199264526,
+      "learning_rate": 9.911599764737632e-06,
+      "loss": 0.8099,
+      "step": 230
+    },
+    {
+      "epoch": 0.09240542881894312,
+      "grad_norm": 0.24602769315242767,
+      "learning_rate": 9.899544079783606e-06,
+      "loss": 0.8052,
+      "step": 240
+    },
+    {
+      "epoch": 0.0962556550197324,
+      "grad_norm": 0.2779531180858612,
+      "learning_rate": 9.886726327312338e-06,
+      "loss": 0.8237,
+      "step": 250
+    },
+    {
+      "epoch": 0.1001058812205217,
+      "grad_norm": 0.23343147337436676,
+      "learning_rate": 9.87314850097724e-06,
+      "loss": 0.8165,
+      "step": 260
+    },
+    {
+      "epoch": 0.103956107421311,
+      "grad_norm": 0.24984779953956604,
+      "learning_rate": 9.858812712652452e-06,
+      "loss": 0.8063,
+      "step": 270
+    },
+    {
+      "epoch": 0.10780633362210029,
+      "grad_norm": 0.24367475509643555,
+      "learning_rate": 9.843721192104341e-06,
+      "loss": 0.8042,
+      "step": 280
+    },
+    {
+      "epoch": 0.11165655982288959,
+      "grad_norm": 0.2383926957845688,
+      "learning_rate": 9.827876286644698e-06,
+      "loss": 0.8235,
+      "step": 290
+    },
+    {
+      "epoch": 0.11550678602367889,
+      "grad_norm": 0.2280806452035904,
+      "learning_rate": 9.811280460765643e-06,
+      "loss": 0.7862,
+      "step": 300
+    },
+    {
+      "epoch": 0.11935701222446819,
+      "grad_norm": 0.23371094465255737,
+      "learning_rate": 9.793936295756292e-06,
+      "loss": 0.7972,
+      "step": 310
+    },
+    {
+      "epoch": 0.12320723842525748,
+      "grad_norm": 0.23976285755634308,
+      "learning_rate": 9.775846489301279e-06,
+      "loss": 0.798,
+      "step": 320
+    },
+    {
+      "epoch": 0.12705746462604678,
+      "grad_norm": 0.24407920241355896,
+      "learning_rate": 9.757013855061157e-06,
+      "loss": 0.7978,
+      "step": 330
+    },
+    {
+      "epoch": 0.13090769082683607,
+      "grad_norm": 0.2385709434747696,
+      "learning_rate": 9.737441322234758e-06,
+      "loss": 0.8089,
+      "step": 340
+    },
+    {
+      "epoch": 0.13475791702762538,
+      "grad_norm": 0.24808724224567413,
+      "learning_rate": 9.717131935103605e-06,
+      "loss": 0.8091,
+      "step": 350
+    },
+    {
+      "epoch": 0.13860814322841467,
+      "grad_norm": 0.2395351082086563,
+      "learning_rate": 9.696088852558398e-06,
+      "loss": 0.7966,
+      "step": 360
+    },
+    {
+      "epoch": 0.14245836942920398,
+      "grad_norm": 0.2530341148376465,
+      "learning_rate": 9.674315347607686e-06,
+      "loss": 0.8223,
+      "step": 370
+    },
+    {
+      "epoch": 0.14630859562999327,
+      "grad_norm": 0.2417709082365036,
+      "learning_rate": 9.651814806868796e-06,
+      "loss": 0.7832,
+      "step": 380
+    },
+    {
+      "epoch": 0.15015882183078255,
+      "grad_norm": 0.22917094826698303,
+      "learning_rate": 9.62859073004107e-06,
+      "loss": 0.7868,
+      "step": 390
+    },
+    {
+      "epoch": 0.15400904803157187,
+      "grad_norm": 0.23599404096603394,
+      "learning_rate": 9.604646729361547e-06,
+      "loss": 0.7885,
+      "step": 400
+    },
+    {
+      "epoch": 0.15785927423236115,
+      "grad_norm": 0.2259943187236786,
+      "learning_rate": 9.579986529043099e-06,
+      "loss": 0.7771,
+      "step": 410
+    },
+    {
+      "epoch": 0.16170950043315044,
+      "grad_norm": 0.23437830805778503,
+      "learning_rate": 9.554613964695189e-06,
+      "loss": 0.7979,
+      "step": 420
+    },
+    {
+      "epoch": 0.16555972663393975,
+      "grad_norm": 0.2439284473657608,
+      "learning_rate": 9.528532982727281e-06,
+      "loss": 0.7891,
+      "step": 430
+    },
+    {
+      "epoch": 0.16940995283472904,
+      "grad_norm": 0.23302191495895386,
+      "learning_rate": 9.501747639735018e-06,
+      "loss": 0.7849,
+      "step": 440
+    },
+    {
+      "epoch": 0.17326017903551832,
+      "grad_norm": 0.2382429987192154,
+      "learning_rate": 9.474262101869272e-06,
+      "loss": 0.7745,
+      "step": 450
+    },
+    {
+      "epoch": 0.17711040523630764,
+      "grad_norm": 0.24534501135349274,
+      "learning_rate": 9.446080644188141e-06,
+      "loss": 0.7815,
+      "step": 460
+    },
+    {
+      "epoch": 0.18096063143709692,
+      "grad_norm": 0.2301216423511505,
+      "learning_rate": 9.417207649992017e-06,
+      "loss": 0.786,
+      "step": 470
+    },
+    {
+      "epoch": 0.18481085763788624,
+      "grad_norm": 0.2358303666114807,
+      "learning_rate": 9.387647610141816e-06,
+      "loss": 0.7851,
+      "step": 480
+    },
+    {
+      "epoch": 0.18866108383867553,
+      "grad_norm": 0.22527718544006348,
+      "learning_rate": 9.35740512236047e-06,
+      "loss": 0.772,
+      "step": 490
+    },
+    {
+      "epoch": 0.1925113100394648,
+      "grad_norm": 0.2324734479188919,
+      "learning_rate": 9.326484890517809e-06,
+      "loss": 0.7807,
+      "step": 500
+    },
+    {
+      "epoch": 0.19636153624025413,
+      "grad_norm": 0.22886815667152405,
+      "learning_rate": 9.294891723898926e-06,
+      "loss": 0.7745,
+      "step": 510
+    },
+    {
+      "epoch": 0.2002117624410434,
+      "grad_norm": 0.23749478161334991,
+      "learning_rate": 9.262630536456153e-06,
+      "loss": 0.8039,
+      "step": 520
+    },
+    {
+      "epoch": 0.2040619886418327,
+      "grad_norm": 0.24932222068309784,
+      "learning_rate": 9.229706346044749e-06,
+      "loss": 0.7746,
+      "step": 530
+    },
+    {
+      "epoch": 0.207912214842622,
+      "grad_norm": 0.25207170844078064,
+      "learning_rate": 9.196124273642422e-06,
+      "loss": 0.7897,
+      "step": 540
+    },
+    {
+      "epoch": 0.2117624410434113,
+      "grad_norm": 0.22541454434394836,
+      "learning_rate": 9.161889542552842e-06,
+      "loss": 0.7942,
+      "step": 550
+    },
+    {
+      "epoch": 0.21561266724420058,
+      "grad_norm": 0.2182478904724121,
+      "learning_rate": 9.127007477593193e-06,
+      "loss": 0.7956,
+      "step": 560
+    },
+    {
+      "epoch": 0.2194628934449899,
+      "grad_norm": 0.2560698688030243,
+      "learning_rate": 9.091483504265974e-06,
+      "loss": 0.8019,
+      "step": 570
+    },
+    {
+      "epoch": 0.22331311964577918,
+      "grad_norm": 0.2510928809642792,
+      "learning_rate": 9.055323147915116e-06,
+      "loss": 0.7848,
+      "step": 580
+    },
+    {
+      "epoch": 0.2271633458465685,
+      "grad_norm": 0.24459414184093475,
+      "learning_rate": 9.018532032866585e-06,
+      "loss": 0.7928,
+      "step": 590
+    },
+    {
+      "epoch": 0.23101357204735778,
+      "grad_norm": 0.21618451178073883,
+      "learning_rate": 8.981115881553579e-06,
+      "loss": 0.7847,
+      "step": 600
+    },
+    {
+      "epoch": 0.23486379824814707,
+      "grad_norm": 0.2531362473964691,
+      "learning_rate": 8.943080513626476e-06,
+      "loss": 0.8052,
+      "step": 610
+    },
+    {
+      "epoch": 0.23871402444893638,
+      "grad_norm": 0.23429495096206665,
+      "learning_rate": 8.904431845047641e-06,
+      "loss": 0.7901,
+      "step": 620
+    },
+    {
+      "epoch": 0.24256425064972567,
+      "grad_norm": 0.2453940063714981,
+      "learning_rate": 8.86517588717128e-06,
+      "loss": 0.8057,
+      "step": 630
+    },
+    {
+      "epoch": 0.24641447685051496,
+      "grad_norm": 0.23376856744289398,
+      "learning_rate": 8.82531874580844e-06,
+      "loss": 0.7983,
+      "step": 640
+    },
+    {
+      "epoch": 0.25026470305130427,
+      "grad_norm": 0.2224569171667099,
+      "learning_rate": 8.784866620277314e-06,
+      "loss": 0.7546,
+      "step": 650
+    },
+    {
+      "epoch": 0.25411492925209356,
+      "grad_norm": 0.250343382358551,
+      "learning_rate": 8.743825802439018e-06,
+      "loss": 0.7858,
+      "step": 660
+    },
+    {
+      "epoch": 0.25796515545288284,
+      "grad_norm": 0.26577088236808777,
+      "learning_rate": 8.702202675718956e-06,
+      "loss": 0.7974,
+      "step": 670
+    },
+    {
+      "epoch": 0.26181538165367213,
+      "grad_norm": 0.22037500143051147,
+      "learning_rate": 8.660003714113959e-06,
+      "loss": 0.7637,
+      "step": 680
+    },
+    {
+      "epoch": 0.26566560785446147,
+      "grad_norm": 0.22400137782096863,
+      "learning_rate": 8.617235481185323e-06,
+      "loss": 0.7873,
+      "step": 690
+    },
+    {
+      "epoch": 0.26951583405525076,
+      "grad_norm": 0.24871869385242462,
+      "learning_rate": 8.573904629037931e-06,
+      "loss": 0.7942,
+      "step": 700
+    },
+    {
+      "epoch": 0.27336606025604004,
+      "grad_norm": 0.2336120456457138,
+      "learning_rate": 8.530017897285585e-06,
+      "loss": 0.7761,
+      "step": 710
+    },
+    {
+      "epoch": 0.27721628645682933,
+      "grad_norm": 0.21100538969039917,
+      "learning_rate": 8.485582112002746e-06,
+      "loss": 0.7797,
+      "step": 720
+    },
+    {
+      "epoch": 0.2810665126576186,
+      "grad_norm": 0.22762100398540497,
+      "learning_rate": 8.440604184662811e-06,
+      "loss": 0.7838,
+      "step": 730
+    },
+    {
+      "epoch": 0.28491673885840796,
+      "grad_norm": 0.22541679441928864,
+      "learning_rate": 8.395091111063118e-06,
+      "loss": 0.7708,
+      "step": 740
+    },
+    {
+      "epoch": 0.28876696505919724,
+      "grad_norm": 0.22223304212093353,
+      "learning_rate": 8.349049970236822e-06,
+      "loss": 0.7732,
+      "step": 750
+    },
+    {
+      "epoch": 0.29261719125998653,
+      "grad_norm": 0.2469939887523651,
+      "learning_rate": 8.302487923351842e-06,
+      "loss": 0.7651,
+      "step": 760
+    },
+    {
+      "epoch": 0.2964674174607758,
+      "grad_norm": 0.22449439764022827,
+      "learning_rate": 8.255412212597021e-06,
+      "loss": 0.7823,
+      "step": 770
+    },
+    {
+      "epoch": 0.3003176436615651,
+      "grad_norm": 0.22803059220314026,
+      "learning_rate": 8.207830160055688e-06,
+      "loss": 0.7802,
+      "step": 780
+    },
+    {
+      "epoch": 0.3041678698623544,
+      "grad_norm": 0.23653031885623932,
+      "learning_rate": 8.159749166566786e-06,
+      "loss": 0.7633,
+      "step": 790
+    },
+    {
+      "epoch": 0.30801809606314373,
+      "grad_norm": 0.2243533432483673,
+      "learning_rate": 8.11117671057377e-06,
+      "loss": 0.7907,
+      "step": 800
+    },
+    {
+      "epoch": 0.311868322263933,
+      "grad_norm": 0.23717503249645233,
+      "learning_rate": 8.062120346961404e-06,
+      "loss": 0.7825,
+      "step": 810
+    },
+    {
+      "epoch": 0.3157185484647223,
+      "grad_norm": 0.22248737514019012,
+      "learning_rate": 8.012587705880705e-06,
+      "loss": 0.7757,
+      "step": 820
+    },
+    {
+      "epoch": 0.3195687746655116,
+      "grad_norm": 0.23530352115631104,
+      "learning_rate": 7.962586491562146e-06,
+      "loss": 0.762,
+      "step": 830
+    },
+    {
+      "epoch": 0.3234190008663009,
+      "grad_norm": 0.23507733643054962,
+      "learning_rate": 7.912124481117357e-06,
+      "loss": 0.7681,
+      "step": 840
+    },
+    {
+      "epoch": 0.3272692270670902,
+      "grad_norm": 0.2354765385389328,
+      "learning_rate": 7.861209523329494e-06,
+      "loss": 0.7813,
+      "step": 850
+    },
+    {
+      "epoch": 0.3311194532678795,
+      "grad_norm": 0.22397348284721375,
+      "learning_rate": 7.809849537432432e-06,
+      "loss": 0.7611,
+      "step": 860
+    },
+    {
+      "epoch": 0.3349696794686688,
+      "grad_norm": 0.2261405736207962,
+      "learning_rate": 7.758052511879035e-06,
+      "loss": 0.7734,
+      "step": 870
+    },
+    {
+      "epoch": 0.3388199056694581,
+      "grad_norm": 0.21974846720695496,
+      "learning_rate": 7.705826503098642e-06,
+      "loss": 0.7689,
+      "step": 880
+    },
+    {
+      "epoch": 0.34267013187024736,
+      "grad_norm": 0.22751405835151672,
+      "learning_rate": 7.653179634243979e-06,
+      "loss": 0.7695,
+      "step": 890
+    },
+    {
+      "epoch": 0.34652035807103665,
+      "grad_norm": 0.21724854409694672,
+      "learning_rate": 7.600120093927696e-06,
+      "loss": 0.7719,
+      "step": 900
+    },
+    {
+      "epoch": 0.350370584271826,
+      "grad_norm": 0.23327471315860748,
+      "learning_rate": 7.546656134948729e-06,
+      "loss": 0.7615,
+      "step": 910
+    },
+    {
+      "epoch": 0.3542208104726153,
+      "grad_norm": 0.23115581274032593,
+      "learning_rate": 7.492796073008661e-06,
+      "loss": 0.7627,
+      "step": 920
+    },
+    {
+      "epoch": 0.35807103667340456,
+      "grad_norm": 0.24455492198467255,
+      "learning_rate": 7.438548285418326e-06,
+      "loss": 0.7482,
+      "step": 930
+    },
+    {
+      "epoch": 0.36192126287419385,
+      "grad_norm": 0.24940283596515656,
+      "learning_rate": 7.3839212097948e-06,
+      "loss": 0.7831,
+      "step": 940
+    },
+    {
+      "epoch": 0.36577148907498314,
+      "grad_norm": 0.23512661457061768,
+      "learning_rate": 7.328923342749039e-06,
+      "loss": 0.7984,
+      "step": 950
+    },
+    {
+      "epoch": 0.3696217152757725,
+      "grad_norm": 0.24270224571228027,
+      "learning_rate": 7.273563238564319e-06,
+      "loss": 0.7826,
+      "step": 960
+    },
+    {
+      "epoch": 0.37347194147656176,
+      "grad_norm": 0.22906149923801422,
+      "learning_rate": 7.217849507865724e-06,
+      "loss": 0.7613,
+      "step": 970
+    },
+    {
+      "epoch": 0.37732216767735105,
+      "grad_norm": 0.2247680127620697,
+      "learning_rate": 7.161790816280858e-06,
+      "loss": 0.7687,
+      "step": 980
+    },
+    {
+      "epoch": 0.38117239387814034,
+      "grad_norm": 0.23405665159225464,
+      "learning_rate": 7.1053958830920125e-06,
+      "loss": 0.7507,
+      "step": 990
+    },
+    {
+      "epoch": 0.3850226200789296,
+      "grad_norm": 0.24462975561618805,
+      "learning_rate": 7.0486734798799715e-06,
+      "loss": 0.7717,
+      "step": 1000
+    },
+    {
+      "epoch": 0.3888728462797189,
+      "grad_norm": 0.23130956292152405,
+      "learning_rate": 6.991632429159708e-06,
+      "loss": 0.7942,
+      "step": 1010
+    },
+    {
+      "epoch": 0.39272307248050825,
+      "grad_norm": 0.2568194270133972,
+      "learning_rate": 6.934281603008132e-06,
+      "loss": 0.7832,
+      "step": 1020
+    },
+    {
+      "epoch": 0.39657329868129754,
+      "grad_norm": 0.23410667479038239,
+      "learning_rate": 6.876629921684143e-06,
+      "loss": 0.7816,
+      "step": 1030
+    },
+    {
+      "epoch": 0.4004235248820868,
+      "grad_norm": 0.23690448701381683,
+      "learning_rate": 6.818686352241199e-06,
+      "loss": 0.7716,
+      "step": 1040
+    },
+    {
+      "epoch": 0.4042737510828761,
+      "grad_norm": 0.2172718644142151,
+      "learning_rate": 6.760459907132573e-06,
+      "loss": 0.766,
+      "step": 1050
+    },
+    {
+      "epoch": 0.4081239772836654,
+      "grad_norm": 0.21788206696510315,
+      "learning_rate": 6.701959642809589e-06,
+      "loss": 0.7724,
+      "step": 1060
+    },
+    {
+      "epoch": 0.41197420348445474,
+      "grad_norm": 0.22874051332473755,
+      "learning_rate": 6.643194658312987e-06,
+      "loss": 0.7656,
+      "step": 1070
+    },
+    {
+      "epoch": 0.415824429685244,
+      "grad_norm": 0.21988199651241302,
+      "learning_rate": 6.584174093857676e-06,
+      "loss": 0.7889,
+      "step": 1080
+    },
+    {
+      "epoch": 0.4196746558860333,
+      "grad_norm": 0.26075708866119385,
+      "learning_rate": 6.5249071294110735e-06,
+      "loss": 0.7715,
+      "step": 1090
+    },
+    {
+      "epoch": 0.4235248820868226,
+      "grad_norm": 0.23880866169929504,
+      "learning_rate": 6.465402983265274e-06,
+      "loss": 0.7847,
+      "step": 1100
+    },
+    {
+      "epoch": 0.4273751082876119,
+      "grad_norm": 0.24006547033786774,
+      "learning_rate": 6.405670910603243e-06,
+      "loss": 0.7803,
+      "step": 1110
+    },
+    {
+      "epoch": 0.43122533448840117,
+      "grad_norm": 0.25401726365089417,
+      "learning_rate": 6.345720202059291e-06,
+      "loss": 0.7764,
+      "step": 1120
+    },
+    {
+      "epoch": 0.4350755606891905,
+      "grad_norm": 0.23378121852874756,
+      "learning_rate": 6.285560182274008e-06,
+      "loss": 0.755,
+      "step": 1130
+    },
+    {
+      "epoch": 0.4389257868899798,
+      "grad_norm": 0.22900429368019104,
+      "learning_rate": 6.225200208443936e-06,
+      "loss": 0.7771,
+      "step": 1140
+    },
+    {
+      "epoch": 0.4427760130907691,
+      "grad_norm": 0.2159281224012375,
+      "learning_rate": 6.16464966886616e-06,
+      "loss": 0.7662,
+      "step": 1150
+    },
+    {
+      "epoch": 0.44662623929155837,
+      "grad_norm": 0.2625824511051178,
+      "learning_rate": 6.103917981478065e-06,
+      "loss": 0.7904,
+      "step": 1160
+    },
+    {
+      "epoch": 0.45047646549234766,
+      "grad_norm": 0.221640944480896,
+      "learning_rate": 6.043014592392489e-06,
+      "loss": 0.801,
+      "step": 1170
+    },
+    {
+      "epoch": 0.454326691693137,
+      "grad_norm": 0.22183653712272644,
+      "learning_rate": 5.981948974428485e-06,
+      "loss": 0.777,
+      "step": 1180
+    },
+    {
+      "epoch": 0.4581769178939263,
+      "grad_norm": 0.24143798649311066,
+      "learning_rate": 5.920730625637934e-06,
+      "loss": 0.7533,
+      "step": 1190
+    },
+    {
+      "epoch": 0.46202714409471557,
+      "grad_norm": 0.2433205395936966,
+      "learning_rate": 5.8593690678282355e-06,
+      "loss": 0.7669,
+      "step": 1200
+    },
+    {
+      "epoch": 0.46587737029550486,
+      "grad_norm": 0.24136380851268768,
+      "learning_rate": 5.797873845081301e-06,
+      "loss": 0.771,
+      "step": 1210
+    },
+    {
+      "epoch": 0.46972759649629414,
+      "grad_norm": 0.23282776772975922,
+      "learning_rate": 5.736254522269075e-06,
+      "loss": 0.7878,
+      "step": 1220
+    },
+    {
+      "epoch": 0.47357782269708343,
+      "grad_norm": 0.23304307460784912,
+      "learning_rate": 5.6745206835658386e-06,
+      "loss": 0.7892,
+      "step": 1230
+    },
+    {
+      "epoch": 0.47742804889787277,
+      "grad_norm": 0.22460979223251343,
+      "learning_rate": 5.612681930957495e-06,
+      "loss": 0.7615,
+      "step": 1240
+    },
+    {
+      "epoch": 0.48127827509866206,
+      "grad_norm": 0.2179747372865677,
+      "learning_rate": 5.550747882748095e-06,
+      "loss": 0.7572,
+      "step": 1250
+    },
+    {
+      "epoch": 0.48512850129945134,
+      "grad_norm": 0.24613367021083832,
+      "learning_rate": 5.4887281720638195e-06,
+      "loss": 0.7646,
+      "step": 1260
+    },
+    {
+      "epoch": 0.48897872750024063,
+      "grad_norm": 0.26133018732070923,
+      "learning_rate": 5.42663244535466e-06,
+      "loss": 0.7481,
+      "step": 1270
+    },
+    {
+      "epoch": 0.4928289537010299,
+      "grad_norm": 0.24708575010299683,
+      "learning_rate": 5.3644703608940205e-06,
+      "loss": 0.7585,
+      "step": 1280
+    },
+    {
+      "epoch": 0.49667917990181926,
+      "grad_norm": 0.23367755115032196,
+      "learning_rate": 5.302251587276491e-06,
+      "loss": 0.7692,
+      "step": 1290
+    },
+    {
+      "epoch": 0.5005294061026085,
+      "grad_norm": 0.22166359424591064,
+      "learning_rate": 5.2399858019140005e-06,
+      "loss": 0.7696,
+      "step": 1300
+    },
+    {
+      "epoch": 0.5043796323033978,
+      "grad_norm": 0.22790922224521637,
+      "learning_rate": 5.177682689530615e-06,
+      "loss": 0.7849,
+      "step": 1310
+    },
+    {
+      "epoch": 0.5082298585041871,
+      "grad_norm": 0.23763158917427063,
+      "learning_rate": 5.11535194065619e-06,
+      "loss": 0.7511,
+      "step": 1320
+    },
+    {
+      "epoch": 0.5120800847049765,
+      "grad_norm": 0.25230172276496887,
+      "learning_rate": 5.053003250119113e-06,
+      "loss": 0.7608,
+      "step": 1330
+    },
+    {
+      "epoch": 0.5159303109057657,
+      "grad_norm": 0.23419001698493958,
+      "learning_rate": 4.9906463155383925e-06,
+      "loss": 0.77,
+      "step": 1340
+    },
+    {
+      "epoch": 0.519780537106555,
+      "grad_norm": 0.2280416339635849,
+      "learning_rate": 4.9282908358153046e-06,
+      "loss": 0.7637,
+      "step": 1350
+    },
+    {
+      "epoch": 0.5236307633073443,
+      "grad_norm": 0.2580273449420929,
+      "learning_rate": 4.8659465096248345e-06,
+      "loss": 0.7786,
+      "step": 1360
+    },
+    {
+      "epoch": 0.5274809895081336,
+      "grad_norm": 0.24924366176128387,
+      "learning_rate": 4.803623033907169e-06,
+      "loss": 0.7773,
+      "step": 1370
+    },
+    {
+      "epoch": 0.5313312157089229,
+      "grad_norm": 0.2223501354455948,
+      "learning_rate": 4.741330102359441e-06,
+      "loss": 0.7835,
+      "step": 1380
+    },
+    {
+      "epoch": 0.5351814419097122,
+      "grad_norm": 0.24978893995285034,
+      "learning_rate": 4.679077403927996e-06,
+      "loss": 0.7634,
+      "step": 1390
+    },
+    {
+      "epoch": 0.5390316681105015,
+      "grad_norm": 0.2292538583278656,
+      "learning_rate": 4.616874621301373e-06,
+      "loss": 0.7637,
+      "step": 1400
+    },
+    {
+      "epoch": 0.5428818943112907,
+      "grad_norm": 0.23414044082164764,
+      "learning_rate": 4.554731429404293e-06,
+      "loss": 0.7774,
+      "step": 1410
+    },
+    {
+      "epoch": 0.5467321205120801,
+      "grad_norm": 0.25060349702835083,
+      "learning_rate": 4.492657493892822e-06,
+      "loss": 0.7609,
+      "step": 1420
+    },
+    {
+      "epoch": 0.5505823467128694,
+      "grad_norm": 0.26405593752861023,
+      "learning_rate": 4.430662469650999e-06,
+      "loss": 0.7769,
+      "step": 1430
+    },
+    {
+      "epoch": 0.5544325729136587,
+      "grad_norm": 0.25376251339912415,
+      "learning_rate": 4.368755999289118e-06,
+      "loss": 0.7817,
+      "step": 1440
+    },
+    {
+      "epoch": 0.558282799114448,
+      "grad_norm": 0.2516332268714905,
+      "learning_rate": 4.306947711643949e-06,
+      "loss": 0.7646,
+      "step": 1450
+    },
+    {
+      "epoch": 0.5621330253152372,
+      "grad_norm": 0.24023941159248352,
+      "learning_rate": 4.245247220281068e-06,
+      "loss": 0.7684,
+      "step": 1460
+    },
+    {
+      "epoch": 0.5659832515160266,
+      "grad_norm": 0.23387639224529266,
+      "learning_rate": 4.183664121999587e-06,
+      "loss": 0.7647,
+      "step": 1470
+    },
+    {
+      "epoch": 0.5698334777168159,
+      "grad_norm": 0.23394618928432465,
+      "learning_rate": 4.122207995339474e-06,
+      "loss": 0.7633,
+      "step": 1480
+    },
+    {
+      "epoch": 0.5736837039176051,
+      "grad_norm": 0.2251158207654953,
+      "learning_rate": 4.060888399091734e-06,
+      "loss": 0.7665,
+      "step": 1490
+    },
+    {
+      "epoch": 0.5775339301183945,
+      "grad_norm": 0.23100876808166504,
+      "learning_rate": 3.999714870811639e-06,
+      "loss": 0.7774,
+      "step": 1500
+    },
+    {
+      "epoch": 0.5813841563191837,
+      "grad_norm": 0.22849220037460327,
+      "learning_rate": 3.938696925335275e-06,
+      "loss": 0.7691,
+      "step": 1510
+    },
+    {
+      "epoch": 0.5852343825199731,
+      "grad_norm": 0.23012445867061615,
+      "learning_rate": 3.87784405329962e-06,
+      "loss": 0.7776,
+      "step": 1520
+    },
+    {
+      "epoch": 0.5890846087207623,
+      "grad_norm": 0.23360022902488708,
+      "learning_rate": 3.817165719666382e-06,
+      "loss": 0.7783,
+      "step": 1530
+    },
+    {
+      "epoch": 0.5929348349215516,
+      "grad_norm": 0.2263231724500656,
+      "learning_rate": 3.756671362249841e-06,
+      "loss": 0.7828,
+      "step": 1540
+    },
+    {
+      "epoch": 0.596785061122341,
+      "grad_norm": 0.23590710759162903,
+      "learning_rate": 3.6963703902488907e-06,
+      "loss": 0.7692,
+      "step": 1550
+    },
+    {
+      "epoch": 0.6006352873231302,
+      "grad_norm": 0.22778210043907166,
+      "learning_rate": 3.6362721827835655e-06,
+      "loss": 0.7855,
+      "step": 1560
+    },
+    {
+      "epoch": 0.6044855135239195,
+      "grad_norm": 0.2428649216890335,
+      "learning_rate": 3.576386087436212e-06,
+      "loss": 0.7811,
+      "step": 1570
+    },
+    {
+      "epoch": 0.6083357397247088,
+      "grad_norm": 0.25078338384628296,
+      "learning_rate": 3.5167214187975913e-06,
+      "loss": 0.7713,
+      "step": 1580
+    },
+    {
+      "epoch": 0.6121859659254981,
+      "grad_norm": 0.20475414395332336,
+      "learning_rate": 3.457287457018086e-06,
+      "loss": 0.7745,
+      "step": 1590
+    },
+    {
+      "epoch": 0.6160361921262875,
+      "grad_norm": 0.22371430695056915,
+      "learning_rate": 3.3980934463643003e-06,
+      "loss": 0.7616,
+      "step": 1600
+    },
+    {
+      "epoch": 0.6198864183270767,
+      "grad_norm": 0.2155885100364685,
+      "learning_rate": 3.3391485937812053e-06,
+      "loss": 0.7447,
+      "step": 1610
+    },
+    {
+      "epoch": 0.623736644527866,
+      "grad_norm": 0.23904988169670105,
+      "learning_rate": 3.280462067460115e-06,
+      "loss": 0.7421,
+      "step": 1620
+    },
+    {
+      "epoch": 0.6275868707286553,
+      "grad_norm": 0.22936026751995087,
+      "learning_rate": 3.222042995412669e-06,
+      "loss": 0.7742,
+      "step": 1630
+    },
+    {
+      "epoch": 0.6314370969294446,
+      "grad_norm": 0.21054501831531525,
+      "learning_rate": 3.163900464051089e-06,
+      "loss": 0.764,
+      "step": 1640
+    },
+    {
+      "epoch": 0.635287323130234,
+      "grad_norm": 0.23206982016563416,
+      "learning_rate": 3.1060435167748826e-06,
+      "loss": 0.7733,
+      "step": 1650
+    },
+    {
+      "epoch": 0.6391375493310232,
+      "grad_norm": 0.23157179355621338,
+      "learning_rate": 3.048481152564253e-06,
+      "loss": 0.7538,
+      "step": 1660
+    },
+    {
+      "epoch": 0.6429877755318125,
+      "grad_norm": 0.2459339201450348,
+      "learning_rate": 2.9912223245804017e-06,
+      "loss": 0.778,
+      "step": 1670
+    },
+    {
+      "epoch": 0.6468380017326018,
+      "grad_norm": 0.234876811504364,
+      "learning_rate": 2.934275938772981e-06,
+      "loss": 0.7705,
+      "step": 1680
+    },
+    {
+      "epoch": 0.6506882279333911,
+      "grad_norm": 0.2555837631225586,
+      "learning_rate": 2.8776508524948676e-06,
+      "loss": 0.7619,
+      "step": 1690
+    },
+    {
+      "epoch": 0.6545384541341804,
+      "grad_norm": 0.213650643825531,
+      "learning_rate": 2.821355873124503e-06,
+      "loss": 0.7588,
+      "step": 1700
+    },
+    {
+      "epoch": 0.6583886803349697,
+      "grad_norm": 0.23964336514472961,
+      "learning_rate": 2.7653997566960144e-06,
+      "loss": 0.7756,
+      "step": 1710
+    },
+    {
+      "epoch": 0.662238906535759,
+      "grad_norm": 0.2439669370651245,
+      "learning_rate": 2.709791206537309e-06,
+      "loss": 0.7647,
+      "step": 1720
+    },
+    {
+      "epoch": 0.6660891327365482,
+      "grad_norm": 0.2375977635383606,
+      "learning_rate": 2.6545388719163747e-06,
+      "loss": 0.769,
+      "step": 1730
+    },
+    {
+      "epoch": 0.6699393589373376,
+      "grad_norm": 0.22826997935771942,
+      "learning_rate": 2.599651346695979e-06,
+      "loss": 0.7681,
+      "step": 1740
+    },
+    {
+      "epoch": 0.6737895851381268,
+      "grad_norm": 0.23960073292255402,
+      "learning_rate": 2.545137167997007e-06,
+      "loss": 0.7646,
+      "step": 1750
+    },
+    {
+      "epoch": 0.6776398113389162,
+      "grad_norm": 0.21956123411655426,
+      "learning_rate": 2.4910048148705972e-06,
+      "loss": 0.7515,
+      "step": 1760
+    },
+    {
+      "epoch": 0.6814900375397055,
+      "grad_norm": 0.2495088130235672,
+      "learning_rate": 2.4372627069793327e-06,
+      "loss": 0.7969,
+      "step": 1770
+    },
+    {
+      "epoch": 0.6853402637404947,
+      "grad_norm": 0.23745761811733246,
+      "learning_rate": 2.3839192032876486e-06,
+      "loss": 0.7818,
+      "step": 1780
+    },
+    {
+      "epoch": 0.6891904899412841,
+      "grad_norm": 0.22445443272590637,
+      "learning_rate": 2.3309826007617047e-06,
+      "loss": 0.7925,
+      "step": 1790
+    },
+    {
+      "epoch": 0.6930407161420733,
+      "grad_norm": 0.22874614596366882,
+      "learning_rate": 2.278461133078881e-06,
+      "loss": 0.7572,
+      "step": 1800
+    },
+    {
+      "epoch": 0.6968909423428626,
+      "grad_norm": 0.23675689101219177,
+      "learning_rate": 2.2263629693471258e-06,
+      "loss": 0.7473,
+      "step": 1810
+    },
+    {
+      "epoch": 0.700741168543652,
+      "grad_norm": 0.22549811005592346,
+      "learning_rate": 2.1746962128343414e-06,
+      "loss": 0.7711,
+      "step": 1820
+    },
+    {
+      "epoch": 0.7045913947444412,
+      "grad_norm": 0.23147058486938477,
+      "learning_rate": 2.1234688997080183e-06,
+      "loss": 0.7609,
+      "step": 1830
+    },
+    {
+      "epoch": 0.7084416209452306,
+      "grad_norm": 0.24992084503173828,
+      "learning_rate": 2.0726889977853053e-06,
+      "loss": 0.7653,
+      "step": 1840
+    },
+    {
+      "epoch": 0.7122918471460198,
+      "grad_norm": 0.2453038990497589,
+      "learning_rate": 2.022364405293703e-06,
+      "loss": 0.7793,
+      "step": 1850
+    },
+    {
+      "epoch": 0.7161420733468091,
+      "grad_norm": 0.23005327582359314,
+      "learning_rate": 1.9725029496425814e-06,
+      "loss": 0.7705,
+      "step": 1860
+    },
+    {
+      "epoch": 0.7199922995475985,
+      "grad_norm": 0.24262535572052002,
+      "learning_rate": 1.923112386205727e-06,
+      "loss": 0.767,
+      "step": 1870
+    },
+    {
+      "epoch": 0.7238425257483877,
+      "grad_norm": 0.22224922478199005,
+      "learning_rate": 1.874200397115079e-06,
+      "loss": 0.7756,
+      "step": 1880
+    },
+    {
+      "epoch": 0.727692751949177,
+      "grad_norm": 0.23713743686676025,
+      "learning_rate": 1.825774590065857e-06,
+      "loss": 0.7593,
+      "step": 1890
+    },
+    {
+      "epoch": 0.7315429781499663,
+      "grad_norm": 0.23296743631362915,
+      "learning_rate": 1.7778424971332813e-06,
+      "loss": 0.7664,
+      "step": 1900
+    },
+    {
+      "epoch": 0.7353932043507556,
+      "grad_norm": 0.2260030061006546,
+      "learning_rate": 1.7304115736010407e-06,
+      "loss": 0.7631,
+      "step": 1910
+    },
+    {
+      "epoch": 0.739243430551545,
+      "grad_norm": 0.22838300466537476,
+      "learning_rate": 1.6834891968017108e-06,
+      "loss": 0.7583,
+      "step": 1920
+    },
+    {
+      "epoch": 0.7430936567523342,
+      "grad_norm": 0.23753949999809265,
+      "learning_rate": 1.6370826649692884e-06,
+      "loss": 0.7742,
+      "step": 1930
+    },
+    {
+      "epoch": 0.7469438829531235,
+      "grad_norm": 0.23847892880439758,
+      "learning_rate": 1.5911991961040463e-06,
+      "loss": 0.7837,
+      "step": 1940
+    },
+    {
+      "epoch": 0.7507941091539128,
+      "grad_norm": 0.24488599598407745,
+      "learning_rate": 1.545845926849851e-06,
+      "loss": 0.7725,
+      "step": 1950
+    },
+    {
+      "epoch": 0.7546443353547021,
+      "grad_norm": 0.2450975626707077,
+      "learning_rate": 1.5010299113841397e-06,
+      "loss": 0.7875,
+      "step": 1960
+    },
+    {
+      "epoch": 0.7584945615554913,
+      "grad_norm": 0.22975526750087738,
+      "learning_rate": 1.4567581203207238e-06,
+      "loss": 0.7785,
+      "step": 1970
+    },
+    {
+      "epoch": 0.7623447877562807,
+      "grad_norm": 0.24324552714824677,
+      "learning_rate": 1.413037439625593e-06,
+      "loss": 0.7746,
+      "step": 1980
+    },
+    {
+      "epoch": 0.76619501395707,
+      "grad_norm": 0.25075972080230713,
+      "learning_rate": 1.3698746695458847e-06,
+      "loss": 0.7719,
+      "step": 1990
+    },
+    {
+      "epoch": 0.7700452401578592,
+      "grad_norm": 0.22820743918418884,
+      "learning_rate": 1.3272765235521745e-06,
+      "loss": 0.7622,
+      "step": 2000
+    },
+    {
+      "epoch": 0.7738954663586486,
+      "grad_norm": 0.22995060682296753,
+      "learning_rate": 1.2852496272942877e-06,
+      "loss": 0.7705,
+      "step": 2010
+    },
+    {
+      "epoch": 0.7777456925594378,
+      "grad_norm": 0.21890415251255035,
+      "learning_rate": 1.2438005175707396e-06,
+      "loss": 0.7579,
+      "step": 2020
+    },
+    {
+      "epoch": 0.7815959187602272,
+      "grad_norm": 0.228641539812088,
+      "learning_rate": 1.2029356413120297e-06,
+      "loss": 0.7424,
+      "step": 2030
+    },
+    {
+      "epoch": 0.7854461449610165,
+      "grad_norm": 0.22487632930278778,
+      "learning_rate": 1.1626613545778786e-06,
+      "loss": 0.76,
+      "step": 2040
+    },
+    {
+      "epoch": 0.7892963711618057,
+      "grad_norm": 0.21648256480693817,
+      "learning_rate": 1.1229839215686294e-06,
+      "loss": 0.7648,
+      "step": 2050
+    },
+    {
+      "epoch": 0.7931465973625951,
+      "grad_norm": 0.21019747853279114,
+      "learning_rate": 1.0839095136509188e-06,
+      "loss": 0.7484,
+      "step": 2060
+    },
+    {
+      "epoch": 0.7969968235633843,
+      "grad_norm": 0.24321739375591278,
+      "learning_rate": 1.045444208397791e-06,
+      "loss": 0.7706,
+      "step": 2070
+    },
+    {
+      "epoch": 0.8008470497641736,
+      "grad_norm": 0.22482147812843323,
+      "learning_rate": 1.007593988643402e-06,
+      "loss": 0.7519,
+      "step": 2080
+    },
+    {
+      "epoch": 0.804697275964963,
+      "grad_norm": 0.22992874681949615,
+      "learning_rate": 9.703647415524613e-07,
+      "loss": 0.7741,
+      "step": 2090
+    },
+    {
+      "epoch": 0.8085475021657522,
+      "grad_norm": 0.23315118253231049,
+      "learning_rate": 9.337622577045485e-07,
+      "loss": 0.7623,
+      "step": 2100
+    },
+    {
+      "epoch": 0.8123977283665416,
+      "grad_norm": 0.2260569930076599,
+      "learning_rate": 8.977922301934588e-07,
+      "loss": 0.7728,
+      "step": 2110
+    },
+    {
+      "epoch": 0.8162479545673308,
+      "grad_norm": 0.23224502801895142,
+      "learning_rate": 8.624602537417009e-07,
+      "loss": 0.7747,
+      "step": 2120
+    },
+    {
+      "epoch": 0.8200981807681201,
+      "grad_norm": 0.24186664819717407,
+      "learning_rate": 8.277718238303123e-07,
+      "loss": 0.7623,
+      "step": 2130
+    },
+    {
+      "epoch": 0.8239484069689095,
+      "grad_norm": 0.23252788186073303,
+      "learning_rate": 7.937323358440935e-07,
+      "loss": 0.7658,
+      "step": 2140
+    },
+    {
+      "epoch": 0.8277986331696987,
+      "grad_norm": 0.2397419810295105,
+      "learning_rate": 7.603470842324223e-07,
+      "loss": 0.7577,
+      "step": 2150
+    },
+    {
+      "epoch": 0.831648859370488,
+      "grad_norm": 0.22339706122875214,
+      "learning_rate": 7.276212616857575e-07,
+      "loss": 0.7745,
+      "step": 2160
+    },
+    {
+      "epoch": 0.8354990855712773,
+      "grad_norm": 0.23098993301391602,
+      "learning_rate": 6.955599583279837e-07,
+      "loss": 0.7731,
+      "step": 2170
+    },
+    {
+      "epoch": 0.8393493117720666,
+      "grad_norm": 0.23366037011146545,
+      "learning_rate": 6.641681609246981e-07,
+      "loss": 0.7654,
+      "step": 2180
+    },
+    {
+      "epoch": 0.8431995379728559,
+      "grad_norm": 0.2179335206747055,
+      "learning_rate": 6.334507521075711e-07,
+      "loss": 0.7715,
+      "step": 2190
+    },
+    {
+      "epoch": 0.8470497641736452,
+      "grad_norm": 0.24422211945056915,
+      "learning_rate": 6.034125096149179e-07,
+      "loss": 0.7674,
+      "step": 2200
+    },
+    {
+      "epoch": 0.8508999903744345,
+      "grad_norm": 0.23494933545589447,
+      "learning_rate": 5.740581055485722e-07,
+      "loss": 0.7401,
+      "step": 2210
+    },
+    {
+      "epoch": 0.8547502165752238,
+      "grad_norm": 0.21706435084342957,
+      "learning_rate": 5.453921056471961e-07,
+      "loss": 0.7724,
+      "step": 2220
+    },
+    {
+      "epoch": 0.8586004427760131,
+      "grad_norm": 0.23675943911075592,
+      "learning_rate": 5.174189685761288e-07,
+      "loss": 0.7662,
+      "step": 2230
+    },
+    {
+      "epoch": 0.8624506689768023,
+      "grad_norm": 0.2251240313053131,
+      "learning_rate": 4.901430452338979e-07,
+      "loss": 0.7786,
+      "step": 2240
+    },
+    {
+      "epoch": 0.8663008951775917,
+      "grad_norm": 0.2351454794406891,
+      "learning_rate": 4.6356857807548286e-07,
+      "loss": 0.7603,
+      "step": 2250
+    },
+    {
+      "epoch": 0.870151121378381,
+      "grad_norm": 0.23076549172401428,
+      "learning_rate": 4.376997004524541e-07,
+      "loss": 0.7503,
+      "step": 2260
+    },
+    {
+      "epoch": 0.8740013475791703,
+      "grad_norm": 0.2236543893814087,
+      "learning_rate": 4.12540435970073e-07,
+      "loss": 0.7588,
+      "step": 2270
+    },
+    {
+      "epoch": 0.8778515737799596,
+      "grad_norm": 0.22948215901851654,
+      "learning_rate": 3.880946978614703e-07,
+      "loss": 0.7724,
+      "step": 2280
+    },
+    {
+      "epoch": 0.8817017999807488,
+      "grad_norm": 0.22590669989585876,
+      "learning_rate": 3.643662883789878e-07,
+      "loss": 0.7741,
+      "step": 2290
+    },
+    {
+      "epoch": 0.8855520261815382,
+      "grad_norm": 0.2154102772474289,
+      "learning_rate": 3.41358898202781e-07,
+      "loss": 0.7526,
+      "step": 2300
+    },
+    {
+      "epoch": 0.8894022523823275,
+      "grad_norm": 0.2202003002166748,
+      "learning_rate": 3.190761058667741e-07,
+      "loss": 0.7628,
+      "step": 2310
+    },
+    {
+      "epoch": 0.8932524785831167,
+      "grad_norm": 0.22671908140182495,
+      "learning_rate": 2.9752137720206463e-07,
+      "loss": 0.7688,
+      "step": 2320
+    },
+    {
+      "epoch": 0.8971027047839061,
+      "grad_norm": 0.23143602907657623,
+      "learning_rate": 2.7669806479785067e-07,
+      "loss": 0.7745,
+      "step": 2330
+    },
+    {
+      "epoch": 0.9009529309846953,
+      "grad_norm": 0.24069814383983612,
+      "learning_rate": 2.566094074799763e-07,
+      "loss": 0.7683,
+      "step": 2340
+    },
+    {
+      "epoch": 0.9048031571854847,
+      "grad_norm": 0.22308221459388733,
+      "learning_rate": 2.3725852980716557e-07,
+      "loss": 0.7495,
+      "step": 2350
+    },
+    {
+      "epoch": 0.908653383386274,
+      "grad_norm": 0.2256549745798111,
+      "learning_rate": 2.1864844158503983e-07,
+      "loss": 0.7633,
+      "step": 2360
+    },
+    {
+      "epoch": 0.9125036095870632,
+      "grad_norm": 0.25843337178230286,
+      "learning_rate": 2.007820373979702e-07,
+      "loss": 0.7564,
+      "step": 2370
+    },
+    {
+      "epoch": 0.9163538357878526,
+      "grad_norm": 0.22947582602500916,
+      "learning_rate": 1.8366209615886532e-07,
+      "loss": 0.7633,
+      "step": 2380
+    },
+    {
+      "epoch": 0.9202040619886418,
+      "grad_norm": 0.22793444991111755,
+      "learning_rate": 1.6729128067693613e-07,
+      "loss": 0.7771,
+      "step": 2390
+    },
+    {
+      "epoch": 0.9240542881894311,
+      "grad_norm": 0.24625392258167267,
+      "learning_rate": 1.5167213724353426e-07,
+      "loss": 0.7765,
+      "step": 2400
+    },
+    {
+      "epoch": 0.9279045143902205,
+      "grad_norm": 0.22047585248947144,
+      "learning_rate": 1.3680709523610058e-07,
+      "loss": 0.7601,
+      "step": 2410
+    },
+    {
+      "epoch": 0.9317547405910097,
+      "grad_norm": 0.22457942366600037,
+      "learning_rate": 1.2269846674030418e-07,
+      "loss": 0.7803,
+      "step": 2420
+    },
+    {
+      "epoch": 0.935604966791799,
+      "grad_norm": 0.2315695732831955,
+      "learning_rate": 1.0934844619042595e-07,
+      "loss": 0.7789,
+      "step": 2430
+    },
+    {
+      "epoch": 0.9394551929925883,
+      "grad_norm": 0.2428281009197235,
+      "learning_rate": 9.675911002803961e-08,
+      "loss": 0.7653,
+      "step": 2440
+    },
+    {
+      "epoch": 0.9433054191933776,
+      "grad_norm": 0.24499140679836273,
+      "learning_rate": 8.493241637904381e-08,
+      "loss": 0.7622,
+      "step": 2450
+    },
+    {
+      "epoch": 0.9471556453941669,
+      "grad_norm": 0.2357485145330429,
+      "learning_rate": 7.387020474909689e-08,
+      "loss": 0.7831,
+      "step": 2460
+    },
+    {
+      "epoch": 0.9510058715949562,
+      "grad_norm": 0.24611593782901764,
+      "learning_rate": 6.357419573750856e-08,
+      "loss": 0.7768,
+      "step": 2470
+    },
+    {
+      "epoch": 0.9548560977957455,
+      "grad_norm": 0.22639408707618713,
+      "learning_rate": 5.4045990769613365e-08,
+      "loss": 0.7608,
+      "step": 2480
+    },
+    {
+      "epoch": 0.9587063239965348,
+      "grad_norm": 0.27551180124282837,
+      "learning_rate": 4.5287071847694434e-08,
+      "loss": 0.7651,
+      "step": 2490
+    },
+    {
+      "epoch": 0.9625565501973241,
+      "grad_norm": 0.2283722460269928,
+      "learning_rate": 3.7298801320467817e-08,
+      "loss": 0.7611,
+      "step": 2500
+    },
+    {
+      "epoch": 0.9664067763981133,
+      "grad_norm": 0.2353820502758026,
+      "learning_rate": 3.0082421671192576e-08,
+      "loss": 0.7551,
+      "step": 2510
+    },
+    {
+      "epoch": 0.9702570025989027,
+      "grad_norm": 0.23306728899478912,
+      "learning_rate": 2.3639055324410355e-08,
+      "loss": 0.7771,
+      "step": 2520
+    },
+    {
+      "epoch": 0.974107228799692,
+      "grad_norm": 0.2280379831790924,
+      "learning_rate": 1.7969704471369477e-08,
+      "loss": 0.7595,
+      "step": 2530
+    },
+    {
+      "epoch": 0.9779574550004813,
+      "grad_norm": 0.223310649394989,
+      "learning_rate": 1.3075250914142413e-08,
+      "loss": 0.7637,
+      "step": 2540
+    },
+    {
+      "epoch": 0.9818076812012706,
+      "grad_norm": 0.22073176503181458,
+      "learning_rate": 8.956455928476049e-09,
+      "loss": 0.7711,
+      "step": 2550
+    },
+    {
+      "epoch": 0.9856579074020598,
+      "grad_norm": 0.24051746726036072,
+      "learning_rate": 5.6139601453786275e-09,
+      "loss": 0.763,
+      "step": 2560
+    },
+    {
+      "epoch": 0.9895081336028492,
+      "grad_norm": 0.23105095326900482,
+      "learning_rate": 3.0482834514805602e-09,
+      "loss": 0.7523,
+      "step": 2570
+    },
+    {
+      "epoch": 0.9933583598036385,
+      "grad_norm": 0.20888514816761017,
+      "learning_rate": 1.2598249081713366e-09,
+      "loss": 0.7459,
+      "step": 2580
+    },
+    {
+      "epoch": 0.9972085860044277,
+      "grad_norm": 0.24714966118335724,
+      "learning_rate": 2.488626895302826e-10,
+      "loss": 0.7791,
+      "step": 2590
+    },
+    {
+      "epoch": 0.9999037443449803,
+      "step": 2597,
+      "total_flos": 3.7731510840385864e+18,
+      "train_loss": 0.7967522604262595,
+      "train_runtime": 10301.0692,
+      "train_samples_per_second": 64.547,
+      "train_steps_per_second": 0.252
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 2597,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.7731510840385864e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}

continuous/qwen3/mc2_b16_224/finetune/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a0e28c440ace0c04cdea715a0ffbd6ab0356d742bbb0b044a4a7741609ffc7a5
+size 6136

continuous/qwen3/mc2_b16_224/pretrain/config.json

@@ -0,0 +1,60 @@
+{
+  "_attn_implementation_autoset": true,
+  "add_faster_video": false,
+  "add_time_instruction": false,
+  "architectures": [
+    "Qwen3ForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "bos_token_id": 151643,
+  "eos_token_id": 151645,
+  "faster_token_stride": 10,
+  "force_sample": false,
+  "head_dim": 128,
+  "hidden_act": "silu",
+  "hidden_size": 2048,
+  "image_aspect_ratio": "square",
+  "image_crop_resolution": null,
+  "image_grid_pinpoints": null,
+  "image_split_resolution": null,
+  "initializer_range": 0.02,
+  "intermediate_size": 6144,
+  "max_position_embeddings": 40960,
+  "max_window_layers": 28,
+  "mm_hidden_size": 768,
+  "mm_newline_position": "grid",
+  "mm_patch_merge_type": "flat",
+  "mm_projector_lr": null,
+  "mm_projector_type": "mlp2x_gelu",
+  "mm_resampler_type": null,
+  "mm_spatial_pool_mode": "bilinear",
+  "mm_spatial_pool_stride": null,
+  "mm_tunable_parts": "mm_mlp_adapter",
+  "mm_use_im_patch_token": false,
+  "mm_use_im_start_end": false,
+  "mm_vision_select_feature": "patch",
+  "mm_vision_select_layer": -2,
+  "mm_vision_tower": "open_clip_hub:ViT-B-16",
+  "mm_vision_tower_lr": null,
+  "model_type": "qwen3",
+  "num_attention_heads": 16,
+  "num_hidden_layers": 28,
+  "num_key_value_heads": 8,
+  "pos_skipping_range": 4096,
+  "rms_norm_eps": 1e-06,
+  "rope_scaling": null,
+  "rope_theta": 1000000,
+  "sliding_window": null,
+  "tie_word_embeddings": true,
+  "tokenizer_model_max_length": 8192,
+  "tokenizer_padding_side": "right",
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.51.3",
+  "use_cache": true,
+  "use_mm_proj": true,
+  "use_pos_skipping": false,
+  "use_sliding_window": false,
+  "vision_tower_pretrained": "/cache/ckpt/download/tokenizer/continuous/mc2/mc2_b16_224.pt",
+  "vocab_size": 151936
+}

continuous/qwen3/mc2_b16_224/pretrain/mm_projector.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1bf68d1d0c4b5922deaf08ff584c426fddc10f5950dfaad434ca1161bdb9647f
+size 23087160

continuous/qwen3/mc2_b16_224/pretrain/trainer_state.json

@@ -0,0 +1,421 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.0,
+  "eval_steps": 500,
+  "global_step": 545,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.01834862385321101,
+      "grad_norm": 6.458603382110596,
+      "learning_rate": 0.00032142857142857147,
+      "loss": 7.9624,
+      "step": 10
+    },
+    {
+      "epoch": 0.03669724770642202,
+      "grad_norm": 0.4278964698314667,
+      "learning_rate": 0.0006785714285714287,
+      "loss": 4.9717,
+      "step": 20
+    },
+    {
+      "epoch": 0.05504587155963303,
+      "grad_norm": 0.16866883635520935,
+      "learning_rate": 0.0009999907688176173,
+      "loss": 4.6781,
+      "step": 30
+    },
+    {
+      "epoch": 0.07339449541284404,
+      "grad_norm": 0.26429980993270874,
+      "learning_rate": 0.0009988834393115766,
+      "loss": 4.6027,
+      "step": 40
+    },
+    {
+      "epoch": 0.09174311926605505,
+      "grad_norm": 0.26086485385894775,
+      "learning_rate": 0.0009959345572940772,
+      "loss": 4.5469,
+      "step": 50
+    },
+    {
+      "epoch": 0.11009174311926606,
+      "grad_norm": 0.17103448510169983,
+      "learning_rate": 0.0009911550081155983,
+      "loss": 4.4878,
+      "step": 60
+    },
+    {
+      "epoch": 0.12844036697247707,
+      "grad_norm": 0.22268082201480865,
+      "learning_rate": 0.0009845624347566433,
+      "loss": 4.4251,
+      "step": 70
+    },
+    {
+      "epoch": 0.14678899082568808,
+      "grad_norm": 0.19803553819656372,
+      "learning_rate": 0.0009761811727013547,
+      "loss": 4.3844,
+      "step": 80
+    },
+    {
+      "epoch": 0.1651376146788991,
+      "grad_norm": 0.20772339403629303,
+      "learning_rate": 0.0009660421601067666,
+      "loss": 4.3321,
+      "step": 90
+    },
+    {
+      "epoch": 0.1834862385321101,
+      "grad_norm": 0.2630506157875061,
+      "learning_rate": 0.0009541828235992926,
+      "loss": 4.3009,
+      "step": 100
+    },
+    {
+      "epoch": 0.2018348623853211,
+      "grad_norm": 0.24872377514839172,
+      "learning_rate": 0.0009406469401200151,
+      "loss": 4.2664,
+      "step": 110
+    },
+    {
+      "epoch": 0.22018348623853212,
+      "grad_norm": 0.2282891422510147,
+      "learning_rate": 0.0009254844753287492,
+      "loss": 4.2512,
+      "step": 120
+    },
+    {
+      "epoch": 0.23853211009174313,
+      "grad_norm": 0.19060726463794708,
+      "learning_rate": 0.0009087513991633924,
+      "loss": 4.2159,
+      "step": 130
+    },
+    {
+      "epoch": 0.25688073394495414,
+      "grad_norm": 0.2277657687664032,
+      "learning_rate": 0.0008905094792353917,
+      "loss": 4.1948,
+      "step": 140
+    },
+    {
+      "epoch": 0.27522935779816515,
+      "grad_norm": 0.1947551965713501,
+      "learning_rate": 0.0008708260528239789,
+      "loss": 4.1752,
+      "step": 150
+    },
+    {
+      "epoch": 0.29357798165137616,
+      "grad_norm": 0.23078195750713348,
+      "learning_rate": 0.0008497737783108237,
+      "loss": 4.1311,
+      "step": 160
+    },
+    {
+      "epoch": 0.3119266055045872,
+      "grad_norm": 0.1958288550376892,
+      "learning_rate": 0.0008274303669726426,
+      "loss": 4.0928,
+      "step": 170
+    },
+    {
+      "epoch": 0.3302752293577982,
+      "grad_norm": 0.23148895800113678,
+      "learning_rate": 0.0008038782961218137,
+      "loss": 4.0682,
+      "step": 180
+    },
+    {
+      "epoch": 0.3486238532110092,
+      "grad_norm": 0.1992112100124359,
+      "learning_rate": 0.000779204504653894,
+      "loss": 4.0543,
+      "step": 190
+    },
+    {
+      "epoch": 0.3669724770642202,
+      "grad_norm": 0.2156057208776474,
+      "learning_rate": 0.0007535000721258791,
+      "loss": 4.0289,
+      "step": 200
+    },
+    {
+      "epoch": 0.3853211009174312,
+      "grad_norm": 0.17427928745746613,
+      "learning_rate": 0.00072685988254984,
+      "loss": 4.0168,
+      "step": 210
+    },
+    {
+      "epoch": 0.4036697247706422,
+      "grad_norm": 0.22813354432582855,
+      "learning_rate": 0.0006993822741429906,
+      "loss": 4.0276,
+      "step": 220
+    },
+    {
+      "epoch": 0.42201834862385323,
+      "grad_norm": 0.17621129751205444,
+      "learning_rate": 0.0006711686763270818,
+      "loss": 4.0197,
+      "step": 230
+    },
+    {
+      "epoch": 0.44036697247706424,
+      "grad_norm": 0.19711147248744965,
+      "learning_rate": 0.0006423232353170831,
+      "loss": 4.0053,
+      "step": 240
+    },
+    {
+      "epoch": 0.45871559633027525,
+      "grad_norm": 0.17635241150856018,
+      "learning_rate": 0.0006129524296812335,
+      "loss": 3.9949,
+      "step": 250
+    },
+    {
+      "epoch": 0.47706422018348627,
+      "grad_norm": 0.17844831943511963,
+      "learning_rate": 0.0005831646772915651,
+      "loss": 4.0033,
+      "step": 260
+    },
+    {
+      "epoch": 0.4954128440366973,
+      "grad_norm": 0.1709989458322525,
+      "learning_rate": 0.0005530699351157809,
+      "loss": 3.9839,
+      "step": 270
+    },
+    {
+      "epoch": 0.5137614678899083,
+      "grad_norm": 0.20238558948040009,
+      "learning_rate": 0.0005227792933277943,
+      "loss": 3.9827,
+      "step": 280
+    },
+    {
+      "epoch": 0.5321100917431193,
+      "grad_norm": 0.18733353912830353,
+      "learning_rate": 0.0004924045652352092,
+      "loss": 3.973,
+      "step": 290
+    },
+    {
+      "epoch": 0.5504587155963303,
+      "grad_norm": 0.18176543712615967,
+      "learning_rate": 0.0004620578745374633,
+      "loss": 3.9737,
+      "step": 300
+    },
+    {
+      "epoch": 0.5688073394495413,
+      "grad_norm": 0.19877950847148895,
+      "learning_rate": 0.0004318512414382059,
+      "loss": 3.985,
+      "step": 310
+    },
+    {
+      "epoch": 0.5871559633027523,
+      "grad_norm": 0.16852682828903198,
+      "learning_rate": 0.0004018961691397148,
+      "loss": 3.9733,
+      "step": 320
+    },
+    {
+      "epoch": 0.6055045871559633,
+      "grad_norm": 0.1781880408525467,
+      "learning_rate": 0.0003723032322457458,
+      "loss": 3.9606,
+      "step": 330
+    },
+    {
+      "epoch": 0.6238532110091743,
+      "grad_norm": 0.18147973716259003,
+      "learning_rate": 0.0003431816685921625,
+      "loss": 3.9611,
+      "step": 340
+    },
+    {
+      "epoch": 0.6422018348623854,
+      "grad_norm": 0.1572204828262329,
+      "learning_rate": 0.0003146389760120416,
+      "loss": 3.9627,
+      "step": 350
+    },
+    {
+      "epoch": 0.6605504587155964,
+      "grad_norm": 0.17670013010501862,
+      "learning_rate": 0.00028678051552373485,
+      "loss": 3.9697,
+      "step": 360
+    },
+    {
+      "epoch": 0.6788990825688074,
+      "grad_norm": 0.16905246675014496,
+      "learning_rate": 0.0002597091224066581,
+      "loss": 3.9574,
+      "step": 370
+    },
+    {
+      "epoch": 0.6972477064220184,
+      "grad_norm": 0.1764896661043167,
+      "learning_rate": 0.0002335247266004629,
+      "loss": 3.9695,
+      "step": 380
+    },
+    {
+      "epoch": 0.7155963302752294,
+      "grad_norm": 0.17443664371967316,
+      "learning_rate": 0.000208323983828828,
+      "loss": 3.952,
+      "step": 390
+    },
+    {
+      "epoch": 0.7339449541284404,
+      "grad_norm": 0.18862049281597137,
+      "learning_rate": 0.0001841999188095224,
+      "loss": 3.9484,
+      "step": 400
+    },
+    {
+      "epoch": 0.7522935779816514,
+      "grad_norm": 0.1624613255262375,
+      "learning_rate": 0.00016124158186777676,
+      "loss": 3.9556,
+      "step": 410
+    },
+    {
+      "epoch": 0.7706422018348624,
+      "grad_norm": 0.15761800110340118,
+      "learning_rate": 0.0001395337202205211,
+      "loss": 3.9483,
+      "step": 420
+    },
+    {
+      "epoch": 0.7889908256880734,
+      "grad_norm": 0.1629873812198639,
+      "learning_rate": 0.00011915646514489292,
+      "loss": 3.9481,
+      "step": 430
+    },
+    {
+      "epoch": 0.8073394495412844,
+      "grad_norm": 0.17246125638484955,
+      "learning_rate": 0.00010018503618578818,
+      "loss": 3.9591,
+      "step": 440
+    },
+    {
+      "epoch": 0.8256880733944955,
+      "grad_norm": 0.16985392570495605,
+      "learning_rate": 8.268946349432583e-05,
+      "loss": 3.9474,
+      "step": 450
+    },
+    {
+      "epoch": 0.8440366972477065,
+      "grad_norm": 0.1629709005355835,
+      "learning_rate": 6.673432932217117e-05,
+      "loss": 3.9418,
+      "step": 460
+    },
+    {
+      "epoch": 0.8623853211009175,
+      "grad_norm": 0.1529712677001953,
+      "learning_rate": 5.2378529625954684e-05,
+      "loss": 3.93,
+      "step": 470
+    },
+    {
+      "epoch": 0.8807339449541285,
+      "grad_norm": 0.17161276936531067,
+      "learning_rate": 3.9675056661785556e-05,
+      "loss": 3.9372,
+      "step": 480
+    },
+    {
+      "epoch": 0.8990825688073395,
+      "grad_norm": 0.1535739004611969,
+      "learning_rate": 2.8670803372379538e-05,
+      "loss": 3.9422,
+      "step": 490
+    },
+    {
+      "epoch": 0.9174311926605505,
+      "grad_norm": 0.15437234938144684,
+      "learning_rate": 1.9406390288876586e-05,
+      "loss": 3.9527,
+      "step": 500
+    },
+    {
+      "epoch": 0.9357798165137615,
+      "grad_norm": 0.15631718933582306,
+      "learning_rate": 1.191601558631461e-05,
+      "loss": 3.942,
+      "step": 510
+    },
+    {
+      "epoch": 0.9541284403669725,
+      "grad_norm": 0.1545124053955078,
+      "learning_rate": 6.2273288462559305e-06,
+      "loss": 3.9396,
+      "step": 520
+    },
+    {
+      "epoch": 0.9724770642201835,
+      "grad_norm": 0.15018726885318756,
+      "learning_rate": 2.361328992552314e-06,
+      "loss": 3.9397,
+      "step": 530
+    },
+    {
+      "epoch": 0.9908256880733946,
+      "grad_norm": 0.1526820957660675,
+      "learning_rate": 3.322867770044202e-07,
+      "loss": 3.9387,
+      "step": 540
+    },
+    {
+      "epoch": 1.0,
+      "step": 545,
+      "total_flos": 3.0978557415417446e+17,
+      "train_loss": 4.165047941295379,
+      "train_runtime": 5621.6977,
+      "train_samples_per_second": 99.281,
+      "train_steps_per_second": 0.097
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 545,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": false,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.0978557415417446e+17,
+  "train_batch_size": 128,
+  "trial_name": null,
+  "trial_params": null
+}

continuous/qwen3/mc2_b16_384/finetune/added_tokens.json

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

continuous/qwen3/mc2_b16_384/finetune/config.json

@@ -0,0 +1,59 @@
+{
+  "add_faster_video": false,
+  "add_time_instruction": false,
+  "architectures": [
+    "LlavaQwen3ForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "bos_token_id": 151643,
+  "eos_token_id": 151645,
+  "faster_token_stride": 10,
+  "force_sample": false,
+  "head_dim": 128,
+  "hidden_act": "silu",
+  "hidden_size": 2048,
+  "image_aspect_ratio": "square",
+  "image_crop_resolution": null,
+  "image_grid_pinpoints": null,
+  "image_split_resolution": null,
+  "initializer_range": 0.02,
+  "intermediate_size": 6144,
+  "max_position_embeddings": 40960,
+  "max_window_layers": 28,
+  "mm_hidden_size": 768,
+  "mm_newline_position": "grid",
+  "mm_patch_merge_type": "flat",
+  "mm_projector_lr": 0.0001,
+  "mm_projector_type": "mlp2x_gelu",
+  "mm_resampler_type": null,
+  "mm_spatial_pool_mode": "bilinear",
+  "mm_spatial_pool_stride": null,
+  "mm_tunable_parts": "mm_mlp_adapter,mm_language_model",
+  "mm_use_im_patch_token": false,
+  "mm_use_im_start_end": false,
+  "mm_vision_select_feature": "patch",
+  "mm_vision_select_layer": -2,
+  "mm_vision_tower": "open_clip_hub:ViT-B-16",
+  "mm_vision_tower_lr": null,
+  "model_type": "qwen3",
+  "num_attention_heads": 16,
+  "num_hidden_layers": 28,
+  "num_key_value_heads": 8,
+  "pos_skipping_range": 4096,
+  "rms_norm_eps": 1e-06,
+  "rope_scaling": null,
+  "rope_theta": 1000000,
+  "sliding_window": null,
+  "tie_word_embeddings": true,
+  "tokenizer_model_max_length": 8192,
+  "tokenizer_padding_side": "right",
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.51.3",
+  "use_cache": true,
+  "use_mm_proj": true,
+  "use_pos_skipping": false,
+  "use_sliding_window": false,
+  "vision_tower_pretrained": "/cache/ckpt/download/tokenizer/continuous/mc2/mc2_b16_384.pt",
+  "vocab_size": 151936
+}

continuous/qwen3/mc2_b16_384/finetune/generation_config.json

@@ -0,0 +1,13 @@
+{
+  "bos_token_id": 151643,
+  "do_sample": true,
+  "eos_token_id": [
+    151645,
+    151643
+  ],
+  "pad_token_id": 151643,
+  "temperature": 0.6,
+  "top_k": 20,
+  "top_p": 0.95,
+  "transformers_version": "4.51.3"
+}

continuous/qwen3/mc2_b16_384/finetune/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:eb1cd312f3652401650ca515f19da454d2ce26c650e7b2fd2ba209f7615c55e7
+size 4259591928

continuous/qwen3/mc2_b16_384/finetune/special_tokens_map.json

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

continuous/qwen3/mc2_b16_384/finetune/tokenizer.json

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

continuous/qwen3/mc2_b16_384/finetune/tokenizer_config.json

@@ -0,0 +1,241 @@
+{
+  "add_bos_token": false,
+  "add_prefix_space": false,
+  "added_tokens_decoder": {
+    "151643": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151644": {
+      "content": "<|im_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151645": {
+      "content": "<|im_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151646": {
+      "content": "<|object_ref_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151647": {
+      "content": "<|object_ref_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151648": {
+      "content": "<|box_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151649": {
+      "content": "<|box_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151650": {
+      "content": "<|quad_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151651": {
+      "content": "<|quad_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151652": {
+      "content": "<|vision_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151653": {
+      "content": "<|vision_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151654": {
+      "content": "<|vision_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151655": {
+      "content": "<|image_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151656": {
+      "content": "<|video_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151657": {
+      "content": "<tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151658": {
+      "content": "</tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151659": {
+      "content": "<|fim_prefix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151660": {
+      "content": "<|fim_middle|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151661": {
+      "content": "<|fim_suffix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151662": {
+      "content": "<|fim_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151663": {
+      "content": "<|repo_name|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151664": {
+      "content": "<|file_sep|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151665": {
+      "content": "<tool_response>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151666": {
+      "content": "</tool_response>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151667": {
+      "content": "<think>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151668": {
+      "content": "</think>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    }
+  },
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "bos_token": null,
+  "chat_template": "{%- if tools %}\n    {{- '<|im_start|>system\\n' }}\n    {%- if messages[0].role == 'system' %}\n        {{- messages[0].content + '\\n\\n' }}\n    {%- endif %}\n    {{- \"# Tools\\n\\nYou may call one or more functions to assist with the user query.\\n\\nYou are provided with function signatures within <tools></tools> XML tags:\\n<tools>\" }}\n    {%- for tool in tools %}\n        {{- \"\\n\" }}\n        {{- tool | tojson }}\n    {%- endfor %}\n    {{- \"\\n</tools>\\n\\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\\n<tool_call>\\n{\\\"name\\\": <function-name>, \\\"arguments\\\": <args-json-object>}\\n</tool_call><|im_end|>\\n\" }}\n{%- else %}\n    {%- if messages[0].role == 'system' %}\n        {{- '<|im_start|>system\\n' + messages[0].content + '<|im_end|>\\n' }}\n    {%- endif %}\n{%- endif %}\n{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}\n{%- for message in messages[::-1] %}\n    {%- set index = (messages|length - 1) - loop.index0 %}\n    {%- if ns.multi_step_tool and message.role == \"user\" and message.content is string and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}\n        {%- set ns.multi_step_tool = false %}\n        {%- set ns.last_query_index = index %}\n    {%- endif %}\n{%- endfor %}\n{%- for message in messages %}\n    {%- if message.content is string %}\n        {%- set content = message.content %}\n    {%- else %}\n        {%- set content = '' %}\n    {%- endif %}\n    {%- if (message.role == \"user\") or (message.role == \"system\" and not loop.first) %}\n        {{- '<|im_start|>' + message.role + '\\n' + content + '<|im_end|>' + '\\n' }}\n    {%- elif message.role == \"assistant\" %}\n        {%- set reasoning_content = '' %}\n        {%- if message.reasoning_content is string %}\n            {%- set reasoning_content = message.reasoning_content %}\n        {%- else %}\n            {%- if '</think>' in content %}\n                {%- set reasoning_content = content.split('</think>')[0].rstrip('\\n').split('<think>')[-1].lstrip('\\n') %}\n                {%- set content = content.split('</think>')[-1].lstrip('\\n') %}\n            {%- endif %}\n        {%- endif %}\n        {%- if loop.index0 > ns.last_query_index %}\n            {%- if loop.last or (not loop.last and reasoning_content) %}\n                {{- '<|im_start|>' + message.role + '\\n<think>\\n' + reasoning_content.strip('\\n') + '\\n</think>\\n\\n' + content.lstrip('\\n') }}\n            {%- else %}\n                {{- '<|im_start|>' + message.role + '\\n' + content }}\n            {%- endif %}\n        {%- else %}\n            {{- '<|im_start|>' + message.role + '\\n' + content }}\n        {%- endif %}\n        {%- if message.tool_calls %}\n            {%- for tool_call in message.tool_calls %}\n                {%- if (loop.first and content) or (not loop.first) %}\n                    {{- '\\n' }}\n                {%- endif %}\n                {%- if tool_call.function %}\n                    {%- set tool_call = tool_call.function %}\n                {%- endif %}\n                {{- '<tool_call>\\n{\"name\": \"' }}\n                {{- tool_call.name }}\n                {{- '\", \"arguments\": ' }}\n                {%- if tool_call.arguments is string %}\n                    {{- tool_call.arguments }}\n                {%- else %}\n                    {{- tool_call.arguments | tojson }}\n                {%- endif %}\n                {{- '}\\n</tool_call>' }}\n            {%- endfor %}\n        {%- endif %}\n        {{- '<|im_end|>\\n' }}\n    {%- elif message.role == \"tool\" %}\n        {%- if loop.first or (messages[loop.index0 - 1].role != \"tool\") %}\n            {{- '<|im_start|>user' }}\n        {%- endif %}\n        {{- '\\n<tool_response>\\n' }}\n        {{- content }}\n        {{- '\\n</tool_response>' }}\n        {%- if loop.last or (messages[loop.index0 + 1].role != \"tool\") %}\n            {{- '<|im_end|>\\n' }}\n        {%- endif %}\n    {%- endif %}\n{%- endfor %}\n{%- if add_generation_prompt %}\n    {{- '<|im_start|>assistant\\n' }}\n    {%- if enable_thinking is defined and enable_thinking is false %}\n        {{- '<think>\\n\\n</think>\\n\\n' }}\n    {%- endif %}\n{%- endif %}",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|im_end|>",
+  "errors": "replace",
+  "extra_special_tokens": {},
+  "model_max_length": 8192,
+  "pad_token": "<|endoftext|>",
+  "padding_side": "right",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}

continuous/qwen3/mc2_b16_384/finetune/trainer_state.json

@@ -0,0 +1,1856 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.9999037443449803,
+  "eval_steps": 500,
+  "global_step": 2597,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.003850226200789296,
+      "grad_norm": 8.270811080932617,
+      "learning_rate": 1.153846153846154e-06,
+      "loss": 2.1504,
+      "step": 10
+    },
+    {
+      "epoch": 0.007700452401578592,
+      "grad_norm": 3.2342593669891357,
+      "learning_rate": 2.435897435897436e-06,
+      "loss": 1.7707,
+      "step": 20
+    },
+    {
+      "epoch": 0.01155067860236789,
+      "grad_norm": 2.2352821826934814,
+      "learning_rate": 3.7179487179487184e-06,
+      "loss": 1.4146,
+      "step": 30
+    },
+    {
+      "epoch": 0.015400904803157185,
+      "grad_norm": 0.4332663118839264,
+      "learning_rate": 5e-06,
+      "loss": 1.0931,
+      "step": 40
+    },
+    {
+      "epoch": 0.019251131003946483,
+      "grad_norm": 0.45374199748039246,
+      "learning_rate": 6.282051282051282e-06,
+      "loss": 1.0107,
+      "step": 50
+    },
+    {
+      "epoch": 0.02310135720473578,
+      "grad_norm": 0.23228329420089722,
+      "learning_rate": 7.564102564102564e-06,
+      "loss": 0.9389,
+      "step": 60
+    },
+    {
+      "epoch": 0.026951583405525073,
+      "grad_norm": 0.18729786574840546,
+      "learning_rate": 8.846153846153847e-06,
+      "loss": 0.8994,
+      "step": 70
+    },
+    {
+      "epoch": 0.03080180960631437,
+      "grad_norm": 0.1693103313446045,
+      "learning_rate": 9.999996111488723e-06,
+      "loss": 0.8995,
+      "step": 80
+    },
+    {
+      "epoch": 0.034652035807103666,
+      "grad_norm": 0.163882315158844,
+      "learning_rate": 9.999529497453782e-06,
+      "loss": 0.8759,
+      "step": 90
+    },
+    {
+      "epoch": 0.038502262007892966,
+      "grad_norm": 0.1453651487827301,
+      "learning_rate": 9.998285264324193e-06,
+      "loss": 0.8485,
+      "step": 100
+    },
+    {
+      "epoch": 0.04235248820868226,
+      "grad_norm": 0.16473984718322754,
+      "learning_rate": 9.996263605626052e-06,
+      "loss": 0.829,
+      "step": 110
+    },
+    {
+      "epoch": 0.04620271440947156,
+      "grad_norm": 0.14588318765163422,
+      "learning_rate": 9.99346483580503e-06,
+      "loss": 0.8369,
+      "step": 120
+    },
+    {
+      "epoch": 0.05005294061026085,
+      "grad_norm": 0.13644719123840332,
+      "learning_rate": 9.989889390177467e-06,
+      "loss": 0.826,
+      "step": 130
+    },
+    {
+      "epoch": 0.053903166811050146,
+      "grad_norm": 0.155082106590271,
+      "learning_rate": 9.985537824862647e-06,
+      "loss": 0.8256,
+      "step": 140
+    },
+    {
+      "epoch": 0.057753393011839446,
+      "grad_norm": 0.17970716953277588,
+      "learning_rate": 9.980410816696317e-06,
+      "loss": 0.8207,
+      "step": 150
+    },
+    {
+      "epoch": 0.06160361921262874,
+      "grad_norm": 0.14768345654010773,
+      "learning_rate": 9.97450916312541e-06,
+      "loss": 0.8174,
+      "step": 160
+    },
+    {
+      "epoch": 0.06545384541341803,
+      "grad_norm": 0.12666116654872894,
+      "learning_rate": 9.967833782084004e-06,
+      "loss": 0.8155,
+      "step": 170
+    },
+    {
+      "epoch": 0.06930407161420733,
+      "grad_norm": 0.14374156296253204,
+      "learning_rate": 9.960385711850554e-06,
+      "loss": 0.8361,
+      "step": 180
+    },
+    {
+      "epoch": 0.07315429781499663,
+      "grad_norm": 0.13645002245903015,
+      "learning_rate": 9.952166110886394e-06,
+      "loss": 0.8395,
+      "step": 190
+    },
+    {
+      "epoch": 0.07700452401578593,
+      "grad_norm": 0.15470530092716217,
+      "learning_rate": 9.943176257655567e-06,
+      "loss": 0.8138,
+      "step": 200
+    },
+    {
+      "epoch": 0.08085475021657522,
+      "grad_norm": 0.1387203335762024,
+      "learning_rate": 9.933417550425952e-06,
+      "loss": 0.8266,
+      "step": 210
+    },
+    {
+      "epoch": 0.08470497641736452,
+      "grad_norm": 0.14447826147079468,
+      "learning_rate": 9.9228915070518e-06,
+      "loss": 0.7998,
+      "step": 220
+    },
+    {
+      "epoch": 0.08855520261815382,
+      "grad_norm": 0.14736291766166687,
+      "learning_rate": 9.911599764737632e-06,
+      "loss": 0.8054,
+      "step": 230
+    },
+    {
+      "epoch": 0.09240542881894312,
+      "grad_norm": 0.1352626532316208,
+      "learning_rate": 9.899544079783606e-06,
+      "loss": 0.7926,
+      "step": 240
+    },
+    {
+      "epoch": 0.0962556550197324,
+      "grad_norm": 0.1410941630601883,
+      "learning_rate": 9.886726327312338e-06,
+      "loss": 0.8124,
+      "step": 250
+    },
+    {
+      "epoch": 0.1001058812205217,
+      "grad_norm": 0.13280747830867767,
+      "learning_rate": 9.87314850097724e-06,
+      "loss": 0.8125,
+      "step": 260
+    },
+    {
+      "epoch": 0.103956107421311,
+      "grad_norm": 0.13109755516052246,
+      "learning_rate": 9.858812712652452e-06,
+      "loss": 0.7985,
+      "step": 270
+    },
+    {
+      "epoch": 0.10780633362210029,
+      "grad_norm": 0.1385888159275055,
+      "learning_rate": 9.843721192104341e-06,
+      "loss": 0.8,
+      "step": 280
+    },
+    {
+      "epoch": 0.11165655982288959,
+      "grad_norm": 0.1326172649860382,
+      "learning_rate": 9.827876286644698e-06,
+      "loss": 0.8154,
+      "step": 290
+    },
+    {
+      "epoch": 0.11550678602367889,
+      "grad_norm": 0.12403558194637299,
+      "learning_rate": 9.811280460765643e-06,
+      "loss": 0.7774,
+      "step": 300
+    },
+    {
+      "epoch": 0.11935701222446819,
+      "grad_norm": 0.13086219131946564,
+      "learning_rate": 9.793936295756292e-06,
+      "loss": 0.7924,
+      "step": 310
+    },
+    {
+      "epoch": 0.12320723842525748,
+      "grad_norm": 0.13778066635131836,
+      "learning_rate": 9.775846489301279e-06,
+      "loss": 0.7888,
+      "step": 320
+    },
+    {
+      "epoch": 0.12705746462604678,
+      "grad_norm": 0.13569174706935883,
+      "learning_rate": 9.757013855061157e-06,
+      "loss": 0.7908,
+      "step": 330
+    },
+    {
+      "epoch": 0.13090769082683607,
+      "grad_norm": 0.13307256996631622,
+      "learning_rate": 9.737441322234758e-06,
+      "loss": 0.7999,
+      "step": 340
+    },
+    {
+      "epoch": 0.13475791702762538,
+      "grad_norm": 0.15013322234153748,
+      "learning_rate": 9.717131935103605e-06,
+      "loss": 0.8066,
+      "step": 350
+    },
+    {
+      "epoch": 0.13860814322841467,
+      "grad_norm": 0.13416215777397156,
+      "learning_rate": 9.696088852558398e-06,
+      "loss": 0.7914,
+      "step": 360
+    },
+    {
+      "epoch": 0.14245836942920398,
+      "grad_norm": 0.14122335612773895,
+      "learning_rate": 9.674315347607686e-06,
+      "loss": 0.8162,
+      "step": 370
+    },
+    {
+      "epoch": 0.14630859562999327,
+      "grad_norm": 0.12501060962677002,
+      "learning_rate": 9.651814806868796e-06,
+      "loss": 0.786,
+      "step": 380
+    },
+    {
+      "epoch": 0.15015882183078255,
+      "grad_norm": 0.13875214755535126,
+      "learning_rate": 9.62859073004107e-06,
+      "loss": 0.7877,
+      "step": 390
+    },
+    {
+      "epoch": 0.15400904803157187,
+      "grad_norm": 0.12969240546226501,
+      "learning_rate": 9.604646729361547e-06,
+      "loss": 0.7882,
+      "step": 400
+    },
+    {
+      "epoch": 0.15785927423236115,
+      "grad_norm": 0.12558481097221375,
+      "learning_rate": 9.579986529043099e-06,
+      "loss": 0.7677,
+      "step": 410
+    },
+    {
+      "epoch": 0.16170950043315044,
+      "grad_norm": 0.12914825975894928,
+      "learning_rate": 9.554613964695189e-06,
+      "loss": 0.7898,
+      "step": 420
+    },
+    {
+      "epoch": 0.16555972663393975,
+      "grad_norm": 0.13334406912326813,
+      "learning_rate": 9.528532982727281e-06,
+      "loss": 0.7834,
+      "step": 430
+    },
+    {
+      "epoch": 0.16940995283472904,
+      "grad_norm": 0.13702696561813354,
+      "learning_rate": 9.501747639735018e-06,
+      "loss": 0.7793,
+      "step": 440
+    },
+    {
+      "epoch": 0.17326017903551832,
+      "grad_norm": 0.13706697523593903,
+      "learning_rate": 9.474262101869272e-06,
+      "loss": 0.7649,
+      "step": 450
+    },
+    {
+      "epoch": 0.17711040523630764,
+      "grad_norm": 0.13655002415180206,
+      "learning_rate": 9.446080644188141e-06,
+      "loss": 0.7821,
+      "step": 460
+    },
+    {
+      "epoch": 0.18096063143709692,
+      "grad_norm": 0.13671965897083282,
+      "learning_rate": 9.417207649992017e-06,
+      "loss": 0.7831,
+      "step": 470
+    },
+    {
+      "epoch": 0.18481085763788624,
+      "grad_norm": 0.13236768543720245,
+      "learning_rate": 9.387647610141816e-06,
+      "loss": 0.7765,
+      "step": 480
+    },
+    {
+      "epoch": 0.18866108383867553,
+      "grad_norm": 0.12446891516447067,
+      "learning_rate": 9.35740512236047e-06,
+      "loss": 0.7678,
+      "step": 490
+    },
+    {
+      "epoch": 0.1925113100394648,
+      "grad_norm": 0.12534692883491516,
+      "learning_rate": 9.326484890517809e-06,
+      "loss": 0.7745,
+      "step": 500
+    },
+    {
+      "epoch": 0.19636153624025413,
+      "grad_norm": 0.128121018409729,
+      "learning_rate": 9.294891723898926e-06,
+      "loss": 0.7588,
+      "step": 510
+    },
+    {
+      "epoch": 0.2002117624410434,
+      "grad_norm": 0.12864899635314941,
+      "learning_rate": 9.262630536456153e-06,
+      "loss": 0.7978,
+      "step": 520
+    },
+    {
+      "epoch": 0.2040619886418327,
+      "grad_norm": 0.13816484808921814,
+      "learning_rate": 9.229706346044749e-06,
+      "loss": 0.7648,
+      "step": 530
+    },
+    {
+      "epoch": 0.207912214842622,
+      "grad_norm": 0.13998211920261383,
+      "learning_rate": 9.196124273642422e-06,
+      "loss": 0.7851,
+      "step": 540
+    },
+    {
+      "epoch": 0.2117624410434113,
+      "grad_norm": 0.12150999158620834,
+      "learning_rate": 9.161889542552842e-06,
+      "loss": 0.7937,
+      "step": 550
+    },
+    {
+      "epoch": 0.21561266724420058,
+      "grad_norm": 0.12532921135425568,
+      "learning_rate": 9.127007477593193e-06,
+      "loss": 0.7896,
+      "step": 560
+    },
+    {
+      "epoch": 0.2194628934449899,
+      "grad_norm": 0.1429353654384613,
+      "learning_rate": 9.091483504265974e-06,
+      "loss": 0.7969,
+      "step": 570
+    },
+    {
+      "epoch": 0.22331311964577918,
+      "grad_norm": 0.13457311689853668,
+      "learning_rate": 9.055323147915116e-06,
+      "loss": 0.7743,
+      "step": 580
+    },
+    {
+      "epoch": 0.2271633458465685,
+      "grad_norm": 0.12993083894252777,
+      "learning_rate": 9.018532032866585e-06,
+      "loss": 0.7913,
+      "step": 590
+    },
+    {
+      "epoch": 0.23101357204735778,
+      "grad_norm": 0.1362653225660324,
+      "learning_rate": 8.981115881553579e-06,
+      "loss": 0.7858,
+      "step": 600
+    },
+    {
+      "epoch": 0.23486379824814707,
+      "grad_norm": 0.131048783659935,
+      "learning_rate": 8.943080513626476e-06,
+      "loss": 0.7991,
+      "step": 610
+    },
+    {
+      "epoch": 0.23871402444893638,
+      "grad_norm": 0.12950511276721954,
+      "learning_rate": 8.904431845047641e-06,
+      "loss": 0.7892,
+      "step": 620
+    },
+    {
+      "epoch": 0.24256425064972567,
+      "grad_norm": 0.1350424885749817,
+      "learning_rate": 8.86517588717128e-06,
+      "loss": 0.7928,
+      "step": 630
+    },
+    {
+      "epoch": 0.24641447685051496,
+      "grad_norm": 0.12839165329933167,
+      "learning_rate": 8.82531874580844e-06,
+      "loss": 0.7925,
+      "step": 640
+    },
+    {
+      "epoch": 0.25026470305130427,
+      "grad_norm": 0.1229851096868515,
+      "learning_rate": 8.784866620277314e-06,
+      "loss": 0.7488,
+      "step": 650
+    },
+    {
+      "epoch": 0.25411492925209356,
+      "grad_norm": 0.13565847277641296,
+      "learning_rate": 8.743825802439018e-06,
+      "loss": 0.7686,
+      "step": 660
+    },
+    {
+      "epoch": 0.25796515545288284,
+      "grad_norm": 0.1485622674226761,
+      "learning_rate": 8.702202675718956e-06,
+      "loss": 0.7925,
+      "step": 670
+    },
+    {
+      "epoch": 0.26181538165367213,
+      "grad_norm": 0.12083691358566284,
+      "learning_rate": 8.660003714113959e-06,
+      "loss": 0.7642,
+      "step": 680
+    },
+    {
+      "epoch": 0.26566560785446147,
+      "grad_norm": 0.12809047102928162,
+      "learning_rate": 8.617235481185323e-06,
+      "loss": 0.7753,
+      "step": 690
+    },
+    {
+      "epoch": 0.26951583405525076,
+      "grad_norm": 0.12868186831474304,
+      "learning_rate": 8.573904629037931e-06,
+      "loss": 0.7837,
+      "step": 700
+    },
+    {
+      "epoch": 0.27336606025604004,
+      "grad_norm": 0.12444128096103668,
+      "learning_rate": 8.530017897285585e-06,
+      "loss": 0.7678,
+      "step": 710
+    },
+    {
+      "epoch": 0.27721628645682933,
+      "grad_norm": 0.11730103194713593,
+      "learning_rate": 8.485582112002746e-06,
+      "loss": 0.7731,
+      "step": 720
+    },
+    {
+      "epoch": 0.2810665126576186,
+      "grad_norm": 0.12611621618270874,
+      "learning_rate": 8.440604184662811e-06,
+      "loss": 0.7797,
+      "step": 730
+    },
+    {
+      "epoch": 0.28491673885840796,
+      "grad_norm": 0.13074618577957153,
+      "learning_rate": 8.395091111063118e-06,
+      "loss": 0.7626,
+      "step": 740
+    },
+    {
+      "epoch": 0.28876696505919724,
+      "grad_norm": 0.12045063823461533,
+      "learning_rate": 8.349049970236822e-06,
+      "loss": 0.7658,
+      "step": 750
+    },
+    {
+      "epoch": 0.29261719125998653,
+      "grad_norm": 0.13384932279586792,
+      "learning_rate": 8.302487923351842e-06,
+      "loss": 0.7555,
+      "step": 760
+    },
+    {
+      "epoch": 0.2964674174607758,
+      "grad_norm": 0.12903627753257751,
+      "learning_rate": 8.255412212597021e-06,
+      "loss": 0.7729,
+      "step": 770
+    },
+    {
+      "epoch": 0.3003176436615651,
+      "grad_norm": 0.1376868635416031,
+      "learning_rate": 8.207830160055688e-06,
+      "loss": 0.7691,
+      "step": 780
+    },
+    {
+      "epoch": 0.3041678698623544,
+      "grad_norm": 0.1365862339735031,
+      "learning_rate": 8.159749166566786e-06,
+      "loss": 0.7559,
+      "step": 790
+    },
+    {
+      "epoch": 0.30801809606314373,
+      "grad_norm": 0.13079684972763062,
+      "learning_rate": 8.11117671057377e-06,
+      "loss": 0.7816,
+      "step": 800
+    },
+    {
+      "epoch": 0.311868322263933,
+      "grad_norm": 0.1300382763147354,
+      "learning_rate": 8.062120346961404e-06,
+      "loss": 0.7676,
+      "step": 810
+    },
+    {
+      "epoch": 0.3157185484647223,
+      "grad_norm": 0.1260078102350235,
+      "learning_rate": 8.012587705880705e-06,
+      "loss": 0.7721,
+      "step": 820
+    },
+    {
+      "epoch": 0.3195687746655116,
+      "grad_norm": 0.13013751804828644,
+      "learning_rate": 7.962586491562146e-06,
+      "loss": 0.7588,
+      "step": 830
+    },
+    {
+      "epoch": 0.3234190008663009,
+      "grad_norm": 0.14214059710502625,
+      "learning_rate": 7.912124481117357e-06,
+      "loss": 0.764,
+      "step": 840
+    },
+    {
+      "epoch": 0.3272692270670902,
+      "grad_norm": 0.1227542981505394,
+      "learning_rate": 7.861209523329494e-06,
+      "loss": 0.7667,
+      "step": 850
+    },
+    {
+      "epoch": 0.3311194532678795,
+      "grad_norm": 0.12686562538146973,
+      "learning_rate": 7.809849537432432e-06,
+      "loss": 0.7556,
+      "step": 860
+    },
+    {
+      "epoch": 0.3349696794686688,
+      "grad_norm": 0.12133225798606873,
+      "learning_rate": 7.758052511879035e-06,
+      "loss": 0.7552,
+      "step": 870
+    },
+    {
+      "epoch": 0.3388199056694581,
+      "grad_norm": 0.11846953630447388,
+      "learning_rate": 7.705826503098642e-06,
+      "loss": 0.7631,
+      "step": 880
+    },
+    {
+      "epoch": 0.34267013187024736,
+      "grad_norm": 0.12234468758106232,
+      "learning_rate": 7.653179634243979e-06,
+      "loss": 0.7548,
+      "step": 890
+    },
+    {
+      "epoch": 0.34652035807103665,
+      "grad_norm": 0.11495310813188553,
+      "learning_rate": 7.600120093927696e-06,
+      "loss": 0.7534,
+      "step": 900
+    },
+    {
+      "epoch": 0.350370584271826,
+      "grad_norm": 0.12574291229248047,
+      "learning_rate": 7.546656134948729e-06,
+      "loss": 0.753,
+      "step": 910
+    },
+    {
+      "epoch": 0.3542208104726153,
+      "grad_norm": 0.13088491559028625,
+      "learning_rate": 7.492796073008661e-06,
+      "loss": 0.7468,
+      "step": 920
+    },
+    {
+      "epoch": 0.35807103667340456,
+      "grad_norm": 0.12960480153560638,
+      "learning_rate": 7.438548285418326e-06,
+      "loss": 0.7397,
+      "step": 930
+    },
+    {
+      "epoch": 0.36192126287419385,
+      "grad_norm": 0.1419215053319931,
+      "learning_rate": 7.3839212097948e-06,
+      "loss": 0.7781,
+      "step": 940
+    },
+    {
+      "epoch": 0.36577148907498314,
+      "grad_norm": 0.12144347280263901,
+      "learning_rate": 7.328923342749039e-06,
+      "loss": 0.7867,
+      "step": 950
+    },
+    {
+      "epoch": 0.3696217152757725,
+      "grad_norm": 0.13296520709991455,
+      "learning_rate": 7.273563238564319e-06,
+      "loss": 0.7655,
+      "step": 960
+    },
+    {
+      "epoch": 0.37347194147656176,
+      "grad_norm": 0.1264794021844864,
+      "learning_rate": 7.217849507865724e-06,
+      "loss": 0.7451,
+      "step": 970
+    },
+    {
+      "epoch": 0.37732216767735105,
+      "grad_norm": 0.12319154292345047,
+      "learning_rate": 7.161790816280858e-06,
+      "loss": 0.7576,
+      "step": 980
+    },
+    {
+      "epoch": 0.38117239387814034,
+      "grad_norm": 0.13152700662612915,
+      "learning_rate": 7.1053958830920125e-06,
+      "loss": 0.7375,
+      "step": 990
+    },
+    {
+      "epoch": 0.3850226200789296,
+      "grad_norm": 0.1257735788822174,
+      "learning_rate": 7.0486734798799715e-06,
+      "loss": 0.7665,
+      "step": 1000
+    },
+    {
+      "epoch": 0.3888728462797189,
+      "grad_norm": 0.1299361139535904,
+      "learning_rate": 6.991632429159708e-06,
+      "loss": 0.7856,
+      "step": 1010
+    },
+    {
+      "epoch": 0.39272307248050825,
+      "grad_norm": 0.13672740757465363,
+      "learning_rate": 6.934281603008132e-06,
+      "loss": 0.7616,
+      "step": 1020
+    },
+    {
+      "epoch": 0.39657329868129754,
+      "grad_norm": 0.12965930998325348,
+      "learning_rate": 6.876629921684143e-06,
+      "loss": 0.7733,
+      "step": 1030
+    },
+    {
+      "epoch": 0.4004235248820868,
+      "grad_norm": 0.12508344650268555,
+      "learning_rate": 6.818686352241199e-06,
+      "loss": 0.7704,
+      "step": 1040
+    },
+    {
+      "epoch": 0.4042737510828761,
+      "grad_norm": 0.11504463851451874,
+      "learning_rate": 6.760459907132573e-06,
+      "loss": 0.7582,
+      "step": 1050
+    },
+    {
+      "epoch": 0.4081239772836654,
+      "grad_norm": 0.1380005180835724,
+      "learning_rate": 6.701959642809589e-06,
+      "loss": 0.7629,
+      "step": 1060
+    },
+    {
+      "epoch": 0.41197420348445474,
+      "grad_norm": 0.12178367376327515,
+      "learning_rate": 6.643194658312987e-06,
+      "loss": 0.7474,
+      "step": 1070
+    },
+    {
+      "epoch": 0.415824429685244,
+      "grad_norm": 0.11998729407787323,
+      "learning_rate": 6.584174093857676e-06,
+      "loss": 0.7796,
+      "step": 1080
+    },
+    {
+      "epoch": 0.4196746558860333,
+      "grad_norm": 0.1446264535188675,
+      "learning_rate": 6.5249071294110735e-06,
+      "loss": 0.7563,
+      "step": 1090
+    },
+    {
+      "epoch": 0.4235248820868226,
+      "grad_norm": 0.13725337386131287,
+      "learning_rate": 6.465402983265274e-06,
+      "loss": 0.7749,
+      "step": 1100
+    },
+    {
+      "epoch": 0.4273751082876119,
+      "grad_norm": 0.1291586011648178,
+      "learning_rate": 6.405670910603243e-06,
+      "loss": 0.77,
+      "step": 1110
+    },
+    {
+      "epoch": 0.43122533448840117,
+      "grad_norm": 0.14525939524173737,
+      "learning_rate": 6.345720202059291e-06,
+      "loss": 0.7621,
+      "step": 1120
+    },
+    {
+      "epoch": 0.4350755606891905,
+      "grad_norm": 0.12513136863708496,
+      "learning_rate": 6.285560182274008e-06,
+      "loss": 0.7511,
+      "step": 1130
+    },
+    {
+      "epoch": 0.4389257868899798,
+      "grad_norm": 0.12317145615816116,
+      "learning_rate": 6.225200208443936e-06,
+      "loss": 0.7577,
+      "step": 1140
+    },
+    {
+      "epoch": 0.4427760130907691,
+      "grad_norm": 0.11795306950807571,
+      "learning_rate": 6.16464966886616e-06,
+      "loss": 0.753,
+      "step": 1150
+    },
+    {
+      "epoch": 0.44662623929155837,
+      "grad_norm": 0.14654427766799927,
+      "learning_rate": 6.103917981478065e-06,
+      "loss": 0.7815,
+      "step": 1160
+    },
+    {
+      "epoch": 0.45047646549234766,
+      "grad_norm": 0.11937741190195084,
+      "learning_rate": 6.043014592392489e-06,
+      "loss": 0.787,
+      "step": 1170
+    },
+    {
+      "epoch": 0.454326691693137,
+      "grad_norm": 0.12402702867984772,
+      "learning_rate": 5.981948974428485e-06,
+      "loss": 0.7609,
+      "step": 1180
+    },
+    {
+      "epoch": 0.4581769178939263,
+      "grad_norm": 0.13226568698883057,
+      "learning_rate": 5.920730625637934e-06,
+      "loss": 0.7398,
+      "step": 1190
+    },
+    {
+      "epoch": 0.46202714409471557,
+      "grad_norm": 0.12952344119548798,
+      "learning_rate": 5.8593690678282355e-06,
+      "loss": 0.7509,
+      "step": 1200
+    },
+    {
+      "epoch": 0.46587737029550486,
+      "grad_norm": 0.13042138516902924,
+      "learning_rate": 5.797873845081301e-06,
+      "loss": 0.7553,
+      "step": 1210
+    },
+    {
+      "epoch": 0.46972759649629414,
+      "grad_norm": 0.12683138251304626,
+      "learning_rate": 5.736254522269075e-06,
+      "loss": 0.7693,
+      "step": 1220
+    },
+    {
+      "epoch": 0.47357782269708343,
+      "grad_norm": 0.12486030161380768,
+      "learning_rate": 5.6745206835658386e-06,
+      "loss": 0.7732,
+      "step": 1230
+    },
+    {
+      "epoch": 0.47742804889787277,
+      "grad_norm": 0.12128016352653503,
+      "learning_rate": 5.612681930957495e-06,
+      "loss": 0.7514,
+      "step": 1240
+    },
+    {
+      "epoch": 0.48127827509866206,
+      "grad_norm": 0.12427471578121185,
+      "learning_rate": 5.550747882748095e-06,
+      "loss": 0.7469,
+      "step": 1250
+    },
+    {
+      "epoch": 0.48512850129945134,
+      "grad_norm": 0.12931612133979797,
+      "learning_rate": 5.4887281720638195e-06,
+      "loss": 0.7499,
+      "step": 1260
+    },
+    {
+      "epoch": 0.48897872750024063,
+      "grad_norm": 0.13545168936252594,
+      "learning_rate": 5.42663244535466e-06,
+      "loss": 0.7369,
+      "step": 1270
+    },
+    {
+      "epoch": 0.4928289537010299,
+      "grad_norm": 0.12245013564825058,
+      "learning_rate": 5.3644703608940205e-06,
+      "loss": 0.7443,
+      "step": 1280
+    },
+    {
+      "epoch": 0.49667917990181926,
+      "grad_norm": 0.12406668812036514,
+      "learning_rate": 5.302251587276491e-06,
+      "loss": 0.7668,
+      "step": 1290
+    },
+    {
+      "epoch": 0.5005294061026085,
+      "grad_norm": 0.12936784327030182,
+      "learning_rate": 5.2399858019140005e-06,
+      "loss": 0.7596,
+      "step": 1300
+    },
+    {
+      "epoch": 0.5043796323033978,
+      "grad_norm": 0.1314261108636856,
+      "learning_rate": 5.177682689530615e-06,
+      "loss": 0.7744,
+      "step": 1310
+    },
+    {
+      "epoch": 0.5082298585041871,
+      "grad_norm": 0.1369849145412445,
+      "learning_rate": 5.11535194065619e-06,
+      "loss": 0.7484,
+      "step": 1320
+    },
+    {
+      "epoch": 0.5120800847049765,
+      "grad_norm": 0.13917316496372223,
+      "learning_rate": 5.053003250119113e-06,
+      "loss": 0.7504,
+      "step": 1330
+    },
+    {
+      "epoch": 0.5159303109057657,
+      "grad_norm": 0.1294451802968979,
+      "learning_rate": 4.9906463155383925e-06,
+      "loss": 0.7592,
+      "step": 1340
+    },
+    {
+      "epoch": 0.519780537106555,
+      "grad_norm": 0.12747259438037872,
+      "learning_rate": 4.9282908358153046e-06,
+      "loss": 0.7521,
+      "step": 1350
+    },
+    {
+      "epoch": 0.5236307633073443,
+      "grad_norm": 0.13810330629348755,
+      "learning_rate": 4.8659465096248345e-06,
+      "loss": 0.7611,
+      "step": 1360
+    },
+    {
+      "epoch": 0.5274809895081336,
+      "grad_norm": 0.12704917788505554,
+      "learning_rate": 4.803623033907169e-06,
+      "loss": 0.7658,
+      "step": 1370
+    },
+    {
+      "epoch": 0.5313312157089229,
+      "grad_norm": 0.11905990540981293,
+      "learning_rate": 4.741330102359441e-06,
+      "loss": 0.7719,
+      "step": 1380
+    },
+    {
+      "epoch": 0.5351814419097122,
+      "grad_norm": 0.13607557117938995,
+      "learning_rate": 4.679077403927996e-06,
+      "loss": 0.7457,
+      "step": 1390
+    },
+    {
+      "epoch": 0.5390316681105015,
+      "grad_norm": 0.12807388603687286,
+      "learning_rate": 4.616874621301373e-06,
+      "loss": 0.7523,
+      "step": 1400
+    },
+    {
+      "epoch": 0.5428818943112907,
+      "grad_norm": 0.1255023181438446,
+      "learning_rate": 4.554731429404293e-06,
+      "loss": 0.7695,
+      "step": 1410
+    },
+    {
+      "epoch": 0.5467321205120801,
+      "grad_norm": 0.13200543820858002,
+      "learning_rate": 4.492657493892822e-06,
+      "loss": 0.7514,
+      "step": 1420
+    },
+    {
+      "epoch": 0.5505823467128694,
+      "grad_norm": 0.140672504901886,
+      "learning_rate": 4.430662469650999e-06,
+      "loss": 0.7634,
+      "step": 1430
+    },
+    {
+      "epoch": 0.5544325729136587,
+      "grad_norm": 0.12781327962875366,
+      "learning_rate": 4.368755999289118e-06,
+      "loss": 0.7666,
+      "step": 1440
+    },
+    {
+      "epoch": 0.558282799114448,
+      "grad_norm": 0.14209328591823578,
+      "learning_rate": 4.306947711643949e-06,
+      "loss": 0.7491,
+      "step": 1450
+    },
+    {
+      "epoch": 0.5621330253152372,
+      "grad_norm": 0.13040544092655182,
+      "learning_rate": 4.245247220281068e-06,
+      "loss": 0.7553,
+      "step": 1460
+    },
+    {
+      "epoch": 0.5659832515160266,
+      "grad_norm": 0.12577781081199646,
+      "learning_rate": 4.183664121999587e-06,
+      "loss": 0.7505,
+      "step": 1470
+    },
+    {
+      "epoch": 0.5698334777168159,
+      "grad_norm": 0.13247822225093842,
+      "learning_rate": 4.122207995339474e-06,
+      "loss": 0.7509,
+      "step": 1480
+    },
+    {
+      "epoch": 0.5736837039176051,
+      "grad_norm": 0.12436836212873459,
+      "learning_rate": 4.060888399091734e-06,
+      "loss": 0.7559,
+      "step": 1490
+    },
+    {
+      "epoch": 0.5775339301183945,
+      "grad_norm": 0.12903152406215668,
+      "learning_rate": 3.999714870811639e-06,
+      "loss": 0.7664,
+      "step": 1500
+    },
+    {
+      "epoch": 0.5813841563191837,
+      "grad_norm": 0.12280617654323578,
+      "learning_rate": 3.938696925335275e-06,
+      "loss": 0.7565,
+      "step": 1510
+    },
+    {
+      "epoch": 0.5852343825199731,
+      "grad_norm": 0.1361386924982071,
+      "learning_rate": 3.87784405329962e-06,
+      "loss": 0.7631,
+      "step": 1520
+    },
+    {
+      "epoch": 0.5890846087207623,
+      "grad_norm": 0.13178296387195587,
+      "learning_rate": 3.817165719666382e-06,
+      "loss": 0.7605,
+      "step": 1530
+    },
+    {
+      "epoch": 0.5929348349215516,
+      "grad_norm": 0.12612542510032654,
+      "learning_rate": 3.756671362249841e-06,
+      "loss": 0.7697,
+      "step": 1540
+    },
+    {
+      "epoch": 0.596785061122341,
+      "grad_norm": 0.12480518966913223,
+      "learning_rate": 3.6963703902488907e-06,
+      "loss": 0.7613,
+      "step": 1550
+    },
+    {
+      "epoch": 0.6006352873231302,
+      "grad_norm": 0.12450690567493439,
+      "learning_rate": 3.6362721827835655e-06,
+      "loss": 0.7722,
+      "step": 1560
+    },
+    {
+      "epoch": 0.6044855135239195,
+      "grad_norm": 0.13088077306747437,
+      "learning_rate": 3.576386087436212e-06,
+      "loss": 0.7625,
+      "step": 1570
+    },
+    {
+      "epoch": 0.6083357397247088,
+      "grad_norm": 0.13108126819133759,
+      "learning_rate": 3.5167214187975913e-06,
+      "loss": 0.7641,
+      "step": 1580
+    },
+    {
+      "epoch": 0.6121859659254981,
+      "grad_norm": 0.11436910927295685,
+      "learning_rate": 3.457287457018086e-06,
+      "loss": 0.7629,
+      "step": 1590
+    },
+    {
+      "epoch": 0.6160361921262875,
+      "grad_norm": 0.12197570502758026,
+      "learning_rate": 3.3980934463643003e-06,
+      "loss": 0.7463,
+      "step": 1600
+    },
+    {
+      "epoch": 0.6198864183270767,
+      "grad_norm": 0.11548321694135666,
+      "learning_rate": 3.3391485937812053e-06,
+      "loss": 0.7289,
+      "step": 1610
+    },
+    {
+      "epoch": 0.623736644527866,
+      "grad_norm": 0.12623880803585052,
+      "learning_rate": 3.280462067460115e-06,
+      "loss": 0.7298,
+      "step": 1620
+    },
+    {
+      "epoch": 0.6275868707286553,
+      "grad_norm": 0.1191994771361351,
+      "learning_rate": 3.222042995412669e-06,
+      "loss": 0.7678,
+      "step": 1630
+    },
+    {
+      "epoch": 0.6314370969294446,
+      "grad_norm": 0.11973993480205536,
+      "learning_rate": 3.163900464051089e-06,
+      "loss": 0.7509,
+      "step": 1640
+    },
+    {
+      "epoch": 0.635287323130234,
+      "grad_norm": 0.12807148694992065,
+      "learning_rate": 3.1060435167748826e-06,
+      "loss": 0.7612,
+      "step": 1650
+    },
+    {
+      "epoch": 0.6391375493310232,
+      "grad_norm": 0.12339237332344055,
+      "learning_rate": 3.048481152564253e-06,
+      "loss": 0.7458,
+      "step": 1660
+    },
+    {
+      "epoch": 0.6429877755318125,
+      "grad_norm": 0.1314622163772583,
+      "learning_rate": 2.9912223245804017e-06,
+      "loss": 0.7701,
+      "step": 1670
+    },
+    {
+      "epoch": 0.6468380017326018,
+      "grad_norm": 0.13058611750602722,
+      "learning_rate": 2.934275938772981e-06,
+      "loss": 0.7523,
+      "step": 1680
+    },
+    {
+      "epoch": 0.6506882279333911,
+      "grad_norm": 0.1321302354335785,
+      "learning_rate": 2.8776508524948676e-06,
+      "loss": 0.7429,
+      "step": 1690
+    },
+    {
+      "epoch": 0.6545384541341804,
+      "grad_norm": 0.11570769548416138,
+      "learning_rate": 2.821355873124503e-06,
+      "loss": 0.7392,
+      "step": 1700
+    },
+    {
+      "epoch": 0.6583886803349697,
+      "grad_norm": 0.12870042026042938,
+      "learning_rate": 2.7653997566960144e-06,
+      "loss": 0.7615,
+      "step": 1710
+    },
+    {
+      "epoch": 0.662238906535759,
+      "grad_norm": 0.12542232871055603,
+      "learning_rate": 2.709791206537309e-06,
+      "loss": 0.7489,
+      "step": 1720
+    },
+    {
+      "epoch": 0.6660891327365482,
+      "grad_norm": 0.12854121625423431,
+      "learning_rate": 2.6545388719163747e-06,
+      "loss": 0.7517,
+      "step": 1730
+    },
+    {
+      "epoch": 0.6699393589373376,
+      "grad_norm": 0.12273897230625153,
+      "learning_rate": 2.599651346695979e-06,
+      "loss": 0.7587,
+      "step": 1740
+    },
+    {
+      "epoch": 0.6737895851381268,
+      "grad_norm": 0.13607297837734222,
+      "learning_rate": 2.545137167997007e-06,
+      "loss": 0.7518,
+      "step": 1750
+    },
+    {
+      "epoch": 0.6776398113389162,
+      "grad_norm": 0.11899443715810776,
+      "learning_rate": 2.4910048148705972e-06,
+      "loss": 0.7384,
+      "step": 1760
+    },
+    {
+      "epoch": 0.6814900375397055,
+      "grad_norm": 0.12851440906524658,
+      "learning_rate": 2.4372627069793327e-06,
+      "loss": 0.7809,
+      "step": 1770
+    },
+    {
+      "epoch": 0.6853402637404947,
+      "grad_norm": 0.12435028702020645,
+      "learning_rate": 2.3839192032876486e-06,
+      "loss": 0.7642,
+      "step": 1780
+    },
+    {
+      "epoch": 0.6891904899412841,
+      "grad_norm": 0.11745928227901459,
+      "learning_rate": 2.3309826007617047e-06,
+      "loss": 0.7757,
+      "step": 1790
+    },
+    {
+      "epoch": 0.6930407161420733,
+      "grad_norm": 0.1209171712398529,
+      "learning_rate": 2.278461133078881e-06,
+      "loss": 0.742,
+      "step": 1800
+    },
+    {
+      "epoch": 0.6968909423428626,
+      "grad_norm": 0.12211718410253525,
+      "learning_rate": 2.2263629693471258e-06,
+      "loss": 0.7354,
+      "step": 1810
+    },
+    {
+      "epoch": 0.700741168543652,
+      "grad_norm": 0.1203233078122139,
+      "learning_rate": 2.1746962128343414e-06,
+      "loss": 0.7601,
+      "step": 1820
+    },
+    {
+      "epoch": 0.7045913947444412,
+      "grad_norm": 0.1309351623058319,
+      "learning_rate": 2.1234688997080183e-06,
+      "loss": 0.7575,
+      "step": 1830
+    },
+    {
+      "epoch": 0.7084416209452306,
+      "grad_norm": 0.13817168772220612,
+      "learning_rate": 2.0726889977853053e-06,
+      "loss": 0.754,
+      "step": 1840
+    },
+    {
+      "epoch": 0.7122918471460198,
+      "grad_norm": 0.13249416649341583,
+      "learning_rate": 2.022364405293703e-06,
+      "loss": 0.7644,
+      "step": 1850
+    },
+    {
+      "epoch": 0.7161420733468091,
+      "grad_norm": 0.12247908115386963,
+      "learning_rate": 1.9725029496425814e-06,
+      "loss": 0.7616,
+      "step": 1860
+    },
+    {
+      "epoch": 0.7199922995475985,
+      "grad_norm": 0.12823611497879028,
+      "learning_rate": 1.923112386205727e-06,
+      "loss": 0.7558,
+      "step": 1870
+    },
+    {
+      "epoch": 0.7238425257483877,
+      "grad_norm": 0.1418517529964447,
+      "learning_rate": 1.874200397115079e-06,
+      "loss": 0.7629,
+      "step": 1880
+    },
+    {
+      "epoch": 0.727692751949177,
+      "grad_norm": 0.1251692771911621,
+      "learning_rate": 1.825774590065857e-06,
+      "loss": 0.7494,
+      "step": 1890
+    },
+    {
+      "epoch": 0.7315429781499663,
+      "grad_norm": 0.13196216523647308,
+      "learning_rate": 1.7778424971332813e-06,
+      "loss": 0.7575,
+      "step": 1900
+    },
+    {
+      "epoch": 0.7353932043507556,
+      "grad_norm": 0.12131299078464508,
+      "learning_rate": 1.7304115736010407e-06,
+      "loss": 0.7558,
+      "step": 1910
+    },
+    {
+      "epoch": 0.739243430551545,
+      "grad_norm": 0.12239788472652435,
+      "learning_rate": 1.6834891968017108e-06,
+      "loss": 0.7477,
+      "step": 1920
+    },
+    {
+      "epoch": 0.7430936567523342,
+      "grad_norm": 0.1264759749174118,
+      "learning_rate": 1.6370826649692884e-06,
+      "loss": 0.7622,
+      "step": 1930
+    },
+    {
+      "epoch": 0.7469438829531235,
+      "grad_norm": 0.12783215939998627,
+      "learning_rate": 1.5911991961040463e-06,
+      "loss": 0.7731,
+      "step": 1940
+    },
+    {
+      "epoch": 0.7507941091539128,
+      "grad_norm": 0.12670502066612244,
+      "learning_rate": 1.545845926849851e-06,
+      "loss": 0.7603,
+      "step": 1950
+    },
+    {
+      "epoch": 0.7546443353547021,
+      "grad_norm": 0.13037952780723572,
+      "learning_rate": 1.5010299113841397e-06,
+      "loss": 0.7756,
+      "step": 1960
+    },
+    {
+      "epoch": 0.7584945615554913,
+      "grad_norm": 0.12564264237880707,
+      "learning_rate": 1.4567581203207238e-06,
+      "loss": 0.7719,
+      "step": 1970
+    },
+    {
+      "epoch": 0.7623447877562807,
+      "grad_norm": 0.12609627842903137,
+      "learning_rate": 1.413037439625593e-06,
+      "loss": 0.755,
+      "step": 1980
+    },
+    {
+      "epoch": 0.76619501395707,
+      "grad_norm": 0.1441761553287506,
+      "learning_rate": 1.3698746695458847e-06,
+      "loss": 0.7587,
+      "step": 1990
+    },
+    {
+      "epoch": 0.7700452401578592,
+      "grad_norm": 0.12375761568546295,
+      "learning_rate": 1.3272765235521745e-06,
+      "loss": 0.7584,
+      "step": 2000
+    },
+    {
+      "epoch": 0.7738954663586486,
+      "grad_norm": 0.12393433600664139,
+      "learning_rate": 1.2852496272942877e-06,
+      "loss": 0.7564,
+      "step": 2010
+    },
+    {
+      "epoch": 0.7777456925594378,
+      "grad_norm": 0.12409992516040802,
+      "learning_rate": 1.2438005175707396e-06,
+      "loss": 0.7445,
+      "step": 2020
+    },
+    {
+      "epoch": 0.7815959187602272,
+      "grad_norm": 0.12584196031093597,
+      "learning_rate": 1.2029356413120297e-06,
+      "loss": 0.7294,
+      "step": 2030
+    },
+    {
+      "epoch": 0.7854461449610165,
+      "grad_norm": 0.12385175377130508,
+      "learning_rate": 1.1626613545778786e-06,
+      "loss": 0.7493,
+      "step": 2040
+    },
+    {
+      "epoch": 0.7892963711618057,
+      "grad_norm": 0.11797705292701721,
+      "learning_rate": 1.1229839215686294e-06,
+      "loss": 0.7551,
+      "step": 2050
+    },
+    {
+      "epoch": 0.7931465973625951,
+      "grad_norm": 0.11500206589698792,
+      "learning_rate": 1.0839095136509188e-06,
+      "loss": 0.7365,
+      "step": 2060
+    },
+    {
+      "epoch": 0.7969968235633843,
+      "grad_norm": 0.13311299681663513,
+      "learning_rate": 1.045444208397791e-06,
+      "loss": 0.7486,
+      "step": 2070
+    },
+    {
+      "epoch": 0.8008470497641736,
+      "grad_norm": 0.12979362905025482,
+      "learning_rate": 1.007593988643402e-06,
+      "loss": 0.7357,
+      "step": 2080
+    },
+    {
+      "epoch": 0.804697275964963,
+      "grad_norm": 0.12067045271396637,
+      "learning_rate": 9.703647415524613e-07,
+      "loss": 0.7583,
+      "step": 2090
+    },
+    {
+      "epoch": 0.8085475021657522,
+      "grad_norm": 0.12109290808439255,
+      "learning_rate": 9.337622577045485e-07,
+      "loss": 0.7538,
+      "step": 2100
+    },
+    {
+      "epoch": 0.8123977283665416,
+      "grad_norm": 0.12093615531921387,
+      "learning_rate": 8.977922301934588e-07,
+      "loss": 0.765,
+      "step": 2110
+    },
+    {
+      "epoch": 0.8162479545673308,
+      "grad_norm": 0.12818536162376404,
+      "learning_rate": 8.624602537417009e-07,
+      "loss": 0.762,
+      "step": 2120
+    },
+    {
+      "epoch": 0.8200981807681201,
+      "grad_norm": 0.12509885430335999,
+      "learning_rate": 8.277718238303123e-07,
+      "loss": 0.7479,
+      "step": 2130
+    },
+    {
+      "epoch": 0.8239484069689095,
+      "grad_norm": 0.12512417137622833,
+      "learning_rate": 7.937323358440935e-07,
+      "loss": 0.7587,
+      "step": 2140
+    },
+    {
+      "epoch": 0.8277986331696987,
+      "grad_norm": 0.13549663126468658,
+      "learning_rate": 7.603470842324223e-07,
+      "loss": 0.7436,
+      "step": 2150
+    },
+    {
+      "epoch": 0.831648859370488,
+      "grad_norm": 0.11683791875839233,
+      "learning_rate": 7.276212616857575e-07,
+      "loss": 0.7609,
+      "step": 2160
+    },
+    {
+      "epoch": 0.8354990855712773,
+      "grad_norm": 0.12464386969804764,
+      "learning_rate": 6.955599583279837e-07,
+      "loss": 0.7618,
+      "step": 2170
+    },
+    {
+      "epoch": 0.8393493117720666,
+      "grad_norm": 0.12324350327253342,
+      "learning_rate": 6.641681609246981e-07,
+      "loss": 0.7514,
+      "step": 2180
+    },
+    {
+      "epoch": 0.8431995379728559,
+      "grad_norm": 0.12235836684703827,
+      "learning_rate": 6.334507521075711e-07,
+      "loss": 0.7608,
+      "step": 2190
+    },
+    {
+      "epoch": 0.8470497641736452,
+      "grad_norm": 0.13358792662620544,
+      "learning_rate": 6.034125096149179e-07,
+      "loss": 0.746,
+      "step": 2200
+    },
+    {
+      "epoch": 0.8508999903744345,
+      "grad_norm": 0.1404440850019455,
+      "learning_rate": 5.740581055485722e-07,
+      "loss": 0.7351,
+      "step": 2210
+    },
+    {
+      "epoch": 0.8547502165752238,
+      "grad_norm": 0.11452886462211609,
+      "learning_rate": 5.453921056471961e-07,
+      "loss": 0.7564,
+      "step": 2220
+    },
+    {
+      "epoch": 0.8586004427760131,
+      "grad_norm": 0.12911519408226013,
+      "learning_rate": 5.174189685761288e-07,
+      "loss": 0.7486,
+      "step": 2230
+    },
+    {
+      "epoch": 0.8624506689768023,
+      "grad_norm": 0.12267130613327026,
+      "learning_rate": 4.901430452338979e-07,
+      "loss": 0.7691,
+      "step": 2240
+    },
+    {
+      "epoch": 0.8663008951775917,
+      "grad_norm": 0.12961603701114655,
+      "learning_rate": 4.6356857807548286e-07,
+      "loss": 0.7499,
+      "step": 2250
+    },
+    {
+      "epoch": 0.870151121378381,
+      "grad_norm": 0.12912559509277344,
+      "learning_rate": 4.376997004524541e-07,
+      "loss": 0.7386,
+      "step": 2260
+    },
+    {
+      "epoch": 0.8740013475791703,
+      "grad_norm": 0.13771231472492218,
+      "learning_rate": 4.12540435970073e-07,
+      "loss": 0.7397,
+      "step": 2270
+    },
+    {
+      "epoch": 0.8778515737799596,
+      "grad_norm": 0.11944252252578735,
+      "learning_rate": 3.880946978614703e-07,
+      "loss": 0.7604,
+      "step": 2280
+    },
+    {
+      "epoch": 0.8817017999807488,
+      "grad_norm": 0.12386585026979446,
+      "learning_rate": 3.643662883789878e-07,
+      "loss": 0.7615,
+      "step": 2290
+    },
+    {
+      "epoch": 0.8855520261815382,
+      "grad_norm": 0.1190757155418396,
+      "learning_rate": 3.41358898202781e-07,
+      "loss": 0.744,
+      "step": 2300
+    },
+    {
+      "epoch": 0.8894022523823275,
+      "grad_norm": 0.11846296489238739,
+      "learning_rate": 3.190761058667741e-07,
+      "loss": 0.7444,
+      "step": 2310
+    },
+    {
+      "epoch": 0.8932524785831167,
+      "grad_norm": 0.11411601305007935,
+      "learning_rate": 2.9752137720206463e-07,
+      "loss": 0.7544,
+      "step": 2320
+    },
+    {
+      "epoch": 0.8971027047839061,
+      "grad_norm": 0.12816940248012543,
+      "learning_rate": 2.7669806479785067e-07,
+      "loss": 0.7654,
+      "step": 2330
+    },
+    {
+      "epoch": 0.9009529309846953,
+      "grad_norm": 0.12635554373264313,
+      "learning_rate": 2.566094074799763e-07,
+      "loss": 0.7563,
+      "step": 2340
+    },
+    {
+      "epoch": 0.9048031571854847,
+      "grad_norm": 0.12104842066764832,
+      "learning_rate": 2.3725852980716557e-07,
+      "loss": 0.7384,
+      "step": 2350
+    },
+    {
+      "epoch": 0.908653383386274,
+      "grad_norm": 0.12439543753862381,
+      "learning_rate": 2.1864844158503983e-07,
+      "loss": 0.7557,
+      "step": 2360
+    },
+    {
+      "epoch": 0.9125036095870632,
+      "grad_norm": 0.13675430417060852,
+      "learning_rate": 2.007820373979702e-07,
+      "loss": 0.7434,
+      "step": 2370
+    },
+    {
+      "epoch": 0.9163538357878526,
+      "grad_norm": 0.12440402805805206,
+      "learning_rate": 1.8366209615886532e-07,
+      "loss": 0.7462,
+      "step": 2380
+    },
+    {
+      "epoch": 0.9202040619886418,
+      "grad_norm": 0.12249922007322311,
+      "learning_rate": 1.6729128067693613e-07,
+      "loss": 0.7666,
+      "step": 2390
+    },
+    {
+      "epoch": 0.9240542881894311,
+      "grad_norm": 0.13059799373149872,
+      "learning_rate": 1.5167213724353426e-07,
+      "loss": 0.7566,
+      "step": 2400
+    },
+    {
+      "epoch": 0.9279045143902205,
+      "grad_norm": 0.12229828536510468,
+      "learning_rate": 1.3680709523610058e-07,
+      "loss": 0.7462,
+      "step": 2410
+    },
+    {
+      "epoch": 0.9317547405910097,
+      "grad_norm": 0.11944251507520676,
+      "learning_rate": 1.2269846674030418e-07,
+      "loss": 0.7706,
+      "step": 2420
+    },
+    {
+      "epoch": 0.935604966791799,
+      "grad_norm": 0.13127902150154114,
+      "learning_rate": 1.0934844619042595e-07,
+      "loss": 0.7652,
+      "step": 2430
+    },
+    {
+      "epoch": 0.9394551929925883,
+      "grad_norm": 0.1313180774450302,
+      "learning_rate": 9.675911002803961e-08,
+      "loss": 0.7511,
+      "step": 2440
+    },
+    {
+      "epoch": 0.9433054191933776,
+      "grad_norm": 0.13129884004592896,
+      "learning_rate": 8.493241637904381e-08,
+      "loss": 0.7496,
+      "step": 2450
+    },
+    {
+      "epoch": 0.9471556453941669,
+      "grad_norm": 0.1330460160970688,
+      "learning_rate": 7.387020474909689e-08,
+      "loss": 0.7694,
+      "step": 2460
+    },
+    {
+      "epoch": 0.9510058715949562,
+      "grad_norm": 0.13475367426872253,
+      "learning_rate": 6.357419573750856e-08,
+      "loss": 0.7546,
+      "step": 2470
+    },
+    {
+      "epoch": 0.9548560977957455,
+      "grad_norm": 0.12273130565881729,
+      "learning_rate": 5.4045990769613365e-08,
+      "loss": 0.7498,
+      "step": 2480
+    },
+    {
+      "epoch": 0.9587063239965348,
+      "grad_norm": 0.14283840358257294,
+      "learning_rate": 4.5287071847694434e-08,
+      "loss": 0.7518,
+      "step": 2490
+    },
+    {
+      "epoch": 0.9625565501973241,
+      "grad_norm": 0.12406594306230545,
+      "learning_rate": 3.7298801320467817e-08,
+      "loss": 0.7481,
+      "step": 2500
+    },
+    {
+      "epoch": 0.9664067763981133,
+      "grad_norm": 0.12613433599472046,
+      "learning_rate": 3.0082421671192576e-08,
+      "loss": 0.7431,
+      "step": 2510
+    },
+    {
+      "epoch": 0.9702570025989027,
+      "grad_norm": 0.12013785541057587,
+      "learning_rate": 2.3639055324410355e-08,
+      "loss": 0.7603,
+      "step": 2520
+    },
+    {
+      "epoch": 0.974107228799692,
+      "grad_norm": 0.12596379220485687,
+      "learning_rate": 1.7969704471369477e-08,
+      "loss": 0.7461,
+      "step": 2530
+    },
+    {
+      "epoch": 0.9779574550004813,
+      "grad_norm": 0.11686945706605911,
+      "learning_rate": 1.3075250914142413e-08,
+      "loss": 0.7512,
+      "step": 2540
+    },
+    {
+      "epoch": 0.9818076812012706,
+      "grad_norm": 0.11480376869440079,
+      "learning_rate": 8.956455928476049e-09,
+      "loss": 0.7604,
+      "step": 2550
+    },
+    {
+      "epoch": 0.9856579074020598,
+      "grad_norm": 0.11558600515127182,
+      "learning_rate": 5.6139601453786275e-09,
+      "loss": 0.7417,
+      "step": 2560
+    },
+    {
+      "epoch": 0.9895081336028492,
+      "grad_norm": 0.1309681087732315,
+      "learning_rate": 3.0482834514805602e-09,
+      "loss": 0.7441,
+      "step": 2570
+    },
+    {
+      "epoch": 0.9933583598036385,
+      "grad_norm": 0.11341144889593124,
+      "learning_rate": 1.2598249081713366e-09,
+      "loss": 0.7313,
+      "step": 2580
+    },
+    {
+      "epoch": 0.9972085860044277,
+      "grad_norm": 0.12703518569469452,
+      "learning_rate": 2.488626895302826e-10,
+      "loss": 0.7631,
+      "step": 2590
+    },
+    {
+      "epoch": 0.9999037443449803,
+      "step": 2597,
+      "total_flos": 2.992887459160785e+18,
+      "train_loss": 0.782410487239986,
+      "train_runtime": 13185.5016,
+      "train_samples_per_second": 50.427,
+      "train_steps_per_second": 0.197
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 2597,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.992887459160785e+18,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}

continuous/qwen3/mc2_b16_384/finetune/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cef388169366c924cbf5d58f30b2f78fc4b34f399804507de67468467971d703
+size 6136

continuous/qwen3/mc2_b16_384/pretrain/config.json

@@ -0,0 +1,60 @@
+{
+  "_attn_implementation_autoset": true,
+  "add_faster_video": false,
+  "add_time_instruction": false,
+  "architectures": [
+    "Qwen3ForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "bos_token_id": 151643,
+  "eos_token_id": 151645,
+  "faster_token_stride": 10,
+  "force_sample": false,
+  "head_dim": 128,
+  "hidden_act": "silu",
+  "hidden_size": 2048,
+  "image_aspect_ratio": "square",
+  "image_crop_resolution": null,
+  "image_grid_pinpoints": null,
+  "image_split_resolution": null,
+  "initializer_range": 0.02,
+  "intermediate_size": 6144,
+  "max_position_embeddings": 40960,
+  "max_window_layers": 28,
+  "mm_hidden_size": 768,
+  "mm_newline_position": "grid",
+  "mm_patch_merge_type": "flat",
+  "mm_projector_lr": null,
+  "mm_projector_type": "mlp2x_gelu",
+  "mm_resampler_type": null,
+  "mm_spatial_pool_mode": "bilinear",
+  "mm_spatial_pool_stride": null,
+  "mm_tunable_parts": "mm_mlp_adapter",
+  "mm_use_im_patch_token": false,
+  "mm_use_im_start_end": false,
+  "mm_vision_select_feature": "patch",
+  "mm_vision_select_layer": -2,
+  "mm_vision_tower": "open_clip_hub:ViT-B-16",
+  "mm_vision_tower_lr": null,
+  "model_type": "qwen3",
+  "num_attention_heads": 16,
+  "num_hidden_layers": 28,
+  "num_key_value_heads": 8,
+  "pos_skipping_range": 4096,
+  "rms_norm_eps": 1e-06,
+  "rope_scaling": null,
+  "rope_theta": 1000000,
+  "sliding_window": null,
+  "tie_word_embeddings": true,
+  "tokenizer_model_max_length": 8192,
+  "tokenizer_padding_side": "right",
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.51.3",
+  "use_cache": true,
+  "use_mm_proj": true,
+  "use_pos_skipping": false,
+  "use_sliding_window": false,
+  "vision_tower_pretrained": "/cache/ckpt/download/tokenizer/continuous/mc2/mc2_b16_384.pt",
+  "vocab_size": 151936
+}

continuous/qwen3/mc2_b16_384/pretrain/mm_projector.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9a3c6a5e031c77f928d46cd5001a389a10240528906544100fdc7c7b786554a4
+size 23087160

continuous/qwen3/mc2_b16_384/pretrain/trainer_state.json

@@ -0,0 +1,421 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.0,
+  "eval_steps": 500,
+  "global_step": 545,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.01834862385321101,
+      "grad_norm": 1.8809576034545898,
+      "learning_rate": 0.00032142857142857147,
+      "loss": 8.2989,
+      "step": 10
+    },
+    {
+      "epoch": 0.03669724770642202,
+      "grad_norm": 0.3237287700176239,
+      "learning_rate": 0.0006785714285714287,
+      "loss": 5.2399,
+      "step": 20
+    },
+    {
+      "epoch": 0.05504587155963303,
+      "grad_norm": 0.10630948841571808,
+      "learning_rate": 0.0009999907688176173,
+      "loss": 4.779,
+      "step": 30
+    },
+    {
+      "epoch": 0.07339449541284404,
+      "grad_norm": 0.08973556756973267,
+      "learning_rate": 0.0009988834393115766,
+      "loss": 4.68,
+      "step": 40
+    },
+    {
+      "epoch": 0.09174311926605505,
+      "grad_norm": 0.04739381745457649,
+      "learning_rate": 0.0009959345572940772,
+      "loss": 4.6296,
+      "step": 50
+    },
+    {
+      "epoch": 0.11009174311926606,
+      "grad_norm": 0.0532786101102829,
+      "learning_rate": 0.0009911550081155983,
+      "loss": 4.5932,
+      "step": 60
+    },
+    {
+      "epoch": 0.12844036697247707,
+      "grad_norm": 0.057843610644340515,
+      "learning_rate": 0.0009845624347566433,
+      "loss": 4.549,
+      "step": 70
+    },
+    {
+      "epoch": 0.14678899082568808,
+      "grad_norm": 0.07807637006044388,
+      "learning_rate": 0.0009761811727013547,
+      "loss": 4.523,
+      "step": 80
+    },
+    {
+      "epoch": 0.1651376146788991,
+      "grad_norm": 0.07449490576982498,
+      "learning_rate": 0.0009660421601067666,
+      "loss": 4.4789,
+      "step": 90
+    },
+    {
+      "epoch": 0.1834862385321101,
+      "grad_norm": 0.10196612030267715,
+      "learning_rate": 0.0009541828235992926,
+      "loss": 4.4525,
+      "step": 100
+    },
+    {
+      "epoch": 0.2018348623853211,
+      "grad_norm": 0.08241499215364456,
+      "learning_rate": 0.0009406469401200151,
+      "loss": 4.4129,
+      "step": 110
+    },
+    {
+      "epoch": 0.22018348623853212,
+      "grad_norm": 0.08423774689435959,
+      "learning_rate": 0.0009254844753287492,
+      "loss": 4.3967,
+      "step": 120
+    },
+    {
+      "epoch": 0.23853211009174313,
+      "grad_norm": 0.06886596232652664,
+      "learning_rate": 0.0009087513991633924,
+      "loss": 4.3595,
+      "step": 130
+    },
+    {
+      "epoch": 0.25688073394495414,
+      "grad_norm": 0.09063743054866791,
+      "learning_rate": 0.0008905094792353917,
+      "loss": 4.3338,
+      "step": 140
+    },
+    {
+      "epoch": 0.27522935779816515,
+      "grad_norm": 0.09054228663444519,
+      "learning_rate": 0.0008708260528239789,
+      "loss": 4.3212,
+      "step": 150
+    },
+    {
+      "epoch": 0.29357798165137616,
+      "grad_norm": 0.08652867376804352,
+      "learning_rate": 0.0008497737783108237,
+      "loss": 4.298,
+      "step": 160
+    },
+    {
+      "epoch": 0.3119266055045872,
+      "grad_norm": 0.06645826250314713,
+      "learning_rate": 0.0008274303669726426,
+      "loss": 4.278,
+      "step": 170
+    },
+    {
+      "epoch": 0.3302752293577982,
+      "grad_norm": 0.08091723918914795,
+      "learning_rate": 0.0008038782961218137,
+      "loss": 4.2579,
+      "step": 180
+    },
+    {
+      "epoch": 0.3486238532110092,
+      "grad_norm": 0.09665943682193756,
+      "learning_rate": 0.000779204504653894,
+      "loss": 4.249,
+      "step": 190
+    },
+    {
+      "epoch": 0.3669724770642202,
+      "grad_norm": 0.07515127956867218,
+      "learning_rate": 0.0007535000721258791,
+      "loss": 4.2268,
+      "step": 200
+    },
+    {
+      "epoch": 0.3853211009174312,
+      "grad_norm": 0.07795866578817368,
+      "learning_rate": 0.00072685988254984,
+      "loss": 4.2144,
+      "step": 210
+    },
+    {
+      "epoch": 0.4036697247706422,
+      "grad_norm": 0.08677603304386139,
+      "learning_rate": 0.0006993822741429906,
+      "loss": 4.2226,
+      "step": 220
+    },
+    {
+      "epoch": 0.42201834862385323,
+      "grad_norm": 0.06349547207355499,
+      "learning_rate": 0.0006711686763270818,
+      "loss": 4.2112,
+      "step": 230
+    },
+    {
+      "epoch": 0.44036697247706424,
+      "grad_norm": 0.06895087659358978,
+      "learning_rate": 0.0006423232353170831,
+      "loss": 4.1895,
+      "step": 240
+    },
+    {
+      "epoch": 0.45871559633027525,
+      "grad_norm": 0.06564758718013763,
+      "learning_rate": 0.0006129524296812335,
+      "loss": 4.1605,
+      "step": 250
+    },
+    {
+      "epoch": 0.47706422018348627,
+      "grad_norm": 0.06338971108198166,
+      "learning_rate": 0.0005831646772915651,
+      "loss": 4.1453,
+      "step": 260
+    },
+    {
+      "epoch": 0.4954128440366973,
+      "grad_norm": 0.06917006522417068,
+      "learning_rate": 0.0005530699351157809,
+      "loss": 4.0997,
+      "step": 270
+    },
+    {
+      "epoch": 0.5137614678899083,
+      "grad_norm": 0.07447806000709534,
+      "learning_rate": 0.0005227792933277943,
+      "loss": 4.0799,
+      "step": 280
+    },
+    {
+      "epoch": 0.5321100917431193,
+      "grad_norm": 0.07382050901651382,
+      "learning_rate": 0.0004924045652352092,
+      "loss": 4.0566,
+      "step": 290
+    },
+    {
+      "epoch": 0.5504587155963303,
+      "grad_norm": 0.07474290579557419,
+      "learning_rate": 0.0004620578745374633,
+      "loss": 4.0427,
+      "step": 300
+    },
+    {
+      "epoch": 0.5688073394495413,
+      "grad_norm": 0.07689139991998672,
+      "learning_rate": 0.0004318512414382059,
+      "loss": 4.047,
+      "step": 310
+    },
+    {
+      "epoch": 0.5871559633027523,
+      "grad_norm": 0.06658244132995605,
+      "learning_rate": 0.0004018961691397148,
+      "loss": 4.0322,
+      "step": 320
+    },
+    {
+      "epoch": 0.6055045871559633,
+      "grad_norm": 0.061979539692401886,
+      "learning_rate": 0.0003723032322457458,
+      "loss": 4.0139,
+      "step": 330
+    },
+    {
+      "epoch": 0.6238532110091743,
+      "grad_norm": 0.08082320541143417,
+      "learning_rate": 0.0003431816685921625,
+      "loss": 4.0103,
+      "step": 340
+    },
+    {
+      "epoch": 0.6422018348623854,
+      "grad_norm": 0.08183998614549637,
+      "learning_rate": 0.0003146389760120416,
+      "loss": 4.0108,
+      "step": 350
+    },
+    {
+      "epoch": 0.6605504587155964,
+      "grad_norm": 0.06483215093612671,
+      "learning_rate": 0.00028678051552373485,
+      "loss": 4.0146,
+      "step": 360
+    },
+    {
+      "epoch": 0.6788990825688074,
+      "grad_norm": 0.06451112031936646,
+      "learning_rate": 0.0002597091224066581,
+      "loss": 3.9985,
+      "step": 370
+    },
+    {
+      "epoch": 0.6972477064220184,
+      "grad_norm": 0.058164965361356735,
+      "learning_rate": 0.0002335247266004629,
+      "loss": 4.0092,
+      "step": 380
+    },
+    {
+      "epoch": 0.7155963302752294,
+      "grad_norm": 0.06486489623785019,
+      "learning_rate": 0.000208323983828828,
+      "loss": 3.9918,
+      "step": 390
+    },
+    {
+      "epoch": 0.7339449541284404,
+      "grad_norm": 0.06293128430843353,
+      "learning_rate": 0.0001841999188095224,
+      "loss": 3.9884,
+      "step": 400
+    },
+    {
+      "epoch": 0.7522935779816514,
+      "grad_norm": 0.05822547897696495,
+      "learning_rate": 0.00016124158186777676,
+      "loss": 3.9935,
+      "step": 410
+    },
+    {
+      "epoch": 0.7706422018348624,
+      "grad_norm": 0.05894167348742485,
+      "learning_rate": 0.0001395337202205211,
+      "loss": 3.9854,
+      "step": 420
+    },
+    {
+      "epoch": 0.7889908256880734,
+      "grad_norm": 0.058482032269239426,
+      "learning_rate": 0.00011915646514489292,
+      "loss": 3.9844,
+      "step": 430
+    },
+    {
+      "epoch": 0.8073394495412844,
+      "grad_norm": 0.06938309967517853,
+      "learning_rate": 0.00010018503618578818,
+      "loss": 3.992,
+      "step": 440
+    },
+    {
+      "epoch": 0.8256880733944955,
+      "grad_norm": 0.061818260699510574,
+      "learning_rate": 8.268946349432583e-05,
+      "loss": 3.9822,
+      "step": 450
+    },
+    {
+      "epoch": 0.8440366972477065,
+      "grad_norm": 0.0555083230137825,
+      "learning_rate": 6.673432932217117e-05,
+      "loss": 3.9767,
+      "step": 460
+    },
+    {
+      "epoch": 0.8623853211009175,
+      "grad_norm": 0.056855209171772,
+      "learning_rate": 5.2378529625954684e-05,
+      "loss": 3.9668,
+      "step": 470
+    },
+    {
+      "epoch": 0.8807339449541285,
+      "grad_norm": 0.06019526347517967,
+      "learning_rate": 3.9675056661785556e-05,
+      "loss": 3.9721,
+      "step": 480
+    },
+    {
+      "epoch": 0.8990825688073395,
+      "grad_norm": 0.054468169808387756,
+      "learning_rate": 2.8670803372379538e-05,
+      "loss": 3.9751,
+      "step": 490
+    },
+    {
+      "epoch": 0.9174311926605505,
+      "grad_norm": 0.055912114679813385,
+      "learning_rate": 1.9406390288876586e-05,
+      "loss": 3.9853,
+      "step": 500
+    },
+    {
+      "epoch": 0.9357798165137615,
+      "grad_norm": 0.05435454100370407,
+      "learning_rate": 1.191601558631461e-05,
+      "loss": 3.9756,
+      "step": 510
+    },
+    {
+      "epoch": 0.9541284403669725,
+      "grad_norm": 0.057322122156620026,
+      "learning_rate": 6.2273288462559305e-06,
+      "loss": 3.9716,
+      "step": 520
+    },
+    {
+      "epoch": 0.9724770642201835,
+      "grad_norm": 0.05582620948553085,
+      "learning_rate": 2.361328992552314e-06,
+      "loss": 3.9741,
+      "step": 530
+    },
+    {
+      "epoch": 0.9908256880733946,
+      "grad_norm": 0.05525941029191017,
+      "learning_rate": 3.322867770044202e-07,
+      "loss": 3.9725,
+      "step": 540
+    },
+    {
+      "epoch": 1.0,
+      "step": 545,
+      "total_flos": 2.873876922939474e+17,
+      "train_loss": 4.268001822375377,
+      "train_runtime": 5729.3558,
+      "train_samples_per_second": 97.415,
+      "train_steps_per_second": 0.095
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 545,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": false,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.873876922939474e+17,
+  "train_batch_size": 32,
+  "trial_name": null,
+  "trial_params": null
+}

continuous/qwen3/mc2_b32_224/finetune/added_tokens.json

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

continuous/qwen3/mc2_b32_224/finetune/config.json

@@ -0,0 +1,59 @@
+{
+  "add_faster_video": false,
+  "add_time_instruction": false,
+  "architectures": [
+    "LlavaQwen3ForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "bos_token_id": 151643,
+  "eos_token_id": 151645,
+  "faster_token_stride": 10,
+  "force_sample": false,
+  "head_dim": 128,
+  "hidden_act": "silu",
+  "hidden_size": 2048,
+  "image_aspect_ratio": "square",
+  "image_crop_resolution": null,
+  "image_grid_pinpoints": null,
+  "image_split_resolution": null,
+  "initializer_range": 0.02,
+  "intermediate_size": 6144,
+  "max_position_embeddings": 40960,
+  "max_window_layers": 28,
+  "mm_hidden_size": 768,
+  "mm_newline_position": "grid",
+  "mm_patch_merge_type": "flat",
+  "mm_projector_lr": 0.0001,
+  "mm_projector_type": "mlp2x_gelu",
+  "mm_resampler_type": null,
+  "mm_spatial_pool_mode": "bilinear",
+  "mm_spatial_pool_stride": null,
+  "mm_tunable_parts": "mm_mlp_adapter,mm_language_model",
+  "mm_use_im_patch_token": false,
+  "mm_use_im_start_end": false,
+  "mm_vision_select_feature": "patch",
+  "mm_vision_select_layer": -2,
+  "mm_vision_tower": "open_clip_hub:ViT-B-32",
+  "mm_vision_tower_lr": null,
+  "model_type": "qwen3",
+  "num_attention_heads": 16,
+  "num_hidden_layers": 28,
+  "num_key_value_heads": 8,
+  "pos_skipping_range": 4096,
+  "rms_norm_eps": 1e-06,
+  "rope_scaling": null,
+  "rope_theta": 1000000,
+  "sliding_window": null,
+  "tie_word_embeddings": true,
+  "tokenizer_model_max_length": 8192,
+  "tokenizer_padding_side": "right",
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.51.3",
+  "use_cache": true,
+  "use_mm_proj": true,
+  "use_pos_skipping": false,
+  "use_sliding_window": false,
+  "vision_tower_pretrained": "/cache/ckpt/download/tokenizer/continuous/mc2/mc2_b32_224.pt",
+  "vocab_size": 151936
+}

continuous/qwen3/mc2_b32_224/finetune/generation_config.json

@@ -0,0 +1,13 @@
+{
+  "bos_token_id": 151643,
+  "do_sample": true,
+  "eos_token_id": [
+    151645,
+    151643
+  ],
+  "pad_token_id": 151643,
+  "temperature": 0.6,
+  "top_k": 20,
+  "top_p": 0.95,
+  "transformers_version": "4.51.3"
+}