robobrain25_4b_full_v1: checkpoint-600

judge4/robobrain25_4b_full_v1/checkpoint-600/config.json

@@ -0,0 +1,67 @@
+{
+  "architectures": [
+    "Qwen3VLForConditionalGeneration"
+  ],
+  "dtype": "bfloat16",
+  "image_token_id": 151655,
+  "model_type": "qwen3_vl",
+  "text_config": {
+    "attention_bias": false,
+    "attention_dropout": 0.0,
+    "bos_token_id": 151643,
+    "dtype": "bfloat16",
+    "eos_token_id": 151645,
+    "head_dim": 128,
+    "hidden_act": "silu",
+    "hidden_size": 2560,
+    "initializer_range": 0.02,
+    "intermediate_size": 9728,
+    "max_position_embeddings": 262144,
+    "model_type": "qwen3_vl_text",
+    "num_attention_heads": 32,
+    "num_hidden_layers": 36,
+    "num_key_value_heads": 8,
+    "pad_token_id": null,
+    "rms_norm_eps": 1e-06,
+    "rope_parameters": {
+      "mrope_interleaved": true,
+      "mrope_section": [
+        24,
+        20,
+        20
+      ],
+      "rope_theta": 5000000,
+      "rope_type": "default"
+    },
+    "tie_word_embeddings": true,
+    "use_cache": true,
+    "vocab_size": 151936
+  },
+  "tie_word_embeddings": true,
+  "transformers_version": "5.10.2",
+  "use_cache": false,
+  "video_token_id": 151656,
+  "vision_config": {
+    "deepstack_visual_indexes": [
+      5,
+      11,
+      17
+    ],
+    "depth": 24,
+    "dtype": "bfloat16",
+    "hidden_act": "gelu_pytorch_tanh",
+    "hidden_size": 1024,
+    "in_channels": 3,
+    "initializer_range": 0.02,
+    "intermediate_size": 4096,
+    "model_type": "qwen3_vl_vision",
+    "num_heads": 16,
+    "num_position_embeddings": 2304,
+    "out_hidden_size": 2560,
+    "patch_size": 16,
+    "spatial_merge_size": 2,
+    "temporal_patch_size": 2
+  },
+  "vision_end_token_id": 151653,
+  "vision_start_token_id": 151652
+}

judge4/robobrain25_4b_full_v1/checkpoint-600/generation_config.json

@@ -0,0 +1,14 @@
+{
+  "bos_token_id": 151643,
+  "do_sample": true,
+  "eos_token_id": [
+    151645,
+    151643
+  ],
+  "pad_token_id": 151643,
+  "repetition_penalty": 1.0,
+  "temperature": 0.7,
+  "top_k": 20,
+  "top_p": 0.8,
+  "transformers_version": "5.10.2"
+}

judge4/robobrain25_4b_full_v1/checkpoint-600/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7e961ddbb8ee442b9f622e52e7ffaca0bf7c772861a6ca9bf41119dd0a390e3d
+size 8875719408

judge4/robobrain25_4b_full_v1/checkpoint-600/trainer_state.json

@@ -0,0 +1,490 @@
+{
+  "best_global_step": 600,
+  "best_metric": 0.9913830968829511,
+  "best_model_checkpoint": "/cache/hyeongjinkim/runs/yesno/robobrain25_4b/checkpoint-600",
+  "epoch": 0.32085561497326204,
+  "eval_steps": 200,
+  "global_step": 600,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0053475935828877,
+      "grad_norm": 81.35555848697916,
+      "learning_rate": 1.5789473684210526e-06,
+      "loss": 0.67803955078125,
+      "step": 10
+    },
+    {
+      "epoch": 0.0106951871657754,
+      "grad_norm": 46.59385136159528,
+      "learning_rate": 3.3333333333333333e-06,
+      "loss": 0.6182425498962403,
+      "step": 20
+    },
+    {
+      "epoch": 0.016042780748663103,
+      "grad_norm": 50.10868928177893,
+      "learning_rate": 5.087719298245615e-06,
+      "loss": 0.3442235946655273,
+      "step": 30
+    },
+    {
+      "epoch": 0.0213903743315508,
+      "grad_norm": 18.419260982343047,
+      "learning_rate": 6.842105263157896e-06,
+      "loss": 0.4191296577453613,
+      "step": 40
+    },
+    {
+      "epoch": 0.026737967914438502,
+      "grad_norm": 15.5507081439074,
+      "learning_rate": 8.596491228070176e-06,
+      "loss": 0.3343177795410156,
+      "step": 50
+    },
+    {
+      "epoch": 0.03208556149732621,
+      "grad_norm": 32.37407267253341,
+      "learning_rate": 9.999969973569808e-06,
+      "loss": 0.27725045680999755,
+      "step": 60
+    },
+    {
+      "epoch": 0.0374331550802139,
+      "grad_norm": 5.294366022502766,
+      "learning_rate": 9.99891908637917e-06,
+      "loss": 0.17519030570983887,
+      "step": 70
+    },
+    {
+      "epoch": 0.0427807486631016,
+      "grad_norm": 28.12894210863794,
+      "learning_rate": 9.996367238293978e-06,
+      "loss": 0.2899880886077881,
+      "step": 80
+    },
+    {
+      "epoch": 0.0481283422459893,
+      "grad_norm": 65.55570652638305,
+      "learning_rate": 9.992315195524705e-06,
+      "loss": 0.21597778797149658,
+      "step": 90
+    },
+    {
+      "epoch": 0.053475935828877004,
+      "grad_norm": 2.841038545896545,
+      "learning_rate": 9.98676417472592e-06,
+      "loss": 0.21681747436523438,
+      "step": 100
+    },
+    {
+      "epoch": 0.058823529411764705,
+      "grad_norm": 29.672372636077025,
+      "learning_rate": 9.979715842630971e-06,
+      "loss": 0.2695551872253418,
+      "step": 110
+    },
+    {
+      "epoch": 0.06417112299465241,
+      "grad_norm": 19.1994697494169,
+      "learning_rate": 9.971172315551539e-06,
+      "loss": 0.28987135887146,
+      "step": 120
+    },
+    {
+      "epoch": 0.06951871657754011,
+      "grad_norm": 28.538835070604733,
+      "learning_rate": 9.961136158742194e-06,
+      "loss": 0.18500001430511476,
+      "step": 130
+    },
+    {
+      "epoch": 0.0748663101604278,
+      "grad_norm": 9.642404272811257,
+      "learning_rate": 9.94961038563017e-06,
+      "loss": 0.2563859224319458,
+      "step": 140
+    },
+    {
+      "epoch": 0.08021390374331551,
+      "grad_norm": 21.754160295577037,
+      "learning_rate": 9.936598456910553e-06,
+      "loss": 0.12378699779510498,
+      "step": 150
+    },
+    {
+      "epoch": 0.0855614973262032,
+      "grad_norm": 51.235199178405615,
+      "learning_rate": 9.922104279507188e-06,
+      "loss": 0.15373640060424804,
+      "step": 160
+    },
+    {
+      "epoch": 0.09090909090909091,
+      "grad_norm": 12.999898710465901,
+      "learning_rate": 9.90613220539959e-06,
+      "loss": 0.2704988718032837,
+      "step": 170
+    },
+    {
+      "epoch": 0.0962566844919786,
+      "grad_norm": 4.191662390639983,
+      "learning_rate": 9.888687030316244e-06,
+      "loss": 0.23029546737670897,
+      "step": 180
+    },
+    {
+      "epoch": 0.10160427807486631,
+      "grad_norm": 0.19977696147271767,
+      "learning_rate": 9.869773992294645e-06,
+      "loss": 0.07535446882247925,
+      "step": 190
+    },
+    {
+      "epoch": 0.10695187165775401,
+      "grad_norm": 13.364810255980874,
+      "learning_rate": 9.849398770108534e-06,
+      "loss": 0.19212841987609863,
+      "step": 200
+    },
+    {
+      "epoch": 0.10695187165775401,
+      "eval_auroc": 0.9721798073211655,
+      "eval_loss": 0.2283112108707428,
+      "eval_runtime": 64.8036,
+      "eval_samples_per_second": 81.832,
+      "eval_steps_per_second": 1.281,
+      "step": 200
+    },
+    {
+      "epoch": 0.11229946524064172,
+      "grad_norm": 14.287945266035386,
+      "learning_rate": 9.82756748156282e-06,
+      "loss": 0.19752094745635987,
+      "step": 210
+    },
+    {
+      "epoch": 0.11764705882352941,
+      "grad_norm": 4.425067711830129,
+      "learning_rate": 9.804286681656661e-06,
+      "loss": 0.14333631992340087,
+      "step": 220
+    },
+    {
+      "epoch": 0.12299465240641712,
+      "grad_norm": 89.557776121957,
+      "learning_rate": 9.779563360615269e-06,
+      "loss": 0.16643595695495605,
+      "step": 230
+    },
+    {
+      "epoch": 0.12834224598930483,
+      "grad_norm": 24.496503277214135,
+      "learning_rate": 9.753404941791063e-06,
+      "loss": 0.2630112648010254,
+      "step": 240
+    },
+    {
+      "epoch": 0.13368983957219252,
+      "grad_norm": 15.014222665772476,
+      "learning_rate": 9.725819279434736e-06,
+      "loss": 0.16255661249160766,
+      "step": 250
+    },
+    {
+      "epoch": 0.13903743315508021,
+      "grad_norm": 16.181421113437565,
+      "learning_rate": 9.696814656336974e-06,
+      "loss": 0.19912848472595215,
+      "step": 260
+    },
+    {
+      "epoch": 0.1443850267379679,
+      "grad_norm": 44.19383781171741,
+      "learning_rate": 9.666399781341487e-06,
+      "loss": 0.15326435565948487,
+      "step": 270
+    },
+    {
+      "epoch": 0.1497326203208556,
+      "grad_norm": 8.755880151547933,
+      "learning_rate": 9.63458378673011e-06,
+      "loss": 0.11530225276947022,
+      "step": 280
+    },
+    {
+      "epoch": 0.15508021390374332,
+      "grad_norm": 35.56686216592682,
+      "learning_rate": 9.60137622548078e-06,
+      "loss": 0.16173202991485597,
+      "step": 290
+    },
+    {
+      "epoch": 0.16042780748663102,
+      "grad_norm": 25.44130299999626,
+      "learning_rate": 9.566787068399182e-06,
+      "loss": 0.13888851404190064,
+      "step": 300
+    },
+    {
+      "epoch": 0.1657754010695187,
+      "grad_norm": 18.064274468866262,
+      "learning_rate": 9.53082670112494e-06,
+      "loss": 0.1708371877670288,
+      "step": 310
+    },
+    {
+      "epoch": 0.1711229946524064,
+      "grad_norm": 5.153801640738739,
+      "learning_rate": 9.493505921013264e-06,
+      "loss": 0.1347058892250061,
+      "step": 320
+    },
+    {
+      "epoch": 0.17647058823529413,
+      "grad_norm": 13.86585419537311,
+      "learning_rate": 9.454835933892961e-06,
+      "loss": 0.13538434505462646,
+      "step": 330
+    },
+    {
+      "epoch": 0.18181818181818182,
+      "grad_norm": 9.682185407017588,
+      "learning_rate": 9.414828350701808e-06,
+      "loss": 0.11364465951919556,
+      "step": 340
+    },
+    {
+      "epoch": 0.18716577540106952,
+      "grad_norm": 25.297181202573153,
+      "learning_rate": 9.37349518400028e-06,
+      "loss": 0.0995937705039978,
+      "step": 350
+    },
+    {
+      "epoch": 0.1925133689839572,
+      "grad_norm": 76.86623504553876,
+      "learning_rate": 9.330848844364709e-06,
+      "loss": 0.1283486604690552,
+      "step": 360
+    },
+    {
+      "epoch": 0.19786096256684493,
+      "grad_norm": 22.232702771809837,
+      "learning_rate": 9.2869021366609e-06,
+      "loss": 0.16085681915283204,
+      "step": 370
+    },
+    {
+      "epoch": 0.20320855614973263,
+      "grad_norm": 1.84593687373222,
+      "learning_rate": 9.241668256199392e-06,
+      "loss": 0.08294941186904907,
+      "step": 380
+    },
+    {
+      "epoch": 0.20855614973262032,
+      "grad_norm": 11.676399713940807,
+      "learning_rate": 9.195160784773471e-06,
+      "loss": 0.11805830001831055,
+      "step": 390
+    },
+    {
+      "epoch": 0.21390374331550802,
+      "grad_norm": 0.06863136435875032,
+      "learning_rate": 9.14739368658114e-06,
+      "loss": 0.0941552460193634,
+      "step": 400
+    },
+    {
+      "epoch": 0.21390374331550802,
+      "eval_auroc": 0.9897083822382569,
+      "eval_loss": 0.21189619600772858,
+      "eval_runtime": 56.0389,
+      "eval_samples_per_second": 94.631,
+      "eval_steps_per_second": 1.481,
+      "step": 400
+    },
+    {
+      "epoch": 0.2192513368983957,
+      "grad_norm": 7.685749423979118,
+      "learning_rate": 9.098381304032267e-06,
+      "loss": 0.13885953426361083,
+      "step": 410
+    },
+    {
+      "epoch": 0.22459893048128343,
+      "grad_norm": 4.047313507008097,
+      "learning_rate": 9.048138353442173e-06,
+      "loss": 0.09946863055229187,
+      "step": 420
+    },
+    {
+      "epoch": 0.22994652406417113,
+      "grad_norm": 5.935897692096215,
+      "learning_rate": 8.996679920612964e-06,
+      "loss": 0.08703446984291077,
+      "step": 430
+    },
+    {
+      "epoch": 0.23529411764705882,
+      "grad_norm": 0.19905556655605508,
+      "learning_rate": 8.944021456303898e-06,
+      "loss": 0.08608518838882447,
+      "step": 440
+    },
+    {
+      "epoch": 0.24064171122994651,
+      "grad_norm": 28.59124601424046,
+      "learning_rate": 8.890178771592198e-06,
+      "loss": 0.08598223328590393,
+      "step": 450
+    },
+    {
+      "epoch": 0.24598930481283424,
+      "grad_norm": 44.28233394509514,
+      "learning_rate": 8.835168033125657e-06,
+      "loss": 0.14374032020568847,
+      "step": 460
+    },
+    {
+      "epoch": 0.25133689839572193,
+      "grad_norm": 4.285444178798713,
+      "learning_rate": 8.779005758268486e-06,
+      "loss": 0.11084957122802734,
+      "step": 470
+    },
+    {
+      "epoch": 0.25668449197860965,
+      "grad_norm": 11.58505654796229,
+      "learning_rate": 8.721708810141862e-06,
+      "loss": 0.08369065523147583,
+      "step": 480
+    },
+    {
+      "epoch": 0.2620320855614973,
+      "grad_norm": 25.45556349454736,
+      "learning_rate": 8.663294392560658e-06,
+      "loss": 0.10972058773040771,
+      "step": 490
+    },
+    {
+      "epoch": 0.26737967914438504,
+      "grad_norm": 4.356905993857335,
+      "learning_rate": 8.603780044867866e-06,
+      "loss": 0.09111945629119873,
+      "step": 500
+    },
+    {
+      "epoch": 0.2727272727272727,
+      "grad_norm": 4.032564959702237,
+      "learning_rate": 8.543183636668302e-06,
+      "loss": 0.0653485357761383,
+      "step": 510
+    },
+    {
+      "epoch": 0.27807486631016043,
+      "grad_norm": 14.348594119432299,
+      "learning_rate": 8.481523362463111e-06,
+      "loss": 0.051397836208343504,
+      "step": 520
+    },
+    {
+      "epoch": 0.28342245989304815,
+      "grad_norm": 18.13990732866606,
+      "learning_rate": 8.418817736186751e-06,
+      "loss": 0.0686537504196167,
+      "step": 530
+    },
+    {
+      "epoch": 0.2887700534759358,
+      "grad_norm": 4.002508870380514,
+      "learning_rate": 8.355085585648054e-06,
+      "loss": 0.12751879692077636,
+      "step": 540
+    },
+    {
+      "epoch": 0.29411764705882354,
+      "grad_norm": 9.286486395473394,
+      "learning_rate": 8.290346046877033e-06,
+      "loss": 0.04066848456859588,
+      "step": 550
+    },
+    {
+      "epoch": 0.2994652406417112,
+      "grad_norm": 7.435719753664092,
+      "learning_rate": 8.224618558379166e-06,
+      "loss": 0.04355362355709076,
+      "step": 560
+    },
+    {
+      "epoch": 0.3048128342245989,
+      "grad_norm": 0.024302497019430814,
+      "learning_rate": 8.157922855298834e-06,
+      "loss": 0.14085142612457274,
+      "step": 570
+    },
+    {
+      "epoch": 0.31016042780748665,
+      "grad_norm": 1.454998652438813,
+      "learning_rate": 8.090278963493714e-06,
+      "loss": 0.06956749558448791,
+      "step": 580
+    },
+    {
+      "epoch": 0.3155080213903743,
+      "grad_norm": 0.02966867777255224,
+      "learning_rate": 8.021707193521865e-06,
+      "loss": 0.03296654522418976,
+      "step": 590
+    },
+    {
+      "epoch": 0.32085561497326204,
+      "grad_norm": 0.6112134916240425,
+      "learning_rate": 7.952228134543357e-06,
+      "loss": 0.09656106233596802,
+      "step": 600
+    },
+    {
+      "epoch": 0.32085561497326204,
+      "eval_auroc": 0.9913830968829511,
+      "eval_loss": 0.1317468285560608,
+      "eval_runtime": 54.6812,
+      "eval_samples_per_second": 96.98,
+      "eval_steps_per_second": 1.518,
+      "step": 600
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 1870,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 200,
+  "stateful_callbacks": {
+    "EarlyStoppingCallback": {
+      "args": {
+        "early_stopping_patience": 5,
+        "early_stopping_threshold": 0.0
+      },
+      "attributes": {
+        "early_stopping_patience_counter": 0
+      }
+    },
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 20892233826304.0,
+  "train_batch_size": 16,
+  "trial_name": null,
+  "trial_params": null
+}

judge4/robobrain25_4b_full_v1/checkpoint-600/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5e2a8177c5f6750417a3ae34fcdf0221b8b841561a714ac4792f19eba0082398
+size 6673

judge4/robobrain25_4b_full_v1/checkpoint-600/zero_to_torch.py

@@ -0,0 +1,54 @@
+#!/usr/bin/env python
+
+# SPDX-License-Identifier: Apache-2.0
+# DeepSpeed Team
+
+import argparse
+
+if __package__:
+    from . import zero_to_fp32
+else:
+    import zero_to_fp32
+
+
+def main(args=None):
+    parser = argparse.ArgumentParser(
+        description="Convert a DeepSpeed ZeRO checkpoint to float32, float16, or bfloat16 PyTorch weights.")
+    parser.add_argument("checkpoint_dir",
+                        type=str,
+                        help="path to the desired checkpoint folder, e.g., path/checkpoint-12")
+    parser.add_argument("output_dir", type=str, help="directory for the converted PyTorch state_dict files")
+    parser.add_argument("--dtype",
+                        type=str,
+                        choices=sorted(zero_to_fp32.OUTPUT_DTYPE_NAMES),
+                        required=True,
+                        help="output tensor dtype")
+    parser.add_argument("--max_shard_size",
+                        type=str,
+                        default="5GB",
+                        help="maximum size of each checkpoint shard, such as 5GB or 500MB")
+    parser.add_argument("--safe_serialization",
+                        default=False,
+                        action='store_true',
+                        help="save with safetensors instead of PyTorch pickle serialization")
+    parser.add_argument("-t",
+                        "--tag",
+                        type=str,
+                        default=None,
+                        help="checkpoint tag used as a unique identifier, e.g., global_step1")
+    parser.add_argument("--exclude_frozen_parameters", action='store_true', help="exclude frozen parameters")
+    parser.add_argument("-d", "--debug", action='store_true', help="enable debug output")
+    parsed_args = parser.parse_args(args)
+
+    zero_to_fp32.debug = parsed_args.debug
+    zero_to_fp32.convert_zero_checkpoint_to_state_dict(parsed_args.checkpoint_dir,
+                                                       parsed_args.output_dir,
+                                                       dtype=parsed_args.dtype,
+                                                       max_shard_size=parsed_args.max_shard_size,
+                                                       safe_serialization=parsed_args.safe_serialization,
+                                                       tag=parsed_args.tag,
+                                                       exclude_frozen_parameters=parsed_args.exclude_frozen_parameters)
+
+
+if __name__ == "__main__":
+    main()