Upload continuous evaluation checkpoints, batch 20

.gitattributes

@@ -158,3 +158,6 @@ continuous/qwen3/pixio_vith16/finetune/tokenizer.json filter=lfs diff=lfs merge=
 continuous/qwen3/pixio_vitl16/finetune/tokenizer.json filter=lfs diff=lfs merge=lfs -text
 continuous/qwen3/raev2_dinov3l_k7/finetune/tokenizer.json filter=lfs diff=lfs merge=lfs -text
 continuous/qwen3/siglip2_b16_224/finetune/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+continuous/qwen3/siglip2_b16_256/finetune/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+continuous/qwen3/siglip2_b16_384/finetune/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+continuous/qwen3/siglip2_b16_512/finetune/tokenizer.json filter=lfs diff=lfs merge=lfs -text

continuous/qwen3/siglip2_b16_256/finetune/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fc81addc5741e53434816a7791826d43fcbc440eaedaff0fbfe6ee38bcd335b0
+size 4272483456

continuous/qwen3/siglip2_b16_256/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_b16_256/finetune/tokenizer.json

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

continuous/qwen3/siglip2_b16_256/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_b16_256/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": 22.325674057006836,
+      "learning_rate": 1.153846153846154e-06,
+      "loss": 2.5709,
+      "step": 10
+    },
+    {
+      "epoch": 0.007700452401578592,
+      "grad_norm": 7.213373184204102,
+      "learning_rate": 2.435897435897436e-06,
+      "loss": 1.9029,
+      "step": 20
+    },
+    {
+      "epoch": 0.01155067860236789,
+      "grad_norm": 4.463220596313477,
+      "learning_rate": 3.7179487179487184e-06,
+      "loss": 1.4756,
+      "step": 30
+    },
+    {
+      "epoch": 0.015400904803157185,
+      "grad_norm": 1.2861229181289673,
+      "learning_rate": 5e-06,
+      "loss": 1.1357,
+      "step": 40
+    },
+    {
+      "epoch": 0.019251131003946483,
+      "grad_norm": 0.7633169889450073,
+      "learning_rate": 6.282051282051282e-06,
+      "loss": 1.0227,
+      "step": 50
+    },
+    {
+      "epoch": 0.02310135720473578,
+      "grad_norm": 0.40343567728996277,
+      "learning_rate": 7.564102564102564e-06,
+      "loss": 0.9316,
+      "step": 60
+    },
+    {
+      "epoch": 0.026951583405525073,
+      "grad_norm": 0.31116607785224915,
+      "learning_rate": 8.846153846153847e-06,
+      "loss": 0.8872,
+      "step": 70
+    },
+    {
+      "epoch": 0.03080180960631437,
+      "grad_norm": 0.3002501428127289,
+      "learning_rate": 9.999996111488723e-06,
+      "loss": 0.8838,
+      "step": 80
+    },
+    {
+      "epoch": 0.034652035807103666,
+      "grad_norm": 0.3025326728820801,
+      "learning_rate": 9.999529497453782e-06,
+      "loss": 0.8532,
+      "step": 90
+    },
+    {
+      "epoch": 0.038502262007892966,
+      "grad_norm": 0.2748545706272125,
+      "learning_rate": 9.998285264324193e-06,
+      "loss": 0.8305,
+      "step": 100
+    },
+    {
+      "epoch": 0.04235248820868226,
+      "grad_norm": 0.27491244673728943,
+      "learning_rate": 9.996263605626052e-06,
+      "loss": 0.8166,
+      "step": 110
+    },
+    {
+      "epoch": 0.04620271440947156,
+      "grad_norm": 0.2842361032962799,
+      "learning_rate": 9.99346483580503e-06,
+      "loss": 0.8268,
+      "step": 120
+    },
+    {
+      "epoch": 0.05005294061026085,
+      "grad_norm": 0.2531988024711609,
+      "learning_rate": 9.989889390177467e-06,
+      "loss": 0.8197,
+      "step": 130
+    },
+    {
+      "epoch": 0.053903166811050146,
+      "grad_norm": 0.3036770224571228,
+      "learning_rate": 9.985537824862647e-06,
+      "loss": 0.8165,
+      "step": 140
+    },
+    {
+      "epoch": 0.057753393011839446,
+      "grad_norm": 0.3014599084854126,
+      "learning_rate": 9.980410816696317e-06,
+      "loss": 0.7983,
+      "step": 150
+    },
+    {
+      "epoch": 0.06160361921262874,
+      "grad_norm": 0.2614928185939789,
+      "learning_rate": 9.97450916312541e-06,
+      "loss": 0.807,
+      "step": 160
+    },
+    {
+      "epoch": 0.06545384541341803,
+      "grad_norm": 0.2369014471769333,
+      "learning_rate": 9.967833782084004e-06,
+      "loss": 0.8091,
+      "step": 170
+    },
+    {
+      "epoch": 0.06930407161420733,
+      "grad_norm": 0.27050548791885376,
+      "learning_rate": 9.960385711850554e-06,
+      "loss": 0.8331,
+      "step": 180
+    },
+    {
+      "epoch": 0.07315429781499663,
+      "grad_norm": 0.2760695219039917,
+      "learning_rate": 9.952166110886394e-06,
+      "loss": 0.825,
+      "step": 190
+    },
+    {
+      "epoch": 0.07700452401578593,
+      "grad_norm": 0.27851760387420654,
+      "learning_rate": 9.943176257655567e-06,
+      "loss": 0.8126,
+      "step": 200
+    },
+    {
+      "epoch": 0.08085475021657522,
+      "grad_norm": 0.26555657386779785,
+      "learning_rate": 9.933417550425952e-06,
+      "loss": 0.8121,
+      "step": 210
+    },
+    {
+      "epoch": 0.08470497641736452,
+      "grad_norm": 0.2724052369594574,
+      "learning_rate": 9.9228915070518e-06,
+      "loss": 0.7928,
+      "step": 220
+    },
+    {
+      "epoch": 0.08855520261815382,
+      "grad_norm": 0.24958202242851257,
+      "learning_rate": 9.911599764737632e-06,
+      "loss": 0.7954,
+      "step": 230
+    },
+    {
+      "epoch": 0.09240542881894312,
+      "grad_norm": 0.25107431411743164,
+      "learning_rate": 9.899544079783606e-06,
+      "loss": 0.79,
+      "step": 240
+    },
+    {
+      "epoch": 0.0962556550197324,
+      "grad_norm": 0.2952631711959839,
+      "learning_rate": 9.886726327312338e-06,
+      "loss": 0.8082,
+      "step": 250
+    },
+    {
+      "epoch": 0.1001058812205217,
+      "grad_norm": 0.246932253241539,
+      "learning_rate": 9.87314850097724e-06,
+      "loss": 0.8013,
+      "step": 260
+    },
+    {
+      "epoch": 0.103956107421311,
+      "grad_norm": 0.24760781228542328,
+      "learning_rate": 9.858812712652452e-06,
+      "loss": 0.7903,
+      "step": 270
+    },
+    {
+      "epoch": 0.10780633362210029,
+      "grad_norm": 0.25343242287635803,
+      "learning_rate": 9.843721192104341e-06,
+      "loss": 0.7899,
+      "step": 280
+    },
+    {
+      "epoch": 0.11165655982288959,
+      "grad_norm": 0.2506624162197113,
+      "learning_rate": 9.827876286644698e-06,
+      "loss": 0.8082,
+      "step": 290
+    },
+    {
+      "epoch": 0.11550678602367889,
+      "grad_norm": 0.2377500981092453,
+      "learning_rate": 9.811280460765643e-06,
+      "loss": 0.7711,
+      "step": 300
+    },
+    {
+      "epoch": 0.11935701222446819,
+      "grad_norm": 0.23943974077701569,
+      "learning_rate": 9.793936295756292e-06,
+      "loss": 0.7808,
+      "step": 310
+    },
+    {
+      "epoch": 0.12320723842525748,
+      "grad_norm": 0.2553127110004425,
+      "learning_rate": 9.775846489301279e-06,
+      "loss": 0.7859,
+      "step": 320
+    },
+    {
+      "epoch": 0.12705746462604678,
+      "grad_norm": 0.2602199912071228,
+      "learning_rate": 9.757013855061157e-06,
+      "loss": 0.7835,
+      "step": 330
+    },
+    {
+      "epoch": 0.13090769082683607,
+      "grad_norm": 0.2433108389377594,
+      "learning_rate": 9.737441322234758e-06,
+      "loss": 0.7937,
+      "step": 340
+    },
+    {
+      "epoch": 0.13475791702762538,
+      "grad_norm": 0.2605005204677582,
+      "learning_rate": 9.717131935103605e-06,
+      "loss": 0.7972,
+      "step": 350
+    },
+    {
+      "epoch": 0.13860814322841467,
+      "grad_norm": 0.25796180963516235,
+      "learning_rate": 9.696088852558398e-06,
+      "loss": 0.7824,
+      "step": 360
+    },
+    {
+      "epoch": 0.14245836942920398,
+      "grad_norm": 0.25043147802352905,
+      "learning_rate": 9.674315347607686e-06,
+      "loss": 0.8108,
+      "step": 370
+    },
+    {
+      "epoch": 0.14630859562999327,
+      "grad_norm": 0.2662828266620636,
+      "learning_rate": 9.651814806868796e-06,
+      "loss": 0.7704,
+      "step": 380
+    },
+    {
+      "epoch": 0.15015882183078255,
+      "grad_norm": 0.24600274860858917,
+      "learning_rate": 9.62859073004107e-06,
+      "loss": 0.7735,
+      "step": 390
+    },
+    {
+      "epoch": 0.15400904803157187,
+      "grad_norm": 0.2365168035030365,
+      "learning_rate": 9.604646729361547e-06,
+      "loss": 0.7729,
+      "step": 400
+    },
+    {
+      "epoch": 0.15785927423236115,
+      "grad_norm": 0.24804677069187164,
+      "learning_rate": 9.579986529043099e-06,
+      "loss": 0.7625,
+      "step": 410
+    },
+    {
+      "epoch": 0.16170950043315044,
+      "grad_norm": 0.25442326068878174,
+      "learning_rate": 9.554613964695189e-06,
+      "loss": 0.7853,
+      "step": 420
+    },
+    {
+      "epoch": 0.16555972663393975,
+      "grad_norm": 0.24929139018058777,
+      "learning_rate": 9.528532982727281e-06,
+      "loss": 0.7775,
+      "step": 430
+    },
+    {
+      "epoch": 0.16940995283472904,
+      "grad_norm": 0.2469961941242218,
+      "learning_rate": 9.501747639735018e-06,
+      "loss": 0.7705,
+      "step": 440
+    },
+    {
+      "epoch": 0.17326017903551832,
+      "grad_norm": 0.2623428404331207,
+      "learning_rate": 9.474262101869272e-06,
+      "loss": 0.7605,
+      "step": 450
+    },
+    {
+      "epoch": 0.17711040523630764,
+      "grad_norm": 0.2498980611562729,
+      "learning_rate": 9.446080644188141e-06,
+      "loss": 0.7679,
+      "step": 460
+    },
+    {
+      "epoch": 0.18096063143709692,
+      "grad_norm": 0.2503974139690399,
+      "learning_rate": 9.417207649992017e-06,
+      "loss": 0.7718,
+      "step": 470
+    },
+    {
+      "epoch": 0.18481085763788624,
+      "grad_norm": 0.252631276845932,
+      "learning_rate": 9.387647610141816e-06,
+      "loss": 0.7699,
+      "step": 480
+    },
+    {
+      "epoch": 0.18866108383867553,
+      "grad_norm": 0.23665153980255127,
+      "learning_rate": 9.35740512236047e-06,
+      "loss": 0.7601,
+      "step": 490
+    },
+    {
+      "epoch": 0.1925113100394648,
+      "grad_norm": 0.2414950430393219,
+      "learning_rate": 9.326484890517809e-06,
+      "loss": 0.7675,
+      "step": 500
+    },
+    {
+      "epoch": 0.19636153624025413,
+      "grad_norm": 0.24909065663814545,
+      "learning_rate": 9.294891723898926e-06,
+      "loss": 0.7586,
+      "step": 510
+    },
+    {
+      "epoch": 0.2002117624410434,
+      "grad_norm": 0.25251463055610657,
+      "learning_rate": 9.262630536456153e-06,
+      "loss": 0.7894,
+      "step": 520
+    },
+    {
+      "epoch": 0.2040619886418327,
+      "grad_norm": 0.2575061023235321,
+      "learning_rate": 9.229706346044749e-06,
+      "loss": 0.7617,
+      "step": 530
+    },
+    {
+      "epoch": 0.207912214842622,
+      "grad_norm": 0.26502957940101624,
+      "learning_rate": 9.196124273642422e-06,
+      "loss": 0.7732,
+      "step": 540
+    },
+    {
+      "epoch": 0.2117624410434113,
+      "grad_norm": 0.23806829750537872,
+      "learning_rate": 9.161889542552842e-06,
+      "loss": 0.7807,
+      "step": 550
+    },
+    {
+      "epoch": 0.21561266724420058,
+      "grad_norm": 0.2305864691734314,
+      "learning_rate": 9.127007477593193e-06,
+      "loss": 0.7791,
+      "step": 560
+    },
+    {
+      "epoch": 0.2194628934449899,
+      "grad_norm": 0.26431789994239807,
+      "learning_rate": 9.091483504265974e-06,
+      "loss": 0.7835,
+      "step": 570
+    },
+    {
+      "epoch": 0.22331311964577918,
+      "grad_norm": 0.2578928470611572,
+      "learning_rate": 9.055323147915116e-06,
+      "loss": 0.7678,
+      "step": 580
+    },
+    {
+      "epoch": 0.2271633458465685,
+      "grad_norm": 0.2555224895477295,
+      "learning_rate": 9.018532032866585e-06,
+      "loss": 0.7753,
+      "step": 590
+    },
+    {
+      "epoch": 0.23101357204735778,
+      "grad_norm": 0.23025396466255188,
+      "learning_rate": 8.981115881553579e-06,
+      "loss": 0.7688,
+      "step": 600
+    },
+    {
+      "epoch": 0.23486379824814707,
+      "grad_norm": 0.2668067216873169,
+      "learning_rate": 8.943080513626476e-06,
+      "loss": 0.788,
+      "step": 610
+    },
+    {
+      "epoch": 0.23871402444893638,
+      "grad_norm": 0.24164004623889923,
+      "learning_rate": 8.904431845047641e-06,
+      "loss": 0.7744,
+      "step": 620
+    },
+    {
+      "epoch": 0.24256425064972567,
+      "grad_norm": 0.2558857798576355,
+      "learning_rate": 8.86517588717128e-06,
+      "loss": 0.7887,
+      "step": 630
+    },
+    {
+      "epoch": 0.24641447685051496,
+      "grad_norm": 0.2417294830083847,
+      "learning_rate": 8.82531874580844e-06,
+      "loss": 0.7847,
+      "step": 640
+    },
+    {
+      "epoch": 0.25026470305130427,
+      "grad_norm": 0.23556189239025116,
+      "learning_rate": 8.784866620277314e-06,
+      "loss": 0.7384,
+      "step": 650
+    },
+    {
+      "epoch": 0.25411492925209356,
+      "grad_norm": 0.25628191232681274,
+      "learning_rate": 8.743825802439018e-06,
+      "loss": 0.769,
+      "step": 660
+    },
+    {
+      "epoch": 0.25796515545288284,
+      "grad_norm": 0.25769975781440735,
+      "learning_rate": 8.702202675718956e-06,
+      "loss": 0.781,
+      "step": 670
+    },
+    {
+      "epoch": 0.26181538165367213,
+      "grad_norm": 0.2402503490447998,
+      "learning_rate": 8.660003714113959e-06,
+      "loss": 0.7465,
+      "step": 680
+    },
+    {
+      "epoch": 0.26566560785446147,
+      "grad_norm": 0.23845496773719788,
+      "learning_rate": 8.617235481185323e-06,
+      "loss": 0.7705,
+      "step": 690
+    },
+    {
+      "epoch": 0.26951583405525076,
+      "grad_norm": 0.25551125407218933,
+      "learning_rate": 8.573904629037931e-06,
+      "loss": 0.7778,
+      "step": 700
+    },
+    {
+      "epoch": 0.27336606025604004,
+      "grad_norm": 0.2853063941001892,
+      "learning_rate": 8.530017897285585e-06,
+      "loss": 0.7606,
+      "step": 710
+    },
+    {
+      "epoch": 0.27721628645682933,
+      "grad_norm": 0.22575555741786957,
+      "learning_rate": 8.485582112002746e-06,
+      "loss": 0.7662,
+      "step": 720
+    },
+    {
+      "epoch": 0.2810665126576186,
+      "grad_norm": 0.2497359961271286,
+      "learning_rate": 8.440604184662811e-06,
+      "loss": 0.7671,
+      "step": 730
+    },
+    {
+      "epoch": 0.28491673885840796,
+      "grad_norm": 0.23865318298339844,
+      "learning_rate": 8.395091111063118e-06,
+      "loss": 0.7544,
+      "step": 740
+    },
+    {
+      "epoch": 0.28876696505919724,
+      "grad_norm": 0.2358783483505249,
+      "learning_rate": 8.349049970236822e-06,
+      "loss": 0.756,
+      "step": 750
+    },
+    {
+      "epoch": 0.29261719125998653,
+      "grad_norm": 0.23552630841732025,
+      "learning_rate": 8.302487923351842e-06,
+      "loss": 0.7482,
+      "step": 760
+    },
+    {
+      "epoch": 0.2964674174607758,
+      "grad_norm": 0.23091401159763336,
+      "learning_rate": 8.255412212597021e-06,
+      "loss": 0.7681,
+      "step": 770
+    },
+    {
+      "epoch": 0.3003176436615651,
+      "grad_norm": 0.23197367787361145,
+      "learning_rate": 8.207830160055688e-06,
+      "loss": 0.7647,
+      "step": 780
+    },
+    {
+      "epoch": 0.3041678698623544,
+      "grad_norm": 0.2511676251888275,
+      "learning_rate": 8.159749166566786e-06,
+      "loss": 0.7469,
+      "step": 790
+    },
+    {
+      "epoch": 0.30801809606314373,
+      "grad_norm": 0.2389538288116455,
+      "learning_rate": 8.11117671057377e-06,
+      "loss": 0.7733,
+      "step": 800
+    },
+    {
+      "epoch": 0.311868322263933,
+      "grad_norm": 0.24781665205955505,
+      "learning_rate": 8.062120346961404e-06,
+      "loss": 0.7627,
+      "step": 810
+    },
+    {
+      "epoch": 0.3157185484647223,
+      "grad_norm": 0.23915305733680725,
+      "learning_rate": 8.012587705880705e-06,
+      "loss": 0.7576,
+      "step": 820
+    },
+    {
+      "epoch": 0.3195687746655116,
+      "grad_norm": 0.24319690465927124,
+      "learning_rate": 7.962586491562146e-06,
+      "loss": 0.7453,
+      "step": 830
+    },
+    {
+      "epoch": 0.3234190008663009,
+      "grad_norm": 0.24663855135440826,
+      "learning_rate": 7.912124481117357e-06,
+      "loss": 0.7511,
+      "step": 840
+    },
+    {
+      "epoch": 0.3272692270670902,
+      "grad_norm": 0.24370235204696655,
+      "learning_rate": 7.861209523329494e-06,
+      "loss": 0.7656,
+      "step": 850
+    },
+    {
+      "epoch": 0.3311194532678795,
+      "grad_norm": 0.23307128250598907,
+      "learning_rate": 7.809849537432432e-06,
+      "loss": 0.7447,
+      "step": 860
+    },
+    {
+      "epoch": 0.3349696794686688,
+      "grad_norm": 0.25284966826438904,
+      "learning_rate": 7.758052511879035e-06,
+      "loss": 0.7559,
+      "step": 870
+    },
+    {
+      "epoch": 0.3388199056694581,
+      "grad_norm": 0.23169557750225067,
+      "learning_rate": 7.705826503098642e-06,
+      "loss": 0.7503,
+      "step": 880
+    },
+    {
+      "epoch": 0.34267013187024736,
+      "grad_norm": 0.23001490533351898,
+      "learning_rate": 7.653179634243979e-06,
+      "loss": 0.7535,
+      "step": 890
+    },
+    {
+      "epoch": 0.34652035807103665,
+      "grad_norm": 0.21794059872627258,
+      "learning_rate": 7.600120093927696e-06,
+      "loss": 0.7542,
+      "step": 900
+    },
+    {
+      "epoch": 0.350370584271826,
+      "grad_norm": 0.24670086801052094,
+      "learning_rate": 7.546656134948729e-06,
+      "loss": 0.743,
+      "step": 910
+    },
+    {
+      "epoch": 0.3542208104726153,
+      "grad_norm": 0.24960315227508545,
+      "learning_rate": 7.492796073008661e-06,
+      "loss": 0.7456,
+      "step": 920
+    },
+    {
+      "epoch": 0.35807103667340456,
+      "grad_norm": 0.2622990310192108,
+      "learning_rate": 7.438548285418326e-06,
+      "loss": 0.7313,
+      "step": 930
+    },
+    {
+      "epoch": 0.36192126287419385,
+      "grad_norm": 0.2569071352481842,
+      "learning_rate": 7.3839212097948e-06,
+      "loss": 0.766,
+      "step": 940
+    },
+    {
+      "epoch": 0.36577148907498314,
+      "grad_norm": 0.24257022142410278,
+      "learning_rate": 7.328923342749039e-06,
+      "loss": 0.7841,
+      "step": 950
+    },
+    {
+      "epoch": 0.3696217152757725,
+      "grad_norm": 0.25453999638557434,
+      "learning_rate": 7.273563238564319e-06,
+      "loss": 0.766,
+      "step": 960
+    },
+    {
+      "epoch": 0.37347194147656176,
+      "grad_norm": 0.2482769936323166,
+      "learning_rate": 7.217849507865724e-06,
+      "loss": 0.7423,
+      "step": 970
+    },
+    {
+      "epoch": 0.37732216767735105,
+      "grad_norm": 0.24652163684368134,
+      "learning_rate": 7.161790816280858e-06,
+      "loss": 0.7533,
+      "step": 980
+    },
+    {
+      "epoch": 0.38117239387814034,
+      "grad_norm": 0.25084182620048523,
+      "learning_rate": 7.1053958830920125e-06,
+      "loss": 0.7312,
+      "step": 990
+    },
+    {
+      "epoch": 0.3850226200789296,
+      "grad_norm": 0.24931836128234863,
+      "learning_rate": 7.0486734798799715e-06,
+      "loss": 0.7548,
+      "step": 1000
+    },
+    {
+      "epoch": 0.3888728462797189,
+      "grad_norm": 0.23999910056591034,
+      "learning_rate": 6.991632429159708e-06,
+      "loss": 0.7783,
+      "step": 1010
+    },
+    {
+      "epoch": 0.39272307248050825,
+      "grad_norm": 0.2649567723274231,
+      "learning_rate": 6.934281603008132e-06,
+      "loss": 0.7652,
+      "step": 1020
+    },
+    {
+      "epoch": 0.39657329868129754,
+      "grad_norm": 0.24615578353405,
+      "learning_rate": 6.876629921684143e-06,
+      "loss": 0.7637,
+      "step": 1030
+    },
+    {
+      "epoch": 0.4004235248820868,
+      "grad_norm": 0.24031710624694824,
+      "learning_rate": 6.818686352241199e-06,
+      "loss": 0.7532,
+      "step": 1040
+    },
+    {
+      "epoch": 0.4042737510828761,
+      "grad_norm": 0.2218167930841446,
+      "learning_rate": 6.760459907132573e-06,
+      "loss": 0.7498,
+      "step": 1050
+    },
+    {
+      "epoch": 0.4081239772836654,
+      "grad_norm": 0.2315954715013504,
+      "learning_rate": 6.701959642809589e-06,
+      "loss": 0.7546,
+      "step": 1060
+    },
+    {
+      "epoch": 0.41197420348445474,
+      "grad_norm": 0.24924255907535553,
+      "learning_rate": 6.643194658312987e-06,
+      "loss": 0.7451,
+      "step": 1070
+    },
+    {
+      "epoch": 0.415824429685244,
+      "grad_norm": 0.22406519949436188,
+      "learning_rate": 6.584174093857676e-06,
+      "loss": 0.7721,
+      "step": 1080
+    },
+    {
+      "epoch": 0.4196746558860333,
+      "grad_norm": 0.2745705842971802,
+      "learning_rate": 6.5249071294110735e-06,
+      "loss": 0.7531,
+      "step": 1090
+    },
+    {
+      "epoch": 0.4235248820868226,
+      "grad_norm": 0.25875258445739746,
+      "learning_rate": 6.465402983265274e-06,
+      "loss": 0.7672,
+      "step": 1100
+    },
+    {
+      "epoch": 0.4273751082876119,
+      "grad_norm": 0.24413785338401794,
+      "learning_rate": 6.405670910603243e-06,
+      "loss": 0.7614,
+      "step": 1110
+    },
+    {
+      "epoch": 0.43122533448840117,
+      "grad_norm": 0.2499418556690216,
+      "learning_rate": 6.345720202059291e-06,
+      "loss": 0.7587,
+      "step": 1120
+    },
+    {
+      "epoch": 0.4350755606891905,
+      "grad_norm": 0.24239076673984528,
+      "learning_rate": 6.285560182274008e-06,
+      "loss": 0.7363,
+      "step": 1130
+    },
+    {
+      "epoch": 0.4389257868899798,
+      "grad_norm": 0.2443019300699234,
+      "learning_rate": 6.225200208443936e-06,
+      "loss": 0.7615,
+      "step": 1140
+    },
+    {
+      "epoch": 0.4427760130907691,
+      "grad_norm": 0.22823528945446014,
+      "learning_rate": 6.16464966886616e-06,
+      "loss": 0.7476,
+      "step": 1150
+    },
+    {
+      "epoch": 0.44662623929155837,
+      "grad_norm": 0.27869200706481934,
+      "learning_rate": 6.103917981478065e-06,
+      "loss": 0.7734,
+      "step": 1160
+    },
+    {
+      "epoch": 0.45047646549234766,
+      "grad_norm": 0.2379787266254425,
+      "learning_rate": 6.043014592392489e-06,
+      "loss": 0.7817,
+      "step": 1170
+    },
+    {
+      "epoch": 0.454326691693137,
+      "grad_norm": 0.22936110198497772,
+      "learning_rate": 5.981948974428485e-06,
+      "loss": 0.7591,
+      "step": 1180
+    },
+    {
+      "epoch": 0.4581769178939263,
+      "grad_norm": 0.25864309072494507,
+      "learning_rate": 5.920730625637934e-06,
+      "loss": 0.7344,
+      "step": 1190
+    },
+    {
+      "epoch": 0.46202714409471557,
+      "grad_norm": 0.25833386182785034,
+      "learning_rate": 5.8593690678282355e-06,
+      "loss": 0.7497,
+      "step": 1200
+    },
+    {
+      "epoch": 0.46587737029550486,
+      "grad_norm": 0.24733944237232208,
+      "learning_rate": 5.797873845081301e-06,
+      "loss": 0.7534,
+      "step": 1210
+    },
+    {
+      "epoch": 0.46972759649629414,
+      "grad_norm": 0.24075521528720856,
+      "learning_rate": 5.736254522269075e-06,
+      "loss": 0.7687,
+      "step": 1220
+    },
+    {
+      "epoch": 0.47357782269708343,
+      "grad_norm": 0.24509303271770477,
+      "learning_rate": 5.6745206835658386e-06,
+      "loss": 0.771,
+      "step": 1230
+    },
+    {
+      "epoch": 0.47742804889787277,
+      "grad_norm": 0.23488254845142365,
+      "learning_rate": 5.612681930957495e-06,
+      "loss": 0.7417,
+      "step": 1240
+    },
+    {
+      "epoch": 0.48127827509866206,
+      "grad_norm": 0.2336212545633316,
+      "learning_rate": 5.550747882748095e-06,
+      "loss": 0.7368,
+      "step": 1250
+    },
+    {
+      "epoch": 0.48512850129945134,
+      "grad_norm": 0.25603893399238586,
+      "learning_rate": 5.4887281720638195e-06,
+      "loss": 0.749,
+      "step": 1260
+    },
+    {
+      "epoch": 0.48897872750024063,
+      "grad_norm": 0.26326969265937805,
+      "learning_rate": 5.42663244535466e-06,
+      "loss": 0.7312,
+      "step": 1270
+    },
+    {
+      "epoch": 0.4928289537010299,
+      "grad_norm": 0.25515881180763245,
+      "learning_rate": 5.3644703608940205e-06,
+      "loss": 0.7404,
+      "step": 1280
+    },
+    {
+      "epoch": 0.49667917990181926,
+      "grad_norm": 0.24312014877796173,
+      "learning_rate": 5.302251587276491e-06,
+      "loss": 0.7507,
+      "step": 1290
+    },
+    {
+      "epoch": 0.5005294061026085,
+      "grad_norm": 0.2301429659128189,
+      "learning_rate": 5.2399858019140005e-06,
+      "loss": 0.7514,
+      "step": 1300
+    },
+    {
+      "epoch": 0.5043796323033978,
+      "grad_norm": 0.2335745096206665,
+      "learning_rate": 5.177682689530615e-06,
+      "loss": 0.7683,
+      "step": 1310
+    },
+    {
+      "epoch": 0.5082298585041871,
+      "grad_norm": 0.25138232111930847,
+      "learning_rate": 5.11535194065619e-06,
+      "loss": 0.7323,
+      "step": 1320
+    },
+    {
+      "epoch": 0.5120800847049765,
+      "grad_norm": 0.25443217158317566,
+      "learning_rate": 5.053003250119113e-06,
+      "loss": 0.7405,
+      "step": 1330
+    },
+    {
+      "epoch": 0.5159303109057657,
+      "grad_norm": 0.246122807264328,
+      "learning_rate": 4.9906463155383925e-06,
+      "loss": 0.7527,
+      "step": 1340
+    },
+    {
+      "epoch": 0.519780537106555,
+      "grad_norm": 0.23790699243545532,
+      "learning_rate": 4.9282908358153046e-06,
+      "loss": 0.7476,
+      "step": 1350
+    },
+    {
+      "epoch": 0.5236307633073443,
+      "grad_norm": 0.26790881156921387,
+      "learning_rate": 4.8659465096248345e-06,
+      "loss": 0.7583,
+      "step": 1360
+    },
+    {
+      "epoch": 0.5274809895081336,
+      "grad_norm": 0.24705077707767487,
+      "learning_rate": 4.803623033907169e-06,
+      "loss": 0.7571,
+      "step": 1370
+    },
+    {
+      "epoch": 0.5313312157089229,
+      "grad_norm": 0.2280808836221695,
+      "learning_rate": 4.741330102359441e-06,
+      "loss": 0.766,
+      "step": 1380
+    },
+    {
+      "epoch": 0.5351814419097122,
+      "grad_norm": 0.259689599275589,
+      "learning_rate": 4.679077403927996e-06,
+      "loss": 0.7448,
+      "step": 1390
+    },
+    {
+      "epoch": 0.5390316681105015,
+      "grad_norm": 0.23548489809036255,
+      "learning_rate": 4.616874621301373e-06,
+      "loss": 0.746,
+      "step": 1400
+    },
+    {
+      "epoch": 0.5428818943112907,
+      "grad_norm": 0.24700024724006653,
+      "learning_rate": 4.554731429404293e-06,
+      "loss": 0.761,
+      "step": 1410
+    },
+    {
+      "epoch": 0.5467321205120801,
+      "grad_norm": 0.25937360525131226,
+      "learning_rate": 4.492657493892822e-06,
+      "loss": 0.7414,
+      "step": 1420
+    },
+    {
+      "epoch": 0.5505823467128694,
+      "grad_norm": 0.27232789993286133,
+      "learning_rate": 4.430662469650999e-06,
+      "loss": 0.7581,
+      "step": 1430
+    },
+    {
+      "epoch": 0.5544325729136587,
+      "grad_norm": 0.26728764176368713,
+      "learning_rate": 4.368755999289118e-06,
+      "loss": 0.7645,
+      "step": 1440
+    },
+    {
+      "epoch": 0.558282799114448,
+      "grad_norm": 0.2662217319011688,
+      "learning_rate": 4.306947711643949e-06,
+      "loss": 0.7464,
+      "step": 1450
+    },
+    {
+      "epoch": 0.5621330253152372,
+      "grad_norm": 0.24922913312911987,
+      "learning_rate": 4.245247220281068e-06,
+      "loss": 0.7507,
+      "step": 1460
+    },
+    {
+      "epoch": 0.5659832515160266,
+      "grad_norm": 0.23917795717716217,
+      "learning_rate": 4.183664121999587e-06,
+      "loss": 0.7446,
+      "step": 1470
+    },
+    {
+      "epoch": 0.5698334777168159,
+      "grad_norm": 0.2477523535490036,
+      "learning_rate": 4.122207995339474e-06,
+      "loss": 0.7436,
+      "step": 1480
+    },
+    {
+      "epoch": 0.5736837039176051,
+      "grad_norm": 0.22891761362552643,
+      "learning_rate": 4.060888399091734e-06,
+      "loss": 0.7478,
+      "step": 1490
+    },
+    {
+      "epoch": 0.5775339301183945,
+      "grad_norm": 0.23203203082084656,
+      "learning_rate": 3.999714870811639e-06,
+      "loss": 0.7599,
+      "step": 1500
+    },
+    {
+      "epoch": 0.5813841563191837,
+      "grad_norm": 0.23997482657432556,
+      "learning_rate": 3.938696925335275e-06,
+      "loss": 0.7511,
+      "step": 1510
+    },
+    {
+      "epoch": 0.5852343825199731,
+      "grad_norm": 0.24647574126720428,
+      "learning_rate": 3.87784405329962e-06,
+      "loss": 0.7593,
+      "step": 1520
+    },
+    {
+      "epoch": 0.5890846087207623,
+      "grad_norm": 0.23370209336280823,
+      "learning_rate": 3.817165719666382e-06,
+      "loss": 0.7593,
+      "step": 1530
+    },
+    {
+      "epoch": 0.5929348349215516,
+      "grad_norm": 0.24609887599945068,
+      "learning_rate": 3.756671362249841e-06,
+      "loss": 0.7664,
+      "step": 1540
+    },
+    {
+      "epoch": 0.596785061122341,
+      "grad_norm": 0.23501643538475037,
+      "learning_rate": 3.6963703902488907e-06,
+      "loss": 0.7501,
+      "step": 1550
+    },
+    {
+      "epoch": 0.6006352873231302,
+      "grad_norm": 0.23786117136478424,
+      "learning_rate": 3.6362721827835655e-06,
+      "loss": 0.7647,
+      "step": 1560
+    },
+    {
+      "epoch": 0.6044855135239195,
+      "grad_norm": 0.24694815278053284,
+      "learning_rate": 3.576386087436212e-06,
+      "loss": 0.7625,
+      "step": 1570
+    },
+    {
+      "epoch": 0.6083357397247088,
+      "grad_norm": 0.24815331399440765,
+      "learning_rate": 3.5167214187975913e-06,
+      "loss": 0.7532,
+      "step": 1580
+    },
+    {
+      "epoch": 0.6121859659254981,
+      "grad_norm": 0.2139756679534912,
+      "learning_rate": 3.457287457018086e-06,
+      "loss": 0.7544,
+      "step": 1590
+    },
+    {
+      "epoch": 0.6160361921262875,
+      "grad_norm": 0.2364327609539032,
+      "learning_rate": 3.3980934463643003e-06,
+      "loss": 0.7447,
+      "step": 1600
+    },
+    {
+      "epoch": 0.6198864183270767,
+      "grad_norm": 0.2212245613336563,
+      "learning_rate": 3.3391485937812053e-06,
+      "loss": 0.7253,
+      "step": 1610
+    },
+    {
+      "epoch": 0.623736644527866,
+      "grad_norm": 0.25001060962677,
+      "learning_rate": 3.280462067460115e-06,
+      "loss": 0.7249,
+      "step": 1620
+    },
+    {
+      "epoch": 0.6275868707286553,
+      "grad_norm": 0.2347634732723236,
+      "learning_rate": 3.222042995412669e-06,
+      "loss": 0.7557,
+      "step": 1630
+    },
+    {
+      "epoch": 0.6314370969294446,
+      "grad_norm": 0.22897182404994965,
+      "learning_rate": 3.163900464051089e-06,
+      "loss": 0.744,
+      "step": 1640
+    },
+    {
+      "epoch": 0.635287323130234,
+      "grad_norm": 0.24808236956596375,
+      "learning_rate": 3.1060435167748826e-06,
+      "loss": 0.7547,
+      "step": 1650
+    },
+    {
+      "epoch": 0.6391375493310232,
+      "grad_norm": 0.23447702825069427,
+      "learning_rate": 3.048481152564253e-06,
+      "loss": 0.7361,
+      "step": 1660
+    },
+    {
+      "epoch": 0.6429877755318125,
+      "grad_norm": 0.25827962160110474,
+      "learning_rate": 2.9912223245804017e-06,
+      "loss": 0.7611,
+      "step": 1670
+    },
+    {
+      "epoch": 0.6468380017326018,
+      "grad_norm": 0.24561026692390442,
+      "learning_rate": 2.934275938772981e-06,
+      "loss": 0.7514,
+      "step": 1680
+    },
+    {
+      "epoch": 0.6506882279333911,
+      "grad_norm": 0.2679162621498108,
+      "learning_rate": 2.8776508524948676e-06,
+      "loss": 0.7437,
+      "step": 1690
+    },
+    {
+      "epoch": 0.6545384541341804,
+      "grad_norm": 0.21846570074558258,
+      "learning_rate": 2.821355873124503e-06,
+      "loss": 0.7392,
+      "step": 1700
+    },
+    {
+      "epoch": 0.6583886803349697,
+      "grad_norm": 0.251023530960083,
+      "learning_rate": 2.7653997566960144e-06,
+      "loss": 0.7578,
+      "step": 1710
+    },
+    {
+      "epoch": 0.662238906535759,
+      "grad_norm": 0.25164997577667236,
+      "learning_rate": 2.709791206537309e-06,
+      "loss": 0.7468,
+      "step": 1720
+    },
+    {
+      "epoch": 0.6660891327365482,
+      "grad_norm": 0.2406506985425949,
+      "learning_rate": 2.6545388719163747e-06,
+      "loss": 0.7476,
+      "step": 1730
+    },
+    {
+      "epoch": 0.6699393589373376,
+      "grad_norm": 0.22837968170642853,
+      "learning_rate": 2.599651346695979e-06,
+      "loss": 0.7492,
+      "step": 1740
+    },
+    {
+      "epoch": 0.6737895851381268,
+      "grad_norm": 0.2502683401107788,
+      "learning_rate": 2.545137167997007e-06,
+      "loss": 0.7472,
+      "step": 1750
+    },
+    {
+      "epoch": 0.6776398113389162,
+      "grad_norm": 0.24016062915325165,
+      "learning_rate": 2.4910048148705972e-06,
+      "loss": 0.7322,
+      "step": 1760
+    },
+    {
+      "epoch": 0.6814900375397055,
+      "grad_norm": 0.2580972909927368,
+      "learning_rate": 2.4372627069793327e-06,
+      "loss": 0.778,
+      "step": 1770
+    },
+    {
+      "epoch": 0.6853402637404947,
+      "grad_norm": 0.2363647073507309,
+      "learning_rate": 2.3839192032876486e-06,
+      "loss": 0.763,
+      "step": 1780
+    },
+    {
+      "epoch": 0.6891904899412841,
+      "grad_norm": 0.23583751916885376,
+      "learning_rate": 2.3309826007617047e-06,
+      "loss": 0.7738,
+      "step": 1790
+    },
+    {
+      "epoch": 0.6930407161420733,
+      "grad_norm": 0.2377644032239914,
+      "learning_rate": 2.278461133078881e-06,
+      "loss": 0.7376,
+      "step": 1800
+    },
+    {
+      "epoch": 0.6968909423428626,
+      "grad_norm": 0.24563556909561157,
+      "learning_rate": 2.2263629693471258e-06,
+      "loss": 0.7269,
+      "step": 1810
+    },
+    {
+      "epoch": 0.700741168543652,
+      "grad_norm": 0.23247992992401123,
+      "learning_rate": 2.1746962128343414e-06,
+      "loss": 0.7517,
+      "step": 1820
+    },
+    {
+      "epoch": 0.7045913947444412,
+      "grad_norm": 0.2280578911304474,
+      "learning_rate": 2.1234688997080183e-06,
+      "loss": 0.7423,
+      "step": 1830
+    },
+    {
+      "epoch": 0.7084416209452306,
+      "grad_norm": 0.2529107928276062,
+      "learning_rate": 2.0726889977853053e-06,
+      "loss": 0.7466,
+      "step": 1840
+    },
+    {
+      "epoch": 0.7122918471460198,
+      "grad_norm": 0.2564008831977844,
+      "learning_rate": 2.022364405293703e-06,
+      "loss": 0.7616,
+      "step": 1850
+    },
+    {
+      "epoch": 0.7161420733468091,
+      "grad_norm": 0.2368805855512619,
+      "learning_rate": 1.9725029496425814e-06,
+      "loss": 0.7524,
+      "step": 1860
+    },
+    {
+      "epoch": 0.7199922995475985,
+      "grad_norm": 0.24253065884113312,
+      "learning_rate": 1.923112386205727e-06,
+      "loss": 0.7485,
+      "step": 1870
+    },
+    {
+      "epoch": 0.7238425257483877,
+      "grad_norm": 0.235978901386261,
+      "learning_rate": 1.874200397115079e-06,
+      "loss": 0.7569,
+      "step": 1880
+    },
+    {
+      "epoch": 0.727692751949177,
+      "grad_norm": 0.2412380874156952,
+      "learning_rate": 1.825774590065857e-06,
+      "loss": 0.7405,
+      "step": 1890
+    },
+    {
+      "epoch": 0.7315429781499663,
+      "grad_norm": 0.24244432151317596,
+      "learning_rate": 1.7778424971332813e-06,
+      "loss": 0.7471,
+      "step": 1900
+    },
+    {
+      "epoch": 0.7353932043507556,
+      "grad_norm": 0.23394449055194855,
+      "learning_rate": 1.7304115736010407e-06,
+      "loss": 0.7444,
+      "step": 1910
+    },
+    {
+      "epoch": 0.739243430551545,
+      "grad_norm": 0.24016313254833221,
+      "learning_rate": 1.6834891968017108e-06,
+      "loss": 0.74,
+      "step": 1920
+    },
+    {
+      "epoch": 0.7430936567523342,
+      "grad_norm": 0.2479017972946167,
+      "learning_rate": 1.6370826649692884e-06,
+      "loss": 0.7542,
+      "step": 1930
+    },
+    {
+      "epoch": 0.7469438829531235,
+      "grad_norm": 0.2528059482574463,
+      "learning_rate": 1.5911991961040463e-06,
+      "loss": 0.7661,
+      "step": 1940
+    },
+    {
+      "epoch": 0.7507941091539128,
+      "grad_norm": 0.2536682188510895,
+      "learning_rate": 1.545845926849851e-06,
+      "loss": 0.7539,
+      "step": 1950
+    },
+    {
+      "epoch": 0.7546443353547021,
+      "grad_norm": 0.25885456800460815,
+      "learning_rate": 1.5010299113841397e-06,
+      "loss": 0.7697,
+      "step": 1960
+    },
+    {
+      "epoch": 0.7584945615554913,
+      "grad_norm": 0.24149109423160553,
+      "learning_rate": 1.4567581203207238e-06,
+      "loss": 0.7591,
+      "step": 1970
+    },
+    {
+      "epoch": 0.7623447877562807,
+      "grad_norm": 0.25305378437042236,
+      "learning_rate": 1.413037439625593e-06,
+      "loss": 0.7563,
+      "step": 1980
+    },
+    {
+      "epoch": 0.76619501395707,
+      "grad_norm": 0.2607440650463104,
+      "learning_rate": 1.3698746695458847e-06,
+      "loss": 0.7533,
+      "step": 1990
+    },
+    {
+      "epoch": 0.7700452401578592,
+      "grad_norm": 0.2451711893081665,
+      "learning_rate": 1.3272765235521745e-06,
+      "loss": 0.7437,
+      "step": 2000
+    },
+    {
+      "epoch": 0.7738954663586486,
+      "grad_norm": 0.24272949993610382,
+      "learning_rate": 1.2852496272942877e-06,
+      "loss": 0.7522,
+      "step": 2010
+    },
+    {
+      "epoch": 0.7777456925594378,
+      "grad_norm": 0.2323727309703827,
+      "learning_rate": 1.2438005175707396e-06,
+      "loss": 0.7372,
+      "step": 2020
+    },
+    {
+      "epoch": 0.7815959187602272,
+      "grad_norm": 0.2400728464126587,
+      "learning_rate": 1.2029356413120297e-06,
+      "loss": 0.7229,
+      "step": 2030
+    },
+    {
+      "epoch": 0.7854461449610165,
+      "grad_norm": 0.2413865625858307,
+      "learning_rate": 1.1626613545778786e-06,
+      "loss": 0.7413,
+      "step": 2040
+    },
+    {
+      "epoch": 0.7892963711618057,
+      "grad_norm": 0.23356689512729645,
+      "learning_rate": 1.1229839215686294e-06,
+      "loss": 0.746,
+      "step": 2050
+    },
+    {
+      "epoch": 0.7931465973625951,
+      "grad_norm": 0.22719332575798035,
+      "learning_rate": 1.0839095136509188e-06,
+      "loss": 0.7278,
+      "step": 2060
+    },
+    {
+      "epoch": 0.7969968235633843,
+      "grad_norm": 0.25770336389541626,
+      "learning_rate": 1.045444208397791e-06,
+      "loss": 0.75,
+      "step": 2070
+    },
+    {
+      "epoch": 0.8008470497641736,
+      "grad_norm": 0.23734548687934875,
+      "learning_rate": 1.007593988643402e-06,
+      "loss": 0.732,
+      "step": 2080
+    },
+    {
+      "epoch": 0.804697275964963,
+      "grad_norm": 0.23477159440517426,
+      "learning_rate": 9.703647415524613e-07,
+      "loss": 0.7557,
+      "step": 2090
+    },
+    {
+      "epoch": 0.8085475021657522,
+      "grad_norm": 0.23390445113182068,
+      "learning_rate": 9.337622577045485e-07,
+      "loss": 0.7425,
+      "step": 2100
+    },
+    {
+      "epoch": 0.8123977283665416,
+      "grad_norm": 0.23517288267612457,
+      "learning_rate": 8.977922301934588e-07,
+      "loss": 0.7557,
+      "step": 2110
+    },
+    {
+      "epoch": 0.8162479545673308,
+      "grad_norm": 0.2359190434217453,
+      "learning_rate": 8.624602537417009e-07,
+      "loss": 0.7576,
+      "step": 2120
+    },
+    {
+      "epoch": 0.8200981807681201,
+      "grad_norm": 0.24864567816257477,
+      "learning_rate": 8.277718238303123e-07,
+      "loss": 0.7431,
+      "step": 2130
+    },
+    {
+      "epoch": 0.8239484069689095,
+      "grad_norm": 0.2409396469593048,
+      "learning_rate": 7.937323358440935e-07,
+      "loss": 0.7456,
+      "step": 2140
+    },
+    {
+      "epoch": 0.8277986331696987,
+      "grad_norm": 0.2533905804157257,
+      "learning_rate": 7.603470842324223e-07,
+      "loss": 0.7384,
+      "step": 2150
+    },
+    {
+      "epoch": 0.831648859370488,
+      "grad_norm": 0.233335480093956,
+      "learning_rate": 7.276212616857575e-07,
+      "loss": 0.7563,
+      "step": 2160
+    },
+    {
+      "epoch": 0.8354990855712773,
+      "grad_norm": 0.23066791892051697,
+      "learning_rate": 6.955599583279837e-07,
+      "loss": 0.7532,
+      "step": 2170
+    },
+    {
+      "epoch": 0.8393493117720666,
+      "grad_norm": 0.2475164532661438,
+      "learning_rate": 6.641681609246981e-07,
+      "loss": 0.7473,
+      "step": 2180
+    },
+    {
+      "epoch": 0.8431995379728559,
+      "grad_norm": 0.2331661880016327,
+      "learning_rate": 6.334507521075711e-07,
+      "loss": 0.7514,
+      "step": 2190
+    },
+    {
+      "epoch": 0.8470497641736452,
+      "grad_norm": 0.25298836827278137,
+      "learning_rate": 6.034125096149179e-07,
+      "loss": 0.7486,
+      "step": 2200
+    },
+    {
+      "epoch": 0.8508999903744345,
+      "grad_norm": 0.24443817138671875,
+      "learning_rate": 5.740581055485722e-07,
+      "loss": 0.7197,
+      "step": 2210
+    },
+    {
+      "epoch": 0.8547502165752238,
+      "grad_norm": 0.22996845841407776,
+      "learning_rate": 5.453921056471961e-07,
+      "loss": 0.7541,
+      "step": 2220
+    },
+    {
+      "epoch": 0.8586004427760131,
+      "grad_norm": 0.2585048973560333,
+      "learning_rate": 5.174189685761288e-07,
+      "loss": 0.7473,
+      "step": 2230
+    },
+    {
+      "epoch": 0.8624506689768023,
+      "grad_norm": 0.23258310556411743,
+      "learning_rate": 4.901430452338979e-07,
+      "loss": 0.7614,
+      "step": 2240
+    },
+    {
+      "epoch": 0.8663008951775917,
+      "grad_norm": 0.2355554699897766,
+      "learning_rate": 4.6356857807548286e-07,
+      "loss": 0.7414,
+      "step": 2250
+    },
+    {
+      "epoch": 0.870151121378381,
+      "grad_norm": 0.2346854954957962,
+      "learning_rate": 4.376997004524541e-07,
+      "loss": 0.7306,
+      "step": 2260
+    },
+    {
+      "epoch": 0.8740013475791703,
+      "grad_norm": 0.2368471324443817,
+      "learning_rate": 4.12540435970073e-07,
+      "loss": 0.7388,
+      "step": 2270
+    },
+    {
+      "epoch": 0.8778515737799596,
+      "grad_norm": 0.23310524225234985,
+      "learning_rate": 3.880946978614703e-07,
+      "loss": 0.7553,
+      "step": 2280
+    },
+    {
+      "epoch": 0.8817017999807488,
+      "grad_norm": 0.2304801493883133,
+      "learning_rate": 3.643662883789878e-07,
+      "loss": 0.7531,
+      "step": 2290
+    },
+    {
+      "epoch": 0.8855520261815382,
+      "grad_norm": 0.22125546634197235,
+      "learning_rate": 3.41358898202781e-07,
+      "loss": 0.7329,
+      "step": 2300
+    },
+    {
+      "epoch": 0.8894022523823275,
+      "grad_norm": 0.22737379372119904,
+      "learning_rate": 3.190761058667741e-07,
+      "loss": 0.7429,
+      "step": 2310
+    },
+    {
+      "epoch": 0.8932524785831167,
+      "grad_norm": 0.23537009954452515,
+      "learning_rate": 2.9752137720206463e-07,
+      "loss": 0.7478,
+      "step": 2320
+    },
+    {
+      "epoch": 0.8971027047839061,
+      "grad_norm": 0.23352354764938354,
+      "learning_rate": 2.7669806479785067e-07,
+      "loss": 0.7556,
+      "step": 2330
+    },
+    {
+      "epoch": 0.9009529309846953,
+      "grad_norm": 0.2456805258989334,
+      "learning_rate": 2.566094074799763e-07,
+      "loss": 0.7472,
+      "step": 2340
+    },
+    {
+      "epoch": 0.9048031571854847,
+      "grad_norm": 0.23326189815998077,
+      "learning_rate": 2.3725852980716557e-07,
+      "loss": 0.7324,
+      "step": 2350
+    },
+    {
+      "epoch": 0.908653383386274,
+      "grad_norm": 0.232341006398201,
+      "learning_rate": 2.1864844158503983e-07,
+      "loss": 0.7427,
+      "step": 2360
+    },
+    {
+      "epoch": 0.9125036095870632,
+      "grad_norm": 0.2706652581691742,
+      "learning_rate": 2.007820373979702e-07,
+      "loss": 0.7346,
+      "step": 2370
+    },
+    {
+      "epoch": 0.9163538357878526,
+      "grad_norm": 0.24787262082099915,
+      "learning_rate": 1.8366209615886532e-07,
+      "loss": 0.7443,
+      "step": 2380
+    },
+    {
+      "epoch": 0.9202040619886418,
+      "grad_norm": 0.2356908917427063,
+      "learning_rate": 1.6729128067693613e-07,
+      "loss": 0.7596,
+      "step": 2390
+    },
+    {
+      "epoch": 0.9240542881894311,
+      "grad_norm": 0.25575998425483704,
+      "learning_rate": 1.5167213724353426e-07,
+      "loss": 0.758,
+      "step": 2400
+    },
+    {
+      "epoch": 0.9279045143902205,
+      "grad_norm": 0.23197045922279358,
+      "learning_rate": 1.3680709523610058e-07,
+      "loss": 0.7381,
+      "step": 2410
+    },
+    {
+      "epoch": 0.9317547405910097,
+      "grad_norm": 0.2328777313232422,
+      "learning_rate": 1.2269846674030418e-07,
+      "loss": 0.761,
+      "step": 2420
+    },
+    {
+      "epoch": 0.935604966791799,
+      "grad_norm": 0.24114975333213806,
+      "learning_rate": 1.0934844619042595e-07,
+      "loss": 0.7603,
+      "step": 2430
+    },
+    {
+      "epoch": 0.9394551929925883,
+      "grad_norm": 0.2477579414844513,
+      "learning_rate": 9.675911002803961e-08,
+      "loss": 0.7459,
+      "step": 2440
+    },
+    {
+      "epoch": 0.9433054191933776,
+      "grad_norm": 0.24621963500976562,
+      "learning_rate": 8.493241637904381e-08,
+      "loss": 0.7419,
+      "step": 2450
+    },
+    {
+      "epoch": 0.9471556453941669,
+      "grad_norm": 0.2435765415430069,
+      "learning_rate": 7.387020474909689e-08,
+      "loss": 0.7659,
+      "step": 2460
+    },
+    {
+      "epoch": 0.9510058715949562,
+      "grad_norm": 0.2520509958267212,
+      "learning_rate": 6.357419573750856e-08,
+      "loss": 0.7588,
+      "step": 2470
+    },
+    {
+      "epoch": 0.9548560977957455,
+      "grad_norm": 0.2387991100549698,
+      "learning_rate": 5.4045990769613365e-08,
+      "loss": 0.7409,
+      "step": 2480
+    },
+    {
+      "epoch": 0.9587063239965348,
+      "grad_norm": 0.28989431262016296,
+      "learning_rate": 4.5287071847694434e-08,
+      "loss": 0.7443,
+      "step": 2490
+    },
+    {
+      "epoch": 0.9625565501973241,
+      "grad_norm": 0.23961113393306732,
+      "learning_rate": 3.7298801320467817e-08,
+      "loss": 0.7415,
+      "step": 2500
+    },
+    {
+      "epoch": 0.9664067763981133,
+      "grad_norm": 0.24300412833690643,
+      "learning_rate": 3.0082421671192576e-08,
+      "loss": 0.7352,
+      "step": 2510
+    },
+    {
+      "epoch": 0.9702570025989027,
+      "grad_norm": 0.2423572540283203,
+      "learning_rate": 2.3639055324410355e-08,
+      "loss": 0.7597,
+      "step": 2520
+    },
+    {
+      "epoch": 0.974107228799692,
+      "grad_norm": 0.23211543262004852,
+      "learning_rate": 1.7969704471369477e-08,
+      "loss": 0.7401,
+      "step": 2530
+    },
+    {
+      "epoch": 0.9779574550004813,
+      "grad_norm": 0.22840356826782227,
+      "learning_rate": 1.3075250914142413e-08,
+      "loss": 0.7455,
+      "step": 2540
+    },
+    {
+      "epoch": 0.9818076812012706,
+      "grad_norm": 0.22685998678207397,
+      "learning_rate": 8.956455928476049e-09,
+      "loss": 0.7524,
+      "step": 2550
+    },
+    {
+      "epoch": 0.9856579074020598,
+      "grad_norm": 0.2342378944158554,
+      "learning_rate": 5.6139601453786275e-09,
+      "loss": 0.7424,
+      "step": 2560
+    },
+    {
+      "epoch": 0.9895081336028492,
+      "grad_norm": 0.22931507229804993,
+      "learning_rate": 3.0482834514805602e-09,
+      "loss": 0.7336,
+      "step": 2570
+    },
+    {
+      "epoch": 0.9933583598036385,
+      "grad_norm": 0.21343012154102325,
+      "learning_rate": 1.2598249081713366e-09,
+      "loss": 0.7253,
+      "step": 2580
+    },
+    {
+      "epoch": 0.9972085860044277,
+      "grad_norm": 0.25895360112190247,
+      "learning_rate": 2.488626895302826e-10,
+      "loss": 0.76,
+      "step": 2590
+    },
+    {
+      "epoch": 0.9999037443449803,
+      "step": 2597,
+      "total_flos": 3.790083715577676e+18,
+      "train_loss": 0.7778543212297195,
+      "train_runtime": 10716.5332,
+      "train_samples_per_second": 62.045,
+      "train_steps_per_second": 0.242
+    }
+  ],
+  "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": 3.790083715577676e+18,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}

continuous/qwen3/siglip2_b16_256/finetune/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4de36105a18eef8c120cad795f18a3ebcea6b8d9569cab63a6c950e5aea595a9
+size 6136

continuous/qwen3/siglip2_b16_256/pretrain/config.json

@@ -0,0 +1,60 @@
+{
+  "_attn_implementation_autoset": true,
+  "add_faster_video": false,
+  "add_time_instruction": false,
+  "architectures": [
+    "Qwen3ForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "bos_token_id": 151643,
+  "eos_token_id": 151645,
+  "faster_token_stride": 10,
+  "force_sample": false,
+  "head_dim": 128,
+  "hidden_act": "silu",
+  "hidden_size": 2048,
+  "image_aspect_ratio": "square",
+  "image_crop_resolution": null,
+  "image_grid_pinpoints": null,
+  "image_split_resolution": null,
+  "initializer_range": 0.02,
+  "intermediate_size": 6144,
+  "max_position_embeddings": 40960,
+  "max_window_layers": 28,
+  "mm_hidden_size": 768,
+  "mm_newline_position": "grid",
+  "mm_patch_merge_type": "flat",
+  "mm_projector_lr": null,
+  "mm_projector_type": "mlp2x_gelu",
+  "mm_resampler_type": null,
+  "mm_spatial_pool_mode": "bilinear",
+  "mm_spatial_pool_stride": null,
+  "mm_tunable_parts": "mm_mlp_adapter",
+  "mm_use_im_patch_token": false,
+  "mm_use_im_start_end": false,
+  "mm_vision_select_feature": "patch",
+  "mm_vision_select_layer": -2,
+  "mm_vision_tower": "open_clip_hub:ViT-B-16-SigLIP2-256",
+  "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_b16_256.npz",
+  "vocab_size": 151936
+}

continuous/qwen3/siglip2_b16_256/pretrain/mm_projector.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4e0791f3d6369c3afe2a82ab9602075a623e53f248f90802c00aa9c5dd1c4e8b
+size 23087160

continuous/qwen3/siglip2_b16_256/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": 2.721109628677368,
+      "learning_rate": 0.00032142857142857147,
+      "loss": 8.9559,
+      "step": 10
+    },
+    {
+      "epoch": 0.03669724770642202,
+      "grad_norm": 1.5921517610549927,
+      "learning_rate": 0.0006785714285714287,
+      "loss": 5.8334,
+      "step": 20
+    },
+    {
+      "epoch": 0.05504587155963303,
+      "grad_norm": 0.2769966423511505,
+      "learning_rate": 0.0009999907688176173,
+      "loss": 4.9896,
+      "step": 30
+    },
+    {
+      "epoch": 0.07339449541284404,
+      "grad_norm": 0.1680532693862915,
+      "learning_rate": 0.0009988834393115766,
+      "loss": 4.744,
+      "step": 40
+    },
+    {
+      "epoch": 0.09174311926605505,
+      "grad_norm": 0.08486717939376831,
+      "learning_rate": 0.0009959345572940772,
+      "loss": 4.669,
+      "step": 50
+    },
+    {
+      "epoch": 0.11009174311926606,
+      "grad_norm": 0.1885014921426773,
+      "learning_rate": 0.0009911550081155983,
+      "loss": 4.6284,
+      "step": 60
+    },
+    {
+      "epoch": 0.12844036697247707,
+      "grad_norm": 0.12540222704410553,
+      "learning_rate": 0.0009845624347566433,
+      "loss": 4.5792,
+      "step": 70
+    },
+    {
+      "epoch": 0.14678899082568808,
+      "grad_norm": 0.22761504352092743,
+      "learning_rate": 0.0009761811727013547,
+      "loss": 4.5406,
+      "step": 80
+    },
+    {
+      "epoch": 0.1651376146788991,
+      "grad_norm": 0.19424667954444885,
+      "learning_rate": 0.0009660421601067666,
+      "loss": 4.4785,
+      "step": 90
+    },
+    {
+      "epoch": 0.1834862385321101,
+      "grad_norm": 0.1409413069486618,
+      "learning_rate": 0.0009541828235992926,
+      "loss": 4.4374,
+      "step": 100
+    },
+    {
+      "epoch": 0.2018348623853211,
+      "grad_norm": 0.14655888080596924,
+      "learning_rate": 0.0009406469401200151,
+      "loss": 4.3894,
+      "step": 110
+    },
+    {
+      "epoch": 0.22018348623853212,
+      "grad_norm": 0.17104056477546692,
+      "learning_rate": 0.0009254844753287492,
+      "loss": 4.3605,
+      "step": 120
+    },
+    {
+      "epoch": 0.23853211009174313,
+      "grad_norm": 0.19066673517227173,
+      "learning_rate": 0.0009087513991633924,
+      "loss": 4.3153,
+      "step": 130
+    },
+    {
+      "epoch": 0.25688073394495414,
+      "grad_norm": 0.1378832757472992,
+      "learning_rate": 0.0008905094792353917,
+      "loss": 4.2837,
+      "step": 140
+    },
+    {
+      "epoch": 0.27522935779816515,
+      "grad_norm": 0.13631513714790344,
+      "learning_rate": 0.0008708260528239789,
+      "loss": 4.2679,
+      "step": 150
+    },
+    {
+      "epoch": 0.29357798165137616,
+      "grad_norm": 0.17127200961112976,
+      "learning_rate": 0.0008497737783108237,
+      "loss": 4.2418,
+      "step": 160
+    },
+    {
+      "epoch": 0.3119266055045872,
+      "grad_norm": 0.161219984292984,
+      "learning_rate": 0.0008274303669726426,
+      "loss": 4.2203,
+      "step": 170
+    },
+    {
+      "epoch": 0.3302752293577982,
+      "grad_norm": 0.15030692517757416,
+      "learning_rate": 0.0008038782961218137,
+      "loss": 4.1988,
+      "step": 180
+    },
+    {
+      "epoch": 0.3486238532110092,
+      "grad_norm": 0.14987820386886597,
+      "learning_rate": 0.000779204504653894,
+      "loss": 4.184,
+      "step": 190
+    },
+    {
+      "epoch": 0.3669724770642202,
+      "grad_norm": 0.14878025650978088,
+      "learning_rate": 0.0007535000721258791,
+      "loss": 4.159,
+      "step": 200
+    },
+    {
+      "epoch": 0.3853211009174312,
+      "grad_norm": 0.1445760875940323,
+      "learning_rate": 0.00072685988254984,
+      "loss": 4.1442,
+      "step": 210
+    },
+    {
+      "epoch": 0.4036697247706422,
+      "grad_norm": 0.13726311922073364,
+      "learning_rate": 0.0006993822741429906,
+      "loss": 4.148,
+      "step": 220
+    },
+    {
+      "epoch": 0.42201834862385323,
+      "grad_norm": 0.15636248886585236,
+      "learning_rate": 0.0006711686763270818,
+      "loss": 4.1365,
+      "step": 230
+    },
+    {
+      "epoch": 0.44036697247706424,
+      "grad_norm": 0.160902738571167,
+      "learning_rate": 0.0006423232353170831,
+      "loss": 4.1121,
+      "step": 240
+    },
+    {
+      "epoch": 0.45871559633027525,
+      "grad_norm": 0.13684585690498352,
+      "learning_rate": 0.0006129524296812335,
+      "loss": 4.075,
+      "step": 250
+    },
+    {
+      "epoch": 0.47706422018348627,
+      "grad_norm": 0.1399865597486496,
+      "learning_rate": 0.0005831646772915651,
+      "loss": 4.0582,
+      "step": 260
+    },
+    {
+      "epoch": 0.4954128440366973,
+      "grad_norm": 0.16995331645011902,
+      "learning_rate": 0.0005530699351157809,
+      "loss": 4.0166,
+      "step": 270
+    },
+    {
+      "epoch": 0.5137614678899083,
+      "grad_norm": 0.1377951055765152,
+      "learning_rate": 0.0005227792933277943,
+      "loss": 3.9939,
+      "step": 280
+    },
+    {
+      "epoch": 0.5321100917431193,
+      "grad_norm": 0.1499321013689041,
+      "learning_rate": 0.0004924045652352092,
+      "loss": 3.9703,
+      "step": 290
+    },
+    {
+      "epoch": 0.5504587155963303,
+      "grad_norm": 0.13068914413452148,
+      "learning_rate": 0.0004620578745374633,
+      "loss": 3.9585,
+      "step": 300
+    },
+    {
+      "epoch": 0.5688073394495413,
+      "grad_norm": 0.15706101059913635,
+      "learning_rate": 0.0004318512414382059,
+      "loss": 3.9617,
+      "step": 310
+    },
+    {
+      "epoch": 0.5871559633027523,
+      "grad_norm": 0.13138192892074585,
+      "learning_rate": 0.0004018961691397148,
+      "loss": 3.9488,
+      "step": 320
+    },
+    {
+      "epoch": 0.6055045871559633,
+      "grad_norm": 0.1253395825624466,
+      "learning_rate": 0.0003723032322457458,
+      "loss": 3.9303,
+      "step": 330
+    },
+    {
+      "epoch": 0.6238532110091743,
+      "grad_norm": 0.13742254674434662,
+      "learning_rate": 0.0003431816685921625,
+      "loss": 3.927,
+      "step": 340
+    },
+    {
+      "epoch": 0.6422018348623854,
+      "grad_norm": 0.1622939109802246,
+      "learning_rate": 0.0003146389760120416,
+      "loss": 3.9247,
+      "step": 350
+    },
+    {
+      "epoch": 0.6605504587155964,
+      "grad_norm": 0.13805417716503143,
+      "learning_rate": 0.00028678051552373485,
+      "loss": 3.9304,
+      "step": 360
+    },
+    {
+      "epoch": 0.6788990825688074,
+      "grad_norm": 0.13340365886688232,
+      "learning_rate": 0.0002597091224066581,
+      "loss": 3.9153,
+      "step": 370
+    },
+    {
+      "epoch": 0.6972477064220184,
+      "grad_norm": 0.12405907362699509,
+      "learning_rate": 0.0002335247266004629,
+      "loss": 3.9219,
+      "step": 380
+    },
+    {
+      "epoch": 0.7155963302752294,
+      "grad_norm": 0.1659265011548996,
+      "learning_rate": 0.000208323983828828,
+      "loss": 3.9083,
+      "step": 390
+    },
+    {
+      "epoch": 0.7339449541284404,
+      "grad_norm": 0.12012692540884018,
+      "learning_rate": 0.0001841999188095224,
+      "loss": 3.906,
+      "step": 400
+    },
+    {
+      "epoch": 0.7522935779816514,
+      "grad_norm": 0.11597146093845367,
+      "learning_rate": 0.00016124158186777676,
+      "loss": 3.9113,
+      "step": 410
+    },
+    {
+      "epoch": 0.7706422018348624,
+      "grad_norm": 0.11750850826501846,
+      "learning_rate": 0.0001395337202205211,
+      "loss": 3.9007,
+      "step": 420
+    },
+    {
+      "epoch": 0.7889908256880734,
+      "grad_norm": 0.11618483066558838,
+      "learning_rate": 0.00011915646514489292,
+      "loss": 3.9011,
+      "step": 430
+    },
+    {
+      "epoch": 0.8073394495412844,
+      "grad_norm": 0.12796686589717865,
+      "learning_rate": 0.00010018503618578818,
+      "loss": 3.9048,
+      "step": 440
+    },
+    {
+      "epoch": 0.8256880733944955,
+      "grad_norm": 0.12116481363773346,
+      "learning_rate": 8.268946349432583e-05,
+      "loss": 3.896,
+      "step": 450
+    },
+    {
+      "epoch": 0.8440366972477065,
+      "grad_norm": 0.12572164833545685,
+      "learning_rate": 6.673432932217117e-05,
+      "loss": 3.8916,
+      "step": 460
+    },
+    {
+      "epoch": 0.8623853211009175,
+      "grad_norm": 0.109195277094841,
+      "learning_rate": 5.2378529625954684e-05,
+      "loss": 3.8809,
+      "step": 470
+    },
+    {
+      "epoch": 0.8807339449541285,
+      "grad_norm": 0.11103541404008865,
+      "learning_rate": 3.9675056661785556e-05,
+      "loss": 3.8882,
+      "step": 480
+    },
+    {
+      "epoch": 0.8990825688073395,
+      "grad_norm": 0.1111704409122467,
+      "learning_rate": 2.8670803372379538e-05,
+      "loss": 3.8893,
+      "step": 490
+    },
+    {
+      "epoch": 0.9174311926605505,
+      "grad_norm": 0.11290071904659271,
+      "learning_rate": 1.9406390288876586e-05,
+      "loss": 3.9017,
+      "step": 500
+    },
+    {
+      "epoch": 0.9357798165137615,
+      "grad_norm": 0.11050901561975479,
+      "learning_rate": 1.191601558631461e-05,
+      "loss": 3.8916,
+      "step": 510
+    },
+    {
+      "epoch": 0.9541284403669725,
+      "grad_norm": 0.11096054315567017,
+      "learning_rate": 6.2273288462559305e-06,
+      "loss": 3.8858,
+      "step": 520
+    },
+    {
+      "epoch": 0.9724770642201835,
+      "grad_norm": 0.11817970871925354,
+      "learning_rate": 2.361328992552314e-06,
+      "loss": 3.8873,
+      "step": 530
+    },
+    {
+      "epoch": 0.9908256880733946,
+      "grad_norm": 0.10840678960084915,
+      "learning_rate": 3.322867770044202e-07,
+      "loss": 3.8857,
+      "step": 540
+    },
+    {
+      "epoch": 1.0,
+      "step": 545,
+      "total_flos": 2.9989926542403174e+17,
+      "train_loss": 4.235761827941334,
+      "train_runtime": 5604.2562,
+      "train_samples_per_second": 99.59,
+      "train_steps_per_second": 0.097
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 545,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": false,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.9989926542403174e+17,
+  "train_batch_size": 64,
+  "trial_name": null,
+  "trial_params": null
+}

continuous/qwen3/siglip2_b16_384/finetune/added_tokens.json

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

continuous/qwen3/siglip2_b16_384/finetune/config.json

@@ -0,0 +1,59 @@
+{
+  "add_faster_video": false,
+  "add_time_instruction": false,
+  "architectures": [
+    "LlavaQwen3ForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "bos_token_id": 151643,
+  "eos_token_id": 151645,
+  "faster_token_stride": 10,
+  "force_sample": false,
+  "head_dim": 128,
+  "hidden_act": "silu",
+  "hidden_size": 2048,
+  "image_aspect_ratio": "square",
+  "image_crop_resolution": null,
+  "image_grid_pinpoints": null,
+  "image_split_resolution": null,
+  "initializer_range": 0.02,
+  "intermediate_size": 6144,
+  "max_position_embeddings": 40960,
+  "max_window_layers": 28,
+  "mm_hidden_size": 768,
+  "mm_newline_position": "grid",
+  "mm_patch_merge_type": "flat",
+  "mm_projector_lr": 0.0001,
+  "mm_projector_type": "mlp2x_gelu",
+  "mm_resampler_type": null,
+  "mm_spatial_pool_mode": "bilinear",
+  "mm_spatial_pool_stride": null,
+  "mm_tunable_parts": "mm_mlp_adapter,mm_language_model",
+  "mm_use_im_patch_token": false,
+  "mm_use_im_start_end": false,
+  "mm_vision_select_feature": "patch",
+  "mm_vision_select_layer": -2,
+  "mm_vision_tower": "open_clip_hub:ViT-B-16-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_b16_384.npz",
+  "vocab_size": 151936
+}

continuous/qwen3/siglip2_b16_384/finetune/generation_config.json

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

continuous/qwen3/siglip2_b16_384/finetune/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5af9c613600baaa72c3e98eb11d772e4428b051295a37ebfc52d68a2dc850a53
+size 4272974976

continuous/qwen3/siglip2_b16_384/finetune/special_tokens_map.json

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

continuous/qwen3/siglip2_b16_384/finetune/tokenizer.json

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

continuous/qwen3/siglip2_b16_384/finetune/tokenizer_config.json

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

continuous/qwen3/siglip2_b16_384/finetune/trainer_state.json

@@ -0,0 +1,1856 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.9999037443449803,
+  "eval_steps": 500,
+  "global_step": 2597,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.003850226200789296,
+      "grad_norm": 8.766497611999512,
+      "learning_rate": 1.153846153846154e-06,
+      "loss": 2.3635,
+      "step": 10
+    },
+    {
+      "epoch": 0.007700452401578592,
+      "grad_norm": 3.5412678718566895,
+      "learning_rate": 2.435897435897436e-06,
+      "loss": 1.8914,
+      "step": 20
+    },
+    {
+      "epoch": 0.01155067860236789,
+      "grad_norm": 2.2382333278656006,
+      "learning_rate": 3.7179487179487184e-06,
+      "loss": 1.5001,
+      "step": 30
+    },
+    {
+      "epoch": 0.015400904803157185,
+      "grad_norm": 0.5654305219650269,
+      "learning_rate": 5e-06,
+      "loss": 1.1737,
+      "step": 40
+    },
+    {
+      "epoch": 0.019251131003946483,
+      "grad_norm": 0.45051735639572144,
+      "learning_rate": 6.282051282051282e-06,
+      "loss": 1.0744,
+      "step": 50
+    },
+    {
+      "epoch": 0.02310135720473578,
+      "grad_norm": 0.23187026381492615,
+      "learning_rate": 7.564102564102564e-06,
+      "loss": 0.998,
+      "step": 60
+    },
+    {
+      "epoch": 0.026951583405525073,
+      "grad_norm": 0.21478913724422455,
+      "learning_rate": 8.846153846153847e-06,
+      "loss": 0.9587,
+      "step": 70
+    },
+    {
+      "epoch": 0.03080180960631437,
+      "grad_norm": 0.17817778885364532,
+      "learning_rate": 9.999996111488723e-06,
+      "loss": 0.9568,
+      "step": 80
+    },
+    {
+      "epoch": 0.034652035807103666,
+      "grad_norm": 0.1880064755678177,
+      "learning_rate": 9.999529497453782e-06,
+      "loss": 0.9319,
+      "step": 90
+    },
+    {
+      "epoch": 0.038502262007892966,
+      "grad_norm": 0.14910034835338593,
+      "learning_rate": 9.998285264324193e-06,
+      "loss": 0.9063,
+      "step": 100
+    },
+    {
+      "epoch": 0.04235248820868226,
+      "grad_norm": 0.15911972522735596,
+      "learning_rate": 9.996263605626052e-06,
+      "loss": 0.882,
+      "step": 110
+    },
+    {
+      "epoch": 0.04620271440947156,
+      "grad_norm": 0.14373700320720673,
+      "learning_rate": 9.99346483580503e-06,
+      "loss": 0.8907,
+      "step": 120
+    },
+    {
+      "epoch": 0.05005294061026085,
+      "grad_norm": 0.13795314729213715,
+      "learning_rate": 9.989889390177467e-06,
+      "loss": 0.8822,
+      "step": 130
+    },
+    {
+      "epoch": 0.053903166811050146,
+      "grad_norm": 0.1580537110567093,
+      "learning_rate": 9.985537824862647e-06,
+      "loss": 0.8768,
+      "step": 140
+    },
+    {
+      "epoch": 0.057753393011839446,
+      "grad_norm": 0.1693747192621231,
+      "learning_rate": 9.980410816696317e-06,
+      "loss": 0.8725,
+      "step": 150
+    },
+    {
+      "epoch": 0.06160361921262874,
+      "grad_norm": 0.15090207755565643,
+      "learning_rate": 9.97450916312541e-06,
+      "loss": 0.8669,
+      "step": 160
+    },
+    {
+      "epoch": 0.06545384541341803,
+      "grad_norm": 0.13166089355945587,
+      "learning_rate": 9.967833782084004e-06,
+      "loss": 0.8637,
+      "step": 170
+    },
+    {
+      "epoch": 0.06930407161420733,
+      "grad_norm": 0.141921728849411,
+      "learning_rate": 9.960385711850554e-06,
+      "loss": 0.889,
+      "step": 180
+    },
+    {
+      "epoch": 0.07315429781499663,
+      "grad_norm": 0.14378538727760315,
+      "learning_rate": 9.952166110886394e-06,
+      "loss": 0.8869,
+      "step": 190
+    },
+    {
+      "epoch": 0.07700452401578593,
+      "grad_norm": 0.1504865288734436,
+      "learning_rate": 9.943176257655567e-06,
+      "loss": 0.8663,
+      "step": 200
+    },
+    {
+      "epoch": 0.08085475021657522,
+      "grad_norm": 0.14597228169441223,
+      "learning_rate": 9.933417550425952e-06,
+      "loss": 0.8709,
+      "step": 210
+    },
+    {
+      "epoch": 0.08470497641736452,
+      "grad_norm": 0.13478966057300568,
+      "learning_rate": 9.9228915070518e-06,
+      "loss": 0.8506,
+      "step": 220
+    },
+    {
+      "epoch": 0.08855520261815382,
+      "grad_norm": 0.15765756368637085,
+      "learning_rate": 9.911599764737632e-06,
+      "loss": 0.8569,
+      "step": 230
+    },
+    {
+      "epoch": 0.09240542881894312,
+      "grad_norm": 0.14720825850963593,
+      "learning_rate": 9.899544079783606e-06,
+      "loss": 0.8432,
+      "step": 240
+    },
+    {
+      "epoch": 0.0962556550197324,
+      "grad_norm": 0.14703981578350067,
+      "learning_rate": 9.886726327312338e-06,
+      "loss": 0.8591,
+      "step": 250
+    },
+    {
+      "epoch": 0.1001058812205217,
+      "grad_norm": 0.1337234079837799,
+      "learning_rate": 9.87314850097724e-06,
+      "loss": 0.863,
+      "step": 260
+    },
+    {
+      "epoch": 0.103956107421311,
+      "grad_norm": 0.13792596757411957,
+      "learning_rate": 9.858812712652452e-06,
+      "loss": 0.8481,
+      "step": 270
+    },
+    {
+      "epoch": 0.10780633362210029,
+      "grad_norm": 0.13864684104919434,
+      "learning_rate": 9.843721192104341e-06,
+      "loss": 0.8522,
+      "step": 280
+    },
+    {
+      "epoch": 0.11165655982288959,
+      "grad_norm": 0.13226495683193207,
+      "learning_rate": 9.827876286644698e-06,
+      "loss": 0.8626,
+      "step": 290
+    },
+    {
+      "epoch": 0.11550678602367889,
+      "grad_norm": 0.13212192058563232,
+      "learning_rate": 9.811280460765643e-06,
+      "loss": 0.8233,
+      "step": 300
+    },
+    {
+      "epoch": 0.11935701222446819,
+      "grad_norm": 0.13428285717964172,
+      "learning_rate": 9.793936295756292e-06,
+      "loss": 0.8417,
+      "step": 310
+    },
+    {
+      "epoch": 0.12320723842525748,
+      "grad_norm": 0.13644687831401825,
+      "learning_rate": 9.775846489301279e-06,
+      "loss": 0.8364,
+      "step": 320
+    },
+    {
+      "epoch": 0.12705746462604678,
+      "grad_norm": 0.1308417022228241,
+      "learning_rate": 9.757013855061157e-06,
+      "loss": 0.8385,
+      "step": 330
+    },
+    {
+      "epoch": 0.13090769082683607,
+      "grad_norm": 0.12911461293697357,
+      "learning_rate": 9.737441322234758e-06,
+      "loss": 0.8481,
+      "step": 340
+    },
+    {
+      "epoch": 0.13475791702762538,
+      "grad_norm": 0.14492258429527283,
+      "learning_rate": 9.717131935103605e-06,
+      "loss": 0.8526,
+      "step": 350
+    },
+    {
+      "epoch": 0.13860814322841467,
+      "grad_norm": 0.13201360404491425,
+      "learning_rate": 9.696088852558398e-06,
+      "loss": 0.8363,
+      "step": 360
+    },
+    {
+      "epoch": 0.14245836942920398,
+      "grad_norm": 0.12838678061962128,
+      "learning_rate": 9.674315347607686e-06,
+      "loss": 0.8614,
+      "step": 370
+    },
+    {
+      "epoch": 0.14630859562999327,
+      "grad_norm": 0.12707985937595367,
+      "learning_rate": 9.651814806868796e-06,
+      "loss": 0.8302,
+      "step": 380
+    },
+    {
+      "epoch": 0.15015882183078255,
+      "grad_norm": 0.14670954644680023,
+      "learning_rate": 9.62859073004107e-06,
+      "loss": 0.836,
+      "step": 390
+    },
+    {
+      "epoch": 0.15400904803157187,
+      "grad_norm": 0.12708036601543427,
+      "learning_rate": 9.604646729361547e-06,
+      "loss": 0.8367,
+      "step": 400
+    },
+    {
+      "epoch": 0.15785927423236115,
+      "grad_norm": 0.1328769475221634,
+      "learning_rate": 9.579986529043099e-06,
+      "loss": 0.8177,
+      "step": 410
+    },
+    {
+      "epoch": 0.16170950043315044,
+      "grad_norm": 0.12348297238349915,
+      "learning_rate": 9.554613964695189e-06,
+      "loss": 0.8336,
+      "step": 420
+    },
+    {
+      "epoch": 0.16555972663393975,
+      "grad_norm": 0.13996286690235138,
+      "learning_rate": 9.528532982727281e-06,
+      "loss": 0.8262,
+      "step": 430
+    },
+    {
+      "epoch": 0.16940995283472904,
+      "grad_norm": 0.14812089502811432,
+      "learning_rate": 9.501747639735018e-06,
+      "loss": 0.8263,
+      "step": 440
+    },
+    {
+      "epoch": 0.17326017903551832,
+      "grad_norm": 0.13950709998607635,
+      "learning_rate": 9.474262101869272e-06,
+      "loss": 0.8086,
+      "step": 450
+    },
+    {
+      "epoch": 0.17711040523630764,
+      "grad_norm": 0.1315620094537735,
+      "learning_rate": 9.446080644188141e-06,
+      "loss": 0.8261,
+      "step": 460
+    },
+    {
+      "epoch": 0.18096063143709692,
+      "grad_norm": 0.14662739634513855,
+      "learning_rate": 9.417207649992017e-06,
+      "loss": 0.8302,
+      "step": 470
+    },
+    {
+      "epoch": 0.18481085763788624,
+      "grad_norm": 0.12621991336345673,
+      "learning_rate": 9.387647610141816e-06,
+      "loss": 0.8241,
+      "step": 480
+    },
+    {
+      "epoch": 0.18866108383867553,
+      "grad_norm": 0.128559872508049,
+      "learning_rate": 9.35740512236047e-06,
+      "loss": 0.8145,
+      "step": 490
+    },
+    {
+      "epoch": 0.1925113100394648,
+      "grad_norm": 0.1283157616853714,
+      "learning_rate": 9.326484890517809e-06,
+      "loss": 0.8163,
+      "step": 500
+    },
+    {
+      "epoch": 0.19636153624025413,
+      "grad_norm": 0.12837550044059753,
+      "learning_rate": 9.294891723898926e-06,
+      "loss": 0.8022,
+      "step": 510
+    },
+    {
+      "epoch": 0.2002117624410434,
+      "grad_norm": 0.12746953964233398,
+      "learning_rate": 9.262630536456153e-06,
+      "loss": 0.8439,
+      "step": 520
+    },
+    {
+      "epoch": 0.2040619886418327,
+      "grad_norm": 0.14754927158355713,
+      "learning_rate": 9.229706346044749e-06,
+      "loss": 0.8099,
+      "step": 530
+    },
+    {
+      "epoch": 0.207912214842622,
+      "grad_norm": 0.14269253611564636,
+      "learning_rate": 9.196124273642422e-06,
+      "loss": 0.8295,
+      "step": 540
+    },
+    {
+      "epoch": 0.2117624410434113,
+      "grad_norm": 0.12189345061779022,
+      "learning_rate": 9.161889542552842e-06,
+      "loss": 0.8335,
+      "step": 550
+    },
+    {
+      "epoch": 0.21561266724420058,
+      "grad_norm": 0.1346394121646881,
+      "learning_rate": 9.127007477593193e-06,
+      "loss": 0.8333,
+      "step": 560
+    },
+    {
+      "epoch": 0.2194628934449899,
+      "grad_norm": 0.14390435814857483,
+      "learning_rate": 9.091483504265974e-06,
+      "loss": 0.8435,
+      "step": 570
+    },
+    {
+      "epoch": 0.22331311964577918,
+      "grad_norm": 0.14086800813674927,
+      "learning_rate": 9.055323147915116e-06,
+      "loss": 0.8173,
+      "step": 580
+    },
+    {
+      "epoch": 0.2271633458465685,
+      "grad_norm": 0.13980339467525482,
+      "learning_rate": 9.018532032866585e-06,
+      "loss": 0.8359,
+      "step": 590
+    },
+    {
+      "epoch": 0.23101357204735778,
+      "grad_norm": 0.1363815814256668,
+      "learning_rate": 8.981115881553579e-06,
+      "loss": 0.8313,
+      "step": 600
+    },
+    {
+      "epoch": 0.23486379824814707,
+      "grad_norm": 0.14619599282741547,
+      "learning_rate": 8.943080513626476e-06,
+      "loss": 0.8416,
+      "step": 610
+    },
+    {
+      "epoch": 0.23871402444893638,
+      "grad_norm": 0.13487179577350616,
+      "learning_rate": 8.904431845047641e-06,
+      "loss": 0.8321,
+      "step": 620
+    },
+    {
+      "epoch": 0.24256425064972567,
+      "grad_norm": 0.1425708383321762,
+      "learning_rate": 8.86517588717128e-06,
+      "loss": 0.8376,
+      "step": 630
+    },
+    {
+      "epoch": 0.24641447685051496,
+      "grad_norm": 0.13107232749462128,
+      "learning_rate": 8.82531874580844e-06,
+      "loss": 0.8311,
+      "step": 640
+    },
+    {
+      "epoch": 0.25026470305130427,
+      "grad_norm": 0.11684010922908783,
+      "learning_rate": 8.784866620277314e-06,
+      "loss": 0.7863,
+      "step": 650
+    },
+    {
+      "epoch": 0.25411492925209356,
+      "grad_norm": 0.1385643035173416,
+      "learning_rate": 8.743825802439018e-06,
+      "loss": 0.8167,
+      "step": 660
+    },
+    {
+      "epoch": 0.25796515545288284,
+      "grad_norm": 0.1397099792957306,
+      "learning_rate": 8.702202675718956e-06,
+      "loss": 0.8299,
+      "step": 670
+    },
+    {
+      "epoch": 0.26181538165367213,
+      "grad_norm": 0.12696576118469238,
+      "learning_rate": 8.660003714113959e-06,
+      "loss": 0.806,
+      "step": 680
+    },
+    {
+      "epoch": 0.26566560785446147,
+      "grad_norm": 0.13636654615402222,
+      "learning_rate": 8.617235481185323e-06,
+      "loss": 0.8191,
+      "step": 690
+    },
+    {
+      "epoch": 0.26951583405525076,
+      "grad_norm": 0.133625328540802,
+      "learning_rate": 8.573904629037931e-06,
+      "loss": 0.8256,
+      "step": 700
+    },
+    {
+      "epoch": 0.27336606025604004,
+      "grad_norm": 0.12326812744140625,
+      "learning_rate": 8.530017897285585e-06,
+      "loss": 0.8091,
+      "step": 710
+    },
+    {
+      "epoch": 0.27721628645682933,
+      "grad_norm": 0.11497246474027634,
+      "learning_rate": 8.485582112002746e-06,
+      "loss": 0.8116,
+      "step": 720
+    },
+    {
+      "epoch": 0.2810665126576186,
+      "grad_norm": 0.1337440311908722,
+      "learning_rate": 8.440604184662811e-06,
+      "loss": 0.8233,
+      "step": 730
+    },
+    {
+      "epoch": 0.28491673885840796,
+      "grad_norm": 0.12975099682807922,
+      "learning_rate": 8.395091111063118e-06,
+      "loss": 0.8041,
+      "step": 740
+    },
+    {
+      "epoch": 0.28876696505919724,
+      "grad_norm": 0.12711885571479797,
+      "learning_rate": 8.349049970236822e-06,
+      "loss": 0.8035,
+      "step": 750
+    },
+    {
+      "epoch": 0.29261719125998653,
+      "grad_norm": 0.12767837941646576,
+      "learning_rate": 8.302487923351842e-06,
+      "loss": 0.7963,
+      "step": 760
+    },
+    {
+      "epoch": 0.2964674174607758,
+      "grad_norm": 0.13174980878829956,
+      "learning_rate": 8.255412212597021e-06,
+      "loss": 0.8147,
+      "step": 770
+    },
+    {
+      "epoch": 0.3003176436615651,
+      "grad_norm": 0.12714515626430511,
+      "learning_rate": 8.207830160055688e-06,
+      "loss": 0.8071,
+      "step": 780
+    },
+    {
+      "epoch": 0.3041678698623544,
+      "grad_norm": 0.13654649257659912,
+      "learning_rate": 8.159749166566786e-06,
+      "loss": 0.7963,
+      "step": 790
+    },
+    {
+      "epoch": 0.30801809606314373,
+      "grad_norm": 0.13397666811943054,
+      "learning_rate": 8.11117671057377e-06,
+      "loss": 0.8232,
+      "step": 800
+    },
+    {
+      "epoch": 0.311868322263933,
+      "grad_norm": 0.12752170860767365,
+      "learning_rate": 8.062120346961404e-06,
+      "loss": 0.81,
+      "step": 810
+    },
+    {
+      "epoch": 0.3157185484647223,
+      "grad_norm": 0.12455415725708008,
+      "learning_rate": 8.012587705880705e-06,
+      "loss": 0.8109,
+      "step": 820
+    },
+    {
+      "epoch": 0.3195687746655116,
+      "grad_norm": 0.13197685778141022,
+      "learning_rate": 7.962586491562146e-06,
+      "loss": 0.7992,
+      "step": 830
+    },
+    {
+      "epoch": 0.3234190008663009,
+      "grad_norm": 0.13197512924671173,
+      "learning_rate": 7.912124481117357e-06,
+      "loss": 0.8035,
+      "step": 840
+    },
+    {
+      "epoch": 0.3272692270670902,
+      "grad_norm": 0.12046568095684052,
+      "learning_rate": 7.861209523329494e-06,
+      "loss": 0.8041,
+      "step": 850
+    },
+    {
+      "epoch": 0.3311194532678795,
+      "grad_norm": 0.13653253018856049,
+      "learning_rate": 7.809849537432432e-06,
+      "loss": 0.7971,
+      "step": 860
+    },
+    {
+      "epoch": 0.3349696794686688,
+      "grad_norm": 0.12424702942371368,
+      "learning_rate": 7.758052511879035e-06,
+      "loss": 0.7957,
+      "step": 870
+    },
+    {
+      "epoch": 0.3388199056694581,
+      "grad_norm": 0.1214323490858078,
+      "learning_rate": 7.705826503098642e-06,
+      "loss": 0.8019,
+      "step": 880
+    },
+    {
+      "epoch": 0.34267013187024736,
+      "grad_norm": 0.1336003839969635,
+      "learning_rate": 7.653179634243979e-06,
+      "loss": 0.7917,
+      "step": 890
+    },
+    {
+      "epoch": 0.34652035807103665,
+      "grad_norm": 0.12333294004201889,
+      "learning_rate": 7.600120093927696e-06,
+      "loss": 0.7968,
+      "step": 900
+    },
+    {
+      "epoch": 0.350370584271826,
+      "grad_norm": 0.12750890851020813,
+      "learning_rate": 7.546656134948729e-06,
+      "loss": 0.7978,
+      "step": 910
+    },
+    {
+      "epoch": 0.3542208104726153,
+      "grad_norm": 0.12951450049877167,
+      "learning_rate": 7.492796073008661e-06,
+      "loss": 0.7842,
+      "step": 920
+    },
+    {
+      "epoch": 0.35807103667340456,
+      "grad_norm": 0.1328691840171814,
+      "learning_rate": 7.438548285418326e-06,
+      "loss": 0.7784,
+      "step": 930
+    },
+    {
+      "epoch": 0.36192126287419385,
+      "grad_norm": 0.14351153373718262,
+      "learning_rate": 7.3839212097948e-06,
+      "loss": 0.8151,
+      "step": 940
+    },
+    {
+      "epoch": 0.36577148907498314,
+      "grad_norm": 0.12357240170240402,
+      "learning_rate": 7.328923342749039e-06,
+      "loss": 0.8256,
+      "step": 950
+    },
+    {
+      "epoch": 0.3696217152757725,
+      "grad_norm": 0.13752807676792145,
+      "learning_rate": 7.273563238564319e-06,
+      "loss": 0.8033,
+      "step": 960
+    },
+    {
+      "epoch": 0.37347194147656176,
+      "grad_norm": 0.12861678004264832,
+      "learning_rate": 7.217849507865724e-06,
+      "loss": 0.784,
+      "step": 970
+    },
+    {
+      "epoch": 0.37732216767735105,
+      "grad_norm": 0.12197241187095642,
+      "learning_rate": 7.161790816280858e-06,
+      "loss": 0.7927,
+      "step": 980
+    },
+    {
+      "epoch": 0.38117239387814034,
+      "grad_norm": 0.1376270353794098,
+      "learning_rate": 7.1053958830920125e-06,
+      "loss": 0.7818,
+      "step": 990
+    },
+    {
+      "epoch": 0.3850226200789296,
+      "grad_norm": 0.12344305962324142,
+      "learning_rate": 7.0486734798799715e-06,
+      "loss": 0.8033,
+      "step": 1000
+    },
+    {
+      "epoch": 0.3888728462797189,
+      "grad_norm": 0.13406606018543243,
+      "learning_rate": 6.991632429159708e-06,
+      "loss": 0.8226,
+      "step": 1010
+    },
+    {
+      "epoch": 0.39272307248050825,
+      "grad_norm": 0.13279324769973755,
+      "learning_rate": 6.934281603008132e-06,
+      "loss": 0.8003,
+      "step": 1020
+    },
+    {
+      "epoch": 0.39657329868129754,
+      "grad_norm": 0.12185411900281906,
+      "learning_rate": 6.876629921684143e-06,
+      "loss": 0.8107,
+      "step": 1030
+    },
+    {
+      "epoch": 0.4004235248820868,
+      "grad_norm": 0.12284626066684723,
+      "learning_rate": 6.818686352241199e-06,
+      "loss": 0.8092,
+      "step": 1040
+    },
+    {
+      "epoch": 0.4042737510828761,
+      "grad_norm": 0.11529895663261414,
+      "learning_rate": 6.760459907132573e-06,
+      "loss": 0.8,
+      "step": 1050
+    },
+    {
+      "epoch": 0.4081239772836654,
+      "grad_norm": 0.12976831197738647,
+      "learning_rate": 6.701959642809589e-06,
+      "loss": 0.8051,
+      "step": 1060
+    },
+    {
+      "epoch": 0.41197420348445474,
+      "grad_norm": 0.12935490906238556,
+      "learning_rate": 6.643194658312987e-06,
+      "loss": 0.7873,
+      "step": 1070
+    },
+    {
+      "epoch": 0.415824429685244,
+      "grad_norm": 0.12145864963531494,
+      "learning_rate": 6.584174093857676e-06,
+      "loss": 0.8198,
+      "step": 1080
+    },
+    {
+      "epoch": 0.4196746558860333,
+      "grad_norm": 0.14266541600227356,
+      "learning_rate": 6.5249071294110735e-06,
+      "loss": 0.7942,
+      "step": 1090
+    },
+    {
+      "epoch": 0.4235248820868226,
+      "grad_norm": 0.1417100876569748,
+      "learning_rate": 6.465402983265274e-06,
+      "loss": 0.8156,
+      "step": 1100
+    },
+    {
+      "epoch": 0.4273751082876119,
+      "grad_norm": 0.13155800104141235,
+      "learning_rate": 6.405670910603243e-06,
+      "loss": 0.808,
+      "step": 1110
+    },
+    {
+      "epoch": 0.43122533448840117,
+      "grad_norm": 0.14628025889396667,
+      "learning_rate": 6.345720202059291e-06,
+      "loss": 0.8029,
+      "step": 1120
+    },
+    {
+      "epoch": 0.4350755606891905,
+      "grad_norm": 0.13024255633354187,
+      "learning_rate": 6.285560182274008e-06,
+      "loss": 0.7913,
+      "step": 1130
+    },
+    {
+      "epoch": 0.4389257868899798,
+      "grad_norm": 0.12493369728326797,
+      "learning_rate": 6.225200208443936e-06,
+      "loss": 0.7952,
+      "step": 1140
+    },
+    {
+      "epoch": 0.4427760130907691,
+      "grad_norm": 0.11819007247686386,
+      "learning_rate": 6.16464966886616e-06,
+      "loss": 0.7924,
+      "step": 1150
+    },
+    {
+      "epoch": 0.44662623929155837,
+      "grad_norm": 0.14608392119407654,
+      "learning_rate": 6.103917981478065e-06,
+      "loss": 0.8179,
+      "step": 1160
+    },
+    {
+      "epoch": 0.45047646549234766,
+      "grad_norm": 0.1280871331691742,
+      "learning_rate": 6.043014592392489e-06,
+      "loss": 0.8257,
+      "step": 1170
+    },
+    {
+      "epoch": 0.454326691693137,
+      "grad_norm": 0.11578552424907684,
+      "learning_rate": 5.981948974428485e-06,
+      "loss": 0.7979,
+      "step": 1180
+    },
+    {
+      "epoch": 0.4581769178939263,
+      "grad_norm": 0.13489389419555664,
+      "learning_rate": 5.920730625637934e-06,
+      "loss": 0.7802,
+      "step": 1190
+    },
+    {
+      "epoch": 0.46202714409471557,
+      "grad_norm": 0.1409725397825241,
+      "learning_rate": 5.8593690678282355e-06,
+      "loss": 0.787,
+      "step": 1200
+    },
+    {
+      "epoch": 0.46587737029550486,
+      "grad_norm": 0.13081054389476776,
+      "learning_rate": 5.797873845081301e-06,
+      "loss": 0.7921,
+      "step": 1210
+    },
+    {
+      "epoch": 0.46972759649629414,
+      "grad_norm": 0.12804190814495087,
+      "learning_rate": 5.736254522269075e-06,
+      "loss": 0.807,
+      "step": 1220
+    },
+    {
+      "epoch": 0.47357782269708343,
+      "grad_norm": 0.12162957340478897,
+      "learning_rate": 5.6745206835658386e-06,
+      "loss": 0.8085,
+      "step": 1230
+    },
+    {
+      "epoch": 0.47742804889787277,
+      "grad_norm": 0.12125927209854126,
+      "learning_rate": 5.612681930957495e-06,
+      "loss": 0.7867,
+      "step": 1240
+    },
+    {
+      "epoch": 0.48127827509866206,
+      "grad_norm": 0.12121114134788513,
+      "learning_rate": 5.550747882748095e-06,
+      "loss": 0.7856,
+      "step": 1250
+    },
+    {
+      "epoch": 0.48512850129945134,
+      "grad_norm": 0.13419754803180695,
+      "learning_rate": 5.4887281720638195e-06,
+      "loss": 0.7845,
+      "step": 1260
+    },
+    {
+      "epoch": 0.48897872750024063,
+      "grad_norm": 0.13723702728748322,
+      "learning_rate": 5.42663244535466e-06,
+      "loss": 0.7759,
+      "step": 1270
+    },
+    {
+      "epoch": 0.4928289537010299,
+      "grad_norm": 0.13546262681484222,
+      "learning_rate": 5.3644703608940205e-06,
+      "loss": 0.7819,
+      "step": 1280
+    },
+    {
+      "epoch": 0.49667917990181926,
+      "grad_norm": 0.12390671670436859,
+      "learning_rate": 5.302251587276491e-06,
+      "loss": 0.802,
+      "step": 1290
+    },
+    {
+      "epoch": 0.5005294061026085,
+      "grad_norm": 0.1310867965221405,
+      "learning_rate": 5.2399858019140005e-06,
+      "loss": 0.7944,
+      "step": 1300
+    },
+    {
+      "epoch": 0.5043796323033978,
+      "grad_norm": 0.13104113936424255,
+      "learning_rate": 5.177682689530615e-06,
+      "loss": 0.8087,
+      "step": 1310
+    },
+    {
+      "epoch": 0.5082298585041871,
+      "grad_norm": 0.14247038960456848,
+      "learning_rate": 5.11535194065619e-06,
+      "loss": 0.786,
+      "step": 1320
+    },
+    {
+      "epoch": 0.5120800847049765,
+      "grad_norm": 0.1437874436378479,
+      "learning_rate": 5.053003250119113e-06,
+      "loss": 0.7886,
+      "step": 1330
+    },
+    {
+      "epoch": 0.5159303109057657,
+      "grad_norm": 0.1297159641981125,
+      "learning_rate": 4.9906463155383925e-06,
+      "loss": 0.7935,
+      "step": 1340
+    },
+    {
+      "epoch": 0.519780537106555,
+      "grad_norm": 0.13293516635894775,
+      "learning_rate": 4.9282908358153046e-06,
+      "loss": 0.7829,
+      "step": 1350
+    },
+    {
+      "epoch": 0.5236307633073443,
+      "grad_norm": 0.1418856829404831,
+      "learning_rate": 4.8659465096248345e-06,
+      "loss": 0.797,
+      "step": 1360
+    },
+    {
+      "epoch": 0.5274809895081336,
+      "grad_norm": 0.13245783746242523,
+      "learning_rate": 4.803623033907169e-06,
+      "loss": 0.8042,
+      "step": 1370
+    },
+    {
+      "epoch": 0.5313312157089229,
+      "grad_norm": 0.12522763013839722,
+      "learning_rate": 4.741330102359441e-06,
+      "loss": 0.8067,
+      "step": 1380
+    },
+    {
+      "epoch": 0.5351814419097122,
+      "grad_norm": 0.13302098214626312,
+      "learning_rate": 4.679077403927996e-06,
+      "loss": 0.7831,
+      "step": 1390
+    },
+    {
+      "epoch": 0.5390316681105015,
+      "grad_norm": 0.13454365730285645,
+      "learning_rate": 4.616874621301373e-06,
+      "loss": 0.7877,
+      "step": 1400
+    },
+    {
+      "epoch": 0.5428818943112907,
+      "grad_norm": 0.13028255105018616,
+      "learning_rate": 4.554731429404293e-06,
+      "loss": 0.8041,
+      "step": 1410
+    },
+    {
+      "epoch": 0.5467321205120801,
+      "grad_norm": 0.1396162062883377,
+      "learning_rate": 4.492657493892822e-06,
+      "loss": 0.7899,
+      "step": 1420
+    },
+    {
+      "epoch": 0.5505823467128694,
+      "grad_norm": 0.14332391321659088,
+      "learning_rate": 4.430662469650999e-06,
+      "loss": 0.7995,
+      "step": 1430
+    },
+    {
+      "epoch": 0.5544325729136587,
+      "grad_norm": 0.1327078938484192,
+      "learning_rate": 4.368755999289118e-06,
+      "loss": 0.8022,
+      "step": 1440
+    },
+    {
+      "epoch": 0.558282799114448,
+      "grad_norm": 0.15169571340084076,
+      "learning_rate": 4.306947711643949e-06,
+      "loss": 0.786,
+      "step": 1450
+    },
+    {
+      "epoch": 0.5621330253152372,
+      "grad_norm": 0.12621429562568665,
+      "learning_rate": 4.245247220281068e-06,
+      "loss": 0.7895,
+      "step": 1460
+    },
+    {
+      "epoch": 0.5659832515160266,
+      "grad_norm": 0.12834370136260986,
+      "learning_rate": 4.183664121999587e-06,
+      "loss": 0.7867,
+      "step": 1470
+    },
+    {
+      "epoch": 0.5698334777168159,
+      "grad_norm": 0.13247255980968475,
+      "learning_rate": 4.122207995339474e-06,
+      "loss": 0.7844,
+      "step": 1480
+    },
+    {
+      "epoch": 0.5736837039176051,
+      "grad_norm": 0.12964721024036407,
+      "learning_rate": 4.060888399091734e-06,
+      "loss": 0.7904,
+      "step": 1490
+    },
+    {
+      "epoch": 0.5775339301183945,
+      "grad_norm": 0.1336209923028946,
+      "learning_rate": 3.999714870811639e-06,
+      "loss": 0.7976,
+      "step": 1500
+    },
+    {
+      "epoch": 0.5813841563191837,
+      "grad_norm": 0.12601768970489502,
+      "learning_rate": 3.938696925335275e-06,
+      "loss": 0.7872,
+      "step": 1510
+    },
+    {
+      "epoch": 0.5852343825199731,
+      "grad_norm": 0.1414192020893097,
+      "learning_rate": 3.87784405329962e-06,
+      "loss": 0.7965,
+      "step": 1520
+    },
+    {
+      "epoch": 0.5890846087207623,
+      "grad_norm": 0.1424041986465454,
+      "learning_rate": 3.817165719666382e-06,
+      "loss": 0.7956,
+      "step": 1530
+    },
+    {
+      "epoch": 0.5929348349215516,
+      "grad_norm": 0.12750843167304993,
+      "learning_rate": 3.756671362249841e-06,
+      "loss": 0.8023,
+      "step": 1540
+    },
+    {
+      "epoch": 0.596785061122341,
+      "grad_norm": 0.1325743943452835,
+      "learning_rate": 3.6963703902488907e-06,
+      "loss": 0.7942,
+      "step": 1550
+    },
+    {
+      "epoch": 0.6006352873231302,
+      "grad_norm": 0.13646931946277618,
+      "learning_rate": 3.6362721827835655e-06,
+      "loss": 0.8095,
+      "step": 1560
+    },
+    {
+      "epoch": 0.6044855135239195,
+      "grad_norm": 0.14857622981071472,
+      "learning_rate": 3.576386087436212e-06,
+      "loss": 0.7948,
+      "step": 1570
+    },
+    {
+      "epoch": 0.6083357397247088,
+      "grad_norm": 0.13300994038581848,
+      "learning_rate": 3.5167214187975913e-06,
+      "loss": 0.7955,
+      "step": 1580
+    },
+    {
+      "epoch": 0.6121859659254981,
+      "grad_norm": 0.11440332233905792,
+      "learning_rate": 3.457287457018086e-06,
+      "loss": 0.7925,
+      "step": 1590
+    },
+    {
+      "epoch": 0.6160361921262875,
+      "grad_norm": 0.13022857904434204,
+      "learning_rate": 3.3980934463643003e-06,
+      "loss": 0.7747,
+      "step": 1600
+    },
+    {
+      "epoch": 0.6198864183270767,
+      "grad_norm": 0.11803010106086731,
+      "learning_rate": 3.3391485937812053e-06,
+      "loss": 0.758,
+      "step": 1610
+    },
+    {
+      "epoch": 0.623736644527866,
+      "grad_norm": 0.1322411745786667,
+      "learning_rate": 3.280462067460115e-06,
+      "loss": 0.7598,
+      "step": 1620
+    },
+    {
+      "epoch": 0.6275868707286553,
+      "grad_norm": 0.1314224898815155,
+      "learning_rate": 3.222042995412669e-06,
+      "loss": 0.796,
+      "step": 1630
+    },
+    {
+      "epoch": 0.6314370969294446,
+      "grad_norm": 0.12177182734012604,
+      "learning_rate": 3.163900464051089e-06,
+      "loss": 0.7776,
+      "step": 1640
+    },
+    {
+      "epoch": 0.635287323130234,
+      "grad_norm": 0.13313783705234528,
+      "learning_rate": 3.1060435167748826e-06,
+      "loss": 0.7872,
+      "step": 1650
+    },
+    {
+      "epoch": 0.6391375493310232,
+      "grad_norm": 0.12848614156246185,
+      "learning_rate": 3.048481152564253e-06,
+      "loss": 0.772,
+      "step": 1660
+    },
+    {
+      "epoch": 0.6429877755318125,
+      "grad_norm": 0.14232799410820007,
+      "learning_rate": 2.9912223245804017e-06,
+      "loss": 0.7946,
+      "step": 1670
+    },
+    {
+      "epoch": 0.6468380017326018,
+      "grad_norm": 0.1331789344549179,
+      "learning_rate": 2.934275938772981e-06,
+      "loss": 0.7799,
+      "step": 1680
+    },
+    {
+      "epoch": 0.6506882279333911,
+      "grad_norm": 0.1478201150894165,
+      "learning_rate": 2.8776508524948676e-06,
+      "loss": 0.7682,
+      "step": 1690
+    },
+    {
+      "epoch": 0.6545384541341804,
+      "grad_norm": 0.12576358020305634,
+      "learning_rate": 2.821355873124503e-06,
+      "loss": 0.7638,
+      "step": 1700
+    },
+    {
+      "epoch": 0.6583886803349697,
+      "grad_norm": 0.14458449184894562,
+      "learning_rate": 2.7653997566960144e-06,
+      "loss": 0.7841,
+      "step": 1710
+    },
+    {
+      "epoch": 0.662238906535759,
+      "grad_norm": 0.13322606682777405,
+      "learning_rate": 2.709791206537309e-06,
+      "loss": 0.7726,
+      "step": 1720
+    },
+    {
+      "epoch": 0.6660891327365482,
+      "grad_norm": 0.1312054693698883,
+      "learning_rate": 2.6545388719163747e-06,
+      "loss": 0.7764,
+      "step": 1730
+    },
+    {
+      "epoch": 0.6699393589373376,
+      "grad_norm": 0.13448017835617065,
+      "learning_rate": 2.599651346695979e-06,
+      "loss": 0.7805,
+      "step": 1740
+    },
+    {
+      "epoch": 0.6737895851381268,
+      "grad_norm": 0.14357765018939972,
+      "learning_rate": 2.545137167997007e-06,
+      "loss": 0.7737,
+      "step": 1750
+    },
+    {
+      "epoch": 0.6776398113389162,
+      "grad_norm": 0.12139645218849182,
+      "learning_rate": 2.4910048148705972e-06,
+      "loss": 0.7598,
+      "step": 1760
+    },
+    {
+      "epoch": 0.6814900375397055,
+      "grad_norm": 0.13442173600196838,
+      "learning_rate": 2.4372627069793327e-06,
+      "loss": 0.8,
+      "step": 1770
+    },
+    {
+      "epoch": 0.6853402637404947,
+      "grad_norm": 0.13472336530685425,
+      "learning_rate": 2.3839192032876486e-06,
+      "loss": 0.7841,
+      "step": 1780
+    },
+    {
+      "epoch": 0.6891904899412841,
+      "grad_norm": 0.1275395303964615,
+      "learning_rate": 2.3309826007617047e-06,
+      "loss": 0.7947,
+      "step": 1790
+    },
+    {
+      "epoch": 0.6930407161420733,
+      "grad_norm": 0.12656980752944946,
+      "learning_rate": 2.278461133078881e-06,
+      "loss": 0.7585,
+      "step": 1800
+    },
+    {
+      "epoch": 0.6968909423428626,
+      "grad_norm": 0.13257886469364166,
+      "learning_rate": 2.2263629693471258e-06,
+      "loss": 0.754,
+      "step": 1810
+    },
+    {
+      "epoch": 0.700741168543652,
+      "grad_norm": 0.13240672647953033,
+      "learning_rate": 2.1746962128343414e-06,
+      "loss": 0.778,
+      "step": 1820
+    },
+    {
+      "epoch": 0.7045913947444412,
+      "grad_norm": 0.13615965843200684,
+      "learning_rate": 2.1234688997080183e-06,
+      "loss": 0.7738,
+      "step": 1830
+    },
+    {
+      "epoch": 0.7084416209452306,
+      "grad_norm": 0.13906416296958923,
+      "learning_rate": 2.0726889977853053e-06,
+      "loss": 0.7708,
+      "step": 1840
+    },
+    {
+      "epoch": 0.7122918471460198,
+      "grad_norm": 0.13431890308856964,
+      "learning_rate": 2.022364405293703e-06,
+      "loss": 0.7792,
+      "step": 1850
+    },
+    {
+      "epoch": 0.7161420733468091,
+      "grad_norm": 0.13306790590286255,
+      "learning_rate": 1.9725029496425814e-06,
+      "loss": 0.7759,
+      "step": 1860
+    },
+    {
+      "epoch": 0.7199922995475985,
+      "grad_norm": 0.1340589076280594,
+      "learning_rate": 1.923112386205727e-06,
+      "loss": 0.7718,
+      "step": 1870
+    },
+    {
+      "epoch": 0.7238425257483877,
+      "grad_norm": 0.1316177099943161,
+      "learning_rate": 1.874200397115079e-06,
+      "loss": 0.7762,
+      "step": 1880
+    },
+    {
+      "epoch": 0.727692751949177,
+      "grad_norm": 0.1352645605802536,
+      "learning_rate": 1.825774590065857e-06,
+      "loss": 0.763,
+      "step": 1890
+    },
+    {
+      "epoch": 0.7315429781499663,
+      "grad_norm": 0.14604684710502625,
+      "learning_rate": 1.7778424971332813e-06,
+      "loss": 0.7706,
+      "step": 1900
+    },
+    {
+      "epoch": 0.7353932043507556,
+      "grad_norm": 0.12245982885360718,
+      "learning_rate": 1.7304115736010407e-06,
+      "loss": 0.7708,
+      "step": 1910
+    },
+    {
+      "epoch": 0.739243430551545,
+      "grad_norm": 0.1327592432498932,
+      "learning_rate": 1.6834891968017108e-06,
+      "loss": 0.7613,
+      "step": 1920
+    },
+    {
+      "epoch": 0.7430936567523342,
+      "grad_norm": 0.13801155984401703,
+      "learning_rate": 1.6370826649692884e-06,
+      "loss": 0.777,
+      "step": 1930
+    },
+    {
+      "epoch": 0.7469438829531235,
+      "grad_norm": 0.1381285935640335,
+      "learning_rate": 1.5911991961040463e-06,
+      "loss": 0.7864,
+      "step": 1940
+    },
+    {
+      "epoch": 0.7507941091539128,
+      "grad_norm": 0.13735860586166382,
+      "learning_rate": 1.545845926849851e-06,
+      "loss": 0.7724,
+      "step": 1950
+    },
+    {
+      "epoch": 0.7546443353547021,
+      "grad_norm": 0.13972720503807068,
+      "learning_rate": 1.5010299113841397e-06,
+      "loss": 0.7874,
+      "step": 1960
+    },
+    {
+      "epoch": 0.7584945615554913,
+      "grad_norm": 0.13210701942443848,
+      "learning_rate": 1.4567581203207238e-06,
+      "loss": 0.7825,
+      "step": 1970
+    },
+    {
+      "epoch": 0.7623447877562807,
+      "grad_norm": 0.13289861381053925,
+      "learning_rate": 1.413037439625593e-06,
+      "loss": 0.767,
+      "step": 1980
+    },
+    {
+      "epoch": 0.76619501395707,
+      "grad_norm": 0.15297633409500122,
+      "learning_rate": 1.3698746695458847e-06,
+      "loss": 0.7694,
+      "step": 1990
+    },
+    {
+      "epoch": 0.7700452401578592,
+      "grad_norm": 0.12530075013637543,
+      "learning_rate": 1.3272765235521745e-06,
+      "loss": 0.7692,
+      "step": 2000
+    },
+    {
+      "epoch": 0.7738954663586486,
+      "grad_norm": 0.13045215606689453,
+      "learning_rate": 1.2852496272942877e-06,
+      "loss": 0.7668,
+      "step": 2010
+    },
+    {
+      "epoch": 0.7777456925594378,
+      "grad_norm": 0.1241084560751915,
+      "learning_rate": 1.2438005175707396e-06,
+      "loss": 0.7537,
+      "step": 2020
+    },
+    {
+      "epoch": 0.7815959187602272,
+      "grad_norm": 0.14476215839385986,
+      "learning_rate": 1.2029356413120297e-06,
+      "loss": 0.7415,
+      "step": 2030
+    },
+    {
+      "epoch": 0.7854461449610165,
+      "grad_norm": 0.14146389067173004,
+      "learning_rate": 1.1626613545778786e-06,
+      "loss": 0.759,
+      "step": 2040
+    },
+    {
+      "epoch": 0.7892963711618057,
+      "grad_norm": 0.1294189840555191,
+      "learning_rate": 1.1229839215686294e-06,
+      "loss": 0.7655,
+      "step": 2050
+    },
+    {
+      "epoch": 0.7931465973625951,
+      "grad_norm": 0.12481428682804108,
+      "learning_rate": 1.0839095136509188e-06,
+      "loss": 0.7468,
+      "step": 2060
+    },
+    {
+      "epoch": 0.7969968235633843,
+      "grad_norm": 0.15507394075393677,
+      "learning_rate": 1.045444208397791e-06,
+      "loss": 0.7604,
+      "step": 2070
+    },
+    {
+      "epoch": 0.8008470497641736,
+      "grad_norm": 0.12914659082889557,
+      "learning_rate": 1.007593988643402e-06,
+      "loss": 0.7438,
+      "step": 2080
+    },
+    {
+      "epoch": 0.804697275964963,
+      "grad_norm": 0.13849639892578125,
+      "learning_rate": 9.703647415524613e-07,
+      "loss": 0.7681,
+      "step": 2090
+    },
+    {
+      "epoch": 0.8085475021657522,
+      "grad_norm": 0.1325039267539978,
+      "learning_rate": 9.337622577045485e-07,
+      "loss": 0.7617,
+      "step": 2100
+    },
+    {
+      "epoch": 0.8123977283665416,
+      "grad_norm": 0.13061149418354034,
+      "learning_rate": 8.977922301934588e-07,
+      "loss": 0.7747,
+      "step": 2110
+    },
+    {
+      "epoch": 0.8162479545673308,
+      "grad_norm": 0.1270773708820343,
+      "learning_rate": 8.624602537417009e-07,
+      "loss": 0.7734,
+      "step": 2120
+    },
+    {
+      "epoch": 0.8200981807681201,
+      "grad_norm": 0.135097935795784,
+      "learning_rate": 8.277718238303123e-07,
+      "loss": 0.7574,
+      "step": 2130
+    },
+    {
+      "epoch": 0.8239484069689095,
+      "grad_norm": 0.13246892392635345,
+      "learning_rate": 7.937323358440935e-07,
+      "loss": 0.7687,
+      "step": 2140
+    },
+    {
+      "epoch": 0.8277986331696987,
+      "grad_norm": 0.1462913453578949,
+      "learning_rate": 7.603470842324223e-07,
+      "loss": 0.7527,
+      "step": 2150
+    },
+    {
+      "epoch": 0.831648859370488,
+      "grad_norm": 0.12965112924575806,
+      "learning_rate": 7.276212616857575e-07,
+      "loss": 0.7698,
+      "step": 2160
+    },
+    {
+      "epoch": 0.8354990855712773,
+      "grad_norm": 0.12776213884353638,
+      "learning_rate": 6.955599583279837e-07,
+      "loss": 0.7716,
+      "step": 2170
+    },
+    {
+      "epoch": 0.8393493117720666,
+      "grad_norm": 0.13238199055194855,
+      "learning_rate": 6.641681609246981e-07,
+      "loss": 0.7588,
+      "step": 2180
+    },
+    {
+      "epoch": 0.8431995379728559,
+      "grad_norm": 0.1374627649784088,
+      "learning_rate": 6.334507521075711e-07,
+      "loss": 0.7712,
+      "step": 2190
+    },
+    {
+      "epoch": 0.8470497641736452,
+      "grad_norm": 0.13433188199996948,
+      "learning_rate": 6.034125096149179e-07,
+      "loss": 0.7542,
+      "step": 2200
+    },
+    {
+      "epoch": 0.8508999903744345,
+      "grad_norm": 0.15401048958301544,
+      "learning_rate": 5.740581055485722e-07,
+      "loss": 0.7417,
+      "step": 2210
+    },
+    {
+      "epoch": 0.8547502165752238,
+      "grad_norm": 0.12312530726194382,
+      "learning_rate": 5.453921056471961e-07,
+      "loss": 0.7662,
+      "step": 2220
+    },
+    {
+      "epoch": 0.8586004427760131,
+      "grad_norm": 0.13862770795822144,
+      "learning_rate": 5.174189685761288e-07,
+      "loss": 0.7566,
+      "step": 2230
+    },
+    {
+      "epoch": 0.8624506689768023,
+      "grad_norm": 0.15337175130844116,
+      "learning_rate": 4.901430452338979e-07,
+      "loss": 0.7756,
+      "step": 2240
+    },
+    {
+      "epoch": 0.8663008951775917,
+      "grad_norm": 0.13048219680786133,
+      "learning_rate": 4.6356857807548286e-07,
+      "loss": 0.7577,
+      "step": 2250
+    },
+    {
+      "epoch": 0.870151121378381,
+      "grad_norm": 0.13689687848091125,
+      "learning_rate": 4.376997004524541e-07,
+      "loss": 0.7443,
+      "step": 2260
+    },
+    {
+      "epoch": 0.8740013475791703,
+      "grad_norm": 0.12887127697467804,
+      "learning_rate": 4.12540435970073e-07,
+      "loss": 0.7496,
+      "step": 2270
+    },
+    {
+      "epoch": 0.8778515737799596,
+      "grad_norm": 0.1253286898136139,
+      "learning_rate": 3.880946978614703e-07,
+      "loss": 0.7689,
+      "step": 2280
+    },
+    {
+      "epoch": 0.8817017999807488,
+      "grad_norm": 0.1271909475326538,
+      "learning_rate": 3.643662883789878e-07,
+      "loss": 0.7693,
+      "step": 2290
+    },
+    {
+      "epoch": 0.8855520261815382,
+      "grad_norm": 0.12707136571407318,
+      "learning_rate": 3.41358898202781e-07,
+      "loss": 0.7505,
+      "step": 2300
+    },
+    {
+      "epoch": 0.8894022523823275,
+      "grad_norm": 0.12200672179460526,
+      "learning_rate": 3.190761058667741e-07,
+      "loss": 0.7516,
+      "step": 2310
+    },
+    {
+      "epoch": 0.8932524785831167,
+      "grad_norm": 0.12649349868297577,
+      "learning_rate": 2.9752137720206463e-07,
+      "loss": 0.7629,
+      "step": 2320
+    },
+    {
+      "epoch": 0.8971027047839061,
+      "grad_norm": 0.13831807672977448,
+      "learning_rate": 2.7669806479785067e-07,
+      "loss": 0.7724,
+      "step": 2330
+    },
+    {
+      "epoch": 0.9009529309846953,
+      "grad_norm": 0.13010922074317932,
+      "learning_rate": 2.566094074799763e-07,
+      "loss": 0.7626,
+      "step": 2340
+    },
+    {
+      "epoch": 0.9048031571854847,
+      "grad_norm": 0.13116800785064697,
+      "learning_rate": 2.3725852980716557e-07,
+      "loss": 0.7469,
+      "step": 2350
+    },
+    {
+      "epoch": 0.908653383386274,
+      "grad_norm": 0.13224147260189056,
+      "learning_rate": 2.1864844158503983e-07,
+      "loss": 0.7631,
+      "step": 2360
+    },
+    {
+      "epoch": 0.9125036095870632,
+      "grad_norm": 0.13986088335514069,
+      "learning_rate": 2.007820373979702e-07,
+      "loss": 0.75,
+      "step": 2370
+    },
+    {
+      "epoch": 0.9163538357878526,
+      "grad_norm": 0.13571591675281525,
+      "learning_rate": 1.8366209615886532e-07,
+      "loss": 0.7541,
+      "step": 2380
+    },
+    {
+      "epoch": 0.9202040619886418,
+      "grad_norm": 0.12901267409324646,
+      "learning_rate": 1.6729128067693613e-07,
+      "loss": 0.7737,
+      "step": 2390
+    },
+    {
+      "epoch": 0.9240542881894311,
+      "grad_norm": 0.13913895189762115,
+      "learning_rate": 1.5167213724353426e-07,
+      "loss": 0.7638,
+      "step": 2400
+    },
+    {
+      "epoch": 0.9279045143902205,
+      "grad_norm": 0.13496410846710205,
+      "learning_rate": 1.3680709523610058e-07,
+      "loss": 0.7502,
+      "step": 2410
+    },
+    {
+      "epoch": 0.9317547405910097,
+      "grad_norm": 0.12399962544441223,
+      "learning_rate": 1.2269846674030418e-07,
+      "loss": 0.7767,
+      "step": 2420
+    },
+    {
+      "epoch": 0.935604966791799,
+      "grad_norm": 0.13872510194778442,
+      "learning_rate": 1.0934844619042595e-07,
+      "loss": 0.7713,
+      "step": 2430
+    },
+    {
+      "epoch": 0.9394551929925883,
+      "grad_norm": 0.1366955041885376,
+      "learning_rate": 9.675911002803961e-08,
+      "loss": 0.7588,
+      "step": 2440
+    },
+    {
+      "epoch": 0.9433054191933776,
+      "grad_norm": 0.14490367472171783,
+      "learning_rate": 8.493241637904381e-08,
+      "loss": 0.7565,
+      "step": 2450
+    },
+    {
+      "epoch": 0.9471556453941669,
+      "grad_norm": 0.13854725658893585,
+      "learning_rate": 7.387020474909689e-08,
+      "loss": 0.7794,
+      "step": 2460
+    },
+    {
+      "epoch": 0.9510058715949562,
+      "grad_norm": 0.12847663462162018,
+      "learning_rate": 6.357419573750856e-08,
+      "loss": 0.7627,
+      "step": 2470
+    },
+    {
+      "epoch": 0.9548560977957455,
+      "grad_norm": 0.1330828070640564,
+      "learning_rate": 5.4045990769613365e-08,
+      "loss": 0.7574,
+      "step": 2480
+    },
+    {
+      "epoch": 0.9587063239965348,
+      "grad_norm": 0.15487876534461975,
+      "learning_rate": 4.5287071847694434e-08,
+      "loss": 0.7562,
+      "step": 2490
+    },
+    {
+      "epoch": 0.9625565501973241,
+      "grad_norm": 0.13544140756130219,
+      "learning_rate": 3.7298801320467817e-08,
+      "loss": 0.7547,
+      "step": 2500
+    },
+    {
+      "epoch": 0.9664067763981133,
+      "grad_norm": 0.13401614129543304,
+      "learning_rate": 3.0082421671192576e-08,
+      "loss": 0.749,
+      "step": 2510
+    },
+    {
+      "epoch": 0.9702570025989027,
+      "grad_norm": 0.12438911944627762,
+      "learning_rate": 2.3639055324410355e-08,
+      "loss": 0.7667,
+      "step": 2520
+    },
+    {
+      "epoch": 0.974107228799692,
+      "grad_norm": 0.1336650252342224,
+      "learning_rate": 1.7969704471369477e-08,
+      "loss": 0.7521,
+      "step": 2530
+    },
+    {
+      "epoch": 0.9779574550004813,
+      "grad_norm": 0.1265072375535965,
+      "learning_rate": 1.3075250914142413e-08,
+      "loss": 0.7601,
+      "step": 2540
+    },
+    {
+      "epoch": 0.9818076812012706,
+      "grad_norm": 0.1250174641609192,
+      "learning_rate": 8.956455928476049e-09,
+      "loss": 0.7712,
+      "step": 2550
+    },
+    {
+      "epoch": 0.9856579074020598,
+      "grad_norm": 0.12185568362474442,
+      "learning_rate": 5.6139601453786275e-09,
+      "loss": 0.7459,
+      "step": 2560
+    },
+    {
+      "epoch": 0.9895081336028492,
+      "grad_norm": 0.13871993124485016,
+      "learning_rate": 3.0482834514805602e-09,
+      "loss": 0.7513,
+      "step": 2570
+    },
+    {
+      "epoch": 0.9933583598036385,
+      "grad_norm": 0.11649906635284424,
+      "learning_rate": 1.2598249081713366e-09,
+      "loss": 0.738,
+      "step": 2580
+    },
+    {
+      "epoch": 0.9972085860044277,
+      "grad_norm": 0.13538645207881927,
+      "learning_rate": 2.488626895302826e-10,
+      "loss": 0.7675,
+      "step": 2590
+    },
+    {
+      "epoch": 0.9999037443449803,
+      "step": 2597,
+      "total_flos": 3.006224099803398e+18,
+      "train_loss": 0.8141947954858685,
+      "train_runtime": 13148.6704,
+      "train_samples_per_second": 50.568,
+      "train_steps_per_second": 0.198
+    }
+  ],
+  "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": 3.006224099803398e+18,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}

continuous/qwen3/siglip2_b16_384/finetune/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:01052b74941f1215adcefdb57f59292610efaf7f0bdfff724a4f157485b2bf7b
+size 6136

continuous/qwen3/siglip2_b16_384/pretrain/config.json

@@ -0,0 +1,60 @@
+{
+  "_attn_implementation_autoset": true,
+  "add_faster_video": false,
+  "add_time_instruction": false,
+  "architectures": [
+    "Qwen3ForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "bos_token_id": 151643,
+  "eos_token_id": 151645,
+  "faster_token_stride": 10,
+  "force_sample": false,
+  "head_dim": 128,
+  "hidden_act": "silu",
+  "hidden_size": 2048,
+  "image_aspect_ratio": "square",
+  "image_crop_resolution": null,
+  "image_grid_pinpoints": null,
+  "image_split_resolution": null,
+  "initializer_range": 0.02,
+  "intermediate_size": 6144,
+  "max_position_embeddings": 40960,
+  "max_window_layers": 28,
+  "mm_hidden_size": 768,
+  "mm_newline_position": "grid",
+  "mm_patch_merge_type": "flat",
+  "mm_projector_lr": null,
+  "mm_projector_type": "mlp2x_gelu",
+  "mm_resampler_type": null,
+  "mm_spatial_pool_mode": "bilinear",
+  "mm_spatial_pool_stride": null,
+  "mm_tunable_parts": "mm_mlp_adapter",
+  "mm_use_im_patch_token": false,
+  "mm_use_im_start_end": false,
+  "mm_vision_select_feature": "patch",
+  "mm_vision_select_layer": -2,
+  "mm_vision_tower": "open_clip_hub:ViT-B-16-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_b16_384.npz",
+  "vocab_size": 151936
+}

continuous/qwen3/siglip2_b16_384/pretrain/mm_projector.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d5bea8fc614c1c4be1514b7e24602e83e1ee067760798094704e12c17cf73c4b
+size 23087160

continuous/qwen3/siglip2_b16_384/pretrain/trainer_state.json

@@ -0,0 +1,421 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.0,
+  "eval_steps": 500,
+  "global_step": 545,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.01834862385321101,
+      "grad_norm": 7.937506198883057,
+      "learning_rate": 0.00032142857142857147,
+      "loss": 9.8801,
+      "step": 10
+    },
+    {
+      "epoch": 0.03669724770642202,
+      "grad_norm": 0.5462223291397095,
+      "learning_rate": 0.0006785714285714287,
+      "loss": 6.6988,
+      "step": 20
+    },
+    {
+      "epoch": 0.05504587155963303,
+      "grad_norm": 0.16430865228176117,
+      "learning_rate": 0.0009999907688176173,
+      "loss": 5.204,
+      "step": 30
+    },
+    {
+      "epoch": 0.07339449541284404,
+      "grad_norm": 0.07302440702915192,
+      "learning_rate": 0.0009988834393115766,
+      "loss": 4.865,
+      "step": 40
+    },
+    {
+      "epoch": 0.09174311926605505,
+      "grad_norm": 0.057309020310640335,
+      "learning_rate": 0.0009959345572940772,
+      "loss": 4.7778,
+      "step": 50
+    },
+    {
+      "epoch": 0.11009174311926606,
+      "grad_norm": 0.057270992547273636,
+      "learning_rate": 0.0009911550081155983,
+      "loss": 4.7311,
+      "step": 60
+    },
+    {
+      "epoch": 0.12844036697247707,
+      "grad_norm": 0.046323101967573166,
+      "learning_rate": 0.0009845624347566433,
+      "loss": 4.687,
+      "step": 70
+    },
+    {
+      "epoch": 0.14678899082568808,
+      "grad_norm": 0.11732953041791916,
+      "learning_rate": 0.0009761811727013547,
+      "loss": 4.6629,
+      "step": 80
+    },
+    {
+      "epoch": 0.1651376146788991,
+      "grad_norm": 0.059729598462581635,
+      "learning_rate": 0.0009660421601067666,
+      "loss": 4.636,
+      "step": 90
+    },
+    {
+      "epoch": 0.1834862385321101,
+      "grad_norm": 0.061038386076688766,
+      "learning_rate": 0.0009541828235992926,
+      "loss": 4.614,
+      "step": 100
+    },
+    {
+      "epoch": 0.2018348623853211,
+      "grad_norm": 0.04457351565361023,
+      "learning_rate": 0.0009406469401200151,
+      "loss": 4.5819,
+      "step": 110
+    },
+    {
+      "epoch": 0.22018348623853212,
+      "grad_norm": 0.2802472412586212,
+      "learning_rate": 0.0009254844753287492,
+      "loss": 4.7207,
+      "step": 120
+    },
+    {
+      "epoch": 0.23853211009174313,
+      "grad_norm": 0.0507575124502182,
+      "learning_rate": 0.0009087513991633924,
+      "loss": 4.7294,
+      "step": 130
+    },
+    {
+      "epoch": 0.25688073394495414,
+      "grad_norm": 0.03221786022186279,
+      "learning_rate": 0.0008905094792353917,
+      "loss": 4.6757,
+      "step": 140
+    },
+    {
+      "epoch": 0.27522935779816515,
+      "grad_norm": 0.026578817516565323,
+      "learning_rate": 0.0008708260528239789,
+      "loss": 4.6406,
+      "step": 150
+    },
+    {
+      "epoch": 0.29357798165137616,
+      "grad_norm": 0.03594611585140228,
+      "learning_rate": 0.0008497737783108237,
+      "loss": 4.6058,
+      "step": 160
+    },
+    {
+      "epoch": 0.3119266055045872,
+      "grad_norm": 0.032315049320459366,
+      "learning_rate": 0.0008274303669726426,
+      "loss": 4.5788,
+      "step": 170
+    },
+    {
+      "epoch": 0.3302752293577982,
+      "grad_norm": 0.028260285034775734,
+      "learning_rate": 0.0008038782961218137,
+      "loss": 4.5532,
+      "step": 180
+    },
+    {
+      "epoch": 0.3486238532110092,
+      "grad_norm": 0.050708819180727005,
+      "learning_rate": 0.000779204504653894,
+      "loss": 4.5355,
+      "step": 190
+    },
+    {
+      "epoch": 0.3669724770642202,
+      "grad_norm": 0.06472963839769363,
+      "learning_rate": 0.0007535000721258791,
+      "loss": 4.5045,
+      "step": 200
+    },
+    {
+      "epoch": 0.3853211009174312,
+      "grad_norm": 0.04333483427762985,
+      "learning_rate": 0.00072685988254984,
+      "loss": 4.4843,
+      "step": 210
+    },
+    {
+      "epoch": 0.4036697247706422,
+      "grad_norm": 0.06802373379468918,
+      "learning_rate": 0.0006993822741429906,
+      "loss": 4.4839,
+      "step": 220
+    },
+    {
+      "epoch": 0.42201834862385323,
+      "grad_norm": 0.05086958408355713,
+      "learning_rate": 0.0006711686763270818,
+      "loss": 4.4699,
+      "step": 230
+    },
+    {
+      "epoch": 0.44036697247706424,
+      "grad_norm": 0.07149030268192291,
+      "learning_rate": 0.0006423232353170831,
+      "loss": 4.4462,
+      "step": 240
+    },
+    {
+      "epoch": 0.45871559633027525,
+      "grad_norm": 0.05015827342867851,
+      "learning_rate": 0.0006129524296812335,
+      "loss": 4.4188,
+      "step": 250
+    },
+    {
+      "epoch": 0.47706422018348627,
+      "grad_norm": 0.06855820119380951,
+      "learning_rate": 0.0005831646772915651,
+      "loss": 4.4248,
+      "step": 260
+    },
+    {
+      "epoch": 0.4954128440366973,
+      "grad_norm": 0.07231036573648453,
+      "learning_rate": 0.0005530699351157809,
+      "loss": 4.3968,
+      "step": 270
+    },
+    {
+      "epoch": 0.5137614678899083,
+      "grad_norm": 0.10077620297670364,
+      "learning_rate": 0.0005227792933277943,
+      "loss": 4.3864,
+      "step": 280
+    },
+    {
+      "epoch": 0.5321100917431193,
+      "grad_norm": 0.072623610496521,
+      "learning_rate": 0.0004924045652352092,
+      "loss": 4.3643,
+      "step": 290
+    },
+    {
+      "epoch": 0.5504587155963303,
+      "grad_norm": 0.06913881748914719,
+      "learning_rate": 0.0004620578745374633,
+      "loss": 4.3578,
+      "step": 300
+    },
+    {
+      "epoch": 0.5688073394495413,
+      "grad_norm": 0.08433078229427338,
+      "learning_rate": 0.0004318512414382059,
+      "loss": 4.3701,
+      "step": 310
+    },
+    {
+      "epoch": 0.5871559633027523,
+      "grad_norm": 0.08106635510921478,
+      "learning_rate": 0.0004018961691397148,
+      "loss": 4.3555,
+      "step": 320
+    },
+    {
+      "epoch": 0.6055045871559633,
+      "grad_norm": 0.06498753279447556,
+      "learning_rate": 0.0003723032322457458,
+      "loss": 4.3343,
+      "step": 330
+    },
+    {
+      "epoch": 0.6238532110091743,
+      "grad_norm": 0.09153219312429428,
+      "learning_rate": 0.0003431816685921625,
+      "loss": 4.3289,
+      "step": 340
+    },
+    {
+      "epoch": 0.6422018348623854,
+      "grad_norm": 0.07757469266653061,
+      "learning_rate": 0.0003146389760120416,
+      "loss": 4.3328,
+      "step": 350
+    },
+    {
+      "epoch": 0.6605504587155964,
+      "grad_norm": 0.09789122641086578,
+      "learning_rate": 0.00028678051552373485,
+      "loss": 4.335,
+      "step": 360
+    },
+    {
+      "epoch": 0.6788990825688074,
+      "grad_norm": 0.06394346803426743,
+      "learning_rate": 0.0002597091224066581,
+      "loss": 4.314,
+      "step": 370
+    },
+    {
+      "epoch": 0.6972477064220184,
+      "grad_norm": 0.05672640725970268,
+      "learning_rate": 0.0002335247266004629,
+      "loss": 4.3279,
+      "step": 380
+    },
+    {
+      "epoch": 0.7155963302752294,
+      "grad_norm": 0.08429772406816483,
+      "learning_rate": 0.000208323983828828,
+      "loss": 4.3093,
+      "step": 390
+    },
+    {
+      "epoch": 0.7339449541284404,
+      "grad_norm": 0.05885216221213341,
+      "learning_rate": 0.0001841999188095224,
+      "loss": 4.307,
+      "step": 400
+    },
+    {
+      "epoch": 0.7522935779816514,
+      "grad_norm": 0.08801107853651047,
+      "learning_rate": 0.00016124158186777676,
+      "loss": 4.3086,
+      "step": 410
+    },
+    {
+      "epoch": 0.7706422018348624,
+      "grad_norm": 0.08172725886106491,
+      "learning_rate": 0.0001395337202205211,
+      "loss": 4.2973,
+      "step": 420
+    },
+    {
+      "epoch": 0.7889908256880734,
+      "grad_norm": 0.07597345113754272,
+      "learning_rate": 0.00011915646514489292,
+      "loss": 4.3019,
+      "step": 430
+    },
+    {
+      "epoch": 0.8073394495412844,
+      "grad_norm": 0.055691882967948914,
+      "learning_rate": 0.00010018503618578818,
+      "loss": 4.3073,
+      "step": 440
+    },
+    {
+      "epoch": 0.8256880733944955,
+      "grad_norm": 0.06062958389520645,
+      "learning_rate": 8.268946349432583e-05,
+      "loss": 4.2992,
+      "step": 450
+    },
+    {
+      "epoch": 0.8440366972477065,
+      "grad_norm": 0.08405902981758118,
+      "learning_rate": 6.673432932217117e-05,
+      "loss": 4.2872,
+      "step": 460
+    },
+    {
+      "epoch": 0.8623853211009175,
+      "grad_norm": 0.06126786768436432,
+      "learning_rate": 5.2378529625954684e-05,
+      "loss": 4.2771,
+      "step": 470
+    },
+    {
+      "epoch": 0.8807339449541285,
+      "grad_norm": 0.05466851219534874,
+      "learning_rate": 3.9675056661785556e-05,
+      "loss": 4.2864,
+      "step": 480
+    },
+    {
+      "epoch": 0.8990825688073395,
+      "grad_norm": 0.0674469992518425,
+      "learning_rate": 2.8670803372379538e-05,
+      "loss": 4.2866,
+      "step": 490
+    },
+    {
+      "epoch": 0.9174311926605505,
+      "grad_norm": 0.05509917810559273,
+      "learning_rate": 1.9406390288876586e-05,
+      "loss": 4.2976,
+      "step": 500
+    },
+    {
+      "epoch": 0.9357798165137615,
+      "grad_norm": 0.05351133272051811,
+      "learning_rate": 1.191601558631461e-05,
+      "loss": 4.2894,
+      "step": 510
+    },
+    {
+      "epoch": 0.9541284403669725,
+      "grad_norm": 0.052223607897758484,
+      "learning_rate": 6.2273288462559305e-06,
+      "loss": 4.2781,
+      "step": 520
+    },
+    {
+      "epoch": 0.9724770642201835,
+      "grad_norm": 0.051613256335258484,
+      "learning_rate": 2.361328992552314e-06,
+      "loss": 4.28,
+      "step": 530
+    },
+    {
+      "epoch": 0.9908256880733946,
+      "grad_norm": 0.05437001213431358,
+      "learning_rate": 3.322867770044202e-07,
+      "loss": 4.2881,
+      "step": 540
+    },
+    {
+      "epoch": 1.0,
+      "step": 545,
+      "total_flos": 2.886683259138212e+17,
+      "train_loss": 4.5994734422876205,
+      "train_runtime": 5678.3345,
+      "train_samples_per_second": 98.291,
+      "train_steps_per_second": 0.096
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 545,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": false,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.886683259138212e+17,
+  "train_batch_size": 32,
+  "trial_name": null,
+  "trial_params": null
+}

continuous/qwen3/siglip2_b16_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_b16_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": 768,
+  "mm_newline_position": "grid",
+  "mm_patch_merge_type": "flat",
+  "mm_projector_lr": 0.0001,
+  "mm_projector_type": "mlp2x_gelu",
+  "mm_resampler_type": null,
+  "mm_spatial_pool_mode": "bilinear",
+  "mm_spatial_pool_stride": null,
+  "mm_tunable_parts": "mm_mlp_adapter,mm_language_model",
+  "mm_use_im_patch_token": false,
+  "mm_use_im_start_end": false,
+  "mm_vision_select_feature": "patch",
+  "mm_vision_select_layer": -2,
+  "mm_vision_tower": "open_clip_hub:ViT-B-16-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_b16_512.npz",
+  "vocab_size": 151936
+}

continuous/qwen3/siglip2_b16_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_b16_512/finetune/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6af32f9ff82758d4ae5ecb0cb67aeed8884ab390689bb8a08a0400f44b937b63
+size 4273663104

continuous/qwen3/siglip2_b16_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_b16_512/finetune/tokenizer.json

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

continuous/qwen3/siglip2_b16_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_b16_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": 8.622236251831055,
+      "learning_rate": 1.153846153846154e-06,
+      "loss": 2.428,
+      "step": 10
+    },
+    {
+      "epoch": 0.007700452401578592,
+      "grad_norm": 3.1859030723571777,
+      "learning_rate": 2.435897435897436e-06,
+      "loss": 1.8624,
+      "step": 20
+    },
+    {
+      "epoch": 0.01155067860236789,
+      "grad_norm": 2.209315299987793,
+      "learning_rate": 3.7179487179487184e-06,
+      "loss": 1.4547,
+      "step": 30
+    },
+    {
+      "epoch": 0.015400904803157185,
+      "grad_norm": 0.4961771070957184,
+      "learning_rate": 5e-06,
+      "loss": 1.1304,
+      "step": 40
+    },
+    {
+      "epoch": 0.019251131003946483,
+      "grad_norm": 0.46895837783813477,
+      "learning_rate": 6.282051282051282e-06,
+      "loss": 1.0335,
+      "step": 50
+    },
+    {
+      "epoch": 0.02310135720473578,
+      "grad_norm": 0.23742976784706116,
+      "learning_rate": 7.564102564102564e-06,
+      "loss": 0.9613,
+      "step": 60
+    },
+    {
+      "epoch": 0.026951583405525073,
+      "grad_norm": 0.1932237446308136,
+      "learning_rate": 8.846153846153847e-06,
+      "loss": 0.9254,
+      "step": 70
+    },
+    {
+      "epoch": 0.03080180960631437,
+      "grad_norm": 0.18635663390159607,
+      "learning_rate": 9.999996111488723e-06,
+      "loss": 0.9238,
+      "step": 80
+    },
+    {
+      "epoch": 0.034652035807103666,
+      "grad_norm": 0.1798827052116394,
+      "learning_rate": 9.999529497453782e-06,
+      "loss": 0.9007,
+      "step": 90
+    },
+    {
+      "epoch": 0.038502262007892966,
+      "grad_norm": 0.16016654670238495,
+      "learning_rate": 9.998285264324193e-06,
+      "loss": 0.8743,
+      "step": 100
+    },
+    {
+      "epoch": 0.04235248820868226,
+      "grad_norm": 0.16860796511173248,
+      "learning_rate": 9.996263605626052e-06,
+      "loss": 0.8515,
+      "step": 110
+    },
+    {
+      "epoch": 0.04620271440947156,
+      "grad_norm": 0.17346420884132385,
+      "learning_rate": 9.99346483580503e-06,
+      "loss": 0.8606,
+      "step": 120
+    },
+    {
+      "epoch": 0.05005294061026085,
+      "grad_norm": 0.15689192712306976,
+      "learning_rate": 9.989889390177467e-06,
+      "loss": 0.8537,
+      "step": 130
+    },
+    {
+      "epoch": 0.053903166811050146,
+      "grad_norm": 0.1647583246231079,
+      "learning_rate": 9.985537824862647e-06,
+      "loss": 0.8497,
+      "step": 140
+    },
+    {
+      "epoch": 0.057753393011839446,
+      "grad_norm": 0.20574039220809937,
+      "learning_rate": 9.980410816696317e-06,
+      "loss": 0.8449,
+      "step": 150
+    },
+    {
+      "epoch": 0.06160361921262874,
+      "grad_norm": 0.16465535759925842,
+      "learning_rate": 9.97450916312541e-06,
+      "loss": 0.8403,
+      "step": 160
+    },
+    {
+      "epoch": 0.06545384541341803,
+      "grad_norm": 0.1415177583694458,
+      "learning_rate": 9.967833782084004e-06,
+      "loss": 0.8371,
+      "step": 170
+    },
+    {
+      "epoch": 0.06930407161420733,
+      "grad_norm": 0.1504281908273697,
+      "learning_rate": 9.960385711850554e-06,
+      "loss": 0.8605,
+      "step": 180
+    },
+    {
+      "epoch": 0.07315429781499663,
+      "grad_norm": 0.15662989020347595,
+      "learning_rate": 9.952166110886394e-06,
+      "loss": 0.859,
+      "step": 190
+    },
+    {
+      "epoch": 0.07700452401578593,
+      "grad_norm": 0.1707412451505661,
+      "learning_rate": 9.943176257655567e-06,
+      "loss": 0.8392,
+      "step": 200
+    },
+    {
+      "epoch": 0.08085475021657522,
+      "grad_norm": 0.16052328050136566,
+      "learning_rate": 9.933417550425952e-06,
+      "loss": 0.8426,
+      "step": 210
+    },
+    {
+      "epoch": 0.08470497641736452,
+      "grad_norm": 0.1564960479736328,
+      "learning_rate": 9.9228915070518e-06,
+      "loss": 0.8196,
+      "step": 220
+    },
+    {
+      "epoch": 0.08855520261815382,
+      "grad_norm": 0.17771422863006592,
+      "learning_rate": 9.911599764737632e-06,
+      "loss": 0.8251,
+      "step": 230
+    },
+    {
+      "epoch": 0.09240542881894312,
+      "grad_norm": 0.16415968537330627,
+      "learning_rate": 9.899544079783606e-06,
+      "loss": 0.8109,
+      "step": 240
+    },
+    {
+      "epoch": 0.0962556550197324,
+      "grad_norm": 0.1658322662115097,
+      "learning_rate": 9.886726327312338e-06,
+      "loss": 0.8258,
+      "step": 250
+    },
+    {
+      "epoch": 0.1001058812205217,
+      "grad_norm": 0.14971713721752167,
+      "learning_rate": 9.87314850097724e-06,
+      "loss": 0.827,
+      "step": 260
+    },
+    {
+      "epoch": 0.103956107421311,
+      "grad_norm": 0.15636467933654785,
+      "learning_rate": 9.858812712652452e-06,
+      "loss": 0.8094,
+      "step": 270
+    },
+    {
+      "epoch": 0.10780633362210029,
+      "grad_norm": 0.14377453923225403,
+      "learning_rate": 9.843721192104341e-06,
+      "loss": 0.8093,
+      "step": 280
+    },
+    {
+      "epoch": 0.11165655982288959,
+      "grad_norm": 0.16836746037006378,
+      "learning_rate": 9.827876286644698e-06,
+      "loss": 0.8184,
+      "step": 290
+    },
+    {
+      "epoch": 0.11550678602367889,
+      "grad_norm": 0.1562587320804596,
+      "learning_rate": 9.811280460765643e-06,
+      "loss": 0.7787,
+      "step": 300
+    },
+    {
+      "epoch": 0.11935701222446819,
+      "grad_norm": 0.1542181521654129,
+      "learning_rate": 9.793936295756292e-06,
+      "loss": 0.7917,
+      "step": 310
+    },
+    {
+      "epoch": 0.12320723842525748,
+      "grad_norm": 0.14966075122356415,
+      "learning_rate": 9.775846489301279e-06,
+      "loss": 0.7891,
+      "step": 320
+    },
+    {
+      "epoch": 0.12705746462604678,
+      "grad_norm": 0.1463336944580078,
+      "learning_rate": 9.757013855061157e-06,
+      "loss": 0.7879,
+      "step": 330
+    },
+    {
+      "epoch": 0.13090769082683607,
+      "grad_norm": 0.17203378677368164,
+      "learning_rate": 9.737441322234758e-06,
+      "loss": 0.7932,
+      "step": 340
+    },
+    {
+      "epoch": 0.13475791702762538,
+      "grad_norm": 0.1663551777601242,
+      "learning_rate": 9.717131935103605e-06,
+      "loss": 0.7989,
+      "step": 350
+    },
+    {
+      "epoch": 0.13860814322841467,
+      "grad_norm": 0.14329446852207184,
+      "learning_rate": 9.696088852558398e-06,
+      "loss": 0.7804,
+      "step": 360
+    },
+    {
+      "epoch": 0.14245836942920398,
+      "grad_norm": 0.16097277402877808,
+      "learning_rate": 9.674315347607686e-06,
+      "loss": 0.8057,
+      "step": 370
+    },
+    {
+      "epoch": 0.14630859562999327,
+      "grad_norm": 0.14663422107696533,
+      "learning_rate": 9.651814806868796e-06,
+      "loss": 0.7739,
+      "step": 380
+    },
+    {
+      "epoch": 0.15015882183078255,
+      "grad_norm": 0.174505814909935,
+      "learning_rate": 9.62859073004107e-06,
+      "loss": 0.7729,
+      "step": 390
+    },
+    {
+      "epoch": 0.15400904803157187,
+      "grad_norm": 0.14290952682495117,
+      "learning_rate": 9.604646729361547e-06,
+      "loss": 0.7738,
+      "step": 400
+    },
+    {
+      "epoch": 0.15785927423236115,
+      "grad_norm": 0.15559697151184082,
+      "learning_rate": 9.579986529043099e-06,
+      "loss": 0.7526,
+      "step": 410
+    },
+    {
+      "epoch": 0.16170950043315044,
+      "grad_norm": 0.14947549998760223,
+      "learning_rate": 9.554613964695189e-06,
+      "loss": 0.775,
+      "step": 420
+    },
+    {
+      "epoch": 0.16555972663393975,
+      "grad_norm": 0.14483171701431274,
+      "learning_rate": 9.528532982727281e-06,
+      "loss": 0.768,
+      "step": 430
+    },
+    {
+      "epoch": 0.16940995283472904,
+      "grad_norm": 0.15229803323745728,
+      "learning_rate": 9.501747639735018e-06,
+      "loss": 0.761,
+      "step": 440
+    },
+    {
+      "epoch": 0.17326017903551832,
+      "grad_norm": 0.14886213839054108,
+      "learning_rate": 9.474262101869272e-06,
+      "loss": 0.7495,
+      "step": 450
+    },
+    {
+      "epoch": 0.17711040523630764,
+      "grad_norm": 0.1468350887298584,
+      "learning_rate": 9.446080644188141e-06,
+      "loss": 0.765,
+      "step": 460
+    },
+    {
+      "epoch": 0.18096063143709692,
+      "grad_norm": 0.15536293387413025,
+      "learning_rate": 9.417207649992017e-06,
+      "loss": 0.7661,
+      "step": 470
+    },
+    {
+      "epoch": 0.18481085763788624,
+      "grad_norm": 0.14100207388401031,
+      "learning_rate": 9.387647610141816e-06,
+      "loss": 0.759,
+      "step": 480
+    },
+    {
+      "epoch": 0.18866108383867553,
+      "grad_norm": 0.1426180899143219,
+      "learning_rate": 9.35740512236047e-06,
+      "loss": 0.753,
+      "step": 490
+    },
+    {
+      "epoch": 0.1925113100394648,
+      "grad_norm": 0.13405033946037292,
+      "learning_rate": 9.326484890517809e-06,
+      "loss": 0.7584,
+      "step": 500
+    },
+    {
+      "epoch": 0.19636153624025413,
+      "grad_norm": 0.14811696112155914,
+      "learning_rate": 9.294891723898926e-06,
+      "loss": 0.7403,
+      "step": 510
+    },
+    {
+      "epoch": 0.2002117624410434,
+      "grad_norm": 0.1378982812166214,
+      "learning_rate": 9.262630536456153e-06,
+      "loss": 0.7802,
+      "step": 520
+    },
+    {
+      "epoch": 0.2040619886418327,
+      "grad_norm": 0.15027783811092377,
+      "learning_rate": 9.229706346044749e-06,
+      "loss": 0.7496,
+      "step": 530
+    },
+    {
+      "epoch": 0.207912214842622,
+      "grad_norm": 0.15850962698459625,
+      "learning_rate": 9.196124273642422e-06,
+      "loss": 0.7668,
+      "step": 540
+    },
+    {
+      "epoch": 0.2117624410434113,
+      "grad_norm": 0.1301824003458023,
+      "learning_rate": 9.161889542552842e-06,
+      "loss": 0.7765,
+      "step": 550
+    },
+    {
+      "epoch": 0.21561266724420058,
+      "grad_norm": 0.1381032019853592,
+      "learning_rate": 9.127007477593193e-06,
+      "loss": 0.7694,
+      "step": 560
+    },
+    {
+      "epoch": 0.2194628934449899,
+      "grad_norm": 0.15555031597614288,
+      "learning_rate": 9.091483504265974e-06,
+      "loss": 0.7758,
+      "step": 570
+    },
+    {
+      "epoch": 0.22331311964577918,
+      "grad_norm": 0.1407177448272705,
+      "learning_rate": 9.055323147915116e-06,
+      "loss": 0.7542,
+      "step": 580
+    },
+    {
+      "epoch": 0.2271633458465685,
+      "grad_norm": 0.14297400414943695,
+      "learning_rate": 9.018532032866585e-06,
+      "loss": 0.7693,
+      "step": 590
+    },
+    {
+      "epoch": 0.23101357204735778,
+      "grad_norm": 0.1447652131319046,
+      "learning_rate": 8.981115881553579e-06,
+      "loss": 0.7658,
+      "step": 600
+    },
+    {
+      "epoch": 0.23486379824814707,
+      "grad_norm": 0.15408623218536377,
+      "learning_rate": 8.943080513626476e-06,
+      "loss": 0.7779,
+      "step": 610
+    },
+    {
+      "epoch": 0.23871402444893638,
+      "grad_norm": 0.1409589797258377,
+      "learning_rate": 8.904431845047641e-06,
+      "loss": 0.7698,
+      "step": 620
+    },
+    {
+      "epoch": 0.24256425064972567,
+      "grad_norm": 0.1448596566915512,
+      "learning_rate": 8.86517588717128e-06,
+      "loss": 0.7737,
+      "step": 630
+    },
+    {
+      "epoch": 0.24641447685051496,
+      "grad_norm": 0.1325070858001709,
+      "learning_rate": 8.82531874580844e-06,
+      "loss": 0.7752,
+      "step": 640
+    },
+    {
+      "epoch": 0.25026470305130427,
+      "grad_norm": 0.1298934817314148,
+      "learning_rate": 8.784866620277314e-06,
+      "loss": 0.7288,
+      "step": 650
+    },
+    {
+      "epoch": 0.25411492925209356,
+      "grad_norm": 0.14275096356868744,
+      "learning_rate": 8.743825802439018e-06,
+      "loss": 0.7442,
+      "step": 660
+    },
+    {
+      "epoch": 0.25796515545288284,
+      "grad_norm": 0.14202426373958588,
+      "learning_rate": 8.702202675718956e-06,
+      "loss": 0.7712,
+      "step": 670
+    },
+    {
+      "epoch": 0.26181538165367213,
+      "grad_norm": 0.14132381975650787,
+      "learning_rate": 8.660003714113959e-06,
+      "loss": 0.7415,
+      "step": 680
+    },
+    {
+      "epoch": 0.26566560785446147,
+      "grad_norm": 0.13455110788345337,
+      "learning_rate": 8.617235481185323e-06,
+      "loss": 0.7538,
+      "step": 690
+    },
+    {
+      "epoch": 0.26951583405525076,
+      "grad_norm": 0.15685854852199554,
+      "learning_rate": 8.573904629037931e-06,
+      "loss": 0.7626,
+      "step": 700
+    },
+    {
+      "epoch": 0.27336606025604004,
+      "grad_norm": 0.13809734582901,
+      "learning_rate": 8.530017897285585e-06,
+      "loss": 0.7475,
+      "step": 710
+    },
+    {
+      "epoch": 0.27721628645682933,
+      "grad_norm": 0.12390965223312378,
+      "learning_rate": 8.485582112002746e-06,
+      "loss": 0.7565,
+      "step": 720
+    },
+    {
+      "epoch": 0.2810665126576186,
+      "grad_norm": 0.15313175320625305,
+      "learning_rate": 8.440604184662811e-06,
+      "loss": 0.7579,
+      "step": 730
+    },
+    {
+      "epoch": 0.28491673885840796,
+      "grad_norm": 0.14157141745090485,
+      "learning_rate": 8.395091111063118e-06,
+      "loss": 0.7427,
+      "step": 740
+    },
+    {
+      "epoch": 0.28876696505919724,
+      "grad_norm": 0.13048799335956573,
+      "learning_rate": 8.349049970236822e-06,
+      "loss": 0.7452,
+      "step": 750
+    },
+    {
+      "epoch": 0.29261719125998653,
+      "grad_norm": 0.1443755179643631,
+      "learning_rate": 8.302487923351842e-06,
+      "loss": 0.7341,
+      "step": 760
+    },
+    {
+      "epoch": 0.2964674174607758,
+      "grad_norm": 0.13631878793239594,
+      "learning_rate": 8.255412212597021e-06,
+      "loss": 0.7542,
+      "step": 770
+    },
+    {
+      "epoch": 0.3003176436615651,
+      "grad_norm": 0.1399456262588501,
+      "learning_rate": 8.207830160055688e-06,
+      "loss": 0.7498,
+      "step": 780
+    },
+    {
+      "epoch": 0.3041678698623544,
+      "grad_norm": 0.14947646856307983,
+      "learning_rate": 8.159749166566786e-06,
+      "loss": 0.7341,
+      "step": 790
+    },
+    {
+      "epoch": 0.30801809606314373,
+      "grad_norm": 0.13226565718650818,
+      "learning_rate": 8.11117671057377e-06,
+      "loss": 0.7584,
+      "step": 800
+    },
+    {
+      "epoch": 0.311868322263933,
+      "grad_norm": 0.13962426781654358,
+      "learning_rate": 8.062120346961404e-06,
+      "loss": 0.7438,
+      "step": 810
+    },
+    {
+      "epoch": 0.3157185484647223,
+      "grad_norm": 0.13481883704662323,
+      "learning_rate": 8.012587705880705e-06,
+      "loss": 0.7503,
+      "step": 820
+    },
+    {
+      "epoch": 0.3195687746655116,
+      "grad_norm": 0.13373404741287231,
+      "learning_rate": 7.962586491562146e-06,
+      "loss": 0.7389,
+      "step": 830
+    },
+    {
+      "epoch": 0.3234190008663009,
+      "grad_norm": 0.1385193169116974,
+      "learning_rate": 7.912124481117357e-06,
+      "loss": 0.7405,
+      "step": 840
+    },
+    {
+      "epoch": 0.3272692270670902,
+      "grad_norm": 0.13182389736175537,
+      "learning_rate": 7.861209523329494e-06,
+      "loss": 0.7463,
+      "step": 850
+    },
+    {
+      "epoch": 0.3311194532678795,
+      "grad_norm": 0.1370089203119278,
+      "learning_rate": 7.809849537432432e-06,
+      "loss": 0.7357,
+      "step": 860
+    },
+    {
+      "epoch": 0.3349696794686688,
+      "grad_norm": 0.13157522678375244,
+      "learning_rate": 7.758052511879035e-06,
+      "loss": 0.7337,
+      "step": 870
+    },
+    {
+      "epoch": 0.3388199056694581,
+      "grad_norm": 0.12768374383449554,
+      "learning_rate": 7.705826503098642e-06,
+      "loss": 0.7405,
+      "step": 880
+    },
+    {
+      "epoch": 0.34267013187024736,
+      "grad_norm": 0.13735149800777435,
+      "learning_rate": 7.653179634243979e-06,
+      "loss": 0.7359,
+      "step": 890
+    },
+    {
+      "epoch": 0.34652035807103665,
+      "grad_norm": 0.1276766061782837,
+      "learning_rate": 7.600120093927696e-06,
+      "loss": 0.7297,
+      "step": 900
+    },
+    {
+      "epoch": 0.350370584271826,
+      "grad_norm": 0.14314362406730652,
+      "learning_rate": 7.546656134948729e-06,
+      "loss": 0.7301,
+      "step": 910
+    },
+    {
+      "epoch": 0.3542208104726153,
+      "grad_norm": 0.13401663303375244,
+      "learning_rate": 7.492796073008661e-06,
+      "loss": 0.7254,
+      "step": 920
+    },
+    {
+      "epoch": 0.35807103667340456,
+      "grad_norm": 0.14241771399974823,
+      "learning_rate": 7.438548285418326e-06,
+      "loss": 0.7186,
+      "step": 930
+    },
+    {
+      "epoch": 0.36192126287419385,
+      "grad_norm": 0.14183223247528076,
+      "learning_rate": 7.3839212097948e-06,
+      "loss": 0.7563,
+      "step": 940
+    },
+    {
+      "epoch": 0.36577148907498314,
+      "grad_norm": 0.1333906203508377,
+      "learning_rate": 7.328923342749039e-06,
+      "loss": 0.7697,
+      "step": 950
+    },
+    {
+      "epoch": 0.3696217152757725,
+      "grad_norm": 0.13790810108184814,
+      "learning_rate": 7.273563238564319e-06,
+      "loss": 0.7456,
+      "step": 960
+    },
+    {
+      "epoch": 0.37347194147656176,
+      "grad_norm": 0.13812369108200073,
+      "learning_rate": 7.217849507865724e-06,
+      "loss": 0.7219,
+      "step": 970
+    },
+    {
+      "epoch": 0.37732216767735105,
+      "grad_norm": 0.13122016191482544,
+      "learning_rate": 7.161790816280858e-06,
+      "loss": 0.7389,
+      "step": 980
+    },
+    {
+      "epoch": 0.38117239387814034,
+      "grad_norm": 0.13433431088924408,
+      "learning_rate": 7.1053958830920125e-06,
+      "loss": 0.7118,
+      "step": 990
+    },
+    {
+      "epoch": 0.3850226200789296,
+      "grad_norm": 0.13221457600593567,
+      "learning_rate": 7.0486734798799715e-06,
+      "loss": 0.7452,
+      "step": 1000
+    },
+    {
+      "epoch": 0.3888728462797189,
+      "grad_norm": 0.14273682236671448,
+      "learning_rate": 6.991632429159708e-06,
+      "loss": 0.766,
+      "step": 1010
+    },
+    {
+      "epoch": 0.39272307248050825,
+      "grad_norm": 0.14686277508735657,
+      "learning_rate": 6.934281603008132e-06,
+      "loss": 0.7379,
+      "step": 1020
+    },
+    {
+      "epoch": 0.39657329868129754,
+      "grad_norm": 0.13202159106731415,
+      "learning_rate": 6.876629921684143e-06,
+      "loss": 0.7504,
+      "step": 1030
+    },
+    {
+      "epoch": 0.4004235248820868,
+      "grad_norm": 0.13106420636177063,
+      "learning_rate": 6.818686352241199e-06,
+      "loss": 0.7463,
+      "step": 1040
+    },
+    {
+      "epoch": 0.4042737510828761,
+      "grad_norm": 0.12878736853599548,
+      "learning_rate": 6.760459907132573e-06,
+      "loss": 0.7345,
+      "step": 1050
+    },
+    {
+      "epoch": 0.4081239772836654,
+      "grad_norm": 0.13919547200202942,
+      "learning_rate": 6.701959642809589e-06,
+      "loss": 0.7383,
+      "step": 1060
+    },
+    {
+      "epoch": 0.41197420348445474,
+      "grad_norm": 0.13864408433437347,
+      "learning_rate": 6.643194658312987e-06,
+      "loss": 0.7222,
+      "step": 1070
+    },
+    {
+      "epoch": 0.415824429685244,
+      "grad_norm": 0.1281714141368866,
+      "learning_rate": 6.584174093857676e-06,
+      "loss": 0.7586,
+      "step": 1080
+    },
+    {
+      "epoch": 0.4196746558860333,
+      "grad_norm": 0.1515379548072815,
+      "learning_rate": 6.5249071294110735e-06,
+      "loss": 0.7325,
+      "step": 1090
+    },
+    {
+      "epoch": 0.4235248820868226,
+      "grad_norm": 0.1574833244085312,
+      "learning_rate": 6.465402983265274e-06,
+      "loss": 0.7497,
+      "step": 1100
+    },
+    {
+      "epoch": 0.4273751082876119,
+      "grad_norm": 0.1405971348285675,
+      "learning_rate": 6.405670910603243e-06,
+      "loss": 0.746,
+      "step": 1110
+    },
+    {
+      "epoch": 0.43122533448840117,
+      "grad_norm": 0.14868763089179993,
+      "learning_rate": 6.345720202059291e-06,
+      "loss": 0.7381,
+      "step": 1120
+    },
+    {
+      "epoch": 0.4350755606891905,
+      "grad_norm": 0.15038363635540009,
+      "learning_rate": 6.285560182274008e-06,
+      "loss": 0.7267,
+      "step": 1130
+    },
+    {
+      "epoch": 0.4389257868899798,
+      "grad_norm": 0.13452117145061493,
+      "learning_rate": 6.225200208443936e-06,
+      "loss": 0.7337,
+      "step": 1140
+    },
+    {
+      "epoch": 0.4427760130907691,
+      "grad_norm": 0.1237647533416748,
+      "learning_rate": 6.16464966886616e-06,
+      "loss": 0.7276,
+      "step": 1150
+    },
+    {
+      "epoch": 0.44662623929155837,
+      "grad_norm": 0.15508653223514557,
+      "learning_rate": 6.103917981478065e-06,
+      "loss": 0.7589,
+      "step": 1160
+    },
+    {
+      "epoch": 0.45047646549234766,
+      "grad_norm": 0.13297973573207855,
+      "learning_rate": 6.043014592392489e-06,
+      "loss": 0.7614,
+      "step": 1170
+    },
+    {
+      "epoch": 0.454326691693137,
+      "grad_norm": 0.12665362656116486,
+      "learning_rate": 5.981948974428485e-06,
+      "loss": 0.7377,
+      "step": 1180
+    },
+    {
+      "epoch": 0.4581769178939263,
+      "grad_norm": 0.13431183993816376,
+      "learning_rate": 5.920730625637934e-06,
+      "loss": 0.7148,
+      "step": 1190
+    },
+    {
+      "epoch": 0.46202714409471557,
+      "grad_norm": 0.13781076669692993,
+      "learning_rate": 5.8593690678282355e-06,
+      "loss": 0.7282,
+      "step": 1200
+    },
+    {
+      "epoch": 0.46587737029550486,
+      "grad_norm": 0.14039716124534607,
+      "learning_rate": 5.797873845081301e-06,
+      "loss": 0.7312,
+      "step": 1210
+    },
+    {
+      "epoch": 0.46972759649629414,
+      "grad_norm": 0.13665497303009033,
+      "learning_rate": 5.736254522269075e-06,
+      "loss": 0.7454,
+      "step": 1220
+    },
+    {
+      "epoch": 0.47357782269708343,
+      "grad_norm": 0.12994715571403503,
+      "learning_rate": 5.6745206835658386e-06,
+      "loss": 0.7493,
+      "step": 1230
+    },
+    {
+      "epoch": 0.47742804889787277,
+      "grad_norm": 0.13454414904117584,
+      "learning_rate": 5.612681930957495e-06,
+      "loss": 0.7278,
+      "step": 1240
+    },
+    {
+      "epoch": 0.48127827509866206,
+      "grad_norm": 0.12686799466609955,
+      "learning_rate": 5.550747882748095e-06,
+      "loss": 0.7204,
+      "step": 1250
+    },
+    {
+      "epoch": 0.48512850129945134,
+      "grad_norm": 0.14200155436992645,
+      "learning_rate": 5.4887281720638195e-06,
+      "loss": 0.7291,
+      "step": 1260
+    },
+    {
+      "epoch": 0.48897872750024063,
+      "grad_norm": 0.15462824702262878,
+      "learning_rate": 5.42663244535466e-06,
+      "loss": 0.7147,
+      "step": 1270
+    },
+    {
+      "epoch": 0.4928289537010299,
+      "grad_norm": 0.13694660365581512,
+      "learning_rate": 5.3644703608940205e-06,
+      "loss": 0.7201,
+      "step": 1280
+    },
+    {
+      "epoch": 0.49667917990181926,
+      "grad_norm": 0.1311800479888916,
+      "learning_rate": 5.302251587276491e-06,
+      "loss": 0.7423,
+      "step": 1290
+    },
+    {
+      "epoch": 0.5005294061026085,
+      "grad_norm": 0.13011717796325684,
+      "learning_rate": 5.2399858019140005e-06,
+      "loss": 0.7345,
+      "step": 1300
+    },
+    {
+      "epoch": 0.5043796323033978,
+      "grad_norm": 0.13660869002342224,
+      "learning_rate": 5.177682689530615e-06,
+      "loss": 0.751,
+      "step": 1310
+    },
+    {
+      "epoch": 0.5082298585041871,
+      "grad_norm": 0.1572946012020111,
+      "learning_rate": 5.11535194065619e-06,
+      "loss": 0.7227,
+      "step": 1320
+    },
+    {
+      "epoch": 0.5120800847049765,
+      "grad_norm": 0.15328024327754974,
+      "learning_rate": 5.053003250119113e-06,
+      "loss": 0.7234,
+      "step": 1330
+    },
+    {
+      "epoch": 0.5159303109057657,
+      "grad_norm": 0.1376146376132965,
+      "learning_rate": 4.9906463155383925e-06,
+      "loss": 0.736,
+      "step": 1340
+    },
+    {
+      "epoch": 0.519780537106555,
+      "grad_norm": 0.13349701464176178,
+      "learning_rate": 4.9282908358153046e-06,
+      "loss": 0.7289,
+      "step": 1350
+    },
+    {
+      "epoch": 0.5236307633073443,
+      "grad_norm": 0.15923210978507996,
+      "learning_rate": 4.8659465096248345e-06,
+      "loss": 0.7332,
+      "step": 1360
+    },
+    {
+      "epoch": 0.5274809895081336,
+      "grad_norm": 0.1336156725883484,
+      "learning_rate": 4.803623033907169e-06,
+      "loss": 0.7392,
+      "step": 1370
+    },
+    {
+      "epoch": 0.5313312157089229,
+      "grad_norm": 0.12731797993183136,
+      "learning_rate": 4.741330102359441e-06,
+      "loss": 0.7482,
+      "step": 1380
+    },
+    {
+      "epoch": 0.5351814419097122,
+      "grad_norm": 0.1390356719493866,
+      "learning_rate": 4.679077403927996e-06,
+      "loss": 0.7195,
+      "step": 1390
+    },
+    {
+      "epoch": 0.5390316681105015,
+      "grad_norm": 0.1362055242061615,
+      "learning_rate": 4.616874621301373e-06,
+      "loss": 0.7289,
+      "step": 1400
+    },
+    {
+      "epoch": 0.5428818943112907,
+      "grad_norm": 0.14385046064853668,
+      "learning_rate": 4.554731429404293e-06,
+      "loss": 0.746,
+      "step": 1410
+    },
+    {
+      "epoch": 0.5467321205120801,
+      "grad_norm": 0.14400991797447205,
+      "learning_rate": 4.492657493892822e-06,
+      "loss": 0.725,
+      "step": 1420
+    },
+    {
+      "epoch": 0.5505823467128694,
+      "grad_norm": 0.144009530544281,
+      "learning_rate": 4.430662469650999e-06,
+      "loss": 0.7381,
+      "step": 1430
+    },
+    {
+      "epoch": 0.5544325729136587,
+      "grad_norm": 0.14204004406929016,
+      "learning_rate": 4.368755999289118e-06,
+      "loss": 0.7434,
+      "step": 1440
+    },
+    {
+      "epoch": 0.558282799114448,
+      "grad_norm": 0.1603064239025116,
+      "learning_rate": 4.306947711643949e-06,
+      "loss": 0.7237,
+      "step": 1450
+    },
+    {
+      "epoch": 0.5621330253152372,
+      "grad_norm": 0.13523414731025696,
+      "learning_rate": 4.245247220281068e-06,
+      "loss": 0.7316,
+      "step": 1460
+    },
+    {
+      "epoch": 0.5659832515160266,
+      "grad_norm": 0.13192573189735413,
+      "learning_rate": 4.183664121999587e-06,
+      "loss": 0.7247,
+      "step": 1470
+    },
+    {
+      "epoch": 0.5698334777168159,
+      "grad_norm": 0.14024165272712708,
+      "learning_rate": 4.122207995339474e-06,
+      "loss": 0.7238,
+      "step": 1480
+    },
+    {
+      "epoch": 0.5736837039176051,
+      "grad_norm": 0.13460655510425568,
+      "learning_rate": 4.060888399091734e-06,
+      "loss": 0.7303,
+      "step": 1490
+    },
+    {
+      "epoch": 0.5775339301183945,
+      "grad_norm": 0.138048455119133,
+      "learning_rate": 3.999714870811639e-06,
+      "loss": 0.7438,
+      "step": 1500
+    },
+    {
+      "epoch": 0.5813841563191837,
+      "grad_norm": 0.12785521149635315,
+      "learning_rate": 3.938696925335275e-06,
+      "loss": 0.7317,
+      "step": 1510
+    },
+    {
+      "epoch": 0.5852343825199731,
+      "grad_norm": 0.14673423767089844,
+      "learning_rate": 3.87784405329962e-06,
+      "loss": 0.7376,
+      "step": 1520
+    },
+    {
+      "epoch": 0.5890846087207623,
+      "grad_norm": 0.1396411657333374,
+      "learning_rate": 3.817165719666382e-06,
+      "loss": 0.7357,
+      "step": 1530
+    },
+    {
+      "epoch": 0.5929348349215516,
+      "grad_norm": 0.13158120214939117,
+      "learning_rate": 3.756671362249841e-06,
+      "loss": 0.748,
+      "step": 1540
+    },
+    {
+      "epoch": 0.596785061122341,
+      "grad_norm": 0.13879969716072083,
+      "learning_rate": 3.6963703902488907e-06,
+      "loss": 0.737,
+      "step": 1550
+    },
+    {
+      "epoch": 0.6006352873231302,
+      "grad_norm": 0.1309322565793991,
+      "learning_rate": 3.6362721827835655e-06,
+      "loss": 0.7458,
+      "step": 1560
+    },
+    {
+      "epoch": 0.6044855135239195,
+      "grad_norm": 0.13760609924793243,
+      "learning_rate": 3.576386087436212e-06,
+      "loss": 0.7373,
+      "step": 1570
+    },
+    {
+      "epoch": 0.6083357397247088,
+      "grad_norm": 0.13804788887500763,
+      "learning_rate": 3.5167214187975913e-06,
+      "loss": 0.7409,
+      "step": 1580
+    },
+    {
+      "epoch": 0.6121859659254981,
+      "grad_norm": 0.12891414761543274,
+      "learning_rate": 3.457287457018086e-06,
+      "loss": 0.7361,
+      "step": 1590
+    },
+    {
+      "epoch": 0.6160361921262875,
+      "grad_norm": 0.12878064811229706,
+      "learning_rate": 3.3980934463643003e-06,
+      "loss": 0.7233,
+      "step": 1600
+    },
+    {
+      "epoch": 0.6198864183270767,
+      "grad_norm": 0.1220131665468216,
+      "learning_rate": 3.3391485937812053e-06,
+      "loss": 0.7029,
+      "step": 1610
+    },
+    {
+      "epoch": 0.623736644527866,
+      "grad_norm": 0.13417959213256836,
+      "learning_rate": 3.280462067460115e-06,
+      "loss": 0.7041,
+      "step": 1620
+    },
+    {
+      "epoch": 0.6275868707286553,
+      "grad_norm": 0.12697157263755798,
+      "learning_rate": 3.222042995412669e-06,
+      "loss": 0.7436,
+      "step": 1630
+    },
+    {
+      "epoch": 0.6314370969294446,
+      "grad_norm": 0.12070801854133606,
+      "learning_rate": 3.163900464051089e-06,
+      "loss": 0.7246,
+      "step": 1640
+    },
+    {
+      "epoch": 0.635287323130234,
+      "grad_norm": 0.13119657337665558,
+      "learning_rate": 3.1060435167748826e-06,
+      "loss": 0.7345,
+      "step": 1650
+    },
+    {
+      "epoch": 0.6391375493310232,
+      "grad_norm": 0.12942147254943848,
+      "learning_rate": 3.048481152564253e-06,
+      "loss": 0.7223,
+      "step": 1660
+    },
+    {
+      "epoch": 0.6429877755318125,
+      "grad_norm": 0.13989965617656708,
+      "learning_rate": 2.9912223245804017e-06,
+      "loss": 0.7474,
+      "step": 1670
+    },
+    {
+      "epoch": 0.6468380017326018,
+      "grad_norm": 0.13832007348537445,
+      "learning_rate": 2.934275938772981e-06,
+      "loss": 0.7259,
+      "step": 1680
+    },
+    {
+      "epoch": 0.6506882279333911,
+      "grad_norm": 0.14065223932266235,
+      "learning_rate": 2.8776508524948676e-06,
+      "loss": 0.7159,
+      "step": 1690
+    },
+    {
+      "epoch": 0.6545384541341804,
+      "grad_norm": 0.1265052855014801,
+      "learning_rate": 2.821355873124503e-06,
+      "loss": 0.7147,
+      "step": 1700
+    },
+    {
+      "epoch": 0.6583886803349697,
+      "grad_norm": 0.13621960580348969,
+      "learning_rate": 2.7653997566960144e-06,
+      "loss": 0.7369,
+      "step": 1710
+    },
+    {
+      "epoch": 0.662238906535759,
+      "grad_norm": 0.13467611372470856,
+      "learning_rate": 2.709791206537309e-06,
+      "loss": 0.7255,
+      "step": 1720
+    },
+    {
+      "epoch": 0.6660891327365482,
+      "grad_norm": 0.13247975707054138,
+      "learning_rate": 2.6545388719163747e-06,
+      "loss": 0.7223,
+      "step": 1730
+    },
+    {
+      "epoch": 0.6699393589373376,
+      "grad_norm": 0.12668092548847198,
+      "learning_rate": 2.599651346695979e-06,
+      "loss": 0.7342,
+      "step": 1740
+    },
+    {
+      "epoch": 0.6737895851381268,
+      "grad_norm": 0.1439962089061737,
+      "learning_rate": 2.545137167997007e-06,
+      "loss": 0.7285,
+      "step": 1750
+    },
+    {
+      "epoch": 0.6776398113389162,
+      "grad_norm": 0.12523972988128662,
+      "learning_rate": 2.4910048148705972e-06,
+      "loss": 0.7121,
+      "step": 1760
+    },
+    {
+      "epoch": 0.6814900375397055,
+      "grad_norm": 0.14067202806472778,
+      "learning_rate": 2.4372627069793327e-06,
+      "loss": 0.7568,
+      "step": 1770
+    },
+    {
+      "epoch": 0.6853402637404947,
+      "grad_norm": 0.13962984085083008,
+      "learning_rate": 2.3839192032876486e-06,
+      "loss": 0.74,
+      "step": 1780
+    },
+    {
+      "epoch": 0.6891904899412841,
+      "grad_norm": 0.12590590119361877,
+      "learning_rate": 2.3309826007617047e-06,
+      "loss": 0.75,
+      "step": 1790
+    },
+    {
+      "epoch": 0.6930407161420733,
+      "grad_norm": 0.1284220814704895,
+      "learning_rate": 2.278461133078881e-06,
+      "loss": 0.7146,
+      "step": 1800
+    },
+    {
+      "epoch": 0.6968909423428626,
+      "grad_norm": 0.1321951299905777,
+      "learning_rate": 2.2263629693471258e-06,
+      "loss": 0.7097,
+      "step": 1810
+    },
+    {
+      "epoch": 0.700741168543652,
+      "grad_norm": 0.12627799808979034,
+      "learning_rate": 2.1746962128343414e-06,
+      "loss": 0.7336,
+      "step": 1820
+    },
+    {
+      "epoch": 0.7045913947444412,
+      "grad_norm": 0.1328195482492447,
+      "learning_rate": 2.1234688997080183e-06,
+      "loss": 0.7318,
+      "step": 1830
+    },
+    {
+      "epoch": 0.7084416209452306,
+      "grad_norm": 0.1458960771560669,
+      "learning_rate": 2.0726889977853053e-06,
+      "loss": 0.7277,
+      "step": 1840
+    },
+    {
+      "epoch": 0.7122918471460198,
+      "grad_norm": 0.13988646864891052,
+      "learning_rate": 2.022364405293703e-06,
+      "loss": 0.7407,
+      "step": 1850
+    },
+    {
+      "epoch": 0.7161420733468091,
+      "grad_norm": 0.13224364817142487,
+      "learning_rate": 1.9725029496425814e-06,
+      "loss": 0.7359,
+      "step": 1860
+    },
+    {
+      "epoch": 0.7199922995475985,
+      "grad_norm": 0.13688234984874725,
+      "learning_rate": 1.923112386205727e-06,
+      "loss": 0.7316,
+      "step": 1870
+    },
+    {
+      "epoch": 0.7238425257483877,
+      "grad_norm": 0.14131563901901245,
+      "learning_rate": 1.874200397115079e-06,
+      "loss": 0.7377,
+      "step": 1880
+    },
+    {
+      "epoch": 0.727692751949177,
+      "grad_norm": 0.13018988072872162,
+      "learning_rate": 1.825774590065857e-06,
+      "loss": 0.7235,
+      "step": 1890
+    },
+    {
+      "epoch": 0.7315429781499663,
+      "grad_norm": 0.13743458688259125,
+      "learning_rate": 1.7778424971332813e-06,
+      "loss": 0.7313,
+      "step": 1900
+    },
+    {
+      "epoch": 0.7353932043507556,
+      "grad_norm": 0.12978999316692352,
+      "learning_rate": 1.7304115736010407e-06,
+      "loss": 0.7315,
+      "step": 1910
+    },
+    {
+      "epoch": 0.739243430551545,
+      "grad_norm": 0.13028721511363983,
+      "learning_rate": 1.6834891968017108e-06,
+      "loss": 0.7236,
+      "step": 1920
+    },
+    {
+      "epoch": 0.7430936567523342,
+      "grad_norm": 0.14371678233146667,
+      "learning_rate": 1.6370826649692884e-06,
+      "loss": 0.7349,
+      "step": 1930
+    },
+    {
+      "epoch": 0.7469438829531235,
+      "grad_norm": 0.14044028520584106,
+      "learning_rate": 1.5911991961040463e-06,
+      "loss": 0.7482,
+      "step": 1940
+    },
+    {
+      "epoch": 0.7507941091539128,
+      "grad_norm": 0.1342523843050003,
+      "learning_rate": 1.545845926849851e-06,
+      "loss": 0.7361,
+      "step": 1950
+    },
+    {
+      "epoch": 0.7546443353547021,
+      "grad_norm": 0.15175294876098633,
+      "learning_rate": 1.5010299113841397e-06,
+      "loss": 0.75,
+      "step": 1960
+    },
+    {
+      "epoch": 0.7584945615554913,
+      "grad_norm": 0.13524343073368073,
+      "learning_rate": 1.4567581203207238e-06,
+      "loss": 0.7475,
+      "step": 1970
+    },
+    {
+      "epoch": 0.7623447877562807,
+      "grad_norm": 0.13587792217731476,
+      "learning_rate": 1.413037439625593e-06,
+      "loss": 0.7287,
+      "step": 1980
+    },
+    {
+      "epoch": 0.76619501395707,
+      "grad_norm": 0.1510016918182373,
+      "learning_rate": 1.3698746695458847e-06,
+      "loss": 0.733,
+      "step": 1990
+    },
+    {
+      "epoch": 0.7700452401578592,
+      "grad_norm": 0.13323615491390228,
+      "learning_rate": 1.3272765235521745e-06,
+      "loss": 0.733,
+      "step": 2000
+    },
+    {
+      "epoch": 0.7738954663586486,
+      "grad_norm": 0.12802083790302277,
+      "learning_rate": 1.2852496272942877e-06,
+      "loss": 0.7321,
+      "step": 2010
+    },
+    {
+      "epoch": 0.7777456925594378,
+      "grad_norm": 0.1393044888973236,
+      "learning_rate": 1.2438005175707396e-06,
+      "loss": 0.7171,
+      "step": 2020
+    },
+    {
+      "epoch": 0.7815959187602272,
+      "grad_norm": 0.13912713527679443,
+      "learning_rate": 1.2029356413120297e-06,
+      "loss": 0.7026,
+      "step": 2030
+    },
+    {
+      "epoch": 0.7854461449610165,
+      "grad_norm": 0.13468778133392334,
+      "learning_rate": 1.1626613545778786e-06,
+      "loss": 0.7215,
+      "step": 2040
+    },
+    {
+      "epoch": 0.7892963711618057,
+      "grad_norm": 0.13181230425834656,
+      "learning_rate": 1.1229839215686294e-06,
+      "loss": 0.7302,
+      "step": 2050
+    },
+    {
+      "epoch": 0.7931465973625951,
+      "grad_norm": 0.12593436241149902,
+      "learning_rate": 1.0839095136509188e-06,
+      "loss": 0.7097,
+      "step": 2060
+    },
+    {
+      "epoch": 0.7969968235633843,
+      "grad_norm": 0.15344317257404327,
+      "learning_rate": 1.045444208397791e-06,
+      "loss": 0.7204,
+      "step": 2070
+    },
+    {
+      "epoch": 0.8008470497641736,
+      "grad_norm": 0.13290542364120483,
+      "learning_rate": 1.007593988643402e-06,
+      "loss": 0.709,
+      "step": 2080
+    },
+    {
+      "epoch": 0.804697275964963,
+      "grad_norm": 0.1279771625995636,
+      "learning_rate": 9.703647415524613e-07,
+      "loss": 0.7332,
+      "step": 2090
+    },
+    {
+      "epoch": 0.8085475021657522,
+      "grad_norm": 0.1265469342470169,
+      "learning_rate": 9.337622577045485e-07,
+      "loss": 0.727,
+      "step": 2100
+    },
+    {
+      "epoch": 0.8123977283665416,
+      "grad_norm": 0.13107852637767792,
+      "learning_rate": 8.977922301934588e-07,
+      "loss": 0.7422,
+      "step": 2110
+    },
+    {
+      "epoch": 0.8162479545673308,
+      "grad_norm": 0.13456086814403534,
+      "learning_rate": 8.624602537417009e-07,
+      "loss": 0.7386,
+      "step": 2120
+    },
+    {
+      "epoch": 0.8200981807681201,
+      "grad_norm": 0.128574401140213,
+      "learning_rate": 8.277718238303123e-07,
+      "loss": 0.7213,
+      "step": 2130
+    },
+    {
+      "epoch": 0.8239484069689095,
+      "grad_norm": 0.13160739839076996,
+      "learning_rate": 7.937323358440935e-07,
+      "loss": 0.7328,
+      "step": 2140
+    },
+    {
+      "epoch": 0.8277986331696987,
+      "grad_norm": 0.14782220125198364,
+      "learning_rate": 7.603470842324223e-07,
+      "loss": 0.7173,
+      "step": 2150
+    },
+    {
+      "epoch": 0.831648859370488,
+      "grad_norm": 0.12510836124420166,
+      "learning_rate": 7.276212616857575e-07,
+      "loss": 0.7349,
+      "step": 2160
+    },
+    {
+      "epoch": 0.8354990855712773,
+      "grad_norm": 0.1314447820186615,
+      "learning_rate": 6.955599583279837e-07,
+      "loss": 0.736,
+      "step": 2170
+    },
+    {
+      "epoch": 0.8393493117720666,
+      "grad_norm": 0.13038775324821472,
+      "learning_rate": 6.641681609246981e-07,
+      "loss": 0.7261,
+      "step": 2180
+    },
+    {
+      "epoch": 0.8431995379728559,
+      "grad_norm": 0.132584810256958,
+      "learning_rate": 6.334507521075711e-07,
+      "loss": 0.7346,
+      "step": 2190
+    },
+    {
+      "epoch": 0.8470497641736452,
+      "grad_norm": 0.137792706489563,
+      "learning_rate": 6.034125096149179e-07,
+      "loss": 0.7215,
+      "step": 2200
+    },
+    {
+      "epoch": 0.8508999903744345,
+      "grad_norm": 0.14896586537361145,
+      "learning_rate": 5.740581055485722e-07,
+      "loss": 0.7076,
+      "step": 2210
+    },
+    {
+      "epoch": 0.8547502165752238,
+      "grad_norm": 0.12186845391988754,
+      "learning_rate": 5.453921056471961e-07,
+      "loss": 0.7305,
+      "step": 2220
+    },
+    {
+      "epoch": 0.8586004427760131,
+      "grad_norm": 0.1443181335926056,
+      "learning_rate": 5.174189685761288e-07,
+      "loss": 0.7234,
+      "step": 2230
+    },
+    {
+      "epoch": 0.8624506689768023,
+      "grad_norm": 0.1287497729063034,
+      "learning_rate": 4.901430452338979e-07,
+      "loss": 0.7437,
+      "step": 2240
+    },
+    {
+      "epoch": 0.8663008951775917,
+      "grad_norm": 0.13642992079257965,
+      "learning_rate": 4.6356857807548286e-07,
+      "loss": 0.7229,
+      "step": 2250
+    },
+    {
+      "epoch": 0.870151121378381,
+      "grad_norm": 0.13144196569919586,
+      "learning_rate": 4.376997004524541e-07,
+      "loss": 0.7112,
+      "step": 2260
+    },
+    {
+      "epoch": 0.8740013475791703,
+      "grad_norm": 0.13032953441143036,
+      "learning_rate": 4.12540435970073e-07,
+      "loss": 0.7121,
+      "step": 2270
+    },
+    {
+      "epoch": 0.8778515737799596,
+      "grad_norm": 0.12505872547626495,
+      "learning_rate": 3.880946978614703e-07,
+      "loss": 0.7363,
+      "step": 2280
+    },
+    {
+      "epoch": 0.8817017999807488,
+      "grad_norm": 0.12613271176815033,
+      "learning_rate": 3.643662883789878e-07,
+      "loss": 0.7325,
+      "step": 2290
+    },
+    {
+      "epoch": 0.8855520261815382,
+      "grad_norm": 0.1264238953590393,
+      "learning_rate": 3.41358898202781e-07,
+      "loss": 0.7171,
+      "step": 2300
+    },
+    {
+      "epoch": 0.8894022523823275,
+      "grad_norm": 0.1271478682756424,
+      "learning_rate": 3.190761058667741e-07,
+      "loss": 0.7181,
+      "step": 2310
+    },
+    {
+      "epoch": 0.8932524785831167,
+      "grad_norm": 0.12122893333435059,
+      "learning_rate": 2.9752137720206463e-07,
+      "loss": 0.7267,
+      "step": 2320
+    },
+    {
+      "epoch": 0.8971027047839061,
+      "grad_norm": 0.13166344165802002,
+      "learning_rate": 2.7669806479785067e-07,
+      "loss": 0.7385,
+      "step": 2330
+    },
+    {
+      "epoch": 0.9009529309846953,
+      "grad_norm": 0.1329558938741684,
+      "learning_rate": 2.566094074799763e-07,
+      "loss": 0.7271,
+      "step": 2340
+    },
+    {
+      "epoch": 0.9048031571854847,
+      "grad_norm": 0.12903714179992676,
+      "learning_rate": 2.3725852980716557e-07,
+      "loss": 0.7152,
+      "step": 2350
+    },
+    {
+      "epoch": 0.908653383386274,
+      "grad_norm": 0.13369542360305786,
+      "learning_rate": 2.1864844158503983e-07,
+      "loss": 0.7284,
+      "step": 2360
+    },
+    {
+      "epoch": 0.9125036095870632,
+      "grad_norm": 0.14874416589736938,
+      "learning_rate": 2.007820373979702e-07,
+      "loss": 0.7157,
+      "step": 2370
+    },
+    {
+      "epoch": 0.9163538357878526,
+      "grad_norm": 0.12981586158275604,
+      "learning_rate": 1.8366209615886532e-07,
+      "loss": 0.7207,
+      "step": 2380
+    },
+    {
+      "epoch": 0.9202040619886418,
+      "grad_norm": 0.1280152052640915,
+      "learning_rate": 1.6729128067693613e-07,
+      "loss": 0.743,
+      "step": 2390
+    },
+    {
+      "epoch": 0.9240542881894311,
+      "grad_norm": 0.13265089690685272,
+      "learning_rate": 1.5167213724353426e-07,
+      "loss": 0.7326,
+      "step": 2400
+    },
+    {
+      "epoch": 0.9279045143902205,
+      "grad_norm": 0.12792818248271942,
+      "learning_rate": 1.3680709523610058e-07,
+      "loss": 0.7176,
+      "step": 2410
+    },
+    {
+      "epoch": 0.9317547405910097,
+      "grad_norm": 0.12365072220563889,
+      "learning_rate": 1.2269846674030418e-07,
+      "loss": 0.7442,
+      "step": 2420
+    },
+    {
+      "epoch": 0.935604966791799,
+      "grad_norm": 0.131765678524971,
+      "learning_rate": 1.0934844619042595e-07,
+      "loss": 0.7394,
+      "step": 2430
+    },
+    {
+      "epoch": 0.9394551929925883,
+      "grad_norm": 0.13961195945739746,
+      "learning_rate": 9.675911002803961e-08,
+      "loss": 0.724,
+      "step": 2440
+    },
+    {
+      "epoch": 0.9433054191933776,
+      "grad_norm": 0.14082501828670502,
+      "learning_rate": 8.493241637904381e-08,
+      "loss": 0.721,
+      "step": 2450
+    },
+    {
+      "epoch": 0.9471556453941669,
+      "grad_norm": 0.14291928708553314,
+      "learning_rate": 7.387020474909689e-08,
+      "loss": 0.7458,
+      "step": 2460
+    },
+    {
+      "epoch": 0.9510058715949562,
+      "grad_norm": 0.13205114006996155,
+      "learning_rate": 6.357419573750856e-08,
+      "loss": 0.7304,
+      "step": 2470
+    },
+    {
+      "epoch": 0.9548560977957455,
+      "grad_norm": 0.134423166513443,
+      "learning_rate": 5.4045990769613365e-08,
+      "loss": 0.7242,
+      "step": 2480
+    },
+    {
+      "epoch": 0.9587063239965348,
+      "grad_norm": 0.15191437304019928,
+      "learning_rate": 4.5287071847694434e-08,
+      "loss": 0.7228,
+      "step": 2490
+    },
+    {
+      "epoch": 0.9625565501973241,
+      "grad_norm": 0.13321591913700104,
+      "learning_rate": 3.7298801320467817e-08,
+      "loss": 0.7234,
+      "step": 2500
+    },
+    {
+      "epoch": 0.9664067763981133,
+      "grad_norm": 0.13030226528644562,
+      "learning_rate": 3.0082421671192576e-08,
+      "loss": 0.7161,
+      "step": 2510
+    },
+    {
+      "epoch": 0.9702570025989027,
+      "grad_norm": 0.12440726161003113,
+      "learning_rate": 2.3639055324410355e-08,
+      "loss": 0.7342,
+      "step": 2520
+    },
+    {
+      "epoch": 0.974107228799692,
+      "grad_norm": 0.12983384728431702,
+      "learning_rate": 1.7969704471369477e-08,
+      "loss": 0.7198,
+      "step": 2530
+    },
+    {
+      "epoch": 0.9779574550004813,
+      "grad_norm": 0.12559209764003754,
+      "learning_rate": 1.3075250914142413e-08,
+      "loss": 0.7261,
+      "step": 2540
+    },
+    {
+      "epoch": 0.9818076812012706,
+      "grad_norm": 0.12556055188179016,
+      "learning_rate": 8.956455928476049e-09,
+      "loss": 0.7372,
+      "step": 2550
+    },
+    {
+      "epoch": 0.9856579074020598,
+      "grad_norm": 0.12110678851604462,
+      "learning_rate": 5.6139601453786275e-09,
+      "loss": 0.7145,
+      "step": 2560
+    },
+    {
+      "epoch": 0.9895081336028492,
+      "grad_norm": 0.13926371932029724,
+      "learning_rate": 3.0482834514805602e-09,
+      "loss": 0.7197,
+      "step": 2570
+    },
+    {
+      "epoch": 0.9933583598036385,
+      "grad_norm": 0.11838061362504959,
+      "learning_rate": 1.2598249081713366e-09,
+      "loss": 0.7033,
+      "step": 2580
+    },
+    {
+      "epoch": 0.9972085860044277,
+      "grad_norm": 0.13590435683727264,
+      "learning_rate": 2.488626895302826e-10,
+      "loss": 0.7376,
+      "step": 2590
+    },
+    {
+      "epoch": 0.9999037443449803,
+      "step": 2597,
+      "total_flos": 3.00690982998691e+18,
+      "train_loss": 0.7656716769779927,
+      "train_runtime": 17590.6193,
+      "train_samples_per_second": 37.799,
+      "train_steps_per_second": 0.148
+    }
+  ],
+  "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": 3.00690982998691e+18,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}

continuous/qwen3/siglip2_b16_512/finetune/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f394ce0c335e841f66e5c855277c4cf3f11f99a09e82ae1692d1c2955fab1cc6
+size 6136

continuous/qwen3/siglip2_b16_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": 768,
+  "mm_newline_position": "grid",
+  "mm_patch_merge_type": "flat",
+  "mm_projector_lr": null,
+  "mm_projector_type": "mlp2x_gelu",
+  "mm_resampler_type": null,
+  "mm_spatial_pool_mode": "bilinear",
+  "mm_spatial_pool_stride": null,
+  "mm_tunable_parts": "mm_mlp_adapter",
+  "mm_use_im_patch_token": false,
+  "mm_use_im_start_end": false,
+  "mm_vision_select_feature": "patch",
+  "mm_vision_select_layer": -2,
+  "mm_vision_tower": "open_clip_hub:ViT-B-16-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_b16_512.npz",
+  "vocab_size": 151936
+}

continuous/qwen3/siglip2_b16_512/pretrain/mm_projector.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0384c29dd76f9361bef3adb1f0b8be6220ea87af339932b379e8fa998eedc04e
+size 23087160

continuous/qwen3/siglip2_b16_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": 0.550396203994751,
+      "learning_rate": 0.00032142857142857147,
+      "loss": 9.5782,
+      "step": 10
+    },
+    {
+      "epoch": 0.03669724770642202,
+      "grad_norm": 0.14447778463363647,
+      "learning_rate": 0.0006785714285714287,
+      "loss": 5.7211,
+      "step": 20
+    },
+    {
+      "epoch": 0.05504587155963303,
+      "grad_norm": 0.034872252494096756,
+      "learning_rate": 0.0009999907688176173,
+      "loss": 4.9456,
+      "step": 30
+    },
+    {
+      "epoch": 0.07339449541284404,
+      "grad_norm": 0.017210006713867188,
+      "learning_rate": 0.0009988834393115766,
+      "loss": 4.7938,
+      "step": 40
+    },
+    {
+      "epoch": 0.09174311926605505,
+      "grad_norm": 0.012165281921625137,
+      "learning_rate": 0.0009959345572940772,
+      "loss": 4.7308,
+      "step": 50
+    },
+    {
+      "epoch": 0.11009174311926606,
+      "grad_norm": 0.012809027917683125,
+      "learning_rate": 0.0009911550081155983,
+      "loss": 4.7017,
+      "step": 60
+    },
+    {
+      "epoch": 0.12844036697247707,
+      "grad_norm": 0.010852714069187641,
+      "learning_rate": 0.0009845624347566433,
+      "loss": 4.6655,
+      "step": 70
+    },
+    {
+      "epoch": 0.14678899082568808,
+      "grad_norm": 0.012630184181034565,
+      "learning_rate": 0.0009761811727013547,
+      "loss": 4.6488,
+      "step": 80
+    },
+    {
+      "epoch": 0.1651376146788991,
+      "grad_norm": 0.011250040493905544,
+      "learning_rate": 0.0009660421601067666,
+      "loss": 4.6236,
+      "step": 90
+    },
+    {
+      "epoch": 0.1834862385321101,
+      "grad_norm": 0.01512724906206131,
+      "learning_rate": 0.0009541828235992926,
+      "loss": 4.6094,
+      "step": 100
+    },
+    {
+      "epoch": 0.2018348623853211,
+      "grad_norm": 0.02023635245859623,
+      "learning_rate": 0.0009406469401200151,
+      "loss": 4.5865,
+      "step": 110
+    },
+    {
+      "epoch": 0.22018348623853212,
+      "grad_norm": 0.031447410583496094,
+      "learning_rate": 0.0009254844753287492,
+      "loss": 4.5716,
+      "step": 120
+    },
+    {
+      "epoch": 0.23853211009174313,
+      "grad_norm": 0.04946161061525345,
+      "learning_rate": 0.0009087513991633924,
+      "loss": 4.5411,
+      "step": 130
+    },
+    {
+      "epoch": 0.25688073394495414,
+      "grad_norm": 0.048852141946554184,
+      "learning_rate": 0.0008905094792353917,
+      "loss": 4.5202,
+      "step": 140
+    },
+    {
+      "epoch": 0.27522935779816515,
+      "grad_norm": 0.05042686313390732,
+      "learning_rate": 0.0008708260528239789,
+      "loss": 4.506,
+      "step": 150
+    },
+    {
+      "epoch": 0.29357798165137616,
+      "grad_norm": 0.03212501108646393,
+      "learning_rate": 0.0008497737783108237,
+      "loss": 4.4787,
+      "step": 160
+    },
+    {
+      "epoch": 0.3119266055045872,
+      "grad_norm": 0.036054350435733795,
+      "learning_rate": 0.0008274303669726426,
+      "loss": 4.4543,
+      "step": 170
+    },
+    {
+      "epoch": 0.3302752293577982,
+      "grad_norm": 0.03531884774565697,
+      "learning_rate": 0.0008038782961218137,
+      "loss": 4.4283,
+      "step": 180
+    },
+    {
+      "epoch": 0.3486238532110092,
+      "grad_norm": 0.034238602966070175,
+      "learning_rate": 0.000779204504653894,
+      "loss": 4.4111,
+      "step": 190
+    },
+    {
+      "epoch": 0.3669724770642202,
+      "grad_norm": 0.03152380883693695,
+      "learning_rate": 0.0007535000721258791,
+      "loss": 4.3798,
+      "step": 200
+    },
+    {
+      "epoch": 0.3853211009174312,
+      "grad_norm": 0.03489081561565399,
+      "learning_rate": 0.00072685988254984,
+      "loss": 4.3592,
+      "step": 210
+    },
+    {
+      "epoch": 0.4036697247706422,
+      "grad_norm": 0.034809697419404984,
+      "learning_rate": 0.0006993822741429906,
+      "loss": 4.3566,
+      "step": 220
+    },
+    {
+      "epoch": 0.42201834862385323,
+      "grad_norm": 0.036431197077035904,
+      "learning_rate": 0.0006711686763270818,
+      "loss": 4.3392,
+      "step": 230
+    },
+    {
+      "epoch": 0.44036697247706424,
+      "grad_norm": 0.03266889229416847,
+      "learning_rate": 0.0006423232353170831,
+      "loss": 4.313,
+      "step": 240
+    },
+    {
+      "epoch": 0.45871559633027525,
+      "grad_norm": 0.03521248325705528,
+      "learning_rate": 0.0006129524296812335,
+      "loss": 4.2834,
+      "step": 250
+    },
+    {
+      "epoch": 0.47706422018348627,
+      "grad_norm": 0.040851376950740814,
+      "learning_rate": 0.0005831646772915651,
+      "loss": 4.2869,
+      "step": 260
+    },
+    {
+      "epoch": 0.4954128440366973,
+      "grad_norm": 0.03484119102358818,
+      "learning_rate": 0.0005530699351157809,
+      "loss": 4.2578,
+      "step": 270
+    },
+    {
+      "epoch": 0.5137614678899083,
+      "grad_norm": 0.024944687262177467,
+      "learning_rate": 0.0005227792933277943,
+      "loss": 4.2447,
+      "step": 280
+    },
+    {
+      "epoch": 0.5321100917431193,
+      "grad_norm": 0.032014552503824234,
+      "learning_rate": 0.0004924045652352092,
+      "loss": 4.2225,
+      "step": 290
+    },
+    {
+      "epoch": 0.5504587155963303,
+      "grad_norm": 0.03163444995880127,
+      "learning_rate": 0.0004620578745374633,
+      "loss": 4.2157,
+      "step": 300
+    },
+    {
+      "epoch": 0.5688073394495413,
+      "grad_norm": 0.04171528294682503,
+      "learning_rate": 0.0004318512414382059,
+      "loss": 4.2241,
+      "step": 310
+    },
+    {
+      "epoch": 0.5871559633027523,
+      "grad_norm": 0.027618147432804108,
+      "learning_rate": 0.0004018961691397148,
+      "loss": 4.2105,
+      "step": 320
+    },
+    {
+      "epoch": 0.6055045871559633,
+      "grad_norm": 0.026053696870803833,
+      "learning_rate": 0.0003723032322457458,
+      "loss": 4.1868,
+      "step": 330
+    },
+    {
+      "epoch": 0.6238532110091743,
+      "grad_norm": 0.032688312232494354,
+      "learning_rate": 0.0003431816685921625,
+      "loss": 4.1866,
+      "step": 340
+    },
+    {
+      "epoch": 0.6422018348623854,
+      "grad_norm": 0.02713450789451599,
+      "learning_rate": 0.0003146389760120416,
+      "loss": 4.1861,
+      "step": 350
+    },
+    {
+      "epoch": 0.6605504587155964,
+      "grad_norm": 0.029250716790556908,
+      "learning_rate": 0.00028678051552373485,
+      "loss": 4.1907,
+      "step": 360
+    },
+    {
+      "epoch": 0.6788990825688074,
+      "grad_norm": 0.02836327813565731,
+      "learning_rate": 0.0002597091224066581,
+      "loss": 4.1703,
+      "step": 370
+    },
+    {
+      "epoch": 0.6972477064220184,
+      "grad_norm": 0.0288862194865942,
+      "learning_rate": 0.0002335247266004629,
+      "loss": 4.1819,
+      "step": 380
+    },
+    {
+      "epoch": 0.7155963302752294,
+      "grad_norm": 0.0262719988822937,
+      "learning_rate": 0.000208323983828828,
+      "loss": 4.1622,
+      "step": 390
+    },
+    {
+      "epoch": 0.7339449541284404,
+      "grad_norm": 0.02566775120794773,
+      "learning_rate": 0.0001841999188095224,
+      "loss": 4.1625,
+      "step": 400
+    },
+    {
+      "epoch": 0.7522935779816514,
+      "grad_norm": 0.02320670895278454,
+      "learning_rate": 0.00016124158186777676,
+      "loss": 4.1597,
+      "step": 410
+    },
+    {
+      "epoch": 0.7706422018348624,
+      "grad_norm": 0.026464169844985008,
+      "learning_rate": 0.0001395337202205211,
+      "loss": 4.15,
+      "step": 420
+    },
+    {
+      "epoch": 0.7889908256880734,
+      "grad_norm": 0.025871647521853447,
+      "learning_rate": 0.00011915646514489292,
+      "loss": 4.157,
+      "step": 430
+    },
+    {
+      "epoch": 0.8073394495412844,
+      "grad_norm": 0.026653289794921875,
+      "learning_rate": 0.00010018503618578818,
+      "loss": 4.1604,
+      "step": 440
+    },
+    {
+      "epoch": 0.8256880733944955,
+      "grad_norm": 0.021817583590745926,
+      "learning_rate": 8.268946349432583e-05,
+      "loss": 4.1499,
+      "step": 450
+    },
+    {
+      "epoch": 0.8440366972477065,
+      "grad_norm": 0.023918623104691505,
+      "learning_rate": 6.673432932217117e-05,
+      "loss": 4.1408,
+      "step": 460
+    },
+    {
+      "epoch": 0.8623853211009175,
+      "grad_norm": 0.02682694047689438,
+      "learning_rate": 5.2378529625954684e-05,
+      "loss": 4.1313,
+      "step": 470
+    },
+    {
+      "epoch": 0.8807339449541285,
+      "grad_norm": 0.024093734100461006,
+      "learning_rate": 3.9675056661785556e-05,
+      "loss": 4.1408,
+      "step": 480
+    },
+    {
+      "epoch": 0.8990825688073395,
+      "grad_norm": 0.020671311765909195,
+      "learning_rate": 2.8670803372379538e-05,
+      "loss": 4.1389,
+      "step": 490
+    },
+    {
+      "epoch": 0.9174311926605505,
+      "grad_norm": 0.021499987691640854,
+      "learning_rate": 1.9406390288876586e-05,
+      "loss": 4.1537,
+      "step": 500
+    },
+    {
+      "epoch": 0.9357798165137615,
+      "grad_norm": 0.021211478859186172,
+      "learning_rate": 1.191601558631461e-05,
+      "loss": 4.1401,
+      "step": 510
+    },
+    {
+      "epoch": 0.9541284403669725,
+      "grad_norm": 0.020458856597542763,
+      "learning_rate": 6.2273288462559305e-06,
+      "loss": 4.1324,
+      "step": 520
+    },
+    {
+      "epoch": 0.9724770642201835,
+      "grad_norm": 0.021325435489416122,
+      "learning_rate": 2.361328992552314e-06,
+      "loss": 4.1327,
+      "step": 530
+    },
+    {
+      "epoch": 0.9908256880733946,
+      "grad_norm": 0.02238606847822666,
+      "learning_rate": 3.322867770044202e-07,
+      "loss": 4.1384,
+      "step": 540
+    },
+    {
+      "epoch": 1.0,
+      "step": 545,
+      "total_flos": 2.643662056099676e+17,
+      "train_loss": 4.453979723169169,
+      "train_runtime": 7854.938,
+      "train_samples_per_second": 71.054,
+      "train_steps_per_second": 0.069
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 545,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": false,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.643662056099676e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}

continuous/qwen3/siglip2_b32_256/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_b32_256/finetune/config.json

@@ -0,0 +1,59 @@
+{
+  "add_faster_video": false,
+  "add_time_instruction": false,
+  "architectures": [
+    "LlavaQwen3ForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "bos_token_id": 151643,
+  "eos_token_id": 151645,
+  "faster_token_stride": 10,
+  "force_sample": false,
+  "head_dim": 128,
+  "hidden_act": "silu",
+  "hidden_size": 2048,
+  "image_aspect_ratio": "square",
+  "image_crop_resolution": null,
+  "image_grid_pinpoints": null,
+  "image_split_resolution": null,
+  "initializer_range": 0.02,
+  "intermediate_size": 6144,
+  "max_position_embeddings": 40960,
+  "max_window_layers": 28,
+  "mm_hidden_size": 768,
+  "mm_newline_position": "grid",
+  "mm_patch_merge_type": "flat",
+  "mm_projector_lr": 0.0001,
+  "mm_projector_type": "mlp2x_gelu",
+  "mm_resampler_type": null,
+  "mm_spatial_pool_mode": "bilinear",
+  "mm_spatial_pool_stride": null,
+  "mm_tunable_parts": "mm_mlp_adapter,mm_language_model",
+  "mm_use_im_patch_token": false,
+  "mm_use_im_start_end": false,
+  "mm_vision_select_feature": "patch",
+  "mm_vision_select_layer": -2,
+  "mm_vision_tower": "open_clip_hub:ViT-B-32-SigLIP2-256",
+  "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_b32_256.npz",
+  "vocab_size": 151936
+}

continuous/qwen3/siglip2_b32_256/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"
+}