Upload continuous evaluation checkpoints, batch 24

.gitattributes

@@ -170,3 +170,6 @@ continuous/qwen3/siglip2_l16_512/finetune/tokenizer.json filter=lfs diff=lfs mer
 continuous/qwen3/siglip2_sm14_224/finetune/tokenizer.json filter=lfs diff=lfs merge=lfs -text
 continuous/qwen3/siglip2_sm14_384/finetune/tokenizer.json filter=lfs diff=lfs merge=lfs -text
 continuous/qwen3/siglip2_sm16_256/finetune/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+continuous/qwen3/siglip2_sm16_384/finetune/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+continuous/qwen3/siglip2_sm16_512/finetune/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+continuous/qwen3/webssl_dino1b_full2b_224/finetune/tokenizer.json filter=lfs diff=lfs merge=lfs -text

continuous/qwen3/siglip2_sm16_384/finetune/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:25b08062308f27e60763ba16614eb850fe259aebbcfd4443d4daee10c12a87eb
+size 4946306704

continuous/qwen3/siglip2_sm16_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/siglip2_sm16_384/finetune/tokenizer.json

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

continuous/qwen3/siglip2_sm16_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/siglip2_sm16_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": 8.444092750549316,
+      "learning_rate": 1.153846153846154e-06,
+      "loss": 2.3187,
+      "step": 10
+    },
+    {
+      "epoch": 0.007700452401578592,
+      "grad_norm": 3.1813998222351074,
+      "learning_rate": 2.435897435897436e-06,
+      "loss": 1.8446,
+      "step": 20
+    },
+    {
+      "epoch": 0.01155067860236789,
+      "grad_norm": 2.2102952003479004,
+      "learning_rate": 3.7179487179487184e-06,
+      "loss": 1.4281,
+      "step": 30
+    },
+    {
+      "epoch": 0.015400904803157185,
+      "grad_norm": 0.47708484530448914,
+      "learning_rate": 5e-06,
+      "loss": 1.082,
+      "step": 40
+    },
+    {
+      "epoch": 0.019251131003946483,
+      "grad_norm": 0.4934893250465393,
+      "learning_rate": 6.282051282051282e-06,
+      "loss": 0.9916,
+      "step": 50
+    },
+    {
+      "epoch": 0.02310135720473578,
+      "grad_norm": 0.2275119423866272,
+      "learning_rate": 7.564102564102564e-06,
+      "loss": 0.912,
+      "step": 60
+    },
+    {
+      "epoch": 0.026951583405525073,
+      "grad_norm": 0.18443241715431213,
+      "learning_rate": 8.846153846153847e-06,
+      "loss": 0.8722,
+      "step": 70
+    },
+    {
+      "epoch": 0.03080180960631437,
+      "grad_norm": 0.17515334486961365,
+      "learning_rate": 9.999996111488723e-06,
+      "loss": 0.8712,
+      "step": 80
+    },
+    {
+      "epoch": 0.034652035807103666,
+      "grad_norm": 0.16495226323604584,
+      "learning_rate": 9.999529497453782e-06,
+      "loss": 0.8488,
+      "step": 90
+    },
+    {
+      "epoch": 0.038502262007892966,
+      "grad_norm": 0.15063345432281494,
+      "learning_rate": 9.998285264324193e-06,
+      "loss": 0.8138,
+      "step": 100
+    },
+    {
+      "epoch": 0.04235248820868226,
+      "grad_norm": 0.15115422010421753,
+      "learning_rate": 9.996263605626052e-06,
+      "loss": 0.7969,
+      "step": 110
+    },
+    {
+      "epoch": 0.04620271440947156,
+      "grad_norm": 0.1476246416568756,
+      "learning_rate": 9.99346483580503e-06,
+      "loss": 0.8062,
+      "step": 120
+    },
+    {
+      "epoch": 0.05005294061026085,
+      "grad_norm": 0.13586124777793884,
+      "learning_rate": 9.989889390177467e-06,
+      "loss": 0.7932,
+      "step": 130
+    },
+    {
+      "epoch": 0.053903166811050146,
+      "grad_norm": 0.1597914695739746,
+      "learning_rate": 9.985537824862647e-06,
+      "loss": 0.7963,
+      "step": 140
+    },
+    {
+      "epoch": 0.057753393011839446,
+      "grad_norm": 0.17970605194568634,
+      "learning_rate": 9.980410816696317e-06,
+      "loss": 0.7862,
+      "step": 150
+    },
+    {
+      "epoch": 0.06160361921262874,
+      "grad_norm": 0.14739163219928741,
+      "learning_rate": 9.97450916312541e-06,
+      "loss": 0.7832,
+      "step": 160
+    },
+    {
+      "epoch": 0.06545384541341803,
+      "grad_norm": 0.13291090726852417,
+      "learning_rate": 9.967833782084004e-06,
+      "loss": 0.7842,
+      "step": 170
+    },
+    {
+      "epoch": 0.06930407161420733,
+      "grad_norm": 0.1440795212984085,
+      "learning_rate": 9.960385711850554e-06,
+      "loss": 0.8026,
+      "step": 180
+    },
+    {
+      "epoch": 0.07315429781499663,
+      "grad_norm": 0.14271526038646698,
+      "learning_rate": 9.952166110886394e-06,
+      "loss": 0.8049,
+      "step": 190
+    },
+    {
+      "epoch": 0.07700452401578593,
+      "grad_norm": 0.14573128521442413,
+      "learning_rate": 9.943176257655567e-06,
+      "loss": 0.7809,
+      "step": 200
+    },
+    {
+      "epoch": 0.08085475021657522,
+      "grad_norm": 0.13388106226921082,
+      "learning_rate": 9.933417550425952e-06,
+      "loss": 0.7941,
+      "step": 210
+    },
+    {
+      "epoch": 0.08470497641736452,
+      "grad_norm": 0.13028627634048462,
+      "learning_rate": 9.9228915070518e-06,
+      "loss": 0.7648,
+      "step": 220
+    },
+    {
+      "epoch": 0.08855520261815382,
+      "grad_norm": 0.1532939225435257,
+      "learning_rate": 9.911599764737632e-06,
+      "loss": 0.7711,
+      "step": 230
+    },
+    {
+      "epoch": 0.09240542881894312,
+      "grad_norm": 0.1331004798412323,
+      "learning_rate": 9.899544079783606e-06,
+      "loss": 0.7577,
+      "step": 240
+    },
+    {
+      "epoch": 0.0962556550197324,
+      "grad_norm": 0.14370179176330566,
+      "learning_rate": 9.886726327312338e-06,
+      "loss": 0.7793,
+      "step": 250
+    },
+    {
+      "epoch": 0.1001058812205217,
+      "grad_norm": 0.13295480608940125,
+      "learning_rate": 9.87314850097724e-06,
+      "loss": 0.779,
+      "step": 260
+    },
+    {
+      "epoch": 0.103956107421311,
+      "grad_norm": 0.13279734551906586,
+      "learning_rate": 9.858812712652452e-06,
+      "loss": 0.7633,
+      "step": 270
+    },
+    {
+      "epoch": 0.10780633362210029,
+      "grad_norm": 0.1349678486585617,
+      "learning_rate": 9.843721192104341e-06,
+      "loss": 0.7665,
+      "step": 280
+    },
+    {
+      "epoch": 0.11165655982288959,
+      "grad_norm": 0.1322297751903534,
+      "learning_rate": 9.827876286644698e-06,
+      "loss": 0.7809,
+      "step": 290
+    },
+    {
+      "epoch": 0.11550678602367889,
+      "grad_norm": 0.12256079912185669,
+      "learning_rate": 9.811280460765643e-06,
+      "loss": 0.7433,
+      "step": 300
+    },
+    {
+      "epoch": 0.11935701222446819,
+      "grad_norm": 0.13502788543701172,
+      "learning_rate": 9.793936295756292e-06,
+      "loss": 0.7556,
+      "step": 310
+    },
+    {
+      "epoch": 0.12320723842525748,
+      "grad_norm": 0.12807384133338928,
+      "learning_rate": 9.775846489301279e-06,
+      "loss": 0.7592,
+      "step": 320
+    },
+    {
+      "epoch": 0.12705746462604678,
+      "grad_norm": 0.13768544793128967,
+      "learning_rate": 9.757013855061157e-06,
+      "loss": 0.7593,
+      "step": 330
+    },
+    {
+      "epoch": 0.13090769082683607,
+      "grad_norm": 0.13503430783748627,
+      "learning_rate": 9.737441322234758e-06,
+      "loss": 0.7663,
+      "step": 340
+    },
+    {
+      "epoch": 0.13475791702762538,
+      "grad_norm": 0.14634670317173004,
+      "learning_rate": 9.717131935103605e-06,
+      "loss": 0.7733,
+      "step": 350
+    },
+    {
+      "epoch": 0.13860814322841467,
+      "grad_norm": 0.1398102194070816,
+      "learning_rate": 9.696088852558398e-06,
+      "loss": 0.7577,
+      "step": 360
+    },
+    {
+      "epoch": 0.14245836942920398,
+      "grad_norm": 0.13184933364391327,
+      "learning_rate": 9.674315347607686e-06,
+      "loss": 0.783,
+      "step": 370
+    },
+    {
+      "epoch": 0.14630859562999327,
+      "grad_norm": 0.1229320764541626,
+      "learning_rate": 9.651814806868796e-06,
+      "loss": 0.7533,
+      "step": 380
+    },
+    {
+      "epoch": 0.15015882183078255,
+      "grad_norm": 0.13577893376350403,
+      "learning_rate": 9.62859073004107e-06,
+      "loss": 0.752,
+      "step": 390
+    },
+    {
+      "epoch": 0.15400904803157187,
+      "grad_norm": 0.13020505011081696,
+      "learning_rate": 9.604646729361547e-06,
+      "loss": 0.753,
+      "step": 400
+    },
+    {
+      "epoch": 0.15785927423236115,
+      "grad_norm": 0.13277286291122437,
+      "learning_rate": 9.579986529043099e-06,
+      "loss": 0.7335,
+      "step": 410
+    },
+    {
+      "epoch": 0.16170950043315044,
+      "grad_norm": 0.12725995481014252,
+      "learning_rate": 9.554613964695189e-06,
+      "loss": 0.7572,
+      "step": 420
+    },
+    {
+      "epoch": 0.16555972663393975,
+      "grad_norm": 0.12992346286773682,
+      "learning_rate": 9.528532982727281e-06,
+      "loss": 0.7512,
+      "step": 430
+    },
+    {
+      "epoch": 0.16940995283472904,
+      "grad_norm": 0.1455039530992508,
+      "learning_rate": 9.501747639735018e-06,
+      "loss": 0.7438,
+      "step": 440
+    },
+    {
+      "epoch": 0.17326017903551832,
+      "grad_norm": 0.13229188323020935,
+      "learning_rate": 9.474262101869272e-06,
+      "loss": 0.7312,
+      "step": 450
+    },
+    {
+      "epoch": 0.17711040523630764,
+      "grad_norm": 0.1307026892900467,
+      "learning_rate": 9.446080644188141e-06,
+      "loss": 0.7493,
+      "step": 460
+    },
+    {
+      "epoch": 0.18096063143709692,
+      "grad_norm": 0.13047805428504944,
+      "learning_rate": 9.417207649992017e-06,
+      "loss": 0.7496,
+      "step": 470
+    },
+    {
+      "epoch": 0.18481085763788624,
+      "grad_norm": 0.12318141013383865,
+      "learning_rate": 9.387647610141816e-06,
+      "loss": 0.7416,
+      "step": 480
+    },
+    {
+      "epoch": 0.18866108383867553,
+      "grad_norm": 0.1233312264084816,
+      "learning_rate": 9.35740512236047e-06,
+      "loss": 0.7357,
+      "step": 490
+    },
+    {
+      "epoch": 0.1925113100394648,
+      "grad_norm": 0.11497997492551804,
+      "learning_rate": 9.326484890517809e-06,
+      "loss": 0.7424,
+      "step": 500
+    },
+    {
+      "epoch": 0.19636153624025413,
+      "grad_norm": 0.1268492192029953,
+      "learning_rate": 9.294891723898926e-06,
+      "loss": 0.726,
+      "step": 510
+    },
+    {
+      "epoch": 0.2002117624410434,
+      "grad_norm": 0.12616607546806335,
+      "learning_rate": 9.262630536456153e-06,
+      "loss": 0.7643,
+      "step": 520
+    },
+    {
+      "epoch": 0.2040619886418327,
+      "grad_norm": 0.13867098093032837,
+      "learning_rate": 9.229706346044749e-06,
+      "loss": 0.7346,
+      "step": 530
+    },
+    {
+      "epoch": 0.207912214842622,
+      "grad_norm": 0.13543157279491425,
+      "learning_rate": 9.196124273642422e-06,
+      "loss": 0.7514,
+      "step": 540
+    },
+    {
+      "epoch": 0.2117624410434113,
+      "grad_norm": 0.12138901650905609,
+      "learning_rate": 9.161889542552842e-06,
+      "loss": 0.7619,
+      "step": 550
+    },
+    {
+      "epoch": 0.21561266724420058,
+      "grad_norm": 0.1291125863790512,
+      "learning_rate": 9.127007477593193e-06,
+      "loss": 0.7545,
+      "step": 560
+    },
+    {
+      "epoch": 0.2194628934449899,
+      "grad_norm": 0.1391144096851349,
+      "learning_rate": 9.091483504265974e-06,
+      "loss": 0.7604,
+      "step": 570
+    },
+    {
+      "epoch": 0.22331311964577918,
+      "grad_norm": 0.1353050023317337,
+      "learning_rate": 9.055323147915116e-06,
+      "loss": 0.7399,
+      "step": 580
+    },
+    {
+      "epoch": 0.2271633458465685,
+      "grad_norm": 0.13196918368339539,
+      "learning_rate": 9.018532032866585e-06,
+      "loss": 0.7522,
+      "step": 590
+    },
+    {
+      "epoch": 0.23101357204735778,
+      "grad_norm": 0.13812850415706635,
+      "learning_rate": 8.981115881553579e-06,
+      "loss": 0.7499,
+      "step": 600
+    },
+    {
+      "epoch": 0.23486379824814707,
+      "grad_norm": 0.13404883444309235,
+      "learning_rate": 8.943080513626476e-06,
+      "loss": 0.7632,
+      "step": 610
+    },
+    {
+      "epoch": 0.23871402444893638,
+      "grad_norm": 0.1320178210735321,
+      "learning_rate": 8.904431845047641e-06,
+      "loss": 0.754,
+      "step": 620
+    },
+    {
+      "epoch": 0.24256425064972567,
+      "grad_norm": 0.12813596427440643,
+      "learning_rate": 8.86517588717128e-06,
+      "loss": 0.7576,
+      "step": 630
+    },
+    {
+      "epoch": 0.24641447685051496,
+      "grad_norm": 0.1248459592461586,
+      "learning_rate": 8.82531874580844e-06,
+      "loss": 0.7604,
+      "step": 640
+    },
+    {
+      "epoch": 0.25026470305130427,
+      "grad_norm": 0.12299128621816635,
+      "learning_rate": 8.784866620277314e-06,
+      "loss": 0.7153,
+      "step": 650
+    },
+    {
+      "epoch": 0.25411492925209356,
+      "grad_norm": 0.13738276064395905,
+      "learning_rate": 8.743825802439018e-06,
+      "loss": 0.7298,
+      "step": 660
+    },
+    {
+      "epoch": 0.25796515545288284,
+      "grad_norm": 0.1280859261751175,
+      "learning_rate": 8.702202675718956e-06,
+      "loss": 0.7562,
+      "step": 670
+    },
+    {
+      "epoch": 0.26181538165367213,
+      "grad_norm": 0.12109211087226868,
+      "learning_rate": 8.660003714113959e-06,
+      "loss": 0.727,
+      "step": 680
+    },
+    {
+      "epoch": 0.26566560785446147,
+      "grad_norm": 0.12776783108711243,
+      "learning_rate": 8.617235481185323e-06,
+      "loss": 0.7394,
+      "step": 690
+    },
+    {
+      "epoch": 0.26951583405525076,
+      "grad_norm": 0.1336740255355835,
+      "learning_rate": 8.573904629037931e-06,
+      "loss": 0.7484,
+      "step": 700
+    },
+    {
+      "epoch": 0.27336606025604004,
+      "grad_norm": 0.12742848694324493,
+      "learning_rate": 8.530017897285585e-06,
+      "loss": 0.7329,
+      "step": 710
+    },
+    {
+      "epoch": 0.27721628645682933,
+      "grad_norm": 0.1256263107061386,
+      "learning_rate": 8.485582112002746e-06,
+      "loss": 0.7431,
+      "step": 720
+    },
+    {
+      "epoch": 0.2810665126576186,
+      "grad_norm": 0.13134102523326874,
+      "learning_rate": 8.440604184662811e-06,
+      "loss": 0.7426,
+      "step": 730
+    },
+    {
+      "epoch": 0.28491673885840796,
+      "grad_norm": 0.1301165223121643,
+      "learning_rate": 8.395091111063118e-06,
+      "loss": 0.7281,
+      "step": 740
+    },
+    {
+      "epoch": 0.28876696505919724,
+      "grad_norm": 0.11764539778232574,
+      "learning_rate": 8.349049970236822e-06,
+      "loss": 0.7301,
+      "step": 750
+    },
+    {
+      "epoch": 0.29261719125998653,
+      "grad_norm": 0.12088257819414139,
+      "learning_rate": 8.302487923351842e-06,
+      "loss": 0.7198,
+      "step": 760
+    },
+    {
+      "epoch": 0.2964674174607758,
+      "grad_norm": 0.13478848338127136,
+      "learning_rate": 8.255412212597021e-06,
+      "loss": 0.7404,
+      "step": 770
+    },
+    {
+      "epoch": 0.3003176436615651,
+      "grad_norm": 0.13754799962043762,
+      "learning_rate": 8.207830160055688e-06,
+      "loss": 0.7361,
+      "step": 780
+    },
+    {
+      "epoch": 0.3041678698623544,
+      "grad_norm": 0.14285042881965637,
+      "learning_rate": 8.159749166566786e-06,
+      "loss": 0.7202,
+      "step": 790
+    },
+    {
+      "epoch": 0.30801809606314373,
+      "grad_norm": 0.12756261229515076,
+      "learning_rate": 8.11117671057377e-06,
+      "loss": 0.7443,
+      "step": 800
+    },
+    {
+      "epoch": 0.311868322263933,
+      "grad_norm": 0.12580090761184692,
+      "learning_rate": 8.062120346961404e-06,
+      "loss": 0.73,
+      "step": 810
+    },
+    {
+      "epoch": 0.3157185484647223,
+      "grad_norm": 0.11821761727333069,
+      "learning_rate": 8.012587705880705e-06,
+      "loss": 0.7355,
+      "step": 820
+    },
+    {
+      "epoch": 0.3195687746655116,
+      "grad_norm": 0.12697190046310425,
+      "learning_rate": 7.962586491562146e-06,
+      "loss": 0.7243,
+      "step": 830
+    },
+    {
+      "epoch": 0.3234190008663009,
+      "grad_norm": 0.13195104897022247,
+      "learning_rate": 7.912124481117357e-06,
+      "loss": 0.726,
+      "step": 840
+    },
+    {
+      "epoch": 0.3272692270670902,
+      "grad_norm": 0.12173645943403244,
+      "learning_rate": 7.861209523329494e-06,
+      "loss": 0.7323,
+      "step": 850
+    },
+    {
+      "epoch": 0.3311194532678795,
+      "grad_norm": 0.1277943253517151,
+      "learning_rate": 7.809849537432432e-06,
+      "loss": 0.7206,
+      "step": 860
+    },
+    {
+      "epoch": 0.3349696794686688,
+      "grad_norm": 0.11526156961917877,
+      "learning_rate": 7.758052511879035e-06,
+      "loss": 0.7197,
+      "step": 870
+    },
+    {
+      "epoch": 0.3388199056694581,
+      "grad_norm": 0.11613860726356506,
+      "learning_rate": 7.705826503098642e-06,
+      "loss": 0.7259,
+      "step": 880
+    },
+    {
+      "epoch": 0.34267013187024736,
+      "grad_norm": 0.12724216282367706,
+      "learning_rate": 7.653179634243979e-06,
+      "loss": 0.7212,
+      "step": 890
+    },
+    {
+      "epoch": 0.34652035807103665,
+      "grad_norm": 0.12039519846439362,
+      "learning_rate": 7.600120093927696e-06,
+      "loss": 0.7152,
+      "step": 900
+    },
+    {
+      "epoch": 0.350370584271826,
+      "grad_norm": 0.1433740109205246,
+      "learning_rate": 7.546656134948729e-06,
+      "loss": 0.7149,
+      "step": 910
+    },
+    {
+      "epoch": 0.3542208104726153,
+      "grad_norm": 0.1256803721189499,
+      "learning_rate": 7.492796073008661e-06,
+      "loss": 0.7122,
+      "step": 920
+    },
+    {
+      "epoch": 0.35807103667340456,
+      "grad_norm": 0.12741079926490784,
+      "learning_rate": 7.438548285418326e-06,
+      "loss": 0.7032,
+      "step": 930
+    },
+    {
+      "epoch": 0.36192126287419385,
+      "grad_norm": 0.13575974106788635,
+      "learning_rate": 7.3839212097948e-06,
+      "loss": 0.7419,
+      "step": 940
+    },
+    {
+      "epoch": 0.36577148907498314,
+      "grad_norm": 0.1188625693321228,
+      "learning_rate": 7.328923342749039e-06,
+      "loss": 0.7548,
+      "step": 950
+    },
+    {
+      "epoch": 0.3696217152757725,
+      "grad_norm": 0.13311919569969177,
+      "learning_rate": 7.273563238564319e-06,
+      "loss": 0.7299,
+      "step": 960
+    },
+    {
+      "epoch": 0.37347194147656176,
+      "grad_norm": 0.1257806271314621,
+      "learning_rate": 7.217849507865724e-06,
+      "loss": 0.7064,
+      "step": 970
+    },
+    {
+      "epoch": 0.37732216767735105,
+      "grad_norm": 0.1240515187382698,
+      "learning_rate": 7.161790816280858e-06,
+      "loss": 0.7245,
+      "step": 980
+    },
+    {
+      "epoch": 0.38117239387814034,
+      "grad_norm": 0.1299629956483841,
+      "learning_rate": 7.1053958830920125e-06,
+      "loss": 0.6958,
+      "step": 990
+    },
+    {
+      "epoch": 0.3850226200789296,
+      "grad_norm": 0.12459651380777359,
+      "learning_rate": 7.0486734798799715e-06,
+      "loss": 0.7305,
+      "step": 1000
+    },
+    {
+      "epoch": 0.3888728462797189,
+      "grad_norm": 0.12910540401935577,
+      "learning_rate": 6.991632429159708e-06,
+      "loss": 0.7518,
+      "step": 1010
+    },
+    {
+      "epoch": 0.39272307248050825,
+      "grad_norm": 0.13408590853214264,
+      "learning_rate": 6.934281603008132e-06,
+      "loss": 0.723,
+      "step": 1020
+    },
+    {
+      "epoch": 0.39657329868129754,
+      "grad_norm": 0.12534502148628235,
+      "learning_rate": 6.876629921684143e-06,
+      "loss": 0.7359,
+      "step": 1030
+    },
+    {
+      "epoch": 0.4004235248820868,
+      "grad_norm": 0.1278948038816452,
+      "learning_rate": 6.818686352241199e-06,
+      "loss": 0.7321,
+      "step": 1040
+    },
+    {
+      "epoch": 0.4042737510828761,
+      "grad_norm": 0.11457230895757675,
+      "learning_rate": 6.760459907132573e-06,
+      "loss": 0.7205,
+      "step": 1050
+    },
+    {
+      "epoch": 0.4081239772836654,
+      "grad_norm": 0.12193015962839127,
+      "learning_rate": 6.701959642809589e-06,
+      "loss": 0.7226,
+      "step": 1060
+    },
+    {
+      "epoch": 0.41197420348445474,
+      "grad_norm": 0.12780991196632385,
+      "learning_rate": 6.643194658312987e-06,
+      "loss": 0.7068,
+      "step": 1070
+    },
+    {
+      "epoch": 0.415824429685244,
+      "grad_norm": 0.11574273556470871,
+      "learning_rate": 6.584174093857676e-06,
+      "loss": 0.7449,
+      "step": 1080
+    },
+    {
+      "epoch": 0.4196746558860333,
+      "grad_norm": 0.1380888670682907,
+      "learning_rate": 6.5249071294110735e-06,
+      "loss": 0.7167,
+      "step": 1090
+    },
+    {
+      "epoch": 0.4235248820868226,
+      "grad_norm": 0.13941815495491028,
+      "learning_rate": 6.465402983265274e-06,
+      "loss": 0.7345,
+      "step": 1100
+    },
+    {
+      "epoch": 0.4273751082876119,
+      "grad_norm": 0.12940993905067444,
+      "learning_rate": 6.405670910603243e-06,
+      "loss": 0.7308,
+      "step": 1110
+    },
+    {
+      "epoch": 0.43122533448840117,
+      "grad_norm": 0.1315070241689682,
+      "learning_rate": 6.345720202059291e-06,
+      "loss": 0.722,
+      "step": 1120
+    },
+    {
+      "epoch": 0.4350755606891905,
+      "grad_norm": 0.12647753953933716,
+      "learning_rate": 6.285560182274008e-06,
+      "loss": 0.713,
+      "step": 1130
+    },
+    {
+      "epoch": 0.4389257868899798,
+      "grad_norm": 0.12512950599193573,
+      "learning_rate": 6.225200208443936e-06,
+      "loss": 0.7198,
+      "step": 1140
+    },
+    {
+      "epoch": 0.4427760130907691,
+      "grad_norm": 0.12069609016180038,
+      "learning_rate": 6.16464966886616e-06,
+      "loss": 0.7119,
+      "step": 1150
+    },
+    {
+      "epoch": 0.44662623929155837,
+      "grad_norm": 0.14929576218128204,
+      "learning_rate": 6.103917981478065e-06,
+      "loss": 0.7435,
+      "step": 1160
+    },
+    {
+      "epoch": 0.45047646549234766,
+      "grad_norm": 0.11929406225681305,
+      "learning_rate": 6.043014592392489e-06,
+      "loss": 0.7482,
+      "step": 1170
+    },
+    {
+      "epoch": 0.454326691693137,
+      "grad_norm": 0.11561467498540878,
+      "learning_rate": 5.981948974428485e-06,
+      "loss": 0.7225,
+      "step": 1180
+    },
+    {
+      "epoch": 0.4581769178939263,
+      "grad_norm": 0.12507545948028564,
+      "learning_rate": 5.920730625637934e-06,
+      "loss": 0.6999,
+      "step": 1190
+    },
+    {
+      "epoch": 0.46202714409471557,
+      "grad_norm": 0.13469937443733215,
+      "learning_rate": 5.8593690678282355e-06,
+      "loss": 0.7132,
+      "step": 1200
+    },
+    {
+      "epoch": 0.46587737029550486,
+      "grad_norm": 0.1284819096326828,
+      "learning_rate": 5.797873845081301e-06,
+      "loss": 0.716,
+      "step": 1210
+    },
+    {
+      "epoch": 0.46972759649629414,
+      "grad_norm": 0.12720425426959991,
+      "learning_rate": 5.736254522269075e-06,
+      "loss": 0.7303,
+      "step": 1220
+    },
+    {
+      "epoch": 0.47357782269708343,
+      "grad_norm": 0.13054780662059784,
+      "learning_rate": 5.6745206835658386e-06,
+      "loss": 0.7352,
+      "step": 1230
+    },
+    {
+      "epoch": 0.47742804889787277,
+      "grad_norm": 0.1215982586145401,
+      "learning_rate": 5.612681930957495e-06,
+      "loss": 0.7129,
+      "step": 1240
+    },
+    {
+      "epoch": 0.48127827509866206,
+      "grad_norm": 0.11867617070674896,
+      "learning_rate": 5.550747882748095e-06,
+      "loss": 0.7057,
+      "step": 1250
+    },
+    {
+      "epoch": 0.48512850129945134,
+      "grad_norm": 0.1330382078886032,
+      "learning_rate": 5.4887281720638195e-06,
+      "loss": 0.7134,
+      "step": 1260
+    },
+    {
+      "epoch": 0.48897872750024063,
+      "grad_norm": 0.14069683849811554,
+      "learning_rate": 5.42663244535466e-06,
+      "loss": 0.6999,
+      "step": 1270
+    },
+    {
+      "epoch": 0.4928289537010299,
+      "grad_norm": 0.12917953729629517,
+      "learning_rate": 5.3644703608940205e-06,
+      "loss": 0.7055,
+      "step": 1280
+    },
+    {
+      "epoch": 0.49667917990181926,
+      "grad_norm": 0.1264776885509491,
+      "learning_rate": 5.302251587276491e-06,
+      "loss": 0.7268,
+      "step": 1290
+    },
+    {
+      "epoch": 0.5005294061026085,
+      "grad_norm": 0.12720006704330444,
+      "learning_rate": 5.2399858019140005e-06,
+      "loss": 0.7196,
+      "step": 1300
+    },
+    {
+      "epoch": 0.5043796323033978,
+      "grad_norm": 0.125726580619812,
+      "learning_rate": 5.177682689530615e-06,
+      "loss": 0.7386,
+      "step": 1310
+    },
+    {
+      "epoch": 0.5082298585041871,
+      "grad_norm": 0.13661082088947296,
+      "learning_rate": 5.11535194065619e-06,
+      "loss": 0.7063,
+      "step": 1320
+    },
+    {
+      "epoch": 0.5120800847049765,
+      "grad_norm": 0.14301863312721252,
+      "learning_rate": 5.053003250119113e-06,
+      "loss": 0.7083,
+      "step": 1330
+    },
+    {
+      "epoch": 0.5159303109057657,
+      "grad_norm": 0.1295623928308487,
+      "learning_rate": 4.9906463155383925e-06,
+      "loss": 0.7202,
+      "step": 1340
+    },
+    {
+      "epoch": 0.519780537106555,
+      "grad_norm": 0.12498000264167786,
+      "learning_rate": 4.9282908358153046e-06,
+      "loss": 0.715,
+      "step": 1350
+    },
+    {
+      "epoch": 0.5236307633073443,
+      "grad_norm": 0.1384839415550232,
+      "learning_rate": 4.8659465096248345e-06,
+      "loss": 0.7187,
+      "step": 1360
+    },
+    {
+      "epoch": 0.5274809895081336,
+      "grad_norm": 0.12472275644540787,
+      "learning_rate": 4.803623033907169e-06,
+      "loss": 0.723,
+      "step": 1370
+    },
+    {
+      "epoch": 0.5313312157089229,
+      "grad_norm": 0.1126171350479126,
+      "learning_rate": 4.741330102359441e-06,
+      "loss": 0.7331,
+      "step": 1380
+    },
+    {
+      "epoch": 0.5351814419097122,
+      "grad_norm": 0.1305978000164032,
+      "learning_rate": 4.679077403927996e-06,
+      "loss": 0.7051,
+      "step": 1390
+    },
+    {
+      "epoch": 0.5390316681105015,
+      "grad_norm": 0.12433719635009766,
+      "learning_rate": 4.616874621301373e-06,
+      "loss": 0.7134,
+      "step": 1400
+    },
+    {
+      "epoch": 0.5428818943112907,
+      "grad_norm": 0.12938237190246582,
+      "learning_rate": 4.554731429404293e-06,
+      "loss": 0.7308,
+      "step": 1410
+    },
+    {
+      "epoch": 0.5467321205120801,
+      "grad_norm": 0.13605360686779022,
+      "learning_rate": 4.492657493892822e-06,
+      "loss": 0.7091,
+      "step": 1420
+    },
+    {
+      "epoch": 0.5505823467128694,
+      "grad_norm": 0.13924220204353333,
+      "learning_rate": 4.430662469650999e-06,
+      "loss": 0.7237,
+      "step": 1430
+    },
+    {
+      "epoch": 0.5544325729136587,
+      "grad_norm": 0.1344563364982605,
+      "learning_rate": 4.368755999289118e-06,
+      "loss": 0.7284,
+      "step": 1440
+    },
+    {
+      "epoch": 0.558282799114448,
+      "grad_norm": 0.13022124767303467,
+      "learning_rate": 4.306947711643949e-06,
+      "loss": 0.7095,
+      "step": 1450
+    },
+    {
+      "epoch": 0.5621330253152372,
+      "grad_norm": 0.12979701161384583,
+      "learning_rate": 4.245247220281068e-06,
+      "loss": 0.7161,
+      "step": 1460
+    },
+    {
+      "epoch": 0.5659832515160266,
+      "grad_norm": 0.12816222012043,
+      "learning_rate": 4.183664121999587e-06,
+      "loss": 0.7095,
+      "step": 1470
+    },
+    {
+      "epoch": 0.5698334777168159,
+      "grad_norm": 0.1287643313407898,
+      "learning_rate": 4.122207995339474e-06,
+      "loss": 0.7083,
+      "step": 1480
+    },
+    {
+      "epoch": 0.5736837039176051,
+      "grad_norm": 0.12115758657455444,
+      "learning_rate": 4.060888399091734e-06,
+      "loss": 0.7148,
+      "step": 1490
+    },
+    {
+      "epoch": 0.5775339301183945,
+      "grad_norm": 0.126042902469635,
+      "learning_rate": 3.999714870811639e-06,
+      "loss": 0.7291,
+      "step": 1500
+    },
+    {
+      "epoch": 0.5813841563191837,
+      "grad_norm": 0.12046265602111816,
+      "learning_rate": 3.938696925335275e-06,
+      "loss": 0.7172,
+      "step": 1510
+    },
+    {
+      "epoch": 0.5852343825199731,
+      "grad_norm": 0.13409367203712463,
+      "learning_rate": 3.87784405329962e-06,
+      "loss": 0.7225,
+      "step": 1520
+    },
+    {
+      "epoch": 0.5890846087207623,
+      "grad_norm": 0.1282576024532318,
+      "learning_rate": 3.817165719666382e-06,
+      "loss": 0.7208,
+      "step": 1530
+    },
+    {
+      "epoch": 0.5929348349215516,
+      "grad_norm": 0.12446697056293488,
+      "learning_rate": 3.756671362249841e-06,
+      "loss": 0.7332,
+      "step": 1540
+    },
+    {
+      "epoch": 0.596785061122341,
+      "grad_norm": 0.12534303963184357,
+      "learning_rate": 3.6963703902488907e-06,
+      "loss": 0.722,
+      "step": 1550
+    },
+    {
+      "epoch": 0.6006352873231302,
+      "grad_norm": 0.1238546147942543,
+      "learning_rate": 3.6362721827835655e-06,
+      "loss": 0.7304,
+      "step": 1560
+    },
+    {
+      "epoch": 0.6044855135239195,
+      "grad_norm": 0.12800800800323486,
+      "learning_rate": 3.576386087436212e-06,
+      "loss": 0.7219,
+      "step": 1570
+    },
+    {
+      "epoch": 0.6083357397247088,
+      "grad_norm": 0.1279337853193283,
+      "learning_rate": 3.5167214187975913e-06,
+      "loss": 0.7256,
+      "step": 1580
+    },
+    {
+      "epoch": 0.6121859659254981,
+      "grad_norm": 0.12042578309774399,
+      "learning_rate": 3.457287457018086e-06,
+      "loss": 0.7214,
+      "step": 1590
+    },
+    {
+      "epoch": 0.6160361921262875,
+      "grad_norm": 0.11824890226125717,
+      "learning_rate": 3.3980934463643003e-06,
+      "loss": 0.7089,
+      "step": 1600
+    },
+    {
+      "epoch": 0.6198864183270767,
+      "grad_norm": 0.11608986556529999,
+      "learning_rate": 3.3391485937812053e-06,
+      "loss": 0.6865,
+      "step": 1610
+    },
+    {
+      "epoch": 0.623736644527866,
+      "grad_norm": 0.12301156669855118,
+      "learning_rate": 3.280462067460115e-06,
+      "loss": 0.6905,
+      "step": 1620
+    },
+    {
+      "epoch": 0.6275868707286553,
+      "grad_norm": 0.12057819962501526,
+      "learning_rate": 3.222042995412669e-06,
+      "loss": 0.729,
+      "step": 1630
+    },
+    {
+      "epoch": 0.6314370969294446,
+      "grad_norm": 0.11592449992895126,
+      "learning_rate": 3.163900464051089e-06,
+      "loss": 0.709,
+      "step": 1640
+    },
+    {
+      "epoch": 0.635287323130234,
+      "grad_norm": 0.12263872474431992,
+      "learning_rate": 3.1060435167748826e-06,
+      "loss": 0.7204,
+      "step": 1650
+    },
+    {
+      "epoch": 0.6391375493310232,
+      "grad_norm": 0.12730717658996582,
+      "learning_rate": 3.048481152564253e-06,
+      "loss": 0.7088,
+      "step": 1660
+    },
+    {
+      "epoch": 0.6429877755318125,
+      "grad_norm": 0.12752953171730042,
+      "learning_rate": 2.9912223245804017e-06,
+      "loss": 0.733,
+      "step": 1670
+    },
+    {
+      "epoch": 0.6468380017326018,
+      "grad_norm": 0.12439528852701187,
+      "learning_rate": 2.934275938772981e-06,
+      "loss": 0.7107,
+      "step": 1680
+    },
+    {
+      "epoch": 0.6506882279333911,
+      "grad_norm": 0.13246221840381622,
+      "learning_rate": 2.8776508524948676e-06,
+      "loss": 0.7004,
+      "step": 1690
+    },
+    {
+      "epoch": 0.6545384541341804,
+      "grad_norm": 0.11799187958240509,
+      "learning_rate": 2.821355873124503e-06,
+      "loss": 0.6991,
+      "step": 1700
+    },
+    {
+      "epoch": 0.6583886803349697,
+      "grad_norm": 0.1264214813709259,
+      "learning_rate": 2.7653997566960144e-06,
+      "loss": 0.7215,
+      "step": 1710
+    },
+    {
+      "epoch": 0.662238906535759,
+      "grad_norm": 0.1261773556470871,
+      "learning_rate": 2.709791206537309e-06,
+      "loss": 0.7108,
+      "step": 1720
+    },
+    {
+      "epoch": 0.6660891327365482,
+      "grad_norm": 0.12162455171346664,
+      "learning_rate": 2.6545388719163747e-06,
+      "loss": 0.7074,
+      "step": 1730
+    },
+    {
+      "epoch": 0.6699393589373376,
+      "grad_norm": 0.11728589236736298,
+      "learning_rate": 2.599651346695979e-06,
+      "loss": 0.7192,
+      "step": 1740
+    },
+    {
+      "epoch": 0.6737895851381268,
+      "grad_norm": 0.13194774091243744,
+      "learning_rate": 2.545137167997007e-06,
+      "loss": 0.7142,
+      "step": 1750
+    },
+    {
+      "epoch": 0.6776398113389162,
+      "grad_norm": 0.11907165497541428,
+      "learning_rate": 2.4910048148705972e-06,
+      "loss": 0.6968,
+      "step": 1760
+    },
+    {
+      "epoch": 0.6814900375397055,
+      "grad_norm": 0.13482016324996948,
+      "learning_rate": 2.4372627069793327e-06,
+      "loss": 0.7416,
+      "step": 1770
+    },
+    {
+      "epoch": 0.6853402637404947,
+      "grad_norm": 0.12820713222026825,
+      "learning_rate": 2.3839192032876486e-06,
+      "loss": 0.7243,
+      "step": 1780
+    },
+    {
+      "epoch": 0.6891904899412841,
+      "grad_norm": 0.11868866533041,
+      "learning_rate": 2.3309826007617047e-06,
+      "loss": 0.7362,
+      "step": 1790
+    },
+    {
+      "epoch": 0.6930407161420733,
+      "grad_norm": 0.12178785353899002,
+      "learning_rate": 2.278461133078881e-06,
+      "loss": 0.7001,
+      "step": 1800
+    },
+    {
+      "epoch": 0.6968909423428626,
+      "grad_norm": 0.12117744982242584,
+      "learning_rate": 2.2263629693471258e-06,
+      "loss": 0.695,
+      "step": 1810
+    },
+    {
+      "epoch": 0.700741168543652,
+      "grad_norm": 0.11947214603424072,
+      "learning_rate": 2.1746962128343414e-06,
+      "loss": 0.7181,
+      "step": 1820
+    },
+    {
+      "epoch": 0.7045913947444412,
+      "grad_norm": 0.12162123620510101,
+      "learning_rate": 2.1234688997080183e-06,
+      "loss": 0.7172,
+      "step": 1830
+    },
+    {
+      "epoch": 0.7084416209452306,
+      "grad_norm": 0.13054846227169037,
+      "learning_rate": 2.0726889977853053e-06,
+      "loss": 0.713,
+      "step": 1840
+    },
+    {
+      "epoch": 0.7122918471460198,
+      "grad_norm": 0.1310279220342636,
+      "learning_rate": 2.022364405293703e-06,
+      "loss": 0.7265,
+      "step": 1850
+    },
+    {
+      "epoch": 0.7161420733468091,
+      "grad_norm": 0.1209600418806076,
+      "learning_rate": 1.9725029496425814e-06,
+      "loss": 0.7208,
+      "step": 1860
+    },
+    {
+      "epoch": 0.7199922995475985,
+      "grad_norm": 0.12491311132907867,
+      "learning_rate": 1.923112386205727e-06,
+      "loss": 0.7173,
+      "step": 1870
+    },
+    {
+      "epoch": 0.7238425257483877,
+      "grad_norm": 0.1263144165277481,
+      "learning_rate": 1.874200397115079e-06,
+      "loss": 0.7234,
+      "step": 1880
+    },
+    {
+      "epoch": 0.727692751949177,
+      "grad_norm": 0.1382729560136795,
+      "learning_rate": 1.825774590065857e-06,
+      "loss": 0.7082,
+      "step": 1890
+    },
+    {
+      "epoch": 0.7315429781499663,
+      "grad_norm": 0.13018561899662018,
+      "learning_rate": 1.7778424971332813e-06,
+      "loss": 0.716,
+      "step": 1900
+    },
+    {
+      "epoch": 0.7353932043507556,
+      "grad_norm": 0.11882250010967255,
+      "learning_rate": 1.7304115736010407e-06,
+      "loss": 0.7164,
+      "step": 1910
+    },
+    {
+      "epoch": 0.739243430551545,
+      "grad_norm": 0.12416566908359528,
+      "learning_rate": 1.6834891968017108e-06,
+      "loss": 0.7083,
+      "step": 1920
+    },
+    {
+      "epoch": 0.7430936567523342,
+      "grad_norm": 0.12922653555870056,
+      "learning_rate": 1.6370826649692884e-06,
+      "loss": 0.7183,
+      "step": 1930
+    },
+    {
+      "epoch": 0.7469438829531235,
+      "grad_norm": 0.12399854511022568,
+      "learning_rate": 1.5911991961040463e-06,
+      "loss": 0.7343,
+      "step": 1940
+    },
+    {
+      "epoch": 0.7507941091539128,
+      "grad_norm": 0.12674225866794586,
+      "learning_rate": 1.545845926849851e-06,
+      "loss": 0.7204,
+      "step": 1950
+    },
+    {
+      "epoch": 0.7546443353547021,
+      "grad_norm": 0.1303989142179489,
+      "learning_rate": 1.5010299113841397e-06,
+      "loss": 0.7356,
+      "step": 1960
+    },
+    {
+      "epoch": 0.7584945615554913,
+      "grad_norm": 0.11998962610960007,
+      "learning_rate": 1.4567581203207238e-06,
+      "loss": 0.7327,
+      "step": 1970
+    },
+    {
+      "epoch": 0.7623447877562807,
+      "grad_norm": 0.1229470744729042,
+      "learning_rate": 1.413037439625593e-06,
+      "loss": 0.7144,
+      "step": 1980
+    },
+    {
+      "epoch": 0.76619501395707,
+      "grad_norm": 0.13213655352592468,
+      "learning_rate": 1.3698746695458847e-06,
+      "loss": 0.7182,
+      "step": 1990
+    },
+    {
+      "epoch": 0.7700452401578592,
+      "grad_norm": 0.12135055661201477,
+      "learning_rate": 1.3272765235521745e-06,
+      "loss": 0.7183,
+      "step": 2000
+    },
+    {
+      "epoch": 0.7738954663586486,
+      "grad_norm": 0.11908351629972458,
+      "learning_rate": 1.2852496272942877e-06,
+      "loss": 0.7176,
+      "step": 2010
+    },
+    {
+      "epoch": 0.7777456925594378,
+      "grad_norm": 0.12859134376049042,
+      "learning_rate": 1.2438005175707396e-06,
+      "loss": 0.7012,
+      "step": 2020
+    },
+    {
+      "epoch": 0.7815959187602272,
+      "grad_norm": 0.12366722524166107,
+      "learning_rate": 1.2029356413120297e-06,
+      "loss": 0.6881,
+      "step": 2030
+    },
+    {
+      "epoch": 0.7854461449610165,
+      "grad_norm": 0.12688297033309937,
+      "learning_rate": 1.1626613545778786e-06,
+      "loss": 0.7066,
+      "step": 2040
+    },
+    {
+      "epoch": 0.7892963711618057,
+      "grad_norm": 0.12214566767215729,
+      "learning_rate": 1.1229839215686294e-06,
+      "loss": 0.7146,
+      "step": 2050
+    },
+    {
+      "epoch": 0.7931465973625951,
+      "grad_norm": 0.11938678473234177,
+      "learning_rate": 1.0839095136509188e-06,
+      "loss": 0.6938,
+      "step": 2060
+    },
+    {
+      "epoch": 0.7969968235633843,
+      "grad_norm": 0.13014480471611023,
+      "learning_rate": 1.045444208397791e-06,
+      "loss": 0.7043,
+      "step": 2070
+    },
+    {
+      "epoch": 0.8008470497641736,
+      "grad_norm": 0.13512416183948517,
+      "learning_rate": 1.007593988643402e-06,
+      "loss": 0.6937,
+      "step": 2080
+    },
+    {
+      "epoch": 0.804697275964963,
+      "grad_norm": 0.12127909064292908,
+      "learning_rate": 9.703647415524613e-07,
+      "loss": 0.7179,
+      "step": 2090
+    },
+    {
+      "epoch": 0.8085475021657522,
+      "grad_norm": 0.1203254759311676,
+      "learning_rate": 9.337622577045485e-07,
+      "loss": 0.7101,
+      "step": 2100
+    },
+    {
+      "epoch": 0.8123977283665416,
+      "grad_norm": 0.12062784284353256,
+      "learning_rate": 8.977922301934588e-07,
+      "loss": 0.7257,
+      "step": 2110
+    },
+    {
+      "epoch": 0.8162479545673308,
+      "grad_norm": 0.1279744952917099,
+      "learning_rate": 8.624602537417009e-07,
+      "loss": 0.7252,
+      "step": 2120
+    },
+    {
+      "epoch": 0.8200981807681201,
+      "grad_norm": 0.12613505125045776,
+      "learning_rate": 8.277718238303123e-07,
+      "loss": 0.7058,
+      "step": 2130
+    },
+    {
+      "epoch": 0.8239484069689095,
+      "grad_norm": 0.12035037577152252,
+      "learning_rate": 7.937323358440935e-07,
+      "loss": 0.7182,
+      "step": 2140
+    },
+    {
+      "epoch": 0.8277986331696987,
+      "grad_norm": 0.13535834848880768,
+      "learning_rate": 7.603470842324223e-07,
+      "loss": 0.7006,
+      "step": 2150
+    },
+    {
+      "epoch": 0.831648859370488,
+      "grad_norm": 0.11819159984588623,
+      "learning_rate": 7.276212616857575e-07,
+      "loss": 0.719,
+      "step": 2160
+    },
+    {
+      "epoch": 0.8354990855712773,
+      "grad_norm": 0.12447500973939896,
+      "learning_rate": 6.955599583279837e-07,
+      "loss": 0.7198,
+      "step": 2170
+    },
+    {
+      "epoch": 0.8393493117720666,
+      "grad_norm": 0.1222880482673645,
+      "learning_rate": 6.641681609246981e-07,
+      "loss": 0.7125,
+      "step": 2180
+    },
+    {
+      "epoch": 0.8431995379728559,
+      "grad_norm": 0.12282800674438477,
+      "learning_rate": 6.334507521075711e-07,
+      "loss": 0.7193,
+      "step": 2190
+    },
+    {
+      "epoch": 0.8470497641736452,
+      "grad_norm": 0.13311302661895752,
+      "learning_rate": 6.034125096149179e-07,
+      "loss": 0.7071,
+      "step": 2200
+    },
+    {
+      "epoch": 0.8508999903744345,
+      "grad_norm": 0.14969627559185028,
+      "learning_rate": 5.740581055485722e-07,
+      "loss": 0.6937,
+      "step": 2210
+    },
+    {
+      "epoch": 0.8547502165752238,
+      "grad_norm": 0.11962267011404037,
+      "learning_rate": 5.453921056471961e-07,
+      "loss": 0.7165,
+      "step": 2220
+    },
+    {
+      "epoch": 0.8586004427760131,
+      "grad_norm": 0.1297527402639389,
+      "learning_rate": 5.174189685761288e-07,
+      "loss": 0.7092,
+      "step": 2230
+    },
+    {
+      "epoch": 0.8624506689768023,
+      "grad_norm": 0.12166012823581696,
+      "learning_rate": 4.901430452338979e-07,
+      "loss": 0.7297,
+      "step": 2240
+    },
+    {
+      "epoch": 0.8663008951775917,
+      "grad_norm": 0.1218472495675087,
+      "learning_rate": 4.6356857807548286e-07,
+      "loss": 0.7082,
+      "step": 2250
+    },
+    {
+      "epoch": 0.870151121378381,
+      "grad_norm": 0.12963467836380005,
+      "learning_rate": 4.376997004524541e-07,
+      "loss": 0.696,
+      "step": 2260
+    },
+    {
+      "epoch": 0.8740013475791703,
+      "grad_norm": 0.12335466593503952,
+      "learning_rate": 4.12540435970073e-07,
+      "loss": 0.697,
+      "step": 2270
+    },
+    {
+      "epoch": 0.8778515737799596,
+      "grad_norm": 0.1200144961476326,
+      "learning_rate": 3.880946978614703e-07,
+      "loss": 0.7231,
+      "step": 2280
+    },
+    {
+      "epoch": 0.8817017999807488,
+      "grad_norm": 0.11697900295257568,
+      "learning_rate": 3.643662883789878e-07,
+      "loss": 0.7167,
+      "step": 2290
+    },
+    {
+      "epoch": 0.8855520261815382,
+      "grad_norm": 0.1233878880739212,
+      "learning_rate": 3.41358898202781e-07,
+      "loss": 0.7007,
+      "step": 2300
+    },
+    {
+      "epoch": 0.8894022523823275,
+      "grad_norm": 0.1196722611784935,
+      "learning_rate": 3.190761058667741e-07,
+      "loss": 0.7044,
+      "step": 2310
+    },
+    {
+      "epoch": 0.8932524785831167,
+      "grad_norm": 0.11713335663080215,
+      "learning_rate": 2.9752137720206463e-07,
+      "loss": 0.7112,
+      "step": 2320
+    },
+    {
+      "epoch": 0.8971027047839061,
+      "grad_norm": 0.12375761568546295,
+      "learning_rate": 2.7669806479785067e-07,
+      "loss": 0.723,
+      "step": 2330
+    },
+    {
+      "epoch": 0.9009529309846953,
+      "grad_norm": 0.12560828030109406,
+      "learning_rate": 2.566094074799763e-07,
+      "loss": 0.7131,
+      "step": 2340
+    },
+    {
+      "epoch": 0.9048031571854847,
+      "grad_norm": 0.11854120343923569,
+      "learning_rate": 2.3725852980716557e-07,
+      "loss": 0.7006,
+      "step": 2350
+    },
+    {
+      "epoch": 0.908653383386274,
+      "grad_norm": 0.12732408940792084,
+      "learning_rate": 2.1864844158503983e-07,
+      "loss": 0.7133,
+      "step": 2360
+    },
+    {
+      "epoch": 0.9125036095870632,
+      "grad_norm": 0.1379638910293579,
+      "learning_rate": 2.007820373979702e-07,
+      "loss": 0.7008,
+      "step": 2370
+    },
+    {
+      "epoch": 0.9163538357878526,
+      "grad_norm": 0.12194670736789703,
+      "learning_rate": 1.8366209615886532e-07,
+      "loss": 0.7066,
+      "step": 2380
+    },
+    {
+      "epoch": 0.9202040619886418,
+      "grad_norm": 0.12399723380804062,
+      "learning_rate": 1.6729128067693613e-07,
+      "loss": 0.7275,
+      "step": 2390
+    },
+    {
+      "epoch": 0.9240542881894311,
+      "grad_norm": 0.12406569719314575,
+      "learning_rate": 1.5167213724353426e-07,
+      "loss": 0.7172,
+      "step": 2400
+    },
+    {
+      "epoch": 0.9279045143902205,
+      "grad_norm": 0.12342435121536255,
+      "learning_rate": 1.3680709523610058e-07,
+      "loss": 0.7029,
+      "step": 2410
+    },
+    {
+      "epoch": 0.9317547405910097,
+      "grad_norm": 0.11704636365175247,
+      "learning_rate": 1.2269846674030418e-07,
+      "loss": 0.7275,
+      "step": 2420
+    },
+    {
+      "epoch": 0.935604966791799,
+      "grad_norm": 0.1284426748752594,
+      "learning_rate": 1.0934844619042595e-07,
+      "loss": 0.7238,
+      "step": 2430
+    },
+    {
+      "epoch": 0.9394551929925883,
+      "grad_norm": 0.12693189084529877,
+      "learning_rate": 9.675911002803961e-08,
+      "loss": 0.7091,
+      "step": 2440
+    },
+    {
+      "epoch": 0.9433054191933776,
+      "grad_norm": 0.12666544318199158,
+      "learning_rate": 8.493241637904381e-08,
+      "loss": 0.7055,
+      "step": 2450
+    },
+    {
+      "epoch": 0.9471556453941669,
+      "grad_norm": 0.13084998726844788,
+      "learning_rate": 7.387020474909689e-08,
+      "loss": 0.7296,
+      "step": 2460
+    },
+    {
+      "epoch": 0.9510058715949562,
+      "grad_norm": 0.12405373156070709,
+      "learning_rate": 6.357419573750856e-08,
+      "loss": 0.7152,
+      "step": 2470
+    },
+    {
+      "epoch": 0.9548560977957455,
+      "grad_norm": 0.12202118337154388,
+      "learning_rate": 5.4045990769613365e-08,
+      "loss": 0.7094,
+      "step": 2480
+    },
+    {
+      "epoch": 0.9587063239965348,
+      "grad_norm": 0.1408575177192688,
+      "learning_rate": 4.5287071847694434e-08,
+      "loss": 0.7071,
+      "step": 2490
+    },
+    {
+      "epoch": 0.9625565501973241,
+      "grad_norm": 0.12194312363862991,
+      "learning_rate": 3.7298801320467817e-08,
+      "loss": 0.7077,
+      "step": 2500
+    },
+    {
+      "epoch": 0.9664067763981133,
+      "grad_norm": 0.12331145256757736,
+      "learning_rate": 3.0082421671192576e-08,
+      "loss": 0.7008,
+      "step": 2510
+    },
+    {
+      "epoch": 0.9702570025989027,
+      "grad_norm": 0.1170383095741272,
+      "learning_rate": 2.3639055324410355e-08,
+      "loss": 0.7204,
+      "step": 2520
+    },
+    {
+      "epoch": 0.974107228799692,
+      "grad_norm": 0.12283165007829666,
+      "learning_rate": 1.7969704471369477e-08,
+      "loss": 0.7049,
+      "step": 2530
+    },
+    {
+      "epoch": 0.9779574550004813,
+      "grad_norm": 0.11516356468200684,
+      "learning_rate": 1.3075250914142413e-08,
+      "loss": 0.7123,
+      "step": 2540
+    },
+    {
+      "epoch": 0.9818076812012706,
+      "grad_norm": 0.1187368705868721,
+      "learning_rate": 8.956455928476049e-09,
+      "loss": 0.7218,
+      "step": 2550
+    },
+    {
+      "epoch": 0.9856579074020598,
+      "grad_norm": 0.11360736936330795,
+      "learning_rate": 5.6139601453786275e-09,
+      "loss": 0.6988,
+      "step": 2560
+    },
+    {
+      "epoch": 0.9895081336028492,
+      "grad_norm": 0.127064511179924,
+      "learning_rate": 3.0482834514805602e-09,
+      "loss": 0.7033,
+      "step": 2570
+    },
+    {
+      "epoch": 0.9933583598036385,
+      "grad_norm": 0.11328203231096268,
+      "learning_rate": 1.2598249081713366e-09,
+      "loss": 0.6874,
+      "step": 2580
+    },
+    {
+      "epoch": 0.9972085860044277,
+      "grad_norm": 0.12906111776828766,
+      "learning_rate": 2.488626895302826e-10,
+      "loss": 0.7229,
+      "step": 2590
+    },
+    {
+      "epoch": 0.9999037443449803,
+      "step": 2597,
+      "total_flos": 3.6756222906046874e+18,
+      "train_loss": 0.7460494926446761,
+      "train_runtime": 13378.4219,
+      "train_samples_per_second": 49.7,
+      "train_steps_per_second": 0.194
+    }
+  ],
+  "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": 3.6756222906046874e+18,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}

continuous/qwen3/siglip2_sm16_384/finetune/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:976c36676e86ccce4728615eaad7529159374e47c295d2ab04d55e296bba1e6a
+size 6136

continuous/qwen3/siglip2_sm16_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": 1152,
+  "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-SO400M-16-SigLIP2-384",
+  "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/siglip2/siglip2_sm16_384.npz",
+  "vocab_size": 151936
+}

continuous/qwen3/siglip2_sm16_384/pretrain/mm_projector.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d2616f41f1a611eab65a5d1842da012382e55ef1716e762b6b6bcf24b62fe95f
+size 26232888

continuous/qwen3/siglip2_sm16_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": 9.592612266540527,
+      "learning_rate": 0.00032142857142857147,
+      "loss": 9.7188,
+      "step": 10
+    },
+    {
+      "epoch": 0.03669724770642202,
+      "grad_norm": 1.3810558319091797,
+      "learning_rate": 0.0006785714285714287,
+      "loss": 7.0882,
+      "step": 20
+    },
+    {
+      "epoch": 0.05504587155963303,
+      "grad_norm": 0.1066640168428421,
+      "learning_rate": 0.0009999907688176173,
+      "loss": 5.2205,
+      "step": 30
+    },
+    {
+      "epoch": 0.07339449541284404,
+      "grad_norm": 0.06683636456727982,
+      "learning_rate": 0.0009988834393115766,
+      "loss": 4.8659,
+      "step": 40
+    },
+    {
+      "epoch": 0.09174311926605505,
+      "grad_norm": 0.04423782229423523,
+      "learning_rate": 0.0009959345572940772,
+      "loss": 4.7612,
+      "step": 50
+    },
+    {
+      "epoch": 0.11009174311926606,
+      "grad_norm": 0.030366962775588036,
+      "learning_rate": 0.0009911550081155983,
+      "loss": 4.7081,
+      "step": 60
+    },
+    {
+      "epoch": 0.12844036697247707,
+      "grad_norm": 0.030705856159329414,
+      "learning_rate": 0.0009845624347566433,
+      "loss": 4.6572,
+      "step": 70
+    },
+    {
+      "epoch": 0.14678899082568808,
+      "grad_norm": 0.051416344940662384,
+      "learning_rate": 0.0009761811727013547,
+      "loss": 4.6286,
+      "step": 80
+    },
+    {
+      "epoch": 0.1651376146788991,
+      "grad_norm": 0.07534493505954742,
+      "learning_rate": 0.0009660421601067666,
+      "loss": 4.595,
+      "step": 90
+    },
+    {
+      "epoch": 0.1834862385321101,
+      "grad_norm": 0.05507621541619301,
+      "learning_rate": 0.0009541828235992926,
+      "loss": 4.5743,
+      "step": 100
+    },
+    {
+      "epoch": 0.2018348623853211,
+      "grad_norm": 0.04051034525036812,
+      "learning_rate": 0.0009406469401200151,
+      "loss": 4.5372,
+      "step": 110
+    },
+    {
+      "epoch": 0.22018348623853212,
+      "grad_norm": 0.06599336117506027,
+      "learning_rate": 0.0009254844753287492,
+      "loss": 4.5032,
+      "step": 120
+    },
+    {
+      "epoch": 0.23853211009174313,
+      "grad_norm": 0.08082321286201477,
+      "learning_rate": 0.0009087513991633924,
+      "loss": 4.4357,
+      "step": 130
+    },
+    {
+      "epoch": 0.25688073394495414,
+      "grad_norm": 0.07353227585554123,
+      "learning_rate": 0.0008905094792353917,
+      "loss": 4.3865,
+      "step": 140
+    },
+    {
+      "epoch": 0.27522935779816515,
+      "grad_norm": 0.08034230023622513,
+      "learning_rate": 0.0008708260528239789,
+      "loss": 4.3464,
+      "step": 150
+    },
+    {
+      "epoch": 0.29357798165137616,
+      "grad_norm": 0.06664060056209564,
+      "learning_rate": 0.0008497737783108237,
+      "loss": 4.3042,
+      "step": 160
+    },
+    {
+      "epoch": 0.3119266055045872,
+      "grad_norm": 0.10047152638435364,
+      "learning_rate": 0.0008274303669726426,
+      "loss": 4.2695,
+      "step": 170
+    },
+    {
+      "epoch": 0.3302752293577982,
+      "grad_norm": 0.09838051348924637,
+      "learning_rate": 0.0008038782961218137,
+      "loss": 4.2412,
+      "step": 180
+    },
+    {
+      "epoch": 0.3486238532110092,
+      "grad_norm": 0.08758968859910965,
+      "learning_rate": 0.000779204504653894,
+      "loss": 4.2201,
+      "step": 190
+    },
+    {
+      "epoch": 0.3669724770642202,
+      "grad_norm": 0.0763658657670021,
+      "learning_rate": 0.0007535000721258791,
+      "loss": 4.1833,
+      "step": 200
+    },
+    {
+      "epoch": 0.3853211009174312,
+      "grad_norm": 0.08120562881231308,
+      "learning_rate": 0.00072685988254984,
+      "loss": 4.1596,
+      "step": 210
+    },
+    {
+      "epoch": 0.4036697247706422,
+      "grad_norm": 0.0851607695221901,
+      "learning_rate": 0.0006993822741429906,
+      "loss": 4.163,
+      "step": 220
+    },
+    {
+      "epoch": 0.42201834862385323,
+      "grad_norm": 0.08280778676271439,
+      "learning_rate": 0.0006711686763270818,
+      "loss": 4.1436,
+      "step": 230
+    },
+    {
+      "epoch": 0.44036697247706424,
+      "grad_norm": 0.0693846344947815,
+      "learning_rate": 0.0006423232353170831,
+      "loss": 4.1214,
+      "step": 240
+    },
+    {
+      "epoch": 0.45871559633027525,
+      "grad_norm": 0.07106713950634003,
+      "learning_rate": 0.0006129524296812335,
+      "loss": 4.0909,
+      "step": 250
+    },
+    {
+      "epoch": 0.47706422018348627,
+      "grad_norm": 0.07372285425662994,
+      "learning_rate": 0.0005831646772915651,
+      "loss": 4.0887,
+      "step": 260
+    },
+    {
+      "epoch": 0.4954128440366973,
+      "grad_norm": 0.10800884664058685,
+      "learning_rate": 0.0005530699351157809,
+      "loss": 4.0522,
+      "step": 270
+    },
+    {
+      "epoch": 0.5137614678899083,
+      "grad_norm": 0.08527794480323792,
+      "learning_rate": 0.0005227792933277943,
+      "loss": 4.0257,
+      "step": 280
+    },
+    {
+      "epoch": 0.5321100917431193,
+      "grad_norm": 0.0866655483841896,
+      "learning_rate": 0.0004924045652352092,
+      "loss": 3.9898,
+      "step": 290
+    },
+    {
+      "epoch": 0.5504587155963303,
+      "grad_norm": 0.08388351649045944,
+      "learning_rate": 0.0004620578745374633,
+      "loss": 3.9646,
+      "step": 300
+    },
+    {
+      "epoch": 0.5688073394495413,
+      "grad_norm": 0.10302125662565231,
+      "learning_rate": 0.0004318512414382059,
+      "loss": 3.9605,
+      "step": 310
+    },
+    {
+      "epoch": 0.5871559633027523,
+      "grad_norm": 0.08074595779180527,
+      "learning_rate": 0.0004018961691397148,
+      "loss": 3.9388,
+      "step": 320
+    },
+    {
+      "epoch": 0.6055045871559633,
+      "grad_norm": 0.08374352753162384,
+      "learning_rate": 0.0003723032322457458,
+      "loss": 3.9071,
+      "step": 330
+    },
+    {
+      "epoch": 0.6238532110091743,
+      "grad_norm": 0.09239687025547028,
+      "learning_rate": 0.0003431816685921625,
+      "loss": 3.8983,
+      "step": 340
+    },
+    {
+      "epoch": 0.6422018348623854,
+      "grad_norm": 0.0881769210100174,
+      "learning_rate": 0.0003146389760120416,
+      "loss": 3.8916,
+      "step": 350
+    },
+    {
+      "epoch": 0.6605504587155964,
+      "grad_norm": 0.08691006153821945,
+      "learning_rate": 0.00028678051552373485,
+      "loss": 3.8938,
+      "step": 360
+    },
+    {
+      "epoch": 0.6788990825688074,
+      "grad_norm": 0.07743919640779495,
+      "learning_rate": 0.0002597091224066581,
+      "loss": 3.8741,
+      "step": 370
+    },
+    {
+      "epoch": 0.6972477064220184,
+      "grad_norm": 0.07442434132099152,
+      "learning_rate": 0.0002335247266004629,
+      "loss": 3.8815,
+      "step": 380
+    },
+    {
+      "epoch": 0.7155963302752294,
+      "grad_norm": 0.07868624478578568,
+      "learning_rate": 0.000208323983828828,
+      "loss": 3.8661,
+      "step": 390
+    },
+    {
+      "epoch": 0.7339449541284404,
+      "grad_norm": 0.06836798787117004,
+      "learning_rate": 0.0001841999188095224,
+      "loss": 3.858,
+      "step": 400
+    },
+    {
+      "epoch": 0.7522935779816514,
+      "grad_norm": 0.06908497959375381,
+      "learning_rate": 0.00016124158186777676,
+      "loss": 3.8643,
+      "step": 410
+    },
+    {
+      "epoch": 0.7706422018348624,
+      "grad_norm": 0.06951092928647995,
+      "learning_rate": 0.0001395337202205211,
+      "loss": 3.8538,
+      "step": 420
+    },
+    {
+      "epoch": 0.7889908256880734,
+      "grad_norm": 0.07305663824081421,
+      "learning_rate": 0.00011915646514489292,
+      "loss": 3.8551,
+      "step": 430
+    },
+    {
+      "epoch": 0.8073394495412844,
+      "grad_norm": 0.07419716566801071,
+      "learning_rate": 0.00010018503618578818,
+      "loss": 3.8583,
+      "step": 440
+    },
+    {
+      "epoch": 0.8256880733944955,
+      "grad_norm": 0.06949955224990845,
+      "learning_rate": 8.268946349432583e-05,
+      "loss": 3.8486,
+      "step": 450
+    },
+    {
+      "epoch": 0.8440366972477065,
+      "grad_norm": 0.06778975576162338,
+      "learning_rate": 6.673432932217117e-05,
+      "loss": 3.8452,
+      "step": 460
+    },
+    {
+      "epoch": 0.8623853211009175,
+      "grad_norm": 0.07071290165185928,
+      "learning_rate": 5.2378529625954684e-05,
+      "loss": 3.831,
+      "step": 470
+    },
+    {
+      "epoch": 0.8807339449541285,
+      "grad_norm": 0.06646542996168137,
+      "learning_rate": 3.9675056661785556e-05,
+      "loss": 3.8405,
+      "step": 480
+    },
+    {
+      "epoch": 0.8990825688073395,
+      "grad_norm": 0.06297536939382553,
+      "learning_rate": 2.8670803372379538e-05,
+      "loss": 3.8379,
+      "step": 490
+    },
+    {
+      "epoch": 0.9174311926605505,
+      "grad_norm": 0.06396061927080154,
+      "learning_rate": 1.9406390288876586e-05,
+      "loss": 3.8495,
+      "step": 500
+    },
+    {
+      "epoch": 0.9357798165137615,
+      "grad_norm": 0.061966102570295334,
+      "learning_rate": 1.191601558631461e-05,
+      "loss": 3.8408,
+      "step": 510
+    },
+    {
+      "epoch": 0.9541284403669725,
+      "grad_norm": 0.06180514022707939,
+      "learning_rate": 6.2273288462559305e-06,
+      "loss": 3.8346,
+      "step": 520
+    },
+    {
+      "epoch": 0.9724770642201835,
+      "grad_norm": 0.06313981860876083,
+      "learning_rate": 2.361328992552314e-06,
+      "loss": 3.8379,
+      "step": 530
+    },
+    {
+      "epoch": 0.9908256880733946,
+      "grad_norm": 0.06104377284646034,
+      "learning_rate": 3.322867770044202e-07,
+      "loss": 3.8368,
+      "step": 540
+    },
+    {
+      "epoch": 1.0,
+      "step": 545,
+      "total_flos": 3.529463200607109e+17,
+      "train_loss": 4.289576269727235,
+      "train_runtime": 5750.7732,
+      "train_samples_per_second": 97.053,
+      "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": 3.529463200607109e+17,
+  "train_batch_size": 32,
+  "trial_name": null,
+  "trial_params": null
+}

continuous/qwen3/siglip2_sm16_512/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/siglip2_sm16_512/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": 1152,
+  "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-SO400M-16-SigLIP2-512",
+  "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/siglip2/siglip2_sm16_512.npz",
+  "vocab_size": 151936
+}

continuous/qwen3/siglip2_sm16_512/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/siglip2_sm16_512/finetune/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cb258c5968eec12286e412c383f3cba72f98d7c9f90e99427525c0b08f6aaf9d
+size 4947338896

continuous/qwen3/siglip2_sm16_512/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/siglip2_sm16_512/finetune/tokenizer.json

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

continuous/qwen3/siglip2_sm16_512/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/siglip2_sm16_512/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": 7.69282341003418,
+      "learning_rate": 1.153846153846154e-06,
+      "loss": 2.1217,
+      "step": 10
+    },
+    {
+      "epoch": 0.007700452401578592,
+      "grad_norm": 3.0307016372680664,
+      "learning_rate": 2.435897435897436e-06,
+      "loss": 1.7634,
+      "step": 20
+    },
+    {
+      "epoch": 0.01155067860236789,
+      "grad_norm": 2.170640230178833,
+      "learning_rate": 3.7179487179487184e-06,
+      "loss": 1.4329,
+      "step": 30
+    },
+    {
+      "epoch": 0.015400904803157185,
+      "grad_norm": 0.40722253918647766,
+      "learning_rate": 5e-06,
+      "loss": 1.1109,
+      "step": 40
+    },
+    {
+      "epoch": 0.019251131003946483,
+      "grad_norm": 0.44781020283699036,
+      "learning_rate": 6.282051282051282e-06,
+      "loss": 1.0382,
+      "step": 50
+    },
+    {
+      "epoch": 0.02310135720473578,
+      "grad_norm": 0.21879667043685913,
+      "learning_rate": 7.564102564102564e-06,
+      "loss": 0.9703,
+      "step": 60
+    },
+    {
+      "epoch": 0.026951583405525073,
+      "grad_norm": 0.18965379893779755,
+      "learning_rate": 8.846153846153847e-06,
+      "loss": 0.9337,
+      "step": 70
+    },
+    {
+      "epoch": 0.03080180960631437,
+      "grad_norm": 0.1643671840429306,
+      "learning_rate": 9.999996111488723e-06,
+      "loss": 0.9352,
+      "step": 80
+    },
+    {
+      "epoch": 0.034652035807103666,
+      "grad_norm": 0.1724574863910675,
+      "learning_rate": 9.999529497453782e-06,
+      "loss": 0.9113,
+      "step": 90
+    },
+    {
+      "epoch": 0.038502262007892966,
+      "grad_norm": 0.15496373176574707,
+      "learning_rate": 9.998285264324193e-06,
+      "loss": 0.8863,
+      "step": 100
+    },
+    {
+      "epoch": 0.04235248820868226,
+      "grad_norm": 0.14994128048419952,
+      "learning_rate": 9.996263605626052e-06,
+      "loss": 0.8651,
+      "step": 110
+    },
+    {
+      "epoch": 0.04620271440947156,
+      "grad_norm": 0.15502619743347168,
+      "learning_rate": 9.99346483580503e-06,
+      "loss": 0.8733,
+      "step": 120
+    },
+    {
+      "epoch": 0.05005294061026085,
+      "grad_norm": 0.1448432356119156,
+      "learning_rate": 9.989889390177467e-06,
+      "loss": 0.8648,
+      "step": 130
+    },
+    {
+      "epoch": 0.053903166811050146,
+      "grad_norm": 0.16513553261756897,
+      "learning_rate": 9.985537824862647e-06,
+      "loss": 0.8607,
+      "step": 140
+    },
+    {
+      "epoch": 0.057753393011839446,
+      "grad_norm": 0.17246918380260468,
+      "learning_rate": 9.980410816696317e-06,
+      "loss": 0.857,
+      "step": 150
+    },
+    {
+      "epoch": 0.06160361921262874,
+      "grad_norm": 0.14631138741970062,
+      "learning_rate": 9.97450916312541e-06,
+      "loss": 0.8507,
+      "step": 160
+    },
+    {
+      "epoch": 0.06545384541341803,
+      "grad_norm": 0.13120506703853607,
+      "learning_rate": 9.967833782084004e-06,
+      "loss": 0.8475,
+      "step": 170
+    },
+    {
+      "epoch": 0.06930407161420733,
+      "grad_norm": 0.1390198916196823,
+      "learning_rate": 9.960385711850554e-06,
+      "loss": 0.873,
+      "step": 180
+    },
+    {
+      "epoch": 0.07315429781499663,
+      "grad_norm": 0.15418776869773865,
+      "learning_rate": 9.952166110886394e-06,
+      "loss": 0.8709,
+      "step": 190
+    },
+    {
+      "epoch": 0.07700452401578593,
+      "grad_norm": 0.15659844875335693,
+      "learning_rate": 9.943176257655567e-06,
+      "loss": 0.85,
+      "step": 200
+    },
+    {
+      "epoch": 0.08085475021657522,
+      "grad_norm": 0.13864581286907196,
+      "learning_rate": 9.933417550425952e-06,
+      "loss": 0.8559,
+      "step": 210
+    },
+    {
+      "epoch": 0.08470497641736452,
+      "grad_norm": 0.13657830655574799,
+      "learning_rate": 9.9228915070518e-06,
+      "loss": 0.8339,
+      "step": 220
+    },
+    {
+      "epoch": 0.08855520261815382,
+      "grad_norm": 0.1602155566215515,
+      "learning_rate": 9.911599764737632e-06,
+      "loss": 0.84,
+      "step": 230
+    },
+    {
+      "epoch": 0.09240542881894312,
+      "grad_norm": 0.14449453353881836,
+      "learning_rate": 9.899544079783606e-06,
+      "loss": 0.8268,
+      "step": 240
+    },
+    {
+      "epoch": 0.0962556550197324,
+      "grad_norm": 0.15164409577846527,
+      "learning_rate": 9.886726327312338e-06,
+      "loss": 0.8427,
+      "step": 250
+    },
+    {
+      "epoch": 0.1001058812205217,
+      "grad_norm": 0.1417578160762787,
+      "learning_rate": 9.87314850097724e-06,
+      "loss": 0.8465,
+      "step": 260
+    },
+    {
+      "epoch": 0.103956107421311,
+      "grad_norm": 0.14316003024578094,
+      "learning_rate": 9.858812712652452e-06,
+      "loss": 0.8317,
+      "step": 270
+    },
+    {
+      "epoch": 0.10780633362210029,
+      "grad_norm": 0.13424164056777954,
+      "learning_rate": 9.843721192104341e-06,
+      "loss": 0.8352,
+      "step": 280
+    },
+    {
+      "epoch": 0.11165655982288959,
+      "grad_norm": 0.14523132145404816,
+      "learning_rate": 9.827876286644698e-06,
+      "loss": 0.8463,
+      "step": 290
+    },
+    {
+      "epoch": 0.11550678602367889,
+      "grad_norm": 0.1312931925058365,
+      "learning_rate": 9.811280460765643e-06,
+      "loss": 0.8086,
+      "step": 300
+    },
+    {
+      "epoch": 0.11935701222446819,
+      "grad_norm": 0.1309727430343628,
+      "learning_rate": 9.793936295756292e-06,
+      "loss": 0.8249,
+      "step": 310
+    },
+    {
+      "epoch": 0.12320723842525748,
+      "grad_norm": 0.13561148941516876,
+      "learning_rate": 9.775846489301279e-06,
+      "loss": 0.8199,
+      "step": 320
+    },
+    {
+      "epoch": 0.12705746462604678,
+      "grad_norm": 0.13507939875125885,
+      "learning_rate": 9.757013855061157e-06,
+      "loss": 0.8226,
+      "step": 330
+    },
+    {
+      "epoch": 0.13090769082683607,
+      "grad_norm": 0.13643309473991394,
+      "learning_rate": 9.737441322234758e-06,
+      "loss": 0.831,
+      "step": 340
+    },
+    {
+      "epoch": 0.13475791702762538,
+      "grad_norm": 0.15719856321811676,
+      "learning_rate": 9.717131935103605e-06,
+      "loss": 0.8366,
+      "step": 350
+    },
+    {
+      "epoch": 0.13860814322841467,
+      "grad_norm": 0.12920133769512177,
+      "learning_rate": 9.696088852558398e-06,
+      "loss": 0.8204,
+      "step": 360
+    },
+    {
+      "epoch": 0.14245836942920398,
+      "grad_norm": 0.1389956921339035,
+      "learning_rate": 9.674315347607686e-06,
+      "loss": 0.8444,
+      "step": 370
+    },
+    {
+      "epoch": 0.14630859562999327,
+      "grad_norm": 0.1355723738670349,
+      "learning_rate": 9.651814806868796e-06,
+      "loss": 0.813,
+      "step": 380
+    },
+    {
+      "epoch": 0.15015882183078255,
+      "grad_norm": 0.15223298966884613,
+      "learning_rate": 9.62859073004107e-06,
+      "loss": 0.8189,
+      "step": 390
+    },
+    {
+      "epoch": 0.15400904803157187,
+      "grad_norm": 0.13950768113136292,
+      "learning_rate": 9.604646729361547e-06,
+      "loss": 0.8178,
+      "step": 400
+    },
+    {
+      "epoch": 0.15785927423236115,
+      "grad_norm": 0.14391587674617767,
+      "learning_rate": 9.579986529043099e-06,
+      "loss": 0.7996,
+      "step": 410
+    },
+    {
+      "epoch": 0.16170950043315044,
+      "grad_norm": 0.13525289297103882,
+      "learning_rate": 9.554613964695189e-06,
+      "loss": 0.8153,
+      "step": 420
+    },
+    {
+      "epoch": 0.16555972663393975,
+      "grad_norm": 0.1329314261674881,
+      "learning_rate": 9.528532982727281e-06,
+      "loss": 0.808,
+      "step": 430
+    },
+    {
+      "epoch": 0.16940995283472904,
+      "grad_norm": 0.1505863219499588,
+      "learning_rate": 9.501747639735018e-06,
+      "loss": 0.8074,
+      "step": 440
+    },
+    {
+      "epoch": 0.17326017903551832,
+      "grad_norm": 0.14331106841564178,
+      "learning_rate": 9.474262101869272e-06,
+      "loss": 0.7907,
+      "step": 450
+    },
+    {
+      "epoch": 0.17711040523630764,
+      "grad_norm": 0.13675405085086823,
+      "learning_rate": 9.446080644188141e-06,
+      "loss": 0.8068,
+      "step": 460
+    },
+    {
+      "epoch": 0.18096063143709692,
+      "grad_norm": 0.1486693173646927,
+      "learning_rate": 9.417207649992017e-06,
+      "loss": 0.8108,
+      "step": 470
+    },
+    {
+      "epoch": 0.18481085763788624,
+      "grad_norm": 0.13473084568977356,
+      "learning_rate": 9.387647610141816e-06,
+      "loss": 0.8046,
+      "step": 480
+    },
+    {
+      "epoch": 0.18866108383867553,
+      "grad_norm": 0.12831509113311768,
+      "learning_rate": 9.35740512236047e-06,
+      "loss": 0.7943,
+      "step": 490
+    },
+    {
+      "epoch": 0.1925113100394648,
+      "grad_norm": 0.12404825538396835,
+      "learning_rate": 9.326484890517809e-06,
+      "loss": 0.7965,
+      "step": 500
+    },
+    {
+      "epoch": 0.19636153624025413,
+      "grad_norm": 0.13987357914447784,
+      "learning_rate": 9.294891723898926e-06,
+      "loss": 0.7826,
+      "step": 510
+    },
+    {
+      "epoch": 0.2002117624410434,
+      "grad_norm": 0.1357942819595337,
+      "learning_rate": 9.262630536456153e-06,
+      "loss": 0.8232,
+      "step": 520
+    },
+    {
+      "epoch": 0.2040619886418327,
+      "grad_norm": 0.1518590748310089,
+      "learning_rate": 9.229706346044749e-06,
+      "loss": 0.7892,
+      "step": 530
+    },
+    {
+      "epoch": 0.207912214842622,
+      "grad_norm": 0.1501801759004593,
+      "learning_rate": 9.196124273642422e-06,
+      "loss": 0.8088,
+      "step": 540
+    },
+    {
+      "epoch": 0.2117624410434113,
+      "grad_norm": 0.1279003769159317,
+      "learning_rate": 9.161889542552842e-06,
+      "loss": 0.8125,
+      "step": 550
+    },
+    {
+      "epoch": 0.21561266724420058,
+      "grad_norm": 0.1337757706642151,
+      "learning_rate": 9.127007477593193e-06,
+      "loss": 0.8125,
+      "step": 560
+    },
+    {
+      "epoch": 0.2194628934449899,
+      "grad_norm": 0.1441097855567932,
+      "learning_rate": 9.091483504265974e-06,
+      "loss": 0.8217,
+      "step": 570
+    },
+    {
+      "epoch": 0.22331311964577918,
+      "grad_norm": 0.13492406904697418,
+      "learning_rate": 9.055323147915116e-06,
+      "loss": 0.7971,
+      "step": 580
+    },
+    {
+      "epoch": 0.2271633458465685,
+      "grad_norm": 0.14721833169460297,
+      "learning_rate": 9.018532032866585e-06,
+      "loss": 0.8133,
+      "step": 590
+    },
+    {
+      "epoch": 0.23101357204735778,
+      "grad_norm": 0.14932498335838318,
+      "learning_rate": 8.981115881553579e-06,
+      "loss": 0.8073,
+      "step": 600
+    },
+    {
+      "epoch": 0.23486379824814707,
+      "grad_norm": 0.14673320949077606,
+      "learning_rate": 8.943080513626476e-06,
+      "loss": 0.8175,
+      "step": 610
+    },
+    {
+      "epoch": 0.23871402444893638,
+      "grad_norm": 0.13756655156612396,
+      "learning_rate": 8.904431845047641e-06,
+      "loss": 0.8096,
+      "step": 620
+    },
+    {
+      "epoch": 0.24256425064972567,
+      "grad_norm": 0.14141610264778137,
+      "learning_rate": 8.86517588717128e-06,
+      "loss": 0.8145,
+      "step": 630
+    },
+    {
+      "epoch": 0.24641447685051496,
+      "grad_norm": 0.13106046617031097,
+      "learning_rate": 8.82531874580844e-06,
+      "loss": 0.8071,
+      "step": 640
+    },
+    {
+      "epoch": 0.25026470305130427,
+      "grad_norm": 0.13474811613559723,
+      "learning_rate": 8.784866620277314e-06,
+      "loss": 0.7647,
+      "step": 650
+    },
+    {
+      "epoch": 0.25411492925209356,
+      "grad_norm": 0.14584143459796906,
+      "learning_rate": 8.743825802439018e-06,
+      "loss": 0.7904,
+      "step": 660
+    },
+    {
+      "epoch": 0.25796515545288284,
+      "grad_norm": 0.15183709561824799,
+      "learning_rate": 8.702202675718956e-06,
+      "loss": 0.8055,
+      "step": 670
+    },
+    {
+      "epoch": 0.26181538165367213,
+      "grad_norm": 0.14225754141807556,
+      "learning_rate": 8.660003714113959e-06,
+      "loss": 0.7804,
+      "step": 680
+    },
+    {
+      "epoch": 0.26566560785446147,
+      "grad_norm": 0.13437888026237488,
+      "learning_rate": 8.617235481185323e-06,
+      "loss": 0.7918,
+      "step": 690
+    },
+    {
+      "epoch": 0.26951583405525076,
+      "grad_norm": 0.14489242434501648,
+      "learning_rate": 8.573904629037931e-06,
+      "loss": 0.7973,
+      "step": 700
+    },
+    {
+      "epoch": 0.27336606025604004,
+      "grad_norm": 0.13477161526679993,
+      "learning_rate": 8.530017897285585e-06,
+      "loss": 0.7817,
+      "step": 710
+    },
+    {
+      "epoch": 0.27721628645682933,
+      "grad_norm": 0.1338420957326889,
+      "learning_rate": 8.485582112002746e-06,
+      "loss": 0.7837,
+      "step": 720
+    },
+    {
+      "epoch": 0.2810665126576186,
+      "grad_norm": 0.1491350680589676,
+      "learning_rate": 8.440604184662811e-06,
+      "loss": 0.7884,
+      "step": 730
+    },
+    {
+      "epoch": 0.28491673885840796,
+      "grad_norm": 0.14853794872760773,
+      "learning_rate": 8.395091111063118e-06,
+      "loss": 0.7685,
+      "step": 740
+    },
+    {
+      "epoch": 0.28876696505919724,
+      "grad_norm": 0.13475355505943298,
+      "learning_rate": 8.349049970236822e-06,
+      "loss": 0.7659,
+      "step": 750
+    },
+    {
+      "epoch": 0.29261719125998653,
+      "grad_norm": 0.14394941926002502,
+      "learning_rate": 8.302487923351842e-06,
+      "loss": 0.7541,
+      "step": 760
+    },
+    {
+      "epoch": 0.2964674174607758,
+      "grad_norm": 0.14894339442253113,
+      "learning_rate": 8.255412212597021e-06,
+      "loss": 0.771,
+      "step": 770
+    },
+    {
+      "epoch": 0.3003176436615651,
+      "grad_norm": 0.1496269255876541,
+      "learning_rate": 8.207830160055688e-06,
+      "loss": 0.7634,
+      "step": 780
+    },
+    {
+      "epoch": 0.3041678698623544,
+      "grad_norm": 0.17681491374969482,
+      "learning_rate": 8.159749166566786e-06,
+      "loss": 0.7478,
+      "step": 790
+    },
+    {
+      "epoch": 0.30801809606314373,
+      "grad_norm": 0.13611136376857758,
+      "learning_rate": 8.11117671057377e-06,
+      "loss": 0.7693,
+      "step": 800
+    },
+    {
+      "epoch": 0.311868322263933,
+      "grad_norm": 0.13902996480464935,
+      "learning_rate": 8.062120346961404e-06,
+      "loss": 0.7527,
+      "step": 810
+    },
+    {
+      "epoch": 0.3157185484647223,
+      "grad_norm": 0.13181741535663605,
+      "learning_rate": 8.012587705880705e-06,
+      "loss": 0.756,
+      "step": 820
+    },
+    {
+      "epoch": 0.3195687746655116,
+      "grad_norm": 0.14604318141937256,
+      "learning_rate": 7.962586491562146e-06,
+      "loss": 0.7429,
+      "step": 830
+    },
+    {
+      "epoch": 0.3234190008663009,
+      "grad_norm": 0.14757570624351501,
+      "learning_rate": 7.912124481117357e-06,
+      "loss": 0.7447,
+      "step": 840
+    },
+    {
+      "epoch": 0.3272692270670902,
+      "grad_norm": 0.12778425216674805,
+      "learning_rate": 7.861209523329494e-06,
+      "loss": 0.7489,
+      "step": 850
+    },
+    {
+      "epoch": 0.3311194532678795,
+      "grad_norm": 0.14695270359516144,
+      "learning_rate": 7.809849537432432e-06,
+      "loss": 0.7379,
+      "step": 860
+    },
+    {
+      "epoch": 0.3349696794686688,
+      "grad_norm": 0.13082924485206604,
+      "learning_rate": 7.758052511879035e-06,
+      "loss": 0.7352,
+      "step": 870
+    },
+    {
+      "epoch": 0.3388199056694581,
+      "grad_norm": 0.12940150499343872,
+      "learning_rate": 7.705826503098642e-06,
+      "loss": 0.741,
+      "step": 880
+    },
+    {
+      "epoch": 0.34267013187024736,
+      "grad_norm": 0.1312270313501358,
+      "learning_rate": 7.653179634243979e-06,
+      "loss": 0.7357,
+      "step": 890
+    },
+    {
+      "epoch": 0.34652035807103665,
+      "grad_norm": 0.12723399698734283,
+      "learning_rate": 7.600120093927696e-06,
+      "loss": 0.7296,
+      "step": 900
+    },
+    {
+      "epoch": 0.350370584271826,
+      "grad_norm": 0.14529801905155182,
+      "learning_rate": 7.546656134948729e-06,
+      "loss": 0.7291,
+      "step": 910
+    },
+    {
+      "epoch": 0.3542208104726153,
+      "grad_norm": 0.1442037969827652,
+      "learning_rate": 7.492796073008661e-06,
+      "loss": 0.7236,
+      "step": 920
+    },
+    {
+      "epoch": 0.35807103667340456,
+      "grad_norm": 0.15021492540836334,
+      "learning_rate": 7.438548285418326e-06,
+      "loss": 0.7162,
+      "step": 930
+    },
+    {
+      "epoch": 0.36192126287419385,
+      "grad_norm": 0.14872850477695465,
+      "learning_rate": 7.3839212097948e-06,
+      "loss": 0.7545,
+      "step": 940
+    },
+    {
+      "epoch": 0.36577148907498314,
+      "grad_norm": 0.12330802530050278,
+      "learning_rate": 7.328923342749039e-06,
+      "loss": 0.7669,
+      "step": 950
+    },
+    {
+      "epoch": 0.3696217152757725,
+      "grad_norm": 0.13883543014526367,
+      "learning_rate": 7.273563238564319e-06,
+      "loss": 0.7429,
+      "step": 960
+    },
+    {
+      "epoch": 0.37347194147656176,
+      "grad_norm": 0.13309960067272186,
+      "learning_rate": 7.217849507865724e-06,
+      "loss": 0.7191,
+      "step": 970
+    },
+    {
+      "epoch": 0.37732216767735105,
+      "grad_norm": 0.1308899074792862,
+      "learning_rate": 7.161790816280858e-06,
+      "loss": 0.736,
+      "step": 980
+    },
+    {
+      "epoch": 0.38117239387814034,
+      "grad_norm": 0.14123553037643433,
+      "learning_rate": 7.1053958830920125e-06,
+      "loss": 0.7075,
+      "step": 990
+    },
+    {
+      "epoch": 0.3850226200789296,
+      "grad_norm": 0.12680956721305847,
+      "learning_rate": 7.0486734798799715e-06,
+      "loss": 0.7413,
+      "step": 1000
+    },
+    {
+      "epoch": 0.3888728462797189,
+      "grad_norm": 0.146232008934021,
+      "learning_rate": 6.991632429159708e-06,
+      "loss": 0.7625,
+      "step": 1010
+    },
+    {
+      "epoch": 0.39272307248050825,
+      "grad_norm": 0.13639914989471436,
+      "learning_rate": 6.934281603008132e-06,
+      "loss": 0.7333,
+      "step": 1020
+    },
+    {
+      "epoch": 0.39657329868129754,
+      "grad_norm": 0.1285993903875351,
+      "learning_rate": 6.876629921684143e-06,
+      "loss": 0.746,
+      "step": 1030
+    },
+    {
+      "epoch": 0.4004235248820868,
+      "grad_norm": 0.1335373967885971,
+      "learning_rate": 6.818686352241199e-06,
+      "loss": 0.7418,
+      "step": 1040
+    },
+    {
+      "epoch": 0.4042737510828761,
+      "grad_norm": 0.12475275993347168,
+      "learning_rate": 6.760459907132573e-06,
+      "loss": 0.7303,
+      "step": 1050
+    },
+    {
+      "epoch": 0.4081239772836654,
+      "grad_norm": 0.12791086733341217,
+      "learning_rate": 6.701959642809589e-06,
+      "loss": 0.733,
+      "step": 1060
+    },
+    {
+      "epoch": 0.41197420348445474,
+      "grad_norm": 0.13180045783519745,
+      "learning_rate": 6.643194658312987e-06,
+      "loss": 0.7173,
+      "step": 1070
+    },
+    {
+      "epoch": 0.415824429685244,
+      "grad_norm": 0.1269163340330124,
+      "learning_rate": 6.584174093857676e-06,
+      "loss": 0.7536,
+      "step": 1080
+    },
+    {
+      "epoch": 0.4196746558860333,
+      "grad_norm": 0.15189269185066223,
+      "learning_rate": 6.5249071294110735e-06,
+      "loss": 0.7256,
+      "step": 1090
+    },
+    {
+      "epoch": 0.4235248820868226,
+      "grad_norm": 0.14127787947654724,
+      "learning_rate": 6.465402983265274e-06,
+      "loss": 0.7441,
+      "step": 1100
+    },
+    {
+      "epoch": 0.4273751082876119,
+      "grad_norm": 0.13507601618766785,
+      "learning_rate": 6.405670910603243e-06,
+      "loss": 0.7399,
+      "step": 1110
+    },
+    {
+      "epoch": 0.43122533448840117,
+      "grad_norm": 0.14485979080200195,
+      "learning_rate": 6.345720202059291e-06,
+      "loss": 0.731,
+      "step": 1120
+    },
+    {
+      "epoch": 0.4350755606891905,
+      "grad_norm": 0.13242067396640778,
+      "learning_rate": 6.285560182274008e-06,
+      "loss": 0.7208,
+      "step": 1130
+    },
+    {
+      "epoch": 0.4389257868899798,
+      "grad_norm": 0.12672993540763855,
+      "learning_rate": 6.225200208443936e-06,
+      "loss": 0.728,
+      "step": 1140
+    },
+    {
+      "epoch": 0.4427760130907691,
+      "grad_norm": 0.12455835938453674,
+      "learning_rate": 6.16464966886616e-06,
+      "loss": 0.7208,
+      "step": 1150
+    },
+    {
+      "epoch": 0.44662623929155837,
+      "grad_norm": 0.15865303575992584,
+      "learning_rate": 6.103917981478065e-06,
+      "loss": 0.7522,
+      "step": 1160
+    },
+    {
+      "epoch": 0.45047646549234766,
+      "grad_norm": 0.13766150176525116,
+      "learning_rate": 6.043014592392489e-06,
+      "loss": 0.7551,
+      "step": 1170
+    },
+    {
+      "epoch": 0.454326691693137,
+      "grad_norm": 0.11921731382608414,
+      "learning_rate": 5.981948974428485e-06,
+      "loss": 0.731,
+      "step": 1180
+    },
+    {
+      "epoch": 0.4581769178939263,
+      "grad_norm": 0.1356174796819687,
+      "learning_rate": 5.920730625637934e-06,
+      "loss": 0.7078,
+      "step": 1190
+    },
+    {
+      "epoch": 0.46202714409471557,
+      "grad_norm": 0.13656339049339294,
+      "learning_rate": 5.8593690678282355e-06,
+      "loss": 0.7205,
+      "step": 1200
+    },
+    {
+      "epoch": 0.46587737029550486,
+      "grad_norm": 0.13170307874679565,
+      "learning_rate": 5.797873845081301e-06,
+      "loss": 0.7243,
+      "step": 1210
+    },
+    {
+      "epoch": 0.46972759649629414,
+      "grad_norm": 0.13415014743804932,
+      "learning_rate": 5.736254522269075e-06,
+      "loss": 0.7374,
+      "step": 1220
+    },
+    {
+      "epoch": 0.47357782269708343,
+      "grad_norm": 0.12880975008010864,
+      "learning_rate": 5.6745206835658386e-06,
+      "loss": 0.7429,
+      "step": 1230
+    },
+    {
+      "epoch": 0.47742804889787277,
+      "grad_norm": 0.12671129405498505,
+      "learning_rate": 5.612681930957495e-06,
+      "loss": 0.7205,
+      "step": 1240
+    },
+    {
+      "epoch": 0.48127827509866206,
+      "grad_norm": 0.12506477534770966,
+      "learning_rate": 5.550747882748095e-06,
+      "loss": 0.7133,
+      "step": 1250
+    },
+    {
+      "epoch": 0.48512850129945134,
+      "grad_norm": 0.13171371817588806,
+      "learning_rate": 5.4887281720638195e-06,
+      "loss": 0.7209,
+      "step": 1260
+    },
+    {
+      "epoch": 0.48897872750024063,
+      "grad_norm": 0.14120160043239594,
+      "learning_rate": 5.42663244535466e-06,
+      "loss": 0.7074,
+      "step": 1270
+    },
+    {
+      "epoch": 0.4928289537010299,
+      "grad_norm": 0.1324440985918045,
+      "learning_rate": 5.3644703608940205e-06,
+      "loss": 0.7128,
+      "step": 1280
+    },
+    {
+      "epoch": 0.49667917990181926,
+      "grad_norm": 0.13463130593299866,
+      "learning_rate": 5.302251587276491e-06,
+      "loss": 0.7346,
+      "step": 1290
+    },
+    {
+      "epoch": 0.5005294061026085,
+      "grad_norm": 0.14038540422916412,
+      "learning_rate": 5.2399858019140005e-06,
+      "loss": 0.7264,
+      "step": 1300
+    },
+    {
+      "epoch": 0.5043796323033978,
+      "grad_norm": 0.13130196928977966,
+      "learning_rate": 5.177682689530615e-06,
+      "loss": 0.7437,
+      "step": 1310
+    },
+    {
+      "epoch": 0.5082298585041871,
+      "grad_norm": 0.14394834637641907,
+      "learning_rate": 5.11535194065619e-06,
+      "loss": 0.7134,
+      "step": 1320
+    },
+    {
+      "epoch": 0.5120800847049765,
+      "grad_norm": 0.1504535973072052,
+      "learning_rate": 5.053003250119113e-06,
+      "loss": 0.7153,
+      "step": 1330
+    },
+    {
+      "epoch": 0.5159303109057657,
+      "grad_norm": 0.1342315971851349,
+      "learning_rate": 4.9906463155383925e-06,
+      "loss": 0.7276,
+      "step": 1340
+    },
+    {
+      "epoch": 0.519780537106555,
+      "grad_norm": 0.12967917323112488,
+      "learning_rate": 4.9282908358153046e-06,
+      "loss": 0.7212,
+      "step": 1350
+    },
+    {
+      "epoch": 0.5236307633073443,
+      "grad_norm": 0.14838355779647827,
+      "learning_rate": 4.8659465096248345e-06,
+      "loss": 0.7254,
+      "step": 1360
+    },
+    {
+      "epoch": 0.5274809895081336,
+      "grad_norm": 0.13574114441871643,
+      "learning_rate": 4.803623033907169e-06,
+      "loss": 0.7293,
+      "step": 1370
+    },
+    {
+      "epoch": 0.5313312157089229,
+      "grad_norm": 0.11819290369749069,
+      "learning_rate": 4.741330102359441e-06,
+      "loss": 0.7391,
+      "step": 1380
+    },
+    {
+      "epoch": 0.5351814419097122,
+      "grad_norm": 0.13712480664253235,
+      "learning_rate": 4.679077403927996e-06,
+      "loss": 0.7116,
+      "step": 1390
+    },
+    {
+      "epoch": 0.5390316681105015,
+      "grad_norm": 0.13123275339603424,
+      "learning_rate": 4.616874621301373e-06,
+      "loss": 0.7205,
+      "step": 1400
+    },
+    {
+      "epoch": 0.5428818943112907,
+      "grad_norm": 0.13591061532497406,
+      "learning_rate": 4.554731429404293e-06,
+      "loss": 0.7367,
+      "step": 1410
+    },
+    {
+      "epoch": 0.5467321205120801,
+      "grad_norm": 0.13752630352973938,
+      "learning_rate": 4.492657493892822e-06,
+      "loss": 0.7166,
+      "step": 1420
+    },
+    {
+      "epoch": 0.5505823467128694,
+      "grad_norm": 0.13902820646762848,
+      "learning_rate": 4.430662469650999e-06,
+      "loss": 0.7289,
+      "step": 1430
+    },
+    {
+      "epoch": 0.5544325729136587,
+      "grad_norm": 0.13922981917858124,
+      "learning_rate": 4.368755999289118e-06,
+      "loss": 0.7342,
+      "step": 1440
+    },
+    {
+      "epoch": 0.558282799114448,
+      "grad_norm": 0.14505617320537567,
+      "learning_rate": 4.306947711643949e-06,
+      "loss": 0.7155,
+      "step": 1450
+    },
+    {
+      "epoch": 0.5621330253152372,
+      "grad_norm": 0.1284371018409729,
+      "learning_rate": 4.245247220281068e-06,
+      "loss": 0.723,
+      "step": 1460
+    },
+    {
+      "epoch": 0.5659832515160266,
+      "grad_norm": 0.12245921045541763,
+      "learning_rate": 4.183664121999587e-06,
+      "loss": 0.7148,
+      "step": 1470
+    },
+    {
+      "epoch": 0.5698334777168159,
+      "grad_norm": 0.1300007551908493,
+      "learning_rate": 4.122207995339474e-06,
+      "loss": 0.7145,
+      "step": 1480
+    },
+    {
+      "epoch": 0.5736837039176051,
+      "grad_norm": 0.1370684653520584,
+      "learning_rate": 4.060888399091734e-06,
+      "loss": 0.7207,
+      "step": 1490
+    },
+    {
+      "epoch": 0.5775339301183945,
+      "grad_norm": 0.12822014093399048,
+      "learning_rate": 3.999714870811639e-06,
+      "loss": 0.7357,
+      "step": 1500
+    },
+    {
+      "epoch": 0.5813841563191837,
+      "grad_norm": 0.12656369805335999,
+      "learning_rate": 3.938696925335275e-06,
+      "loss": 0.7226,
+      "step": 1510
+    },
+    {
+      "epoch": 0.5852343825199731,
+      "grad_norm": 0.13717973232269287,
+      "learning_rate": 3.87784405329962e-06,
+      "loss": 0.7276,
+      "step": 1520
+    },
+    {
+      "epoch": 0.5890846087207623,
+      "grad_norm": 0.13012702763080597,
+      "learning_rate": 3.817165719666382e-06,
+      "loss": 0.7264,
+      "step": 1530
+    },
+    {
+      "epoch": 0.5929348349215516,
+      "grad_norm": 0.12633641064167023,
+      "learning_rate": 3.756671362249841e-06,
+      "loss": 0.739,
+      "step": 1540
+    },
+    {
+      "epoch": 0.596785061122341,
+      "grad_norm": 0.13143609464168549,
+      "learning_rate": 3.6963703902488907e-06,
+      "loss": 0.7272,
+      "step": 1550
+    },
+    {
+      "epoch": 0.6006352873231302,
+      "grad_norm": 0.13014492392539978,
+      "learning_rate": 3.6362721827835655e-06,
+      "loss": 0.7354,
+      "step": 1560
+    },
+    {
+      "epoch": 0.6044855135239195,
+      "grad_norm": 0.13805533945560455,
+      "learning_rate": 3.576386087436212e-06,
+      "loss": 0.7279,
+      "step": 1570
+    },
+    {
+      "epoch": 0.6083357397247088,
+      "grad_norm": 0.13548319041728973,
+      "learning_rate": 3.5167214187975913e-06,
+      "loss": 0.7313,
+      "step": 1580
+    },
+    {
+      "epoch": 0.6121859659254981,
+      "grad_norm": 0.11511724442243576,
+      "learning_rate": 3.457287457018086e-06,
+      "loss": 0.7266,
+      "step": 1590
+    },
+    {
+      "epoch": 0.6160361921262875,
+      "grad_norm": 0.12748992443084717,
+      "learning_rate": 3.3980934463643003e-06,
+      "loss": 0.7132,
+      "step": 1600
+    },
+    {
+      "epoch": 0.6198864183270767,
+      "grad_norm": 0.11877070367336273,
+      "learning_rate": 3.3391485937812053e-06,
+      "loss": 0.6918,
+      "step": 1610
+    },
+    {
+      "epoch": 0.623736644527866,
+      "grad_norm": 0.13848890364170074,
+      "learning_rate": 3.280462067460115e-06,
+      "loss": 0.6961,
+      "step": 1620
+    },
+    {
+      "epoch": 0.6275868707286553,
+      "grad_norm": 0.13019347190856934,
+      "learning_rate": 3.222042995412669e-06,
+      "loss": 0.7338,
+      "step": 1630
+    },
+    {
+      "epoch": 0.6314370969294446,
+      "grad_norm": 0.11594462394714355,
+      "learning_rate": 3.163900464051089e-06,
+      "loss": 0.7137,
+      "step": 1640
+    },
+    {
+      "epoch": 0.635287323130234,
+      "grad_norm": 0.13018640875816345,
+      "learning_rate": 3.1060435167748826e-06,
+      "loss": 0.725,
+      "step": 1650
+    },
+    {
+      "epoch": 0.6391375493310232,
+      "grad_norm": 0.1298709362745285,
+      "learning_rate": 3.048481152564253e-06,
+      "loss": 0.7137,
+      "step": 1660
+    },
+    {
+      "epoch": 0.6429877755318125,
+      "grad_norm": 0.14070317149162292,
+      "learning_rate": 2.9912223245804017e-06,
+      "loss": 0.7375,
+      "step": 1670
+    },
+    {
+      "epoch": 0.6468380017326018,
+      "grad_norm": 0.1320551186800003,
+      "learning_rate": 2.934275938772981e-06,
+      "loss": 0.7158,
+      "step": 1680
+    },
+    {
+      "epoch": 0.6506882279333911,
+      "grad_norm": 0.13418708741664886,
+      "learning_rate": 2.8776508524948676e-06,
+      "loss": 0.7058,
+      "step": 1690
+    },
+    {
+      "epoch": 0.6545384541341804,
+      "grad_norm": 0.12640799582004547,
+      "learning_rate": 2.821355873124503e-06,
+      "loss": 0.7041,
+      "step": 1700
+    },
+    {
+      "epoch": 0.6583886803349697,
+      "grad_norm": 0.1371382176876068,
+      "learning_rate": 2.7653997566960144e-06,
+      "loss": 0.7263,
+      "step": 1710
+    },
+    {
+      "epoch": 0.662238906535759,
+      "grad_norm": 0.13352133333683014,
+      "learning_rate": 2.709791206537309e-06,
+      "loss": 0.7154,
+      "step": 1720
+    },
+    {
+      "epoch": 0.6660891327365482,
+      "grad_norm": 0.1287153661251068,
+      "learning_rate": 2.6545388719163747e-06,
+      "loss": 0.7122,
+      "step": 1730
+    },
+    {
+      "epoch": 0.6699393589373376,
+      "grad_norm": 0.12523669004440308,
+      "learning_rate": 2.599651346695979e-06,
+      "loss": 0.7234,
+      "step": 1740
+    },
+    {
+      "epoch": 0.6737895851381268,
+      "grad_norm": 0.13551196455955505,
+      "learning_rate": 2.545137167997007e-06,
+      "loss": 0.718,
+      "step": 1750
+    },
+    {
+      "epoch": 0.6776398113389162,
+      "grad_norm": 0.12166937440633774,
+      "learning_rate": 2.4910048148705972e-06,
+      "loss": 0.7022,
+      "step": 1760
+    },
+    {
+      "epoch": 0.6814900375397055,
+      "grad_norm": 0.16463300585746765,
+      "learning_rate": 2.4372627069793327e-06,
+      "loss": 0.7462,
+      "step": 1770
+    },
+    {
+      "epoch": 0.6853402637404947,
+      "grad_norm": 0.13410532474517822,
+      "learning_rate": 2.3839192032876486e-06,
+      "loss": 0.7293,
+      "step": 1780
+    },
+    {
+      "epoch": 0.6891904899412841,
+      "grad_norm": 0.12571538984775543,
+      "learning_rate": 2.3309826007617047e-06,
+      "loss": 0.7403,
+      "step": 1790
+    },
+    {
+      "epoch": 0.6930407161420733,
+      "grad_norm": 0.1275581270456314,
+      "learning_rate": 2.278461133078881e-06,
+      "loss": 0.7051,
+      "step": 1800
+    },
+    {
+      "epoch": 0.6968909423428626,
+      "grad_norm": 0.12681931257247925,
+      "learning_rate": 2.2263629693471258e-06,
+      "loss": 0.6997,
+      "step": 1810
+    },
+    {
+      "epoch": 0.700741168543652,
+      "grad_norm": 0.12576428055763245,
+      "learning_rate": 2.1746962128343414e-06,
+      "loss": 0.7225,
+      "step": 1820
+    },
+    {
+      "epoch": 0.7045913947444412,
+      "grad_norm": 0.12487654387950897,
+      "learning_rate": 2.1234688997080183e-06,
+      "loss": 0.7217,
+      "step": 1830
+    },
+    {
+      "epoch": 0.7084416209452306,
+      "grad_norm": 0.1396723985671997,
+      "learning_rate": 2.0726889977853053e-06,
+      "loss": 0.7174,
+      "step": 1840
+    },
+    {
+      "epoch": 0.7122918471460198,
+      "grad_norm": 0.13787128031253815,
+      "learning_rate": 2.022364405293703e-06,
+      "loss": 0.7308,
+      "step": 1850
+    },
+    {
+      "epoch": 0.7161420733468091,
+      "grad_norm": 0.1288633942604065,
+      "learning_rate": 1.9725029496425814e-06,
+      "loss": 0.7249,
+      "step": 1860
+    },
+    {
+      "epoch": 0.7199922995475985,
+      "grad_norm": 0.13202349841594696,
+      "learning_rate": 1.923112386205727e-06,
+      "loss": 0.7216,
+      "step": 1870
+    },
+    {
+      "epoch": 0.7238425257483877,
+      "grad_norm": 0.1295361965894699,
+      "learning_rate": 1.874200397115079e-06,
+      "loss": 0.7275,
+      "step": 1880
+    },
+    {
+      "epoch": 0.727692751949177,
+      "grad_norm": 0.13354672491550446,
+      "learning_rate": 1.825774590065857e-06,
+      "loss": 0.7135,
+      "step": 1890
+    },
+    {
+      "epoch": 0.7315429781499663,
+      "grad_norm": 0.13512201607227325,
+      "learning_rate": 1.7778424971332813e-06,
+      "loss": 0.7205,
+      "step": 1900
+    },
+    {
+      "epoch": 0.7353932043507556,
+      "grad_norm": 0.12618939578533173,
+      "learning_rate": 1.7304115736010407e-06,
+      "loss": 0.7206,
+      "step": 1910
+    },
+    {
+      "epoch": 0.739243430551545,
+      "grad_norm": 0.12947958707809448,
+      "learning_rate": 1.6834891968017108e-06,
+      "loss": 0.7137,
+      "step": 1920
+    },
+    {
+      "epoch": 0.7430936567523342,
+      "grad_norm": 0.13044729828834534,
+      "learning_rate": 1.6370826649692884e-06,
+      "loss": 0.7236,
+      "step": 1930
+    },
+    {
+      "epoch": 0.7469438829531235,
+      "grad_norm": 0.1351284682750702,
+      "learning_rate": 1.5911991961040463e-06,
+      "loss": 0.7388,
+      "step": 1940
+    },
+    {
+      "epoch": 0.7507941091539128,
+      "grad_norm": 0.13178005814552307,
+      "learning_rate": 1.545845926849851e-06,
+      "loss": 0.725,
+      "step": 1950
+    },
+    {
+      "epoch": 0.7546443353547021,
+      "grad_norm": 0.14060914516448975,
+      "learning_rate": 1.5010299113841397e-06,
+      "loss": 0.74,
+      "step": 1960
+    },
+    {
+      "epoch": 0.7584945615554913,
+      "grad_norm": 0.12631021440029144,
+      "learning_rate": 1.4567581203207238e-06,
+      "loss": 0.7369,
+      "step": 1970
+    },
+    {
+      "epoch": 0.7623447877562807,
+      "grad_norm": 0.13136033713817596,
+      "learning_rate": 1.413037439625593e-06,
+      "loss": 0.7183,
+      "step": 1980
+    },
+    {
+      "epoch": 0.76619501395707,
+      "grad_norm": 0.14082825183868408,
+      "learning_rate": 1.3698746695458847e-06,
+      "loss": 0.7225,
+      "step": 1990
+    },
+    {
+      "epoch": 0.7700452401578592,
+      "grad_norm": 0.1294088214635849,
+      "learning_rate": 1.3272765235521745e-06,
+      "loss": 0.7224,
+      "step": 2000
+    },
+    {
+      "epoch": 0.7738954663586486,
+      "grad_norm": 0.1250632405281067,
+      "learning_rate": 1.2852496272942877e-06,
+      "loss": 0.7217,
+      "step": 2010
+    },
+    {
+      "epoch": 0.7777456925594378,
+      "grad_norm": 0.13226376473903656,
+      "learning_rate": 1.2438005175707396e-06,
+      "loss": 0.7054,
+      "step": 2020
+    },
+    {
+      "epoch": 0.7815959187602272,
+      "grad_norm": 0.13443224132061005,
+      "learning_rate": 1.2029356413120297e-06,
+      "loss": 0.6928,
+      "step": 2030
+    },
+    {
+      "epoch": 0.7854461449610165,
+      "grad_norm": 0.13270144164562225,
+      "learning_rate": 1.1626613545778786e-06,
+      "loss": 0.7107,
+      "step": 2040
+    },
+    {
+      "epoch": 0.7892963711618057,
+      "grad_norm": 0.12672492861747742,
+      "learning_rate": 1.1229839215686294e-06,
+      "loss": 0.7189,
+      "step": 2050
+    },
+    {
+      "epoch": 0.7931465973625951,
+      "grad_norm": 0.12657304108142853,
+      "learning_rate": 1.0839095136509188e-06,
+      "loss": 0.6984,
+      "step": 2060
+    },
+    {
+      "epoch": 0.7969968235633843,
+      "grad_norm": 0.14053994417190552,
+      "learning_rate": 1.045444208397791e-06,
+      "loss": 0.7095,
+      "step": 2070
+    },
+    {
+      "epoch": 0.8008470497641736,
+      "grad_norm": 0.12990248203277588,
+      "learning_rate": 1.007593988643402e-06,
+      "loss": 0.6982,
+      "step": 2080
+    },
+    {
+      "epoch": 0.804697275964963,
+      "grad_norm": 0.12582260370254517,
+      "learning_rate": 9.703647415524613e-07,
+      "loss": 0.7222,
+      "step": 2090
+    },
+    {
+      "epoch": 0.8085475021657522,
+      "grad_norm": 0.12614530324935913,
+      "learning_rate": 9.337622577045485e-07,
+      "loss": 0.7149,
+      "step": 2100
+    },
+    {
+      "epoch": 0.8123977283665416,
+      "grad_norm": 0.1269351840019226,
+      "learning_rate": 8.977922301934588e-07,
+      "loss": 0.7308,
+      "step": 2110
+    },
+    {
+      "epoch": 0.8162479545673308,
+      "grad_norm": 0.1311035007238388,
+      "learning_rate": 8.624602537417009e-07,
+      "loss": 0.7287,
+      "step": 2120
+    },
+    {
+      "epoch": 0.8200981807681201,
+      "grad_norm": 0.13031479716300964,
+      "learning_rate": 8.277718238303123e-07,
+      "loss": 0.7101,
+      "step": 2130
+    },
+    {
+      "epoch": 0.8239484069689095,
+      "grad_norm": 0.12159845978021622,
+      "learning_rate": 7.937323358440935e-07,
+      "loss": 0.7225,
+      "step": 2140
+    },
+    {
+      "epoch": 0.8277986331696987,
+      "grad_norm": 0.1481645107269287,
+      "learning_rate": 7.603470842324223e-07,
+      "loss": 0.7048,
+      "step": 2150
+    },
+    {
+      "epoch": 0.831648859370488,
+      "grad_norm": 0.11979886144399643,
+      "learning_rate": 7.276212616857575e-07,
+      "loss": 0.723,
+      "step": 2160
+    },
+    {
+      "epoch": 0.8354990855712773,
+      "grad_norm": 0.12683701515197754,
+      "learning_rate": 6.955599583279837e-07,
+      "loss": 0.7243,
+      "step": 2170
+    },
+    {
+      "epoch": 0.8393493117720666,
+      "grad_norm": 0.12848322093486786,
+      "learning_rate": 6.641681609246981e-07,
+      "loss": 0.7161,
+      "step": 2180
+    },
+    {
+      "epoch": 0.8431995379728559,
+      "grad_norm": 0.1325085461139679,
+      "learning_rate": 6.334507521075711e-07,
+      "loss": 0.7235,
+      "step": 2190
+    },
+    {
+      "epoch": 0.8470497641736452,
+      "grad_norm": 0.13858214020729065,
+      "learning_rate": 6.034125096149179e-07,
+      "loss": 0.7109,
+      "step": 2200
+    },
+    {
+      "epoch": 0.8508999903744345,
+      "grad_norm": 0.150949165225029,
+      "learning_rate": 5.740581055485722e-07,
+      "loss": 0.6982,
+      "step": 2210
+    },
+    {
+      "epoch": 0.8547502165752238,
+      "grad_norm": 0.12205518037080765,
+      "learning_rate": 5.453921056471961e-07,
+      "loss": 0.7205,
+      "step": 2220
+    },
+    {
+      "epoch": 0.8586004427760131,
+      "grad_norm": 0.1375947743654251,
+      "learning_rate": 5.174189685761288e-07,
+      "loss": 0.7132,
+      "step": 2230
+    },
+    {
+      "epoch": 0.8624506689768023,
+      "grad_norm": 0.12930737435817719,
+      "learning_rate": 4.901430452338979e-07,
+      "loss": 0.7336,
+      "step": 2240
+    },
+    {
+      "epoch": 0.8663008951775917,
+      "grad_norm": 0.13644804060459137,
+      "learning_rate": 4.6356857807548286e-07,
+      "loss": 0.7118,
+      "step": 2250
+    },
+    {
+      "epoch": 0.870151121378381,
+      "grad_norm": 0.13539598882198334,
+      "learning_rate": 4.376997004524541e-07,
+      "loss": 0.7007,
+      "step": 2260
+    },
+    {
+      "epoch": 0.8740013475791703,
+      "grad_norm": 0.1274230033159256,
+      "learning_rate": 4.12540435970073e-07,
+      "loss": 0.7014,
+      "step": 2270
+    },
+    {
+      "epoch": 0.8778515737799596,
+      "grad_norm": 0.12634588778018951,
+      "learning_rate": 3.880946978614703e-07,
+      "loss": 0.7261,
+      "step": 2280
+    },
+    {
+      "epoch": 0.8817017999807488,
+      "grad_norm": 0.12383083254098892,
+      "learning_rate": 3.643662883789878e-07,
+      "loss": 0.7211,
+      "step": 2290
+    },
+    {
+      "epoch": 0.8855520261815382,
+      "grad_norm": 0.12589849531650543,
+      "learning_rate": 3.41358898202781e-07,
+      "loss": 0.7058,
+      "step": 2300
+    },
+    {
+      "epoch": 0.8894022523823275,
+      "grad_norm": 0.12301251292228699,
+      "learning_rate": 3.190761058667741e-07,
+      "loss": 0.7082,
+      "step": 2310
+    },
+    {
+      "epoch": 0.8932524785831167,
+      "grad_norm": 0.12378346920013428,
+      "learning_rate": 2.9752137720206463e-07,
+      "loss": 0.7161,
+      "step": 2320
+    },
+    {
+      "epoch": 0.8971027047839061,
+      "grad_norm": 0.12856724858283997,
+      "learning_rate": 2.7669806479785067e-07,
+      "loss": 0.7268,
+      "step": 2330
+    },
+    {
+      "epoch": 0.9009529309846953,
+      "grad_norm": 0.13129015266895294,
+      "learning_rate": 2.566094074799763e-07,
+      "loss": 0.7168,
+      "step": 2340
+    },
+    {
+      "epoch": 0.9048031571854847,
+      "grad_norm": 0.12490590661764145,
+      "learning_rate": 2.3725852980716557e-07,
+      "loss": 0.7047,
+      "step": 2350
+    },
+    {
+      "epoch": 0.908653383386274,
+      "grad_norm": 0.1259477585554123,
+      "learning_rate": 2.1864844158503983e-07,
+      "loss": 0.7178,
+      "step": 2360
+    },
+    {
+      "epoch": 0.9125036095870632,
+      "grad_norm": 0.14469479024410248,
+      "learning_rate": 2.007820373979702e-07,
+      "loss": 0.7059,
+      "step": 2370
+    },
+    {
+      "epoch": 0.9163538357878526,
+      "grad_norm": 0.1277933269739151,
+      "learning_rate": 1.8366209615886532e-07,
+      "loss": 0.7105,
+      "step": 2380
+    },
+    {
+      "epoch": 0.9202040619886418,
+      "grad_norm": 0.13781572878360748,
+      "learning_rate": 1.6729128067693613e-07,
+      "loss": 0.7333,
+      "step": 2390
+    },
+    {
+      "epoch": 0.9240542881894311,
+      "grad_norm": 0.12524011731147766,
+      "learning_rate": 1.5167213724353426e-07,
+      "loss": 0.7218,
+      "step": 2400
+    },
+    {
+      "epoch": 0.9279045143902205,
+      "grad_norm": 0.12497400492429733,
+      "learning_rate": 1.3680709523610058e-07,
+      "loss": 0.7068,
+      "step": 2410
+    },
+    {
+      "epoch": 0.9317547405910097,
+      "grad_norm": 0.12136722356081009,
+      "learning_rate": 1.2269846674030418e-07,
+      "loss": 0.7312,
+      "step": 2420
+    },
+    {
+      "epoch": 0.935604966791799,
+      "grad_norm": 0.13629084825515747,
+      "learning_rate": 1.0934844619042595e-07,
+      "loss": 0.7287,
+      "step": 2430
+    },
+    {
+      "epoch": 0.9394551929925883,
+      "grad_norm": 0.131565123796463,
+      "learning_rate": 9.675911002803961e-08,
+      "loss": 0.7127,
+      "step": 2440
+    },
+    {
+      "epoch": 0.9433054191933776,
+      "grad_norm": 0.13772745430469513,
+      "learning_rate": 8.493241637904381e-08,
+      "loss": 0.7098,
+      "step": 2450
+    },
+    {
+      "epoch": 0.9471556453941669,
+      "grad_norm": 0.14211246371269226,
+      "learning_rate": 7.387020474909689e-08,
+      "loss": 0.7343,
+      "step": 2460
+    },
+    {
+      "epoch": 0.9510058715949562,
+      "grad_norm": 0.1308647245168686,
+      "learning_rate": 6.357419573750856e-08,
+      "loss": 0.7191,
+      "step": 2470
+    },
+    {
+      "epoch": 0.9548560977957455,
+      "grad_norm": 0.12989413738250732,
+      "learning_rate": 5.4045990769613365e-08,
+      "loss": 0.7138,
+      "step": 2480
+    },
+    {
+      "epoch": 0.9587063239965348,
+      "grad_norm": 0.14245612919330597,
+      "learning_rate": 4.5287071847694434e-08,
+      "loss": 0.712,
+      "step": 2490
+    },
+    {
+      "epoch": 0.9625565501973241,
+      "grad_norm": 0.13266925513744354,
+      "learning_rate": 3.7298801320467817e-08,
+      "loss": 0.712,
+      "step": 2500
+    },
+    {
+      "epoch": 0.9664067763981133,
+      "grad_norm": 0.12658116221427917,
+      "learning_rate": 3.0082421671192576e-08,
+      "loss": 0.7048,
+      "step": 2510
+    },
+    {
+      "epoch": 0.9702570025989027,
+      "grad_norm": 0.12447308003902435,
+      "learning_rate": 2.3639055324410355e-08,
+      "loss": 0.7244,
+      "step": 2520
+    },
+    {
+      "epoch": 0.974107228799692,
+      "grad_norm": 0.12757615745067596,
+      "learning_rate": 1.7969704471369477e-08,
+      "loss": 0.7088,
+      "step": 2530
+    },
+    {
+      "epoch": 0.9779574550004813,
+      "grad_norm": 0.12181824445724487,
+      "learning_rate": 1.3075250914142413e-08,
+      "loss": 0.7161,
+      "step": 2540
+    },
+    {
+      "epoch": 0.9818076812012706,
+      "grad_norm": 0.127007395029068,
+      "learning_rate": 8.956455928476049e-09,
+      "loss": 0.7267,
+      "step": 2550
+    },
+    {
+      "epoch": 0.9856579074020598,
+      "grad_norm": 0.11733075231313705,
+      "learning_rate": 5.6139601453786275e-09,
+      "loss": 0.703,
+      "step": 2560
+    },
+    {
+      "epoch": 0.9895081336028492,
+      "grad_norm": 0.13906611502170563,
+      "learning_rate": 3.0482834514805602e-09,
+      "loss": 0.7078,
+      "step": 2570
+    },
+    {
+      "epoch": 0.9933583598036385,
+      "grad_norm": 0.11542243510484695,
+      "learning_rate": 1.2598249081713366e-09,
+      "loss": 0.692,
+      "step": 2580
+    },
+    {
+      "epoch": 0.9972085860044277,
+      "grad_norm": 0.13147349655628204,
+      "learning_rate": 2.488626895302826e-10,
+      "loss": 0.7266,
+      "step": 2590
+    },
+    {
+      "epoch": 0.9999037443449803,
+      "step": 2597,
+      "total_flos": 3.676650883732472e+18,
+      "train_loss": 0.7665747314588263,
+      "train_runtime": 18154.8782,
+      "train_samples_per_second": 36.624,
+      "train_steps_per_second": 0.143
+    }
+  ],
+  "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": 3.676650883732472e+18,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}

continuous/qwen3/siglip2_sm16_512/finetune/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0c133f09d0e3d924c3efafb92186ea4346f18fa578ca4a2cb6a07f1cc54fb6c1
+size 6136

continuous/qwen3/siglip2_sm16_512/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": 1152,
+  "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-SO400M-16-SigLIP2-512",
+  "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/siglip2/siglip2_sm16_512.npz",
+  "vocab_size": 151936
+}

continuous/qwen3/siglip2_sm16_512/pretrain/mm_projector.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c419e402bb66b4d059dc211fb6c435cbfb08597bb12af6da367cb83484c1f164
+size 26232888

continuous/qwen3/siglip2_sm16_512/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": 0.9551379084587097,
+      "learning_rate": 0.00032142857142857147,
+      "loss": 9.1586,
+      "step": 10
+    },
+    {
+      "epoch": 0.03669724770642202,
+      "grad_norm": 0.1993943601846695,
+      "learning_rate": 0.0006785714285714287,
+      "loss": 5.5743,
+      "step": 20
+    },
+    {
+      "epoch": 0.05504587155963303,
+      "grad_norm": 0.035142261534929276,
+      "learning_rate": 0.0009999907688176173,
+      "loss": 4.8999,
+      "step": 30
+    },
+    {
+      "epoch": 0.07339449541284404,
+      "grad_norm": 0.01988411694765091,
+      "learning_rate": 0.0009988834393115766,
+      "loss": 4.7569,
+      "step": 40
+    },
+    {
+      "epoch": 0.09174311926605505,
+      "grad_norm": 0.02365502342581749,
+      "learning_rate": 0.0009959345572940772,
+      "loss": 4.6966,
+      "step": 50
+    },
+    {
+      "epoch": 0.11009174311926606,
+      "grad_norm": 0.01133514754474163,
+      "learning_rate": 0.0009911550081155983,
+      "loss": 4.6654,
+      "step": 60
+    },
+    {
+      "epoch": 0.12844036697247707,
+      "grad_norm": 0.044102948158979416,
+      "learning_rate": 0.0009845624347566433,
+      "loss": 4.628,
+      "step": 70
+    },
+    {
+      "epoch": 0.14678899082568808,
+      "grad_norm": 0.05455225333571434,
+      "learning_rate": 0.0009761811727013547,
+      "loss": 4.6153,
+      "step": 80
+    },
+    {
+      "epoch": 0.1651376146788991,
+      "grad_norm": 0.02840876206755638,
+      "learning_rate": 0.0009660421601067666,
+      "loss": 4.5832,
+      "step": 90
+    },
+    {
+      "epoch": 0.1834862385321101,
+      "grad_norm": 0.03132019564509392,
+      "learning_rate": 0.0009541828235992926,
+      "loss": 4.5657,
+      "step": 100
+    },
+    {
+      "epoch": 0.2018348623853211,
+      "grad_norm": 0.029212117195129395,
+      "learning_rate": 0.0009406469401200151,
+      "loss": 4.5397,
+      "step": 110
+    },
+    {
+      "epoch": 0.22018348623853212,
+      "grad_norm": 0.035299867391586304,
+      "learning_rate": 0.0009254844753287492,
+      "loss": 4.5186,
+      "step": 120
+    },
+    {
+      "epoch": 0.23853211009174313,
+      "grad_norm": 0.03019171766936779,
+      "learning_rate": 0.0009087513991633924,
+      "loss": 4.4844,
+      "step": 130
+    },
+    {
+      "epoch": 0.25688073394495414,
+      "grad_norm": 0.036737799644470215,
+      "learning_rate": 0.0008905094792353917,
+      "loss": 4.4532,
+      "step": 140
+    },
+    {
+      "epoch": 0.27522935779816515,
+      "grad_norm": 0.03188127279281616,
+      "learning_rate": 0.0008708260528239789,
+      "loss": 4.4351,
+      "step": 150
+    },
+    {
+      "epoch": 0.29357798165137616,
+      "grad_norm": 0.031223541125655174,
+      "learning_rate": 0.0008497737783108237,
+      "loss": 4.4052,
+      "step": 160
+    },
+    {
+      "epoch": 0.3119266055045872,
+      "grad_norm": 0.029355280101299286,
+      "learning_rate": 0.0008274303669726426,
+      "loss": 4.3784,
+      "step": 170
+    },
+    {
+      "epoch": 0.3302752293577982,
+      "grad_norm": 0.041189733892679214,
+      "learning_rate": 0.0008038782961218137,
+      "loss": 4.3547,
+      "step": 180
+    },
+    {
+      "epoch": 0.3486238532110092,
+      "grad_norm": 0.037285756319761276,
+      "learning_rate": 0.000779204504653894,
+      "loss": 4.342,
+      "step": 190
+    },
+    {
+      "epoch": 0.3669724770642202,
+      "grad_norm": 0.044834934175014496,
+      "learning_rate": 0.0007535000721258791,
+      "loss": 4.3149,
+      "step": 200
+    },
+    {
+      "epoch": 0.3853211009174312,
+      "grad_norm": 0.04416954889893532,
+      "learning_rate": 0.00072685988254984,
+      "loss": 4.3011,
+      "step": 210
+    },
+    {
+      "epoch": 0.4036697247706422,
+      "grad_norm": 0.03624406456947327,
+      "learning_rate": 0.0006993822741429906,
+      "loss": 4.3071,
+      "step": 220
+    },
+    {
+      "epoch": 0.42201834862385323,
+      "grad_norm": 0.04260483384132385,
+      "learning_rate": 0.0006711686763270818,
+      "loss": 4.2989,
+      "step": 230
+    },
+    {
+      "epoch": 0.44036697247706424,
+      "grad_norm": 0.03445016220211983,
+      "learning_rate": 0.0006423232353170831,
+      "loss": 4.2833,
+      "step": 240
+    },
+    {
+      "epoch": 0.45871559633027525,
+      "grad_norm": 0.029704533517360687,
+      "learning_rate": 0.0006129524296812335,
+      "loss": 4.2615,
+      "step": 250
+    },
+    {
+      "epoch": 0.47706422018348627,
+      "grad_norm": 0.03121739812195301,
+      "learning_rate": 0.0005831646772915651,
+      "loss": 4.2681,
+      "step": 260
+    },
+    {
+      "epoch": 0.4954128440366973,
+      "grad_norm": 0.028113972395658493,
+      "learning_rate": 0.0005530699351157809,
+      "loss": 4.243,
+      "step": 270
+    },
+    {
+      "epoch": 0.5137614678899083,
+      "grad_norm": 0.03399241715669632,
+      "learning_rate": 0.0005227792933277943,
+      "loss": 4.2368,
+      "step": 280
+    },
+    {
+      "epoch": 0.5321100917431193,
+      "grad_norm": 0.030555685982108116,
+      "learning_rate": 0.0004924045652352092,
+      "loss": 4.2188,
+      "step": 290
+    },
+    {
+      "epoch": 0.5504587155963303,
+      "grad_norm": 0.03584282472729683,
+      "learning_rate": 0.0004620578745374633,
+      "loss": 4.2175,
+      "step": 300
+    },
+    {
+      "epoch": 0.5688073394495413,
+      "grad_norm": 0.03778780624270439,
+      "learning_rate": 0.0004318512414382059,
+      "loss": 4.2305,
+      "step": 310
+    },
+    {
+      "epoch": 0.5871559633027523,
+      "grad_norm": 0.030973229557275772,
+      "learning_rate": 0.0004018961691397148,
+      "loss": 4.2186,
+      "step": 320
+    },
+    {
+      "epoch": 0.6055045871559633,
+      "grad_norm": 0.028175104409456253,
+      "learning_rate": 0.0003723032322457458,
+      "loss": 4.1997,
+      "step": 330
+    },
+    {
+      "epoch": 0.6238532110091743,
+      "grad_norm": 0.025974594056606293,
+      "learning_rate": 0.0003431816685921625,
+      "loss": 4.1997,
+      "step": 340
+    },
+    {
+      "epoch": 0.6422018348623854,
+      "grad_norm": 0.029336268082261086,
+      "learning_rate": 0.0003146389760120416,
+      "loss": 4.2006,
+      "step": 350
+    },
+    {
+      "epoch": 0.6605504587155964,
+      "grad_norm": 0.02818327583372593,
+      "learning_rate": 0.00028678051552373485,
+      "loss": 4.2053,
+      "step": 360
+    },
+    {
+      "epoch": 0.6788990825688074,
+      "grad_norm": 0.02655731700360775,
+      "learning_rate": 0.0002597091224066581,
+      "loss": 4.1878,
+      "step": 370
+    },
+    {
+      "epoch": 0.6972477064220184,
+      "grad_norm": 0.027622001245617867,
+      "learning_rate": 0.0002335247266004629,
+      "loss": 4.2019,
+      "step": 380
+    },
+    {
+      "epoch": 0.7155963302752294,
+      "grad_norm": 0.04036848992109299,
+      "learning_rate": 0.000208323983828828,
+      "loss": 4.1828,
+      "step": 390
+    },
+    {
+      "epoch": 0.7339449541284404,
+      "grad_norm": 0.03148654103279114,
+      "learning_rate": 0.0001841999188095224,
+      "loss": 4.1861,
+      "step": 400
+    },
+    {
+      "epoch": 0.7522935779816514,
+      "grad_norm": 0.024086318910121918,
+      "learning_rate": 0.00016124158186777676,
+      "loss": 4.1831,
+      "step": 410
+    },
+    {
+      "epoch": 0.7706422018348624,
+      "grad_norm": 0.025344667956233025,
+      "learning_rate": 0.0001395337202205211,
+      "loss": 4.1732,
+      "step": 420
+    },
+    {
+      "epoch": 0.7889908256880734,
+      "grad_norm": 0.0258070919662714,
+      "learning_rate": 0.00011915646514489292,
+      "loss": 4.1777,
+      "step": 430
+    },
+    {
+      "epoch": 0.8073394495412844,
+      "grad_norm": 0.024478459730744362,
+      "learning_rate": 0.00010018503618578818,
+      "loss": 4.1847,
+      "step": 440
+    },
+    {
+      "epoch": 0.8256880733944955,
+      "grad_norm": 0.02600502222776413,
+      "learning_rate": 8.268946349432583e-05,
+      "loss": 4.1743,
+      "step": 450
+    },
+    {
+      "epoch": 0.8440366972477065,
+      "grad_norm": 0.021571330726146698,
+      "learning_rate": 6.673432932217117e-05,
+      "loss": 4.1656,
+      "step": 460
+    },
+    {
+      "epoch": 0.8623853211009175,
+      "grad_norm": 0.023993484675884247,
+      "learning_rate": 5.2378529625954684e-05,
+      "loss": 4.1549,
+      "step": 470
+    },
+    {
+      "epoch": 0.8807339449541285,
+      "grad_norm": 0.022441335022449493,
+      "learning_rate": 3.9675056661785556e-05,
+      "loss": 4.1648,
+      "step": 480
+    },
+    {
+      "epoch": 0.8990825688073395,
+      "grad_norm": 0.021649949252605438,
+      "learning_rate": 2.8670803372379538e-05,
+      "loss": 4.1656,
+      "step": 490
+    },
+    {
+      "epoch": 0.9174311926605505,
+      "grad_norm": 0.02038056217133999,
+      "learning_rate": 1.9406390288876586e-05,
+      "loss": 4.1796,
+      "step": 500
+    },
+    {
+      "epoch": 0.9357798165137615,
+      "grad_norm": 0.02016083337366581,
+      "learning_rate": 1.191601558631461e-05,
+      "loss": 4.1671,
+      "step": 510
+    },
+    {
+      "epoch": 0.9541284403669725,
+      "grad_norm": 0.021832549944519997,
+      "learning_rate": 6.2273288462559305e-06,
+      "loss": 4.157,
+      "step": 520
+    },
+    {
+      "epoch": 0.9724770642201835,
+      "grad_norm": 0.020971130579710007,
+      "learning_rate": 2.361328992552314e-06,
+      "loss": 4.1608,
+      "step": 530
+    },
+    {
+      "epoch": 0.9908256880733946,
+      "grad_norm": 0.02151712030172348,
+      "learning_rate": 3.322867770044202e-07,
+      "loss": 4.1669,
+      "step": 540
+    },
+    {
+      "epoch": 1.0,
+      "step": 545,
+      "total_flos": 3.232495471553413e+17,
+      "train_loss": 4.430919118758736,
+      "train_runtime": 8253.9424,
+      "train_samples_per_second": 67.62,
+      "train_steps_per_second": 0.066
+    }
+  ],
+  "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": 3.232495471553413e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}

continuous/qwen3/webssl_dino1b_full2b_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/webssl_dino1b_full2b_224/finetune/chat_template.jinja

@@ -0,0 +1,89 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# 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>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\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" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- 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>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if message.content is string %}
+        {%- set content = message.content %}
+    {%- else %}
+        {%- set content = '' %}
+    {%- endif %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}

continuous/qwen3/webssl_dino1b_full2b_224/finetune/config.json

@@ -0,0 +1,89 @@
+{
+  "add_faster_video": false,
+  "add_time_instruction": false,
+  "architectures": [
+    "LlavaQwen3ForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "dtype": "bfloat16",
+  "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,
+  "layer_types": [
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention"
+  ],
+  "max_position_embeddings": 40960,
+  "max_window_layers": 28,
+  "mm_hidden_size": 1536,
+  "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": -1,
+  "mm_vision_tower": "hf:/cache/ckpt/download/tokenizer/continuous/webssl/webssl-dino1b-full2b-224",
+  "mm_vision_tower_lr": null,
+  "model_type": "qwen3",
+  "num_attention_heads": 16,
+  "num_hidden_layers": 28,
+  "num_key_value_heads": 8,
+  "pad_token_id": 151643,
+  "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",
+  "transformers_version": "4.57.6",
+  "use_cache": true,
+  "use_mm_proj": true,
+  "use_pos_skipping": false,
+  "use_sliding_window": false,
+  "vision_tower_pretrained": null,
+  "vocab_size": 151936
+}

continuous/qwen3/webssl_dino1b_full2b_224/finetune/generation_config.json

@@ -0,0 +1,12 @@
+{
+  "do_sample": true,
+  "eos_token_id": [
+    151645,
+    151643
+  ],
+  "pad_token_id": 151643,
+  "temperature": 0.6,
+  "top_k": 20,
+  "top_p": 0.95,
+  "transformers_version": "4.57.6"
+}

continuous/qwen3/webssl_dino1b_full2b_224/finetune/model-00001-of-00002.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a6c1738e361df31f38e9f98032f55bf49025d3555f3b060f0b07a78b0b17fb8e
+size 4992945184

continuous/qwen3/webssl_dino1b_full2b_224/finetune/model-00002-of-00002.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e08389ad557cdc771d7f73ec710308fd3ab4940a4729d0f622382ec034f1c5e1
+size 1369596176

continuous/qwen3/webssl_dino1b_full2b_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/webssl_dino1b_full2b_224/finetune/tokenizer.json

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

continuous/qwen3/webssl_dino1b_full2b_224/finetune/tokenizer_config.json

@@ -0,0 +1,240 @@
+{
+  "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,
+  "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/webssl_dino1b_full2b_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": 16.841554641723633,
+      "learning_rate": 1.153846153846154e-06,
+      "loss": 2.2927,
+      "step": 10
+    },
+    {
+      "epoch": 0.007700452401578592,
+      "grad_norm": 6.223072052001953,
+      "learning_rate": 2.435897435897436e-06,
+      "loss": 1.8552,
+      "step": 20
+    },
+    {
+      "epoch": 0.01155067860236789,
+      "grad_norm": 4.436450958251953,
+      "learning_rate": 3.7179487179487184e-06,
+      "loss": 1.479,
+      "step": 30
+    },
+    {
+      "epoch": 0.015400904803157185,
+      "grad_norm": 0.9747268557548523,
+      "learning_rate": 5e-06,
+      "loss": 1.1437,
+      "step": 40
+    },
+    {
+      "epoch": 0.019251131003946483,
+      "grad_norm": 0.7561534643173218,
+      "learning_rate": 6.282051282051282e-06,
+      "loss": 1.0444,
+      "step": 50
+    },
+    {
+      "epoch": 0.02310135720473578,
+      "grad_norm": 0.37353721261024475,
+      "learning_rate": 7.564102564102564e-06,
+      "loss": 0.9614,
+      "step": 60
+    },
+    {
+      "epoch": 0.026951583405525073,
+      "grad_norm": 0.30472084879875183,
+      "learning_rate": 8.846153846153847e-06,
+      "loss": 0.9167,
+      "step": 70
+    },
+    {
+      "epoch": 0.03080180960631437,
+      "grad_norm": 0.30347684025764465,
+      "learning_rate": 9.999996111488723e-06,
+      "loss": 0.9143,
+      "step": 80
+    },
+    {
+      "epoch": 0.034652035807103666,
+      "grad_norm": 0.2868176996707916,
+      "learning_rate": 9.999529497453782e-06,
+      "loss": 0.883,
+      "step": 90
+    },
+    {
+      "epoch": 0.038502262007892966,
+      "grad_norm": 0.25255417823791504,
+      "learning_rate": 9.998285264324193e-06,
+      "loss": 0.866,
+      "step": 100
+    },
+    {
+      "epoch": 0.04235248820868226,
+      "grad_norm": 0.26879966259002686,
+      "learning_rate": 9.996263605626052e-06,
+      "loss": 0.8475,
+      "step": 110
+    },
+    {
+      "epoch": 0.04620271440947156,
+      "grad_norm": 0.2480972409248352,
+      "learning_rate": 9.99346483580503e-06,
+      "loss": 0.8577,
+      "step": 120
+    },
+    {
+      "epoch": 0.05005294061026085,
+      "grad_norm": 0.24682271480560303,
+      "learning_rate": 9.989889390177467e-06,
+      "loss": 0.8544,
+      "step": 130
+    },
+    {
+      "epoch": 0.053903166811050146,
+      "grad_norm": 0.2768690288066864,
+      "learning_rate": 9.985537824862647e-06,
+      "loss": 0.8486,
+      "step": 140
+    },
+    {
+      "epoch": 0.057753393011839446,
+      "grad_norm": 0.29373252391815186,
+      "learning_rate": 9.980410816696317e-06,
+      "loss": 0.8346,
+      "step": 150
+    },
+    {
+      "epoch": 0.06160361921262874,
+      "grad_norm": 0.2772243320941925,
+      "learning_rate": 9.97450916312541e-06,
+      "loss": 0.8395,
+      "step": 160
+    },
+    {
+      "epoch": 0.06545384541341803,
+      "grad_norm": 0.22518379986286163,
+      "learning_rate": 9.967833782084004e-06,
+      "loss": 0.8404,
+      "step": 170
+    },
+    {
+      "epoch": 0.06930407161420733,
+      "grad_norm": 0.2662390172481537,
+      "learning_rate": 9.960385711850554e-06,
+      "loss": 0.8687,
+      "step": 180
+    },
+    {
+      "epoch": 0.07315429781499663,
+      "grad_norm": 0.26747775077819824,
+      "learning_rate": 9.952166110886394e-06,
+      "loss": 0.8571,
+      "step": 190
+    },
+    {
+      "epoch": 0.07700452401578593,
+      "grad_norm": 0.2915347218513489,
+      "learning_rate": 9.943176257655567e-06,
+      "loss": 0.8469,
+      "step": 200
+    },
+    {
+      "epoch": 0.08085475021657522,
+      "grad_norm": 0.2507801651954651,
+      "learning_rate": 9.933417550425952e-06,
+      "loss": 0.841,
+      "step": 210
+    },
+    {
+      "epoch": 0.08470497641736452,
+      "grad_norm": 0.24764472246170044,
+      "learning_rate": 9.9228915070518e-06,
+      "loss": 0.8284,
+      "step": 220
+    },
+    {
+      "epoch": 0.08855520261815382,
+      "grad_norm": 0.2597410976886749,
+      "learning_rate": 9.911599764737632e-06,
+      "loss": 0.8297,
+      "step": 230
+    },
+    {
+      "epoch": 0.09240542881894312,
+      "grad_norm": 0.2779732644557953,
+      "learning_rate": 9.899544079783606e-06,
+      "loss": 0.8259,
+      "step": 240
+    },
+    {
+      "epoch": 0.0962556550197324,
+      "grad_norm": 0.2884956896305084,
+      "learning_rate": 9.886726327312338e-06,
+      "loss": 0.8403,
+      "step": 250
+    },
+    {
+      "epoch": 0.1001058812205217,
+      "grad_norm": 0.24819380044937134,
+      "learning_rate": 9.87314850097724e-06,
+      "loss": 0.8378,
+      "step": 260
+    },
+    {
+      "epoch": 0.103956107421311,
+      "grad_norm": 0.24854539334774017,
+      "learning_rate": 9.858812712652452e-06,
+      "loss": 0.8263,
+      "step": 270
+    },
+    {
+      "epoch": 0.10780633362210029,
+      "grad_norm": 0.24063651263713837,
+      "learning_rate": 9.843721192104341e-06,
+      "loss": 0.8259,
+      "step": 280
+    },
+    {
+      "epoch": 0.11165655982288959,
+      "grad_norm": 0.24321305751800537,
+      "learning_rate": 9.827876286644698e-06,
+      "loss": 0.844,
+      "step": 290
+    },
+    {
+      "epoch": 0.11550678602367889,
+      "grad_norm": 0.22961346805095673,
+      "learning_rate": 9.811280460765643e-06,
+      "loss": 0.8037,
+      "step": 300
+    },
+    {
+      "epoch": 0.11935701222446819,
+      "grad_norm": 0.23439735174179077,
+      "learning_rate": 9.793936295756292e-06,
+      "loss": 0.8181,
+      "step": 310
+    },
+    {
+      "epoch": 0.12320723842525748,
+      "grad_norm": 0.26188716292381287,
+      "learning_rate": 9.775846489301279e-06,
+      "loss": 0.8174,
+      "step": 320
+    },
+    {
+      "epoch": 0.12705746462604678,
+      "grad_norm": 0.25659945607185364,
+      "learning_rate": 9.757013855061157e-06,
+      "loss": 0.8169,
+      "step": 330
+    },
+    {
+      "epoch": 0.13090769082683607,
+      "grad_norm": 0.2614240050315857,
+      "learning_rate": 9.737441322234758e-06,
+      "loss": 0.8301,
+      "step": 340
+    },
+    {
+      "epoch": 0.13475791702762538,
+      "grad_norm": 0.28240230679512024,
+      "learning_rate": 9.717131935103605e-06,
+      "loss": 0.8284,
+      "step": 350
+    },
+    {
+      "epoch": 0.13860814322841467,
+      "grad_norm": 0.2474726289510727,
+      "learning_rate": 9.696088852558398e-06,
+      "loss": 0.8158,
+      "step": 360
+    },
+    {
+      "epoch": 0.14245836942920398,
+      "grad_norm": 0.24855849146842957,
+      "learning_rate": 9.674315347607686e-06,
+      "loss": 0.8413,
+      "step": 370
+    },
+    {
+      "epoch": 0.14630859562999327,
+      "grad_norm": 0.2547649145126343,
+      "learning_rate": 9.651814806868796e-06,
+      "loss": 0.801,
+      "step": 380
+    },
+    {
+      "epoch": 0.15015882183078255,
+      "grad_norm": 0.2500818967819214,
+      "learning_rate": 9.62859073004107e-06,
+      "loss": 0.8097,
+      "step": 390
+    },
+    {
+      "epoch": 0.15400904803157187,
+      "grad_norm": 0.2496584802865982,
+      "learning_rate": 9.604646729361547e-06,
+      "loss": 0.813,
+      "step": 400
+    },
+    {
+      "epoch": 0.15785927423236115,
+      "grad_norm": 0.2505379617214203,
+      "learning_rate": 9.579986529043099e-06,
+      "loss": 0.8014,
+      "step": 410
+    },
+    {
+      "epoch": 0.16170950043315044,
+      "grad_norm": 0.24915994703769684,
+      "learning_rate": 9.554613964695189e-06,
+      "loss": 0.8194,
+      "step": 420
+    },
+    {
+      "epoch": 0.16555972663393975,
+      "grad_norm": 0.25371381640434265,
+      "learning_rate": 9.528532982727281e-06,
+      "loss": 0.8073,
+      "step": 430
+    },
+    {
+      "epoch": 0.16940995283472904,
+      "grad_norm": 0.2461254894733429,
+      "learning_rate": 9.501747639735018e-06,
+      "loss": 0.8056,
+      "step": 440
+    },
+    {
+      "epoch": 0.17326017903551832,
+      "grad_norm": 0.254752516746521,
+      "learning_rate": 9.474262101869272e-06,
+      "loss": 0.7955,
+      "step": 450
+    },
+    {
+      "epoch": 0.17711040523630764,
+      "grad_norm": 0.24336902797222137,
+      "learning_rate": 9.446080644188141e-06,
+      "loss": 0.7999,
+      "step": 460
+    },
+    {
+      "epoch": 0.18096063143709692,
+      "grad_norm": 0.257129430770874,
+      "learning_rate": 9.417207649992017e-06,
+      "loss": 0.8091,
+      "step": 470
+    },
+    {
+      "epoch": 0.18481085763788624,
+      "grad_norm": 0.24667391180992126,
+      "learning_rate": 9.387647610141816e-06,
+      "loss": 0.8068,
+      "step": 480
+    },
+    {
+      "epoch": 0.18866108383867553,
+      "grad_norm": 0.23147781193256378,
+      "learning_rate": 9.35740512236047e-06,
+      "loss": 0.7926,
+      "step": 490
+    },
+    {
+      "epoch": 0.1925113100394648,
+      "grad_norm": 0.237773135304451,
+      "learning_rate": 9.326484890517809e-06,
+      "loss": 0.7979,
+      "step": 500
+    },
+    {
+      "epoch": 0.19636153624025413,
+      "grad_norm": 0.23258322477340698,
+      "learning_rate": 9.294891723898926e-06,
+      "loss": 0.7941,
+      "step": 510
+    },
+    {
+      "epoch": 0.2002117624410434,
+      "grad_norm": 0.25370046496391296,
+      "learning_rate": 9.262630536456153e-06,
+      "loss": 0.8228,
+      "step": 520
+    },
+    {
+      "epoch": 0.2040619886418327,
+      "grad_norm": 0.2534472942352295,
+      "learning_rate": 9.229706346044749e-06,
+      "loss": 0.795,
+      "step": 530
+    },
+    {
+      "epoch": 0.207912214842622,
+      "grad_norm": 0.2547314763069153,
+      "learning_rate": 9.196124273642422e-06,
+      "loss": 0.809,
+      "step": 540
+    },
+    {
+      "epoch": 0.2117624410434113,
+      "grad_norm": 0.22826513648033142,
+      "learning_rate": 9.161889542552842e-06,
+      "loss": 0.8096,
+      "step": 550
+    },
+    {
+      "epoch": 0.21561266724420058,
+      "grad_norm": 0.22348780930042267,
+      "learning_rate": 9.127007477593193e-06,
+      "loss": 0.816,
+      "step": 560
+    },
+    {
+      "epoch": 0.2194628934449899,
+      "grad_norm": 0.26571881771087646,
+      "learning_rate": 9.091483504265974e-06,
+      "loss": 0.8237,
+      "step": 570
+    },
+    {
+      "epoch": 0.22331311964577918,
+      "grad_norm": 0.25904521346092224,
+      "learning_rate": 9.055323147915116e-06,
+      "loss": 0.8043,
+      "step": 580
+    },
+    {
+      "epoch": 0.2271633458465685,
+      "grad_norm": 0.2496531754732132,
+      "learning_rate": 9.018532032866585e-06,
+      "loss": 0.8134,
+      "step": 590
+    },
+    {
+      "epoch": 0.23101357204735778,
+      "grad_norm": 0.23123985528945923,
+      "learning_rate": 8.981115881553579e-06,
+      "loss": 0.8038,
+      "step": 600
+    },
+    {
+      "epoch": 0.23486379824814707,
+      "grad_norm": 0.27567970752716064,
+      "learning_rate": 8.943080513626476e-06,
+      "loss": 0.8248,
+      "step": 610
+    },
+    {
+      "epoch": 0.23871402444893638,
+      "grad_norm": 0.24130094051361084,
+      "learning_rate": 8.904431845047641e-06,
+      "loss": 0.8099,
+      "step": 620
+    },
+    {
+      "epoch": 0.24256425064972567,
+      "grad_norm": 0.24264828860759735,
+      "learning_rate": 8.86517588717128e-06,
+      "loss": 0.825,
+      "step": 630
+    },
+    {
+      "epoch": 0.24641447685051496,
+      "grad_norm": 0.2315923422574997,
+      "learning_rate": 8.82531874580844e-06,
+      "loss": 0.8147,
+      "step": 640
+    },
+    {
+      "epoch": 0.25026470305130427,
+      "grad_norm": 0.23517341911792755,
+      "learning_rate": 8.784866620277314e-06,
+      "loss": 0.772,
+      "step": 650
+    },
+    {
+      "epoch": 0.25411492925209356,
+      "grad_norm": 0.25642725825309753,
+      "learning_rate": 8.743825802439018e-06,
+      "loss": 0.8089,
+      "step": 660
+    },
+    {
+      "epoch": 0.25796515545288284,
+      "grad_norm": 0.2576068639755249,
+      "learning_rate": 8.702202675718956e-06,
+      "loss": 0.8145,
+      "step": 670
+    },
+    {
+      "epoch": 0.26181538165367213,
+      "grad_norm": 0.23717550933361053,
+      "learning_rate": 8.660003714113959e-06,
+      "loss": 0.7823,
+      "step": 680
+    },
+    {
+      "epoch": 0.26566560785446147,
+      "grad_norm": 0.23880265653133392,
+      "learning_rate": 8.617235481185323e-06,
+      "loss": 0.8087,
+      "step": 690
+    },
+    {
+      "epoch": 0.26951583405525076,
+      "grad_norm": 0.24606555700302124,
+      "learning_rate": 8.573904629037931e-06,
+      "loss": 0.8136,
+      "step": 700
+    },
+    {
+      "epoch": 0.27336606025604004,
+      "grad_norm": 0.24231910705566406,
+      "learning_rate": 8.530017897285585e-06,
+      "loss": 0.7945,
+      "step": 710
+    },
+    {
+      "epoch": 0.27721628645682933,
+      "grad_norm": 0.23077157139778137,
+      "learning_rate": 8.485582112002746e-06,
+      "loss": 0.7973,
+      "step": 720
+    },
+    {
+      "epoch": 0.2810665126576186,
+      "grad_norm": 0.2406284660100937,
+      "learning_rate": 8.440604184662811e-06,
+      "loss": 0.804,
+      "step": 730
+    },
+    {
+      "epoch": 0.28491673885840796,
+      "grad_norm": 0.22809401154518127,
+      "learning_rate": 8.395091111063118e-06,
+      "loss": 0.7918,
+      "step": 740
+    },
+    {
+      "epoch": 0.28876696505919724,
+      "grad_norm": 0.22327549755573273,
+      "learning_rate": 8.349049970236822e-06,
+      "loss": 0.7914,
+      "step": 750
+    },
+    {
+      "epoch": 0.29261719125998653,
+      "grad_norm": 0.24296042323112488,
+      "learning_rate": 8.302487923351842e-06,
+      "loss": 0.7844,
+      "step": 760
+    },
+    {
+      "epoch": 0.2964674174607758,
+      "grad_norm": 0.22919802367687225,
+      "learning_rate": 8.255412212597021e-06,
+      "loss": 0.8006,
+      "step": 770
+    },
+    {
+      "epoch": 0.3003176436615651,
+      "grad_norm": 0.23725295066833496,
+      "learning_rate": 8.207830160055688e-06,
+      "loss": 0.7978,
+      "step": 780
+    },
+    {
+      "epoch": 0.3041678698623544,
+      "grad_norm": 0.25556209683418274,
+      "learning_rate": 8.159749166566786e-06,
+      "loss": 0.7824,
+      "step": 790
+    },
+    {
+      "epoch": 0.30801809606314373,
+      "grad_norm": 0.2343255579471588,
+      "learning_rate": 8.11117671057377e-06,
+      "loss": 0.8097,
+      "step": 800
+    },
+    {
+      "epoch": 0.311868322263933,
+      "grad_norm": 0.23988030850887299,
+      "learning_rate": 8.062120346961404e-06,
+      "loss": 0.8026,
+      "step": 810
+    },
+    {
+      "epoch": 0.3157185484647223,
+      "grad_norm": 0.2356143444776535,
+      "learning_rate": 8.012587705880705e-06,
+      "loss": 0.7945,
+      "step": 820
+    },
+    {
+      "epoch": 0.3195687746655116,
+      "grad_norm": 0.2357935756444931,
+      "learning_rate": 7.962586491562146e-06,
+      "loss": 0.7812,
+      "step": 830
+    },
+    {
+      "epoch": 0.3234190008663009,
+      "grad_norm": 0.24228931963443756,
+      "learning_rate": 7.912124481117357e-06,
+      "loss": 0.7879,
+      "step": 840
+    },
+    {
+      "epoch": 0.3272692270670902,
+      "grad_norm": 0.23523560166358948,
+      "learning_rate": 7.861209523329494e-06,
+      "loss": 0.7985,
+      "step": 850
+    },
+    {
+      "epoch": 0.3311194532678795,
+      "grad_norm": 0.24248653650283813,
+      "learning_rate": 7.809849537432432e-06,
+      "loss": 0.7811,
+      "step": 860
+    },
+    {
+      "epoch": 0.3349696794686688,
+      "grad_norm": 0.24546179175376892,
+      "learning_rate": 7.758052511879035e-06,
+      "loss": 0.7919,
+      "step": 870
+    },
+    {
+      "epoch": 0.3388199056694581,
+      "grad_norm": 0.21898093819618225,
+      "learning_rate": 7.705826503098642e-06,
+      "loss": 0.7873,
+      "step": 880
+    },
+    {
+      "epoch": 0.34267013187024736,
+      "grad_norm": 0.23412802815437317,
+      "learning_rate": 7.653179634243979e-06,
+      "loss": 0.7866,
+      "step": 890
+    },
+    {
+      "epoch": 0.34652035807103665,
+      "grad_norm": 0.21919184923171997,
+      "learning_rate": 7.600120093927696e-06,
+      "loss": 0.7924,
+      "step": 900
+    },
+    {
+      "epoch": 0.350370584271826,
+      "grad_norm": 0.23922601342201233,
+      "learning_rate": 7.546656134948729e-06,
+      "loss": 0.7828,
+      "step": 910
+    },
+    {
+      "epoch": 0.3542208104726153,
+      "grad_norm": 0.24462978541851044,
+      "learning_rate": 7.492796073008661e-06,
+      "loss": 0.7811,
+      "step": 920
+    },
+    {
+      "epoch": 0.35807103667340456,
+      "grad_norm": 0.26148614287376404,
+      "learning_rate": 7.438548285418326e-06,
+      "loss": 0.7664,
+      "step": 930
+    },
+    {
+      "epoch": 0.36192126287419385,
+      "grad_norm": 0.2655225694179535,
+      "learning_rate": 7.3839212097948e-06,
+      "loss": 0.8009,
+      "step": 940
+    },
+    {
+      "epoch": 0.36577148907498314,
+      "grad_norm": 0.23052294552326202,
+      "learning_rate": 7.328923342749039e-06,
+      "loss": 0.8148,
+      "step": 950
+    },
+    {
+      "epoch": 0.3696217152757725,
+      "grad_norm": 0.25780078768730164,
+      "learning_rate": 7.273563238564319e-06,
+      "loss": 0.7988,
+      "step": 960
+    },
+    {
+      "epoch": 0.37347194147656176,
+      "grad_norm": 0.2423023283481598,
+      "learning_rate": 7.217849507865724e-06,
+      "loss": 0.7792,
+      "step": 970
+    },
+    {
+      "epoch": 0.37732216767735105,
+      "grad_norm": 0.23222769796848297,
+      "learning_rate": 7.161790816280858e-06,
+      "loss": 0.7857,
+      "step": 980
+    },
+    {
+      "epoch": 0.38117239387814034,
+      "grad_norm": 0.24056680500507355,
+      "learning_rate": 7.1053958830920125e-06,
+      "loss": 0.772,
+      "step": 990
+    },
+    {
+      "epoch": 0.3850226200789296,
+      "grad_norm": 0.24268104135990143,
+      "learning_rate": 7.0486734798799715e-06,
+      "loss": 0.7876,
+      "step": 1000
+    },
+    {
+      "epoch": 0.3888728462797189,
+      "grad_norm": 0.24655593931674957,
+      "learning_rate": 6.991632429159708e-06,
+      "loss": 0.8113,
+      "step": 1010
+    },
+    {
+      "epoch": 0.39272307248050825,
+      "grad_norm": 0.2544262409210205,
+      "learning_rate": 6.934281603008132e-06,
+      "loss": 0.8022,
+      "step": 1020
+    },
+    {
+      "epoch": 0.39657329868129754,
+      "grad_norm": 0.24109460413455963,
+      "learning_rate": 6.876629921684143e-06,
+      "loss": 0.7985,
+      "step": 1030
+    },
+    {
+      "epoch": 0.4004235248820868,
+      "grad_norm": 0.24664868414402008,
+      "learning_rate": 6.818686352241199e-06,
+      "loss": 0.7894,
+      "step": 1040
+    },
+    {
+      "epoch": 0.4042737510828761,
+      "grad_norm": 0.22157110273838043,
+      "learning_rate": 6.760459907132573e-06,
+      "loss": 0.7847,
+      "step": 1050
+    },
+    {
+      "epoch": 0.4081239772836654,
+      "grad_norm": 0.22760920226573944,
+      "learning_rate": 6.701959642809589e-06,
+      "loss": 0.7924,
+      "step": 1060
+    },
+    {
+      "epoch": 0.41197420348445474,
+      "grad_norm": 0.24312156438827515,
+      "learning_rate": 6.643194658312987e-06,
+      "loss": 0.7857,
+      "step": 1070
+    },
+    {
+      "epoch": 0.415824429685244,
+      "grad_norm": 0.22838635742664337,
+      "learning_rate": 6.584174093857676e-06,
+      "loss": 0.8092,
+      "step": 1080
+    },
+    {
+      "epoch": 0.4196746558860333,
+      "grad_norm": 0.27704912424087524,
+      "learning_rate": 6.5249071294110735e-06,
+      "loss": 0.7885,
+      "step": 1090
+    },
+    {
+      "epoch": 0.4235248820868226,
+      "grad_norm": 0.25486913323402405,
+      "learning_rate": 6.465402983265274e-06,
+      "loss": 0.8026,
+      "step": 1100
+    },
+    {
+      "epoch": 0.4273751082876119,
+      "grad_norm": 0.24362187087535858,
+      "learning_rate": 6.405670910603243e-06,
+      "loss": 0.7986,
+      "step": 1110
+    },
+    {
+      "epoch": 0.43122533448840117,
+      "grad_norm": 0.2596502900123596,
+      "learning_rate": 6.345720202059291e-06,
+      "loss": 0.7951,
+      "step": 1120
+    },
+    {
+      "epoch": 0.4350755606891905,
+      "grad_norm": 0.24332164227962494,
+      "learning_rate": 6.285560182274008e-06,
+      "loss": 0.7738,
+      "step": 1130
+    },
+    {
+      "epoch": 0.4389257868899798,
+      "grad_norm": 0.2406086027622223,
+      "learning_rate": 6.225200208443936e-06,
+      "loss": 0.7972,
+      "step": 1140
+    },
+    {
+      "epoch": 0.4427760130907691,
+      "grad_norm": 0.22681929171085358,
+      "learning_rate": 6.16464966886616e-06,
+      "loss": 0.7854,
+      "step": 1150
+    },
+    {
+      "epoch": 0.44662623929155837,
+      "grad_norm": 0.27006813883781433,
+      "learning_rate": 6.103917981478065e-06,
+      "loss": 0.8077,
+      "step": 1160
+    },
+    {
+      "epoch": 0.45047646549234766,
+      "grad_norm": 0.23780298233032227,
+      "learning_rate": 6.043014592392489e-06,
+      "loss": 0.8204,
+      "step": 1170
+    },
+    {
+      "epoch": 0.454326691693137,
+      "grad_norm": 0.22992080450057983,
+      "learning_rate": 5.981948974428485e-06,
+      "loss": 0.7951,
+      "step": 1180
+    },
+    {
+      "epoch": 0.4581769178939263,
+      "grad_norm": 0.25620514154434204,
+      "learning_rate": 5.920730625637934e-06,
+      "loss": 0.7753,
+      "step": 1190
+    },
+    {
+      "epoch": 0.46202714409471557,
+      "grad_norm": 0.24296902120113373,
+      "learning_rate": 5.8593690678282355e-06,
+      "loss": 0.7847,
+      "step": 1200
+    },
+    {
+      "epoch": 0.46587737029550486,
+      "grad_norm": 0.2554122805595398,
+      "learning_rate": 5.797873845081301e-06,
+      "loss": 0.7888,
+      "step": 1210
+    },
+    {
+      "epoch": 0.46972759649629414,
+      "grad_norm": 0.23443573713302612,
+      "learning_rate": 5.736254522269075e-06,
+      "loss": 0.8053,
+      "step": 1220
+    },
+    {
+      "epoch": 0.47357782269708343,
+      "grad_norm": 0.2356691211462021,
+      "learning_rate": 5.6745206835658386e-06,
+      "loss": 0.8073,
+      "step": 1230
+    },
+    {
+      "epoch": 0.47742804889787277,
+      "grad_norm": 0.23487070202827454,
+      "learning_rate": 5.612681930957495e-06,
+      "loss": 0.7787,
+      "step": 1240
+    },
+    {
+      "epoch": 0.48127827509866206,
+      "grad_norm": 0.23245781660079956,
+      "learning_rate": 5.550747882748095e-06,
+      "loss": 0.7772,
+      "step": 1250
+    },
+    {
+      "epoch": 0.48512850129945134,
+      "grad_norm": 0.2567727267742157,
+      "learning_rate": 5.4887281720638195e-06,
+      "loss": 0.7809,
+      "step": 1260
+    },
+    {
+      "epoch": 0.48897872750024063,
+      "grad_norm": 0.2667885720729828,
+      "learning_rate": 5.42663244535466e-06,
+      "loss": 0.7692,
+      "step": 1270
+    },
+    {
+      "epoch": 0.4928289537010299,
+      "grad_norm": 0.25751590728759766,
+      "learning_rate": 5.3644703608940205e-06,
+      "loss": 0.7762,
+      "step": 1280
+    },
+    {
+      "epoch": 0.49667917990181926,
+      "grad_norm": 0.23840709030628204,
+      "learning_rate": 5.302251587276491e-06,
+      "loss": 0.7866,
+      "step": 1290
+    },
+    {
+      "epoch": 0.5005294061026085,
+      "grad_norm": 0.22708842158317566,
+      "learning_rate": 5.2399858019140005e-06,
+      "loss": 0.7874,
+      "step": 1300
+    },
+    {
+      "epoch": 0.5043796323033978,
+      "grad_norm": 0.22886091470718384,
+      "learning_rate": 5.177682689530615e-06,
+      "loss": 0.8021,
+      "step": 1310
+    },
+    {
+      "epoch": 0.5082298585041871,
+      "grad_norm": 0.24759408831596375,
+      "learning_rate": 5.11535194065619e-06,
+      "loss": 0.7706,
+      "step": 1320
+    },
+    {
+      "epoch": 0.5120800847049765,
+      "grad_norm": 0.26894548535346985,
+      "learning_rate": 5.053003250119113e-06,
+      "loss": 0.7813,
+      "step": 1330
+    },
+    {
+      "epoch": 0.5159303109057657,
+      "grad_norm": 0.23908446729183197,
+      "learning_rate": 4.9906463155383925e-06,
+      "loss": 0.787,
+      "step": 1340
+    },
+    {
+      "epoch": 0.519780537106555,
+      "grad_norm": 0.2348807007074356,
+      "learning_rate": 4.9282908358153046e-06,
+      "loss": 0.778,
+      "step": 1350
+    },
+    {
+      "epoch": 0.5236307633073443,
+      "grad_norm": 0.25548985600471497,
+      "learning_rate": 4.8659465096248345e-06,
+      "loss": 0.7974,
+      "step": 1360
+    },
+    {
+      "epoch": 0.5274809895081336,
+      "grad_norm": 0.24771852791309357,
+      "learning_rate": 4.803623033907169e-06,
+      "loss": 0.7973,
+      "step": 1370
+    },
+    {
+      "epoch": 0.5313312157089229,
+      "grad_norm": 0.22747862339019775,
+      "learning_rate": 4.741330102359441e-06,
+      "loss": 0.8022,
+      "step": 1380
+    },
+    {
+      "epoch": 0.5351814419097122,
+      "grad_norm": 0.2522660791873932,
+      "learning_rate": 4.679077403927996e-06,
+      "loss": 0.7821,
+      "step": 1390
+    },
+    {
+      "epoch": 0.5390316681105015,
+      "grad_norm": 0.24375304579734802,
+      "learning_rate": 4.616874621301373e-06,
+      "loss": 0.7805,
+      "step": 1400
+    },
+    {
+      "epoch": 0.5428818943112907,
+      "grad_norm": 0.24226823449134827,
+      "learning_rate": 4.554731429404293e-06,
+      "loss": 0.7948,
+      "step": 1410
+    },
+    {
+      "epoch": 0.5467321205120801,
+      "grad_norm": 0.25609147548675537,
+      "learning_rate": 4.492657493892822e-06,
+      "loss": 0.7809,
+      "step": 1420
+    },
+    {
+      "epoch": 0.5505823467128694,
+      "grad_norm": 0.28300920128822327,
+      "learning_rate": 4.430662469650999e-06,
+      "loss": 0.7961,
+      "step": 1430
+    },
+    {
+      "epoch": 0.5544325729136587,
+      "grad_norm": 0.25916585326194763,
+      "learning_rate": 4.368755999289118e-06,
+      "loss": 0.7985,
+      "step": 1440
+    },
+    {
+      "epoch": 0.558282799114448,
+      "grad_norm": 0.2610478103160858,
+      "learning_rate": 4.306947711643949e-06,
+      "loss": 0.7852,
+      "step": 1450
+    },
+    {
+      "epoch": 0.5621330253152372,
+      "grad_norm": 0.2483023852109909,
+      "learning_rate": 4.245247220281068e-06,
+      "loss": 0.7857,
+      "step": 1460
+    },
+    {
+      "epoch": 0.5659832515160266,
+      "grad_norm": 0.23252356052398682,
+      "learning_rate": 4.183664121999587e-06,
+      "loss": 0.7846,
+      "step": 1470
+    },
+    {
+      "epoch": 0.5698334777168159,
+      "grad_norm": 0.23354333639144897,
+      "learning_rate": 4.122207995339474e-06,
+      "loss": 0.7806,
+      "step": 1480
+    },
+    {
+      "epoch": 0.5736837039176051,
+      "grad_norm": 0.23356527090072632,
+      "learning_rate": 4.060888399091734e-06,
+      "loss": 0.7857,
+      "step": 1490
+    },
+    {
+      "epoch": 0.5775339301183945,
+      "grad_norm": 0.2344336062669754,
+      "learning_rate": 3.999714870811639e-06,
+      "loss": 0.7916,
+      "step": 1500
+    },
+    {
+      "epoch": 0.5813841563191837,
+      "grad_norm": 0.22988799214363098,
+      "learning_rate": 3.938696925335275e-06,
+      "loss": 0.788,
+      "step": 1510
+    },
+    {
+      "epoch": 0.5852343825199731,
+      "grad_norm": 0.24438662827014923,
+      "learning_rate": 3.87784405329962e-06,
+      "loss": 0.7946,
+      "step": 1520
+    },
+    {
+      "epoch": 0.5890846087207623,
+      "grad_norm": 0.23676881194114685,
+      "learning_rate": 3.817165719666382e-06,
+      "loss": 0.7975,
+      "step": 1530
+    },
+    {
+      "epoch": 0.5929348349215516,
+      "grad_norm": 0.23832564055919647,
+      "learning_rate": 3.756671362249841e-06,
+      "loss": 0.7997,
+      "step": 1540
+    },
+    {
+      "epoch": 0.596785061122341,
+      "grad_norm": 0.23080331087112427,
+      "learning_rate": 3.6963703902488907e-06,
+      "loss": 0.787,
+      "step": 1550
+    },
+    {
+      "epoch": 0.6006352873231302,
+      "grad_norm": 0.22870323061943054,
+      "learning_rate": 3.6362721827835655e-06,
+      "loss": 0.8059,
+      "step": 1560
+    },
+    {
+      "epoch": 0.6044855135239195,
+      "grad_norm": 0.241682231426239,
+      "learning_rate": 3.576386087436212e-06,
+      "loss": 0.7986,
+      "step": 1570
+    },
+    {
+      "epoch": 0.6083357397247088,
+      "grad_norm": 0.25029483437538147,
+      "learning_rate": 3.5167214187975913e-06,
+      "loss": 0.7891,
+      "step": 1580
+    },
+    {
+      "epoch": 0.6121859659254981,
+      "grad_norm": 0.21725353598594666,
+      "learning_rate": 3.457287457018086e-06,
+      "loss": 0.7907,
+      "step": 1590
+    },
+    {
+      "epoch": 0.6160361921262875,
+      "grad_norm": 0.23485542833805084,
+      "learning_rate": 3.3980934463643003e-06,
+      "loss": 0.7782,
+      "step": 1600
+    },
+    {
+      "epoch": 0.6198864183270767,
+      "grad_norm": 0.22257378697395325,
+      "learning_rate": 3.3391485937812053e-06,
+      "loss": 0.762,
+      "step": 1610
+    },
+    {
+      "epoch": 0.623736644527866,
+      "grad_norm": 0.24912196397781372,
+      "learning_rate": 3.280462067460115e-06,
+      "loss": 0.7608,
+      "step": 1620
+    },
+    {
+      "epoch": 0.6275868707286553,
+      "grad_norm": 0.2392888069152832,
+      "learning_rate": 3.222042995412669e-06,
+      "loss": 0.7913,
+      "step": 1630
+    },
+    {
+      "epoch": 0.6314370969294446,
+      "grad_norm": 0.21915405988693237,
+      "learning_rate": 3.163900464051089e-06,
+      "loss": 0.7818,
+      "step": 1640
+    },
+    {
+      "epoch": 0.635287323130234,
+      "grad_norm": 0.23636718094348907,
+      "learning_rate": 3.1060435167748826e-06,
+      "loss": 0.7894,
+      "step": 1650
+    },
+    {
+      "epoch": 0.6391375493310232,
+      "grad_norm": 0.23975181579589844,
+      "learning_rate": 3.048481152564253e-06,
+      "loss": 0.7699,
+      "step": 1660
+    },
+    {
+      "epoch": 0.6429877755318125,
+      "grad_norm": 0.24814823269844055,
+      "learning_rate": 2.9912223245804017e-06,
+      "loss": 0.7918,
+      "step": 1670
+    },
+    {
+      "epoch": 0.6468380017326018,
+      "grad_norm": 0.24476778507232666,
+      "learning_rate": 2.934275938772981e-06,
+      "loss": 0.7895,
+      "step": 1680
+    },
+    {
+      "epoch": 0.6506882279333911,
+      "grad_norm": 0.25311481952667236,
+      "learning_rate": 2.8776508524948676e-06,
+      "loss": 0.7783,
+      "step": 1690
+    },
+    {
+      "epoch": 0.6545384541341804,
+      "grad_norm": 0.21385720372200012,
+      "learning_rate": 2.821355873124503e-06,
+      "loss": 0.7774,
+      "step": 1700
+    },
+    {
+      "epoch": 0.6583886803349697,
+      "grad_norm": 0.2441474199295044,
+      "learning_rate": 2.7653997566960144e-06,
+      "loss": 0.7916,
+      "step": 1710
+    },
+    {
+      "epoch": 0.662238906535759,
+      "grad_norm": 0.24739249050617218,
+      "learning_rate": 2.709791206537309e-06,
+      "loss": 0.7815,
+      "step": 1720
+    },
+    {
+      "epoch": 0.6660891327365482,
+      "grad_norm": 0.23807218670845032,
+      "learning_rate": 2.6545388719163747e-06,
+      "loss": 0.7889,
+      "step": 1730
+    },
+    {
+      "epoch": 0.6699393589373376,
+      "grad_norm": 0.23439769446849823,
+      "learning_rate": 2.599651346695979e-06,
+      "loss": 0.7861,
+      "step": 1740
+    },
+    {
+      "epoch": 0.6737895851381268,
+      "grad_norm": 0.24357961118221283,
+      "learning_rate": 2.545137167997007e-06,
+      "loss": 0.7823,
+      "step": 1750
+    },
+    {
+      "epoch": 0.6776398113389162,
+      "grad_norm": 0.2284420281648636,
+      "learning_rate": 2.4910048148705972e-06,
+      "loss": 0.7693,
+      "step": 1760
+    },
+    {
+      "epoch": 0.6814900375397055,
+      "grad_norm": 0.2600451707839966,
+      "learning_rate": 2.4372627069793327e-06,
+      "loss": 0.8127,
+      "step": 1770
+    },
+    {
+      "epoch": 0.6853402637404947,
+      "grad_norm": 0.24222847819328308,
+      "learning_rate": 2.3839192032876486e-06,
+      "loss": 0.7997,
+      "step": 1780
+    },
+    {
+      "epoch": 0.6891904899412841,
+      "grad_norm": 0.23689132928848267,
+      "learning_rate": 2.3309826007617047e-06,
+      "loss": 0.8093,
+      "step": 1790
+    },
+    {
+      "epoch": 0.6930407161420733,
+      "grad_norm": 0.2341219037771225,
+      "learning_rate": 2.278461133078881e-06,
+      "loss": 0.7747,
+      "step": 1800
+    },
+    {
+      "epoch": 0.6968909423428626,
+      "grad_norm": 0.24800433218479156,
+      "learning_rate": 2.2263629693471258e-06,
+      "loss": 0.7645,
+      "step": 1810
+    },
+    {
+      "epoch": 0.700741168543652,
+      "grad_norm": 0.23428261280059814,
+      "learning_rate": 2.1746962128343414e-06,
+      "loss": 0.7906,
+      "step": 1820
+    },
+    {
+      "epoch": 0.7045913947444412,
+      "grad_norm": 0.22502082586288452,
+      "learning_rate": 2.1234688997080183e-06,
+      "loss": 0.778,
+      "step": 1830
+    },
+    {
+      "epoch": 0.7084416209452306,
+      "grad_norm": 0.24504925310611725,
+      "learning_rate": 2.0726889977853053e-06,
+      "loss": 0.7821,
+      "step": 1840
+    },
+    {
+      "epoch": 0.7122918471460198,
+      "grad_norm": 0.25640496611595154,
+      "learning_rate": 2.022364405293703e-06,
+      "loss": 0.7971,
+      "step": 1850
+    },
+    {
+      "epoch": 0.7161420733468091,
+      "grad_norm": 0.23833368718624115,
+      "learning_rate": 1.9725029496425814e-06,
+      "loss": 0.7862,
+      "step": 1860
+    },
+    {
+      "epoch": 0.7199922995475985,
+      "grad_norm": 0.2403428852558136,
+      "learning_rate": 1.923112386205727e-06,
+      "loss": 0.7849,
+      "step": 1870
+    },
+    {
+      "epoch": 0.7238425257483877,
+      "grad_norm": 0.22354473173618317,
+      "learning_rate": 1.874200397115079e-06,
+      "loss": 0.791,
+      "step": 1880
+    },
+    {
+      "epoch": 0.727692751949177,
+      "grad_norm": 0.24569858610630035,
+      "learning_rate": 1.825774590065857e-06,
+      "loss": 0.7744,
+      "step": 1890
+    },
+    {
+      "epoch": 0.7315429781499663,
+      "grad_norm": 0.24243475496768951,
+      "learning_rate": 1.7778424971332813e-06,
+      "loss": 0.7818,
+      "step": 1900
+    },
+    {
+      "epoch": 0.7353932043507556,
+      "grad_norm": 0.231866717338562,
+      "learning_rate": 1.7304115736010407e-06,
+      "loss": 0.7813,
+      "step": 1910
+    },
+    {
+      "epoch": 0.739243430551545,
+      "grad_norm": 0.23963728547096252,
+      "learning_rate": 1.6834891968017108e-06,
+      "loss": 0.7722,
+      "step": 1920
+    },
+    {
+      "epoch": 0.7430936567523342,
+      "grad_norm": 0.25933918356895447,
+      "learning_rate": 1.6370826649692884e-06,
+      "loss": 0.7943,
+      "step": 1930
+    },
+    {
+      "epoch": 0.7469438829531235,
+      "grad_norm": 0.2415194809436798,
+      "learning_rate": 1.5911991961040463e-06,
+      "loss": 0.8016,
+      "step": 1940
+    },
+    {
+      "epoch": 0.7507941091539128,
+      "grad_norm": 0.2766319811344147,
+      "learning_rate": 1.545845926849851e-06,
+      "loss": 0.7873,
+      "step": 1950
+    },
+    {
+      "epoch": 0.7546443353547021,
+      "grad_norm": 0.255489319562912,
+      "learning_rate": 1.5010299113841397e-06,
+      "loss": 0.8053,
+      "step": 1960
+    },
+    {
+      "epoch": 0.7584945615554913,
+      "grad_norm": 0.2512015104293823,
+      "learning_rate": 1.4567581203207238e-06,
+      "loss": 0.7928,
+      "step": 1970
+    },
+    {
+      "epoch": 0.7623447877562807,
+      "grad_norm": 0.2516551613807678,
+      "learning_rate": 1.413037439625593e-06,
+      "loss": 0.7943,
+      "step": 1980
+    },
+    {
+      "epoch": 0.76619501395707,
+      "grad_norm": 0.25249722599983215,
+      "learning_rate": 1.3698746695458847e-06,
+      "loss": 0.7864,
+      "step": 1990
+    },
+    {
+      "epoch": 0.7700452401578592,
+      "grad_norm": 0.23430432379245758,
+      "learning_rate": 1.3272765235521745e-06,
+      "loss": 0.7775,
+      "step": 2000
+    },
+    {
+      "epoch": 0.7738954663586486,
+      "grad_norm": 0.22957263886928558,
+      "learning_rate": 1.2852496272942877e-06,
+      "loss": 0.7854,
+      "step": 2010
+    },
+    {
+      "epoch": 0.7777456925594378,
+      "grad_norm": 0.22727033495903015,
+      "learning_rate": 1.2438005175707396e-06,
+      "loss": 0.7728,
+      "step": 2020
+    },
+    {
+      "epoch": 0.7815959187602272,
+      "grad_norm": 0.2321804165840149,
+      "learning_rate": 1.2029356413120297e-06,
+      "loss": 0.7608,
+      "step": 2030
+    },
+    {
+      "epoch": 0.7854461449610165,
+      "grad_norm": 0.23626334965229034,
+      "learning_rate": 1.1626613545778786e-06,
+      "loss": 0.7769,
+      "step": 2040
+    },
+    {
+      "epoch": 0.7892963711618057,
+      "grad_norm": 0.2268046885728836,
+      "learning_rate": 1.1229839215686294e-06,
+      "loss": 0.7805,
+      "step": 2050
+    },
+    {
+      "epoch": 0.7931465973625951,
+      "grad_norm": 0.21847327053546906,
+      "learning_rate": 1.0839095136509188e-06,
+      "loss": 0.7685,
+      "step": 2060
+    },
+    {
+      "epoch": 0.7969968235633843,
+      "grad_norm": 0.26072630286216736,
+      "learning_rate": 1.045444208397791e-06,
+      "loss": 0.7899,
+      "step": 2070
+    },
+    {
+      "epoch": 0.8008470497641736,
+      "grad_norm": 0.22972933948040009,
+      "learning_rate": 1.007593988643402e-06,
+      "loss": 0.7679,
+      "step": 2080
+    },
+    {
+      "epoch": 0.804697275964963,
+      "grad_norm": 0.23452307283878326,
+      "learning_rate": 9.703647415524613e-07,
+      "loss": 0.7916,
+      "step": 2090
+    },
+    {
+      "epoch": 0.8085475021657522,
+      "grad_norm": 0.2332160323858261,
+      "learning_rate": 9.337622577045485e-07,
+      "loss": 0.7778,
+      "step": 2100
+    },
+    {
+      "epoch": 0.8123977283665416,
+      "grad_norm": 0.2309354692697525,
+      "learning_rate": 8.977922301934588e-07,
+      "loss": 0.7874,
+      "step": 2110
+    },
+    {
+      "epoch": 0.8162479545673308,
+      "grad_norm": 0.23642677068710327,
+      "learning_rate": 8.624602537417009e-07,
+      "loss": 0.7895,
+      "step": 2120
+    },
+    {
+      "epoch": 0.8200981807681201,
+      "grad_norm": 0.2485097497701645,
+      "learning_rate": 8.277718238303123e-07,
+      "loss": 0.7775,
+      "step": 2130
+    },
+    {
+      "epoch": 0.8239484069689095,
+      "grad_norm": 0.23456016182899475,
+      "learning_rate": 7.937323358440935e-07,
+      "loss": 0.7812,
+      "step": 2140
+    },
+    {
+      "epoch": 0.8277986331696987,
+      "grad_norm": 0.2540385127067566,
+      "learning_rate": 7.603470842324223e-07,
+      "loss": 0.7743,
+      "step": 2150
+    },
+    {
+      "epoch": 0.831648859370488,
+      "grad_norm": 0.22980892658233643,
+      "learning_rate": 7.276212616857575e-07,
+      "loss": 0.7905,
+      "step": 2160
+    },
+    {
+      "epoch": 0.8354990855712773,
+      "grad_norm": 0.23359587788581848,
+      "learning_rate": 6.955599583279837e-07,
+      "loss": 0.7929,
+      "step": 2170
+    },
+    {
+      "epoch": 0.8393493117720666,
+      "grad_norm": 0.2423620969057083,
+      "learning_rate": 6.641681609246981e-07,
+      "loss": 0.7826,
+      "step": 2180
+    },
+    {
+      "epoch": 0.8431995379728559,
+      "grad_norm": 0.22636087238788605,
+      "learning_rate": 6.334507521075711e-07,
+      "loss": 0.7906,
+      "step": 2190
+    },
+    {
+      "epoch": 0.8470497641736452,
+      "grad_norm": 0.2549637258052826,
+      "learning_rate": 6.034125096149179e-07,
+      "loss": 0.7838,
+      "step": 2200
+    },
+    {
+      "epoch": 0.8508999903744345,
+      "grad_norm": 0.2388104945421219,
+      "learning_rate": 5.740581055485722e-07,
+      "loss": 0.7553,
+      "step": 2210
+    },
+    {
+      "epoch": 0.8547502165752238,
+      "grad_norm": 0.22281107306480408,
+      "learning_rate": 5.453921056471961e-07,
+      "loss": 0.7889,
+      "step": 2220
+    },
+    {
+      "epoch": 0.8586004427760131,
+      "grad_norm": 0.23528993129730225,
+      "learning_rate": 5.174189685761288e-07,
+      "loss": 0.7808,
+      "step": 2230
+    },
+    {
+      "epoch": 0.8624506689768023,
+      "grad_norm": 0.2300231009721756,
+      "learning_rate": 4.901430452338979e-07,
+      "loss": 0.7941,
+      "step": 2240
+    },
+    {
+      "epoch": 0.8663008951775917,
+      "grad_norm": 0.23825491964817047,
+      "learning_rate": 4.6356857807548286e-07,
+      "loss": 0.7788,
+      "step": 2250
+    },
+    {
+      "epoch": 0.870151121378381,
+      "grad_norm": 0.2390144318342209,
+      "learning_rate": 4.376997004524541e-07,
+      "loss": 0.7676,
+      "step": 2260
+    },
+    {
+      "epoch": 0.8740013475791703,
+      "grad_norm": 0.2332000434398651,
+      "learning_rate": 4.12540435970073e-07,
+      "loss": 0.779,
+      "step": 2270
+    },
+    {
+      "epoch": 0.8778515737799596,
+      "grad_norm": 0.23488333821296692,
+      "learning_rate": 3.880946978614703e-07,
+      "loss": 0.7867,
+      "step": 2280
+    },
+    {
+      "epoch": 0.8817017999807488,
+      "grad_norm": 0.2334536910057068,
+      "learning_rate": 3.643662883789878e-07,
+      "loss": 0.792,
+      "step": 2290
+    },
+    {
+      "epoch": 0.8855520261815382,
+      "grad_norm": 0.22744931280612946,
+      "learning_rate": 3.41358898202781e-07,
+      "loss": 0.7691,
+      "step": 2300
+    },
+    {
+      "epoch": 0.8894022523823275,
+      "grad_norm": 0.22276608645915985,
+      "learning_rate": 3.190761058667741e-07,
+      "loss": 0.7813,
+      "step": 2310
+    },
+    {
+      "epoch": 0.8932524785831167,
+      "grad_norm": 0.23100407421588898,
+      "learning_rate": 2.9752137720206463e-07,
+      "loss": 0.788,
+      "step": 2320
+    },
+    {
+      "epoch": 0.8971027047839061,
+      "grad_norm": 0.23106254637241364,
+      "learning_rate": 2.7669806479785067e-07,
+      "loss": 0.793,
+      "step": 2330
+    },
+    {
+      "epoch": 0.9009529309846953,
+      "grad_norm": 0.2433275282382965,
+      "learning_rate": 2.566094074799763e-07,
+      "loss": 0.7878,
+      "step": 2340
+    },
+    {
+      "epoch": 0.9048031571854847,
+      "grad_norm": 0.2315477728843689,
+      "learning_rate": 2.3725852980716557e-07,
+      "loss": 0.7668,
+      "step": 2350
+    },
+    {
+      "epoch": 0.908653383386274,
+      "grad_norm": 0.2288062572479248,
+      "learning_rate": 2.1864844158503983e-07,
+      "loss": 0.7809,
+      "step": 2360
+    },
+    {
+      "epoch": 0.9125036095870632,
+      "grad_norm": 0.2542728781700134,
+      "learning_rate": 2.007820373979702e-07,
+      "loss": 0.7715,
+      "step": 2370
+    },
+    {
+      "epoch": 0.9163538357878526,
+      "grad_norm": 0.2367563247680664,
+      "learning_rate": 1.8366209615886532e-07,
+      "loss": 0.7804,
+      "step": 2380
+    },
+    {
+      "epoch": 0.9202040619886418,
+      "grad_norm": 0.2387159913778305,
+      "learning_rate": 1.6729128067693613e-07,
+      "loss": 0.7881,
+      "step": 2390
+    },
+    {
+      "epoch": 0.9240542881894311,
+      "grad_norm": 0.25151440501213074,
+      "learning_rate": 1.5167213724353426e-07,
+      "loss": 0.7914,
+      "step": 2400
+    },
+    {
+      "epoch": 0.9279045143902205,
+      "grad_norm": 0.22943991422653198,
+      "learning_rate": 1.3680709523610058e-07,
+      "loss": 0.7748,
+      "step": 2410
+    },
+    {
+      "epoch": 0.9317547405910097,
+      "grad_norm": 0.22889508306980133,
+      "learning_rate": 1.2269846674030418e-07,
+      "loss": 0.7962,
+      "step": 2420
+    },
+    {
+      "epoch": 0.935604966791799,
+      "grad_norm": 0.23346251249313354,
+      "learning_rate": 1.0934844619042595e-07,
+      "loss": 0.7936,
+      "step": 2430
+    },
+    {
+      "epoch": 0.9394551929925883,
+      "grad_norm": 0.25133898854255676,
+      "learning_rate": 9.675911002803961e-08,
+      "loss": 0.7818,
+      "step": 2440
+    },
+    {
+      "epoch": 0.9433054191933776,
+      "grad_norm": 0.24764913320541382,
+      "learning_rate": 8.493241637904381e-08,
+      "loss": 0.7799,
+      "step": 2450
+    },
+    {
+      "epoch": 0.9471556453941669,
+      "grad_norm": 0.2359425574541092,
+      "learning_rate": 7.387020474909689e-08,
+      "loss": 0.7985,
+      "step": 2460
+    },
+    {
+      "epoch": 0.9510058715949562,
+      "grad_norm": 0.24563314020633698,
+      "learning_rate": 6.357419573750856e-08,
+      "loss": 0.7948,
+      "step": 2470
+    },
+    {
+      "epoch": 0.9548560977957455,
+      "grad_norm": 0.23891755938529968,
+      "learning_rate": 5.4045990769613365e-08,
+      "loss": 0.7774,
+      "step": 2480
+    },
+    {
+      "epoch": 0.9587063239965348,
+      "grad_norm": 0.27896466851234436,
+      "learning_rate": 4.5287071847694434e-08,
+      "loss": 0.7828,
+      "step": 2490
+    },
+    {
+      "epoch": 0.9625565501973241,
+      "grad_norm": 0.23237410187721252,
+      "learning_rate": 3.7298801320467817e-08,
+      "loss": 0.775,
+      "step": 2500
+    },
+    {
+      "epoch": 0.9664067763981133,
+      "grad_norm": 0.23584948480129242,
+      "learning_rate": 3.0082421671192576e-08,
+      "loss": 0.7692,
+      "step": 2510
+    },
+    {
+      "epoch": 0.9702570025989027,
+      "grad_norm": 0.2360149472951889,
+      "learning_rate": 2.3639055324410355e-08,
+      "loss": 0.791,
+      "step": 2520
+    },
+    {
+      "epoch": 0.974107228799692,
+      "grad_norm": 0.2312159240245819,
+      "learning_rate": 1.7969704471369477e-08,
+      "loss": 0.7756,
+      "step": 2530
+    },
+    {
+      "epoch": 0.9779574550004813,
+      "grad_norm": 0.22454580664634705,
+      "learning_rate": 1.3075250914142413e-08,
+      "loss": 0.7791,
+      "step": 2540
+    },
+    {
+      "epoch": 0.9818076812012706,
+      "grad_norm": 0.22735947370529175,
+      "learning_rate": 8.956455928476049e-09,
+      "loss": 0.7882,
+      "step": 2550
+    },
+    {
+      "epoch": 0.9856579074020598,
+      "grad_norm": 0.23890043795108795,
+      "learning_rate": 5.6139601453786275e-09,
+      "loss": 0.7807,
+      "step": 2560
+    },
+    {
+      "epoch": 0.9895081336028492,
+      "grad_norm": 0.23844030499458313,
+      "learning_rate": 3.0482834514805602e-09,
+      "loss": 0.7694,
+      "step": 2570
+    },
+    {
+      "epoch": 0.9933583598036385,
+      "grad_norm": 0.2166026383638382,
+      "learning_rate": 1.2598249081713366e-09,
+      "loss": 0.7646,
+      "step": 2580
+    },
+    {
+      "epoch": 0.9972085860044277,
+      "grad_norm": 0.24653111398220062,
+      "learning_rate": 2.488626895302826e-10,
+      "loss": 0.7951,
+      "step": 2590
+    },
+    {
+      "epoch": 0.9999037443449803,
+      "step": 2597,
+      "total_flos": 6.412320322489418e+18,
+      "train_loss": 0.8115227478579645,
+      "train_runtime": 13239.9652,
+      "train_samples_per_second": 50.219,
+      "train_steps_per_second": 0.196
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 2597,
+  "num_input_tokens_seen": 418855352,
+  "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": 6.412320322489418e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}

continuous/qwen3/webssl_dino1b_full2b_224/finetune/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:205897467e40640b318453340868d509962b1710b11b19466b8f3c1ae68f4a5f
+size 6264

continuous/qwen3/webssl_dino1b_full2b_224/pretrain/config.json

@@ -0,0 +1,89 @@
+{
+  "add_faster_video": false,
+  "add_time_instruction": false,
+  "architectures": [
+    "Qwen3ForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "dtype": "bfloat16",
+  "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,
+  "layer_types": [
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention"
+  ],
+  "max_position_embeddings": 40960,
+  "max_window_layers": 28,
+  "mm_hidden_size": 1536,
+  "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": -1,
+  "mm_vision_tower": "hf:/cache/ckpt/download/tokenizer/continuous/webssl/webssl-dino1b-full2b-224",
+  "mm_vision_tower_lr": null,
+  "model_type": "qwen3",
+  "num_attention_heads": 16,
+  "num_hidden_layers": 28,
+  "num_key_value_heads": 8,
+  "pad_token_id": 151643,
+  "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",
+  "transformers_version": "4.57.6",
+  "use_cache": true,
+  "use_mm_proj": true,
+  "use_pos_skipping": false,
+  "use_sliding_window": false,
+  "vision_tower_pretrained": null,
+  "vocab_size": 151936
+}

continuous/qwen3/webssl_dino1b_full2b_224/pretrain/mm_projector.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8edea4cf4e305e7e642d75b511ab3fda07af245c807df815e2262984e0d69c2c
+size 29378616

continuous/qwen3/webssl_dino1b_full2b_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": 18.761455535888672,
+      "learning_rate": 0.00032142857142857147,
+      "loss": 9.5654,
+      "step": 10
+    },
+    {
+      "epoch": 0.03669724770642202,
+      "grad_norm": 0.3320707082748413,
+      "learning_rate": 0.0006785714285714287,
+      "loss": 5.3418,
+      "step": 20
+    },
+    {
+      "epoch": 0.05504587155963303,
+      "grad_norm": 0.08488818258047104,
+      "learning_rate": 0.0009999907688176173,
+      "loss": 4.7653,
+      "step": 30
+    },
+    {
+      "epoch": 0.07339449541284404,
+      "grad_norm": 0.037156879901885986,
+      "learning_rate": 0.0009988834393115766,
+      "loss": 4.6696,
+      "step": 40
+    },
+    {
+      "epoch": 0.09174311926605505,
+      "grad_norm": 0.15282955765724182,
+      "learning_rate": 0.0009959345572940772,
+      "loss": 4.6208,
+      "step": 50
+    },
+    {
+      "epoch": 0.11009174311926606,
+      "grad_norm": 0.03435467183589935,
+      "learning_rate": 0.0009911550081155983,
+      "loss": 4.6069,
+      "step": 60
+    },
+    {
+      "epoch": 0.12844036697247707,
+      "grad_norm": 0.04698146507143974,
+      "learning_rate": 0.0009845624347566433,
+      "loss": 4.5635,
+      "step": 70
+    },
+    {
+      "epoch": 0.14678899082568808,
+      "grad_norm": 0.06590677052736282,
+      "learning_rate": 0.0009761811727013547,
+      "loss": 4.5319,
+      "step": 80
+    },
+    {
+      "epoch": 0.1651376146788991,
+      "grad_norm": 0.0817234218120575,
+      "learning_rate": 0.0009660421601067666,
+      "loss": 4.4765,
+      "step": 90
+    },
+    {
+      "epoch": 0.1834862385321101,
+      "grad_norm": 0.05971318110823631,
+      "learning_rate": 0.0009541828235992926,
+      "loss": 4.4464,
+      "step": 100
+    },
+    {
+      "epoch": 0.2018348623853211,
+      "grad_norm": 0.05726119130849838,
+      "learning_rate": 0.0009406469401200151,
+      "loss": 4.4029,
+      "step": 110
+    },
+    {
+      "epoch": 0.22018348623853212,
+      "grad_norm": 0.10169172286987305,
+      "learning_rate": 0.0009254844753287492,
+      "loss": 4.3778,
+      "step": 120
+    },
+    {
+      "epoch": 0.23853211009174313,
+      "grad_norm": 0.10542549937963486,
+      "learning_rate": 0.0009087513991633924,
+      "loss": 4.3324,
+      "step": 130
+    },
+    {
+      "epoch": 0.25688073394495414,
+      "grad_norm": 0.09493378549814224,
+      "learning_rate": 0.0008905094792353917,
+      "loss": 4.3057,
+      "step": 140
+    },
+    {
+      "epoch": 0.27522935779816515,
+      "grad_norm": 0.08435375988483429,
+      "learning_rate": 0.0008708260528239789,
+      "loss": 4.2928,
+      "step": 150
+    },
+    {
+      "epoch": 0.29357798165137616,
+      "grad_norm": 0.14110971987247467,
+      "learning_rate": 0.0008497737783108237,
+      "loss": 4.2651,
+      "step": 160
+    },
+    {
+      "epoch": 0.3119266055045872,
+      "grad_norm": 0.081395722925663,
+      "learning_rate": 0.0008274303669726426,
+      "loss": 4.2499,
+      "step": 170
+    },
+    {
+      "epoch": 0.3302752293577982,
+      "grad_norm": 0.0679878517985344,
+      "learning_rate": 0.0008038782961218137,
+      "loss": 4.232,
+      "step": 180
+    },
+    {
+      "epoch": 0.3486238532110092,
+      "grad_norm": 0.06032344698905945,
+      "learning_rate": 0.000779204504653894,
+      "loss": 4.2216,
+      "step": 190
+    },
+    {
+      "epoch": 0.3669724770642202,
+      "grad_norm": 0.061087559908628464,
+      "learning_rate": 0.0007535000721258791,
+      "loss": 4.2008,
+      "step": 200
+    },
+    {
+      "epoch": 0.3853211009174312,
+      "grad_norm": 0.057567231357097626,
+      "learning_rate": 0.00072685988254984,
+      "loss": 4.1887,
+      "step": 210
+    },
+    {
+      "epoch": 0.4036697247706422,
+      "grad_norm": 0.05859406664967537,
+      "learning_rate": 0.0006993822741429906,
+      "loss": 4.2001,
+      "step": 220
+    },
+    {
+      "epoch": 0.42201834862385323,
+      "grad_norm": 0.05695907771587372,
+      "learning_rate": 0.0006711686763270818,
+      "loss": 4.1951,
+      "step": 230
+    },
+    {
+      "epoch": 0.44036697247706424,
+      "grad_norm": 0.06442370265722275,
+      "learning_rate": 0.0006423232353170831,
+      "loss": 4.1826,
+      "step": 240
+    },
+    {
+      "epoch": 0.45871559633027525,
+      "grad_norm": 0.07409843802452087,
+      "learning_rate": 0.0006129524296812335,
+      "loss": 4.1625,
+      "step": 250
+    },
+    {
+      "epoch": 0.47706422018348627,
+      "grad_norm": 0.061325594782829285,
+      "learning_rate": 0.0005831646772915651,
+      "loss": 4.1732,
+      "step": 260
+    },
+    {
+      "epoch": 0.4954128440366973,
+      "grad_norm": 0.06499546766281128,
+      "learning_rate": 0.0005530699351157809,
+      "loss": 4.1553,
+      "step": 270
+    },
+    {
+      "epoch": 0.5137614678899083,
+      "grad_norm": 0.0636533871293068,
+      "learning_rate": 0.0005227792933277943,
+      "loss": 4.1513,
+      "step": 280
+    },
+    {
+      "epoch": 0.5321100917431193,
+      "grad_norm": 0.05742020159959793,
+      "learning_rate": 0.0004924045652352092,
+      "loss": 4.1387,
+      "step": 290
+    },
+    {
+      "epoch": 0.5504587155963303,
+      "grad_norm": 0.07275547087192535,
+      "learning_rate": 0.0004620578745374633,
+      "loss": 4.138,
+      "step": 300
+    },
+    {
+      "epoch": 0.5688073394495413,
+      "grad_norm": 0.050887513905763626,
+      "learning_rate": 0.0004318512414382059,
+      "loss": 4.1521,
+      "step": 310
+    },
+    {
+      "epoch": 0.5871559633027523,
+      "grad_norm": 0.05948026478290558,
+      "learning_rate": 0.0004018961691397148,
+      "loss": 4.1407,
+      "step": 320
+    },
+    {
+      "epoch": 0.6055045871559633,
+      "grad_norm": 0.05940104275941849,
+      "learning_rate": 0.0003723032322457458,
+      "loss": 4.1216,
+      "step": 330
+    },
+    {
+      "epoch": 0.6238532110091743,
+      "grad_norm": 0.0893605500459671,
+      "learning_rate": 0.0003431816685921625,
+      "loss": 4.1282,
+      "step": 340
+    },
+    {
+      "epoch": 0.6422018348623854,
+      "grad_norm": 0.06026393547654152,
+      "learning_rate": 0.0003146389760120416,
+      "loss": 4.1304,
+      "step": 350
+    },
+    {
+      "epoch": 0.6605504587155964,
+      "grad_norm": 0.05902978777885437,
+      "learning_rate": 0.00028678051552373485,
+      "loss": 4.1369,
+      "step": 360
+    },
+    {
+      "epoch": 0.6788990825688074,
+      "grad_norm": 0.05430439114570618,
+      "learning_rate": 0.0002597091224066581,
+      "loss": 4.1174,
+      "step": 370
+    },
+    {
+      "epoch": 0.6972477064220184,
+      "grad_norm": 0.05181219428777695,
+      "learning_rate": 0.0002335247266004629,
+      "loss": 4.1307,
+      "step": 380
+    },
+    {
+      "epoch": 0.7155963302752294,
+      "grad_norm": 0.05691559240221977,
+      "learning_rate": 0.000208323983828828,
+      "loss": 4.1122,
+      "step": 390
+    },
+    {
+      "epoch": 0.7339449541284404,
+      "grad_norm": 0.059544216841459274,
+      "learning_rate": 0.0001841999188095224,
+      "loss": 4.1188,
+      "step": 400
+    },
+    {
+      "epoch": 0.7522935779816514,
+      "grad_norm": 0.05225243419408798,
+      "learning_rate": 0.00016124158186777676,
+      "loss": 4.1159,
+      "step": 410
+    },
+    {
+      "epoch": 0.7706422018348624,
+      "grad_norm": 0.04934331029653549,
+      "learning_rate": 0.0001395337202205211,
+      "loss": 4.1052,
+      "step": 420
+    },
+    {
+      "epoch": 0.7889908256880734,
+      "grad_norm": 0.05131152272224426,
+      "learning_rate": 0.00011915646514489292,
+      "loss": 4.1157,
+      "step": 430
+    },
+    {
+      "epoch": 0.8073394495412844,
+      "grad_norm": 0.06026066467165947,
+      "learning_rate": 0.00010018503618578818,
+      "loss": 4.123,
+      "step": 440
+    },
+    {
+      "epoch": 0.8256880733944955,
+      "grad_norm": 0.04959550127387047,
+      "learning_rate": 8.268946349432583e-05,
+      "loss": 4.1114,
+      "step": 450
+    },
+    {
+      "epoch": 0.8440366972477065,
+      "grad_norm": 0.05523325875401497,
+      "learning_rate": 6.673432932217117e-05,
+      "loss": 4.1011,
+      "step": 460
+    },
+    {
+      "epoch": 0.8623853211009175,
+      "grad_norm": 0.05080917105078697,
+      "learning_rate": 5.2378529625954684e-05,
+      "loss": 4.0926,
+      "step": 470
+    },
+    {
+      "epoch": 0.8807339449541285,
+      "grad_norm": 0.05264311283826828,
+      "learning_rate": 3.9675056661785556e-05,
+      "loss": 4.1007,
+      "step": 480
+    },
+    {
+      "epoch": 0.8990825688073395,
+      "grad_norm": 0.04710981994867325,
+      "learning_rate": 2.8670803372379538e-05,
+      "loss": 4.1018,
+      "step": 490
+    },
+    {
+      "epoch": 0.9174311926605505,
+      "grad_norm": 0.0508495569229126,
+      "learning_rate": 1.9406390288876586e-05,
+      "loss": 4.1121,
+      "step": 500
+    },
+    {
+      "epoch": 0.9357798165137615,
+      "grad_norm": 0.04948792979121208,
+      "learning_rate": 1.191601558631461e-05,
+      "loss": 4.1022,
+      "step": 510
+    },
+    {
+      "epoch": 0.9541284403669725,
+      "grad_norm": 0.04769504442811012,
+      "learning_rate": 6.2273288462559305e-06,
+      "loss": 4.0944,
+      "step": 520
+    },
+    {
+      "epoch": 0.9724770642201835,
+      "grad_norm": 0.048696454614400864,
+      "learning_rate": 2.361328992552314e-06,
+      "loss": 4.0942,
+      "step": 530
+    },
+    {
+      "epoch": 0.9908256880733946,
+      "grad_norm": 0.047999635338783264,
+      "learning_rate": 3.322867770044202e-07,
+      "loss": 4.1012,
+      "step": 540
+    },
+    {
+      "epoch": 1.0,
+      "step": 545,
+      "total_flos": 4.883090487608607e+17,
+      "train_loss": 4.348349062018438,
+      "train_runtime": 5080.0149,
+      "train_samples_per_second": 109.867,
+      "train_steps_per_second": 0.107
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 545,
+  "num_input_tokens_seen": 31896544,
+  "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": 4.883090487608607e+17,
+  "train_batch_size": 32,
+  "trial_name": null,
+  "trial_params": null
+}

continuous/qwen3/webssl_mae1b_full2b_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/webssl_mae1b_full2b_224/finetune/chat_template.jinja

@@ -0,0 +1,89 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# 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>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\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" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- 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>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if message.content is string %}
+        {%- set content = message.content %}
+    {%- else %}
+        {%- set content = '' %}
+    {%- endif %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}

continuous/qwen3/webssl_mae1b_full2b_224/finetune/config.json

@@ -0,0 +1,89 @@
+{
+  "add_faster_video": false,
+  "add_time_instruction": false,
+  "architectures": [
+    "LlavaQwen3ForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "dtype": "bfloat16",
+  "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,
+  "layer_types": [
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention"
+  ],
+  "max_position_embeddings": 40960,
+  "max_window_layers": 28,
+  "mm_hidden_size": 1536,
+  "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": -1,
+  "mm_vision_tower": "hf:/cache/ckpt/download/tokenizer/continuous/webssl/webssl-mae1b-full2b-224",
+  "mm_vision_tower_lr": null,
+  "model_type": "qwen3",
+  "num_attention_heads": 16,
+  "num_hidden_layers": 28,
+  "num_key_value_heads": 8,
+  "pad_token_id": 151643,
+  "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",
+  "transformers_version": "4.57.6",
+  "use_cache": true,
+  "use_mm_proj": true,
+  "use_pos_skipping": false,
+  "use_sliding_window": false,
+  "vision_tower_pretrained": null,
+  "vocab_size": 151936
+}

continuous/qwen3/webssl_mae1b_full2b_224/finetune/generation_config.json

@@ -0,0 +1,12 @@
+{
+  "do_sample": true,
+  "eos_token_id": [
+    151645,
+    151643
+  ],
+  "pad_token_id": 151643,
+  "temperature": 0.6,
+  "top_k": 20,
+  "top_p": 0.95,
+  "transformers_version": "4.57.6"
+}