Blame for speed_fixed_bs16/train.log

[INFO] Running in WANDB offline mode
[INFO] Running in WANDB offline mode
[INFO] Running in WANDB offline mode
[INFO] Running in WANDB offline mode
Trainer._get_train_sampler replaced with custom implementation.
warmup_ratio is deprecated and will be removed in v5.2. Use `warmup_steps` instead.
[2026-09-18 03:39:48,382] [INFO] [real_accelerator.py:222:get_accelerator] Setting ds_accelerator to cuda (auto detect)
Trainer._get_train_sampler replaced with custom implementation.
warmup_ratio is deprecated and will be removed in v5.2. Use `warmup_steps` instead.
Trainer._get_train_sampler replaced with custom implementation.
warmup_ratio is deprecated and will be removed in v5.2. Use `warmup_steps` instead.
[2026-09-18 03:39:48,439] [INFO] [real_accelerator.py:222:get_accelerator] Setting ds_accelerator to cuda (auto detect)
Trainer._get_train_sampler replaced with custom implementation.
warmup_ratio is deprecated and will be removed in v5.2. Use `warmup_steps` instead.
[2026-09-18 03:39:48,457] [INFO] [real_accelerator.py:222:get_accelerator] Setting ds_accelerator to cuda (auto detect)
[2026-09-18 03:39:48,485] [INFO] [real_accelerator.py:222:get_accelerator] Setting ds_accelerator to cuda (auto detect)
[2026-09-18 03:39:48,729] [INFO] [comm.py:658:init_distributed] cdb=None
[2026-09-18 03:39:48,758] [INFO] [comm.py:658:init_distributed] cdb=None
[2026-09-18 03:39:48,785] [INFO] [comm.py:658:init_distributed] cdb=None
[2026-09-18 03:39:48,810] [INFO] [comm.py:658:init_distributed] cdb=None
[2026-09-18 03:39:48,810] [INFO] [comm.py:689:init_distributed] Initializing TorchBackend in DeepSpeed with backend nccl
[TrainingFix] Qwen3.5 FlashAttention MRoPE packing guard enabled
[TrainingFix] Qwen3.5 FlashAttention MRoPE packing guard enabled
[TrainingFix] Qwen3.5 FlashAttention MRoPE packing guard enabled
[TrainingFix] Qwen3.5 FlashAttention MRoPE packing guard enabled

Loading weights:   0%|          | 0/723 [00:00<?, ?it/s]
Loading weights:   0%|          | 0/723 [00:00<?, ?it/s]
Loading weights:   0%|          | 0/723 [00:00<?, ?it/s]
Loading weights:   0%|          | 0/723 [00:00<?, ?it/s]
Loading weights: 100%|██████████| 723/723 [00:00<00:00, 9474.79it/s]

Loading weights: 100%|██████████| 723/723 [00:00<00:00, 11503.19it/s]

Loading weights: 100%|██████████| 723/723 [00:00<00:00, 10532.34it/s]

Loading weights: 100%|██████████| 723/723 [00:00<00:00, 11489.68it/s]
[VGGTEncoder] eval KV-cache window: start=8 recent=48 (total=56 frames)
[FeatureFusion] geometry fusion delta scaled by 0.5
[FeatureFusion] importance gate ENABLED on geometry delta
[FeatureFusion] learnable geometry scale ENABLED (init 0.5)
[VGGTEncoder] eval KV-cache window: start=8 recent=48 (total=56 frames)
[FeatureFusion] geometry fusion delta scaled by 0.5
[FeatureFusion] importance gate ENABLED on geometry delta
[FeatureFusion] learnable geometry scale ENABLED (init 0.5)
[VGGTEncoder] eval KV-cache window: start=8 recent=48 (total=56 frames)
[FeatureFusion] geometry fusion delta scaled by 0.5
[FeatureFusion] importance gate ENABLED on geometry delta
[FeatureFusion] learnable geometry scale ENABLED (init 0.5)
[VGGTEncoder] eval KV-cache window: start=8 recent=48 (total=56 frames)
[FeatureFusion] geometry fusion delta scaled by 0.5
[FeatureFusion] importance gate ENABLED on geometry delta
[FeatureFusion] learnable geometry scale ENABLED (init 0.5)
Loading weights from local directory
Loading weights from local directory
Loading weights from local directory
Loading weights from local directory
>>>>> STOP loss up-weight: 3.0 on token ids [48964]
Qwen3_5Config {
  "architectures": [
    "Qwen3_5ForConditionalGeneration"
  ],
  "dtype": "bfloat16",
  "feature_fusion_method": "deepstack_language_add",
  "fusion_num_layers": 1,
  "geometry_encoder_layers": [
    11,
    17,
    23
  ],
  "geometry_encoder_path": "/mnt/data/vmo-ai-task/dungpq6/model-checkpoint/VGGT-1B",
  "geometry_encoder_streaming": true,
  "geometry_encoder_type": "vggt",
  "geometry_frame_strict": true,
  "geometry_fusion_layers": [
    0,
    1,
    2
  ],
  "geometry_fusion_scale": 0.5,
  "geometry_importance_gate": true,
  "geometry_learnable_scale": true,
  "geometry_merger_type": "mlp",
  "geometry_spatial_bias": false,
  "image_token_id": 248056,
  "include_camera_token": false,
  "model_type": "qwen3_5",
  "pos_encoding_type": "none",
  "reference_frame": "first",
  "supervised_token_loss": true,
  "text_config": {
    "attention_bias": false,
    "attention_dropout": 0.0,
    "attn_output_gate": true,
    "bos_token_id": null,
    "dtype": "bfloat16",
    "eos_token_id": 248044,
    "full_attention_interval": 4,
    "head_dim": 256,
    "hidden_act": "silu",
    "hidden_size": 2560,
    "initializer_range": 0.02,
    "intermediate_size": 9216,
    "layer_types": [
      "linear_attention",
      "linear_attention",
      "linear_attention",
      "full_attention",
      "linear_attention",
      "linear_attention",
      "linear_attention",
      "full_attention",
      "linear_attention",
      "linear_attention",
      "linear_attention",
      "full_attention",
      "linear_attention",
      "linear_attention",
      "linear_attention",
      "full_attention",
      "linear_attention",
      "linear_attention",
      "linear_attention",
      "full_attention",
      "linear_attention",
      "linear_attention",
      "linear_attention",
      "full_attention",
      "linear_attention",
      "linear_attention",
      "linear_attention",
      "full_attention",
      "linear_attention",
      "linear_attention",
      "linear_attention",
      "full_attention"
    ],
    "linear_conv_kernel_dim": 4,
    "linear_key_head_dim": 128,
    "linear_num_key_heads": 16,
    "linear_num_value_heads": 32,
    "linear_value_head_dim": 128,
    "mamba_ssm_dtype": "float32",
    "max_position_embeddings": 262144,
    "mlp_only_layers": [],
    "model_type": "qwen3_5_text",
    "mtp_num_hidden_layers": 1,
    "mtp_use_dedicated_embeddings": false,
    "num_attention_heads": 16,
    "num_hidden_layers": 32,
    "num_key_value_heads": 4,
    "pad_token_id": null,
    "partial_rotary_factor": 0.25,
    "rms_norm_eps": 1e-06,
    "rope_parameters": {
      "mrope_interleaved": true,
      "mrope_section": [
        11,
        11,
        10
      ],
      "partial_rotary_factor": 0.25,
      "rope_theta": 10000000,
      "rope_type": "default"
    },
    "tie_word_embeddings": true,
    "use_cache": true,
    "vocab_size": 248320
  },
  "tie_word_embeddings": true,
  "transformers_version": "5.3.0",
  "use_cache": false,
  "use_geometry_encoder": true,
  "video_token_id": 248057,
  "vision_config": {
    "deepstack_visual_indexes": [],
    "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_5",
    "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": 248054,
  "vision_language_fusion_layers": null,
  "vision_start_token_id": 248053
}

Loading datasets: [{'annotation_path': '/mnt/data/vmo-ai-task/anhdh35/JanusVLN/train_r2r_rxr.json', 'data_path': '/mnt/data/vmo-ai-task/anhdh35/JanusVLN', 'tag': '3d', 'sampling_rate': 1.0, 'dataset_name': 'train_r2r_rxr'}]
>>>>> STOP loss up-weight: 3.0 on token ids [48964]
Qwen3_5Config {
  "architectures": [
    "Qwen3_5ForConditionalGeneration"
  ],
  "dtype": "bfloat16",
  "feature_fusion_method": "deepstack_language_add",
  "fusion_num_layers": 1,
  "geometry_encoder_layers": [
    11,
    17,
    23
  ],
  "geometry_encoder_path": "/mnt/data/vmo-ai-task/dungpq6/model-checkpoint/VGGT-1B",
  "geometry_encoder_streaming": true,
  "geometry_encoder_type": "vggt",
  "geometry_frame_strict": true,
  "geometry_fusion_layers": [
    0,
    1,
    2
  ],
  "geometry_fusion_scale": 0.5,
  "geometry_importance_gate": true,
  "geometry_learnable_scale": true,
  "geometry_merger_type": "mlp",
  "geometry_spatial_bias": false,
  "image_token_id": 248056,
  "include_camera_token": false,
  "model_type": "qwen3_5",
  "pos_encoding_type": "none",
  "reference_frame": "first",
  "supervised_token_loss": true,
  "text_config": {
    "attention_bias": false,
    "attention_dropout": 0.0,
    "attn_output_gate": true,
    "bos_token_id": null,
    "dtype": "bfloat16",
    "eos_token_id": 248044,
    "full_attention_interval": 4,
    "head_dim": 256,
    "hidden_act": "silu",
    "hidden_size": 2560,
    "initializer_range": 0.02,
    "intermediate_size": 9216,
    "layer_types": [
      "linear_attention",
      "linear_attention",
      "linear_attention",
      "full_attention",
      "linear_attention",
      "linear_attention",
      "linear_attention",
      "full_attention",
      "linear_attention",
      "linear_attention",
      "linear_attention",
      "full_attention",
      "linear_attention",
      "linear_attention",
      "linear_attention",
      "full_attention",
      "linear_attention",
      "linear_attention",
      "linear_attention",
      "full_attention",
      "linear_attention",
      "linear_attention",
      "linear_attention",
      "full_attention",
      "linear_attention",
      "linear_attention",
      "linear_attention",
      "full_attention",
      "linear_attention",
      "linear_attention",
      "linear_attention",
      "full_attention"
    ],
    "linear_conv_kernel_dim": 4,
    "linear_key_head_dim": 128,
    "linear_num_key_heads": 16,
    "linear_num_value_heads": 32,
    "linear_value_head_dim": 128,
    "mamba_ssm_dtype": "float32",
    "max_position_embeddings": 262144,
    "mlp_only_layers": [],
    "model_type": "qwen3_5_text",
    "mtp_num_hidden_layers": 1,
    "mtp_use_dedicated_embeddings": false,
    "num_attention_heads": 16,
    "num_hidden_layers": 32,
    "num_key_value_heads": 4,
    "pad_token_id": null,
    "partial_rotary_factor": 0.25,
    "rms_norm_eps": 1e-06,
    "rope_parameters": {
      "mrope_interleaved": true,
      "mrope_section": [
        11,
        11,
        10
      ],
      "partial_rotary_factor": 0.25,
      "rope_theta": 10000000,
      "rope_type": "default"
    },
    "tie_word_embeddings": true,
    "use_cache": true,
    "vocab_size": 248320
  },
  "tie_word_embeddings": true,
  "transformers_version": "5.3.0",
  "use_cache": false,
  "use_geometry_encoder": true,
  "video_token_id": 248057,
  "vision_config": {
    "deepstack_visual_indexes": [],
    "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_5",
    "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": 248054,
  "vision_language_fusion_layers": null,
  "vision_start_token_id": 248053
}

Loading datasets: [{'annotation_path': '/mnt/data/vmo-ai-task/anhdh35/JanusVLN/train_r2r_rxr.json', 'data_path': '/mnt/data/vmo-ai-task/anhdh35/JanusVLN', 'tag': '3d', 'sampling_rate': 1.0, 'dataset_name': 'train_r2r_rxr'}]
>>>>> STOP loss up-weight: 3.0 on token ids [48964]
Vision Module - Attention Blocks:
Trainable Block Indices: None
Non-Trainable Block Indices: [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23]
Merger Module Trainable: True
LLM Module - Embed Tokens Trainable: True
LLM Module - Trainable Layer Indices: [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31]
LLM Module - Non-Trainable Layer Indices: None
Qwen3_5Config {
  "architectures": [
    "Qwen3_5ForConditionalGeneration"
  ],
  "dtype": "bfloat16",
  "feature_fusion_method": "deepstack_language_add",
  "fusion_num_layers": 1,
  "geometry_encoder_layers": [
    11,
    17,
    23
  ],
  "geometry_encoder_path": "/mnt/data/vmo-ai-task/dungpq6/model-checkpoint/VGGT-1B",
  "geometry_encoder_streaming": true,
  "geometry_encoder_type": "vggt",
  "geometry_frame_strict": true,
  "geometry_fusion_layers": [
    0,
    1,
    2
  ],
  "geometry_fusion_scale": 0.5,
  "geometry_importance_gate": true,
  "geometry_learnable_scale": true,
  "geometry_merger_type": "mlp",
  "geometry_spatial_bias": false,
  "image_token_id": 248056,
  "include_camera_token": false,
  "model_type": "qwen3_5",
  "pos_encoding_type": "none",
  "reference_frame": "first",
  "supervised_token_loss": true,
  "text_config": {
    "attention_bias": false,
    "attention_dropout": 0.0,
    "attn_output_gate": true,
    "bos_token_id": null,
    "dtype": "bfloat16",
    "eos_token_id": 248044,
    "full_attention_interval": 4,
    "head_dim": 256,
    "hidden_act": "silu",
    "hidden_size": 2560,
    "initializer_range": 0.02,
    "intermediate_size": 9216,
    "layer_types": [
      "linear_attention",
      "linear_attention",
      "linear_attention",
      "full_attention",
      "linear_attention",
      "linear_attention",
      "linear_attention",
      "full_attention",
      "linear_attention",
      "linear_attention",
      "linear_attention",
      "full_attention",
      "linear_attention",
      "linear_attention",
      "linear_attention",
      "full_attention",
      "linear_attention",
      "linear_attention",
      "linear_attention",
      "full_attention",
      "linear_attention",
      "linear_attention",
      "linear_attention",
      "full_attention",
      "linear_attention",
      "linear_attention",
      "linear_attention",
      "full_attention",
      "linear_attention",
      "linear_attention",
      "linear_attention",
      "full_attention"
    ],
    "linear_conv_kernel_dim": 4,
    "linear_key_head_dim": 128,
    "linear_num_key_heads": 16,
    "linear_num_value_heads": 32,
    "linear_value_head_dim": 128,
    "mamba_ssm_dtype": "float32",
    "max_position_embeddings": 262144,
    "mlp_only_layers": [],
    "model_type": "qwen3_5_text",
    "mtp_num_hidden_layers": 1,
    "mtp_use_dedicated_embeddings": false,
    "num_attention_heads": 16,
    "num_hidden_layers": 32,
    "num_key_value_heads": 4,
    "pad_token_id": null,
    "partial_rotary_factor": 0.25,
    "rms_norm_eps": 1e-06,
    "rope_parameters": {
      "mrope_interleaved": true,
      "mrope_section": [
        11,
        11,
        10
      ],
      "partial_rotary_factor": 0.25,
      "rope_theta": 10000000,
      "rope_type": "default"
    },
    "tie_word_embeddings": true,
    "use_cache": true,
    "vocab_size": 248320
  },
  "tie_word_embeddings": true,
  "transformers_version": "5.3.0",
  "use_cache": false,
  "use_geometry_encoder": true,
  "video_token_id": 248057,
  "vision_config": {
    "deepstack_visual_indexes": [],
    "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_5",
    "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": 248054,
  "vision_language_fusion_layers": null,
  "vision_start_token_id": 248053
}

Loading datasets: [{'annotation_path': '/mnt/data/vmo-ai-task/anhdh35/JanusVLN/train_r2r_rxr.json', 'data_path': '/mnt/data/vmo-ai-task/anhdh35/JanusVLN', 'tag': '3d', 'sampling_rate': 1.0, 'dataset_name': 'train_r2r_rxr'}]
>>>>> STOP loss up-weight: 3.0 on token ids [48964]
Qwen3_5Config {
  "architectures": [
    "Qwen3_5ForConditionalGeneration"
  ],
  "dtype": "bfloat16",
  "feature_fusion_method": "deepstack_language_add",
  "fusion_num_layers": 1,
  "geometry_encoder_layers": [
    11,
    17,
    23
  ],
  "geometry_encoder_path": "/mnt/data/vmo-ai-task/dungpq6/model-checkpoint/VGGT-1B",
  "geometry_encoder_streaming": true,
  "geometry_encoder_type": "vggt",
  "geometry_frame_strict": true,
  "geometry_fusion_layers": [
    0,
    1,
    2
  ],
  "geometry_fusion_scale": 0.5,
  "geometry_importance_gate": true,
  "geometry_learnable_scale": true,
  "geometry_merger_type": "mlp",
  "geometry_spatial_bias": false,
  "image_token_id": 248056,
  "include_camera_token": false,
  "model_type": "qwen3_5",
  "pos_encoding_type": "none",
  "reference_frame": "first",
  "supervised_token_loss": true,
  "text_config": {
    "attention_bias": false,
    "attention_dropout": 0.0,
    "attn_output_gate": true,
    "bos_token_id": null,
    "dtype": "bfloat16",
    "eos_token_id": 248044,
    "full_attention_interval": 4,
    "head_dim": 256,
    "hidden_act": "silu",
    "hidden_size": 2560,
    "initializer_range": 0.02,
    "intermediate_size": 9216,
    "layer_types": [
      "linear_attention",
      "linear_attention",
      "linear_attention",
      "full_attention",
      "linear_attention",
      "linear_attention",
      "linear_attention",
      "full_attention",
      "linear_attention",
      "linear_attention",
      "linear_attention",
      "full_attention",
      "linear_attention",
      "linear_attention",
      "linear_attention",
      "full_attention",
      "linear_attention",
      "linear_attention",
      "linear_attention",
      "full_attention",
      "linear_attention",
      "linear_attention",
      "linear_attention",
      "full_attention",
      "linear_attention",
      "linear_attention",
      "linear_attention",
      "full_attention",
      "linear_attention",
      "linear_attention",
      "linear_attention",
      "full_attention"
    ],
    "linear_conv_kernel_dim": 4,
    "linear_key_head_dim": 128,
    "linear_num_key_heads": 16,
    "linear_num_value_heads": 32,
    "linear_value_head_dim": 128,
    "mamba_ssm_dtype": "float32",
    "max_position_embeddings": 262144,
    "mlp_only_layers": [],
    "model_type": "qwen3_5_text",
    "mtp_num_hidden_layers": 1,
    "mtp_use_dedicated_embeddings": false,
    "num_attention_heads": 16,
    "num_hidden_layers": 32,
    "num_key_value_heads": 4,
    "pad_token_id": null,
    "partial_rotary_factor": 0.25,
    "rms_norm_eps": 1e-06,
    "rope_parameters": {
      "mrope_interleaved": true,
      "mrope_section": [
        11,
        11,
        10
      ],
      "partial_rotary_factor": 0.25,
      "rope_theta": 10000000,
      "rope_type": "default"
    },
    "tie_word_embeddings": true,
    "use_cache": true,
    "vocab_size": 248320
  },
  "tie_word_embeddings": true,
  "transformers_version": "5.3.0",
  "use_cache": false,
  "use_geometry_encoder": true,
  "video_token_id": 248057,
  "vision_config": {
    "deepstack_visual_indexes": [],
    "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_5",
    "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": 248054,
  "vision_language_fusion_layers": null,
  "vision_start_token_id": 248053
}

Loading datasets: [{'annotation_path': '/mnt/data/vmo-ai-task/anhdh35/JanusVLN/train_r2r_rxr.json', 'data_path': '/mnt/data/vmo-ai-task/anhdh35/JanusVLN', 'tag': '3d', 'sampling_rate': 1.0, 'dataset_name': 'train_r2r_rxr'}]
Total training samples: 2429620
Total training samples: 2429620
Total training samples: 2429620
Formatting inputs...Skip in lazy mode
Formatting inputs...Skip in lazy mode
Formatting inputs...Skip in lazy mode
Total training samples: 2429620
The tokenizer has new PAD/BOS/EOS tokens that differ from the model config and generation config. The model config and generation config were aligned accordingly, being updated with the tokenizer's values. Updated tokens: {'eos_token_id': 248046, 'pad_token_id': 248044}.
The tokenizer has new PAD/BOS/EOS tokens that differ from the model config and generation config. The model config and generation config were aligned accordingly, being updated with the tokenizer's values. Updated tokens: {'eos_token_id': 248046, 'pad_token_id': 248044}.
Formatting inputs...Skip in lazy mode
The tokenizer has new PAD/BOS/EOS tokens that differ from the model config and generation config. The model config and generation config were aligned accordingly, being updated with the tokenizer's values. Updated tokens: {'eos_token_id': 248046, 'pad_token_id': 248044}.
The tokenizer has new PAD/BOS/EOS tokens that differ from the model config and generation config. The model config and generation config were aligned accordingly, being updated with the tokenizer's values. Updated tokens: {'eos_token_id': 248046, 'pad_token_id': 248044}.
NCCL version 2.26.2+cuda12.2

[2026-09-18 03:40:48] worker-1:2236406:2237140 [1] ras/client_support.cc:160 NCCL WARN Call to bind failed: Address already in use

[2026-09-18 03:40:48] worker-1:2236405:2237138 [0] ras/client_support.cc:160 NCCL WARN Call to bind failed: Address already in use

[2026-09-18 03:40:48] worker-1:2236408:2237142 [3] ras/client_support.cc:160 NCCL WARN Call to bind failed: Address already in use

[2026-09-18 03:40:48] worker-1:2236407:2237145 [2] ras/client_support.cc:160 NCCL WARN Call to bind failed: Address already in use
[TrainingRuntime] {"rank": "3", "versions": {"torch": "2.7.0+cu128", "transformers": "5.3.0", "deepspeed": "0.16.4", "flash-attn": "2.8.3", "triton": "3.3.0", "flash-linear-attention": "0.5.1", "causal-conv1d": "1.6.2.post1"}, "cuda": "12.8", "world_size": 4, "gpu": "NVIDIA H100 80GB HBM3", "micro_batch": 16, "accumulation": 1, "effective_global_batch": 64, "supervised_token_loss": true, "NCCL_NVLS_ENABLE": "0", "CUDA_LAUNCH_BLOCKING": "0", "deepspeed": "scripts/zero2_h100.json"}
[TrainingMemory] {"rank": "3", "stage": "train_begin", "step": 0, "allocated_GiB": 14.601, "reserved_GiB": 19.906, "peak_allocated_GiB": 18.713, "peak_reserved_GiB": 19.906, "device_free_GiB": 57.864, "device_total_GiB": 79.189}
[TrainingRuntime] {"rank": "1", "versions": {"torch": "2.7.0+cu128", "transformers": "5.3.0", "deepspeed": "0.16.4", "flash-attn": "2.8.3", "triton": "3.3.0", "flash-linear-attention": "0.5.1", "causal-conv1d": "1.6.2.post1"}, "cuda": "12.8", "world_size": 4, "gpu": "NVIDIA H100 80GB HBM3", "micro_batch": 16, "accumulation": 1, "effective_global_batch": 64, "supervised_token_loss": true, "NCCL_NVLS_ENABLE": "0", "CUDA_LAUNCH_BLOCKING": "0", "deepspeed": "scripts/zero2_h100.json"}
[TrainingMemory] {"rank": "1", "stage": "train_begin", "step": 0, "allocated_GiB": 14.601, "reserved_GiB": 19.906, "peak_allocated_GiB": 18.713, "peak_reserved_GiB": 19.906, "device_free_GiB": 57.864, "device_total_GiB": 79.189}
[TrainingRuntime] {"rank": "2", "versions": {"torch": "2.7.0+cu128", "transformers": "5.3.0", "deepspeed": "0.16.4", "flash-attn": "2.8.3", "triton": "3.3.0", "flash-linear-attention": "0.5.1", "causal-conv1d": "1.6.2.post1"}, "cuda": "12.8", "world_size": 4, "gpu": "NVIDIA H100 80GB HBM3", "micro_batch": 16, "accumulation": 1, "effective_global_batch": 64, "supervised_token_loss": true, "NCCL_NVLS_ENABLE": "0", "CUDA_LAUNCH_BLOCKING": "0", "deepspeed": "scripts/zero2_h100.json"}
[TrainingMemory] {"rank": "2", "stage": "train_begin", "step": 0, "allocated_GiB": 14.601, "reserved_GiB": 19.906, "peak_allocated_GiB": 18.713, "peak_reserved_GiB": 19.906, "device_free_GiB": 57.864, "device_total_GiB": 79.189}
[TrainingRuntime] {"rank": "0", "versions": {"torch": "2.7.0+cu128", "transformers": "5.3.0", "deepspeed": "0.16.4", "flash-attn": "2.8.3", "triton": "3.3.0", "flash-linear-attention": "0.5.1", "causal-conv1d": "1.6.2.post1"}, "cuda": "12.8", "world_size": 4, "gpu": "NVIDIA H100 80GB HBM3", "micro_batch": 16, "accumulation": 1, "effective_global_batch": 64, "supervised_token_loss": true, "NCCL_NVLS_ENABLE": "0", "CUDA_LAUNCH_BLOCKING": "0", "deepspeed": "scripts/zero2_h100.json"}
[TrainingMemory] {"rank": "0", "stage": "train_begin", "step": 0, "allocated_GiB": 14.601, "reserved_GiB": 19.906, "peak_allocated_GiB": 14.601, "peak_reserved_GiB": 19.906, "device_free_GiB": 57.864, "device_total_GiB": 79.189}

  0%|          | 0/50 [00:00<?, ?it/s][GeometryBatch] micro_batch=16, mode=per-sample VGGT fallback, input_shapes=[(9, 3, 420, 560), (9, 3, 420, 560), (9, 3, 420, 560), (9, 3, 420, 560), (9, 3, 420, 560), (9, 3, 420, 560), (9, 3, 420, 560), (9, 3, 420, 560), (9, 3, 420, 560), (9, 3, 420, 560), (9, 3, 420, 560), (9, 3, 420, 560), (9, 3, 420, 560), (9, 3, 420, 560), (6, 3, 420, 560), (3, 3, 420, 560)]
/home/dungpq6/Project/SpatialStack/src/qwen_vl/model/geometry_encoders/vggt_encoder.py:337: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead.
  with torch.cuda.amp.autocast(dtype=dtype):
[GeometryBatch] micro_batch=16, mode=per-sample VGGT fallback, input_shapes=[(9, 3, 420, 560), (9, 3, 420, 560), (9, 3, 420, 560), (9, 3, 420, 560), (9, 3, 420, 560), (9, 3, 420, 560), (9, 3, 420, 560), (9, 3, 420, 560), (9, 3, 420, 560), (9, 3, 420, 560), (9, 3, 420, 560), (9, 3, 420, 560), (9, 3, 420, 560), (9, 3, 420, 560), (8, 3, 420, 560), (3, 3, 420, 560)]
/home/dungpq6/Project/SpatialStack/src/qwen_vl/model/geometry_encoders/vggt_encoder.py:337: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead.
  with torch.cuda.amp.autocast(dtype=dtype):
[GeometryBatch] micro_batch=16, mode=per-sample VGGT fallback, input_shapes=[(9, 3, 420, 560), (9, 3, 420, 560), (9, 3, 420, 560), (9, 3, 420, 560), (9, 3, 420, 560), (9, 3, 420, 560), (9, 3, 420, 560), (9, 3, 420, 560), (9, 3, 420, 560), (9, 3, 420, 560), (9, 3, 420, 560), (9, 3, 420, 560), (9, 3, 420, 560), (9, 3, 420, 560), (5, 3, 420, 560), (4, 3, 420, 560)]
/home/dungpq6/Project/SpatialStack/src/qwen_vl/model/geometry_encoders/vggt_encoder.py:337: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead.
  with torch.cuda.amp.autocast(dtype=dtype):
ERROR: Unexpected bus error encountered in worker. This might be caused by insufficient shared memory (shm).
ERROR: Unexpected bus error encountered in worker. This might be caused by insufficient shared memory (shm).
ERROR: Unexpected bus error encountered in worker. This might be caused by insufficient shared memory (shm).
ERROR: Unexpected bus error encountered in worker. This might be caused by insufficient shared memory (shm).
ERROR: Unexpected bus error encountered in worker. This might be caused by insufficient shared memory (shm).
ERROR: Unexpected bus error encountered in worker. This might be caused by insufficient shared memory (shm).
ERROR: Unexpected bus error encountered in worker. This might be caused by insufficient shared memory (shm).
ERROR: Unexpected bus error encountered in worker. This might be caused by insufficient shared memory (shm).
ERROR: Unexpected bus error encountered in worker. This might be caused by insufficient shared memory (shm).
ERROR: Unexpected bus error encountered in worker. This might be caused by insufficient shared memory (shm).
ERROR: Unexpected bus error encountered in worker. This might be caused by insufficient shared memory (shm).
ERROR: Unexpected bus error encountered in worker. This might be caused by insufficient shared memory (shm).
ERROR: Unexpected bus error encountered in worker. This might be caused by insufficient shared memory (shm).
ERROR: Unexpected bus error encountered in worker. This might be caused by insufficient shared memory (shm).
ERROR: Unexpected bus error encountered in worker. This might be caused by insufficient shared memory (shm).
ERROR: Unexpected bus error encountered in worker. This might be caused by insufficient shared memory (shm).
ERROR: Unexpected bus error encountered in worker. This might be caused by insufficient shared memory (shm).
ERROR: Unexpected bus error encountered in worker. This might be caused by insufficient shared memory (shm).
[rank0]: Traceback (most recent call last):
[rank0]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/utils/data/dataloader.py", line 1284, in _try_get_data
[rank0]:     data = self._data_queue.get(timeout=timeout)
[rank0]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank0]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/queue.py", line 180, in get
[rank0]:     self.not_empty.wait(remaining)
[rank0]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/threading.py", line 359, in wait
[rank0]:     gotit = waiter.acquire(True, timeout)
[rank0]:             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank0]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/utils/data/_utils/signal_handling.py", line 73, in handler
[rank0]:     _error_if_any_worker_fails()
[rank0]: RuntimeError: DataLoader worker (pid 2237269) is killed by signal: Bus error. It is possible that dataloader's workers are out of shared memory. Please try to raise your shared memory limit.

[rank0]: The above exception was the direct cause of the following exception:

[rank0]: Traceback (most recent call last):
[rank0]:   File "/home/dungpq6/Project/SpatialStack/src/qwen_vl/train/train_qwen.py", line 414, in <module>
[rank0]:     train(attn_implementation="flash_attention_2")
[rank0]:   File "/home/dungpq6/Project/SpatialStack/src/qwen_vl/train/train_qwen.py", line 378, in train
[rank0]:     trainer.train()
[rank0]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/transformers/trainer.py", line 1424, in train
[rank0]:     return inner_training_loop(
[rank0]:            ^^^^^^^^^^^^^^^^^^^^
[rank0]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/transformers/trainer.py", line 1506, in _inner_training_loop
[rank0]:     self._run_epoch(
[rank0]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/transformers/trainer.py", line 1701, in _run_epoch
[rank0]:     batch_samples, num_items_in_batch = self.get_batch_samples(epoch_iterator, num_batches, self.args.device)
[rank0]:                                         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank0]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/transformers/trainer.py", line 2102, in get_batch_samples
[rank0]:     batch_samples.append(next(epoch_iterator))
[rank0]:                          ^^^^^^^^^^^^^^^^^^^^
[rank0]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/accelerate/data_loader.py", line 577, in __iter__
[rank0]:     current_batch = next(dataloader_iter)
[rank0]:                     ^^^^^^^^^^^^^^^^^^^^^
[rank0]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/utils/data/dataloader.py", line 733, in __next__
[rank0]:     data = self._next_data()
[rank0]:            ^^^^^^^^^^^^^^^^^
[rank0]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/utils/data/dataloader.py", line 1491, in _next_data
[rank0]:     idx, data = self._get_data()
[rank0]:                 ^^^^^^^^^^^^^^^^
[rank0]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/utils/data/dataloader.py", line 1443, in _get_data
[rank0]:     success, data = self._try_get_data()
[rank0]:                     ^^^^^^^^^^^^^^^^^^^^
[rank0]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/utils/data/dataloader.py", line 1297, in _try_get_data
[rank0]:     raise RuntimeError(
[rank0]: RuntimeError: DataLoader worker (pid(s) 2237269) exited unexpectedly
Exception in thread Thread-2 (_pin_memory_loop):
Traceback (most recent call last):
  File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/threading.py", line 1075, in _bootstrap_inner
    self.run()
  File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/threading.py", line 1012, in run
    self._target(*self._args, **self._kwargs)
  File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/utils/data/_utils/pin_memory.py", line 61, in _pin_memory_loop
    do_one_step()
  File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/utils/data/_utils/pin_memory.py", line 37, in do_one_step
    r = in_queue.get(timeout=MP_STATUS_CHECK_INTERVAL)
        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/multiprocessing/queues.py", line 122, in get
    return _ForkingPickler.loads(res)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/multiprocessing/reductions.py", line 541, in rebuild_storage_fd
    fd = df.detach()
         ^^^^^^^^^^^
  File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/multiprocessing/resource_sharer.py", line 57, in detach
    with _resource_sharer.get_connection(self._id) as conn:
         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/multiprocessing/resource_sharer.py", line 86, in get_connection
    c = Client(address, authkey=process.current_process().authkey)
        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/multiprocessing/connection.py", line 525, in Client
    answer_challenge(c, authkey)
  File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/multiprocessing/connection.py", line 953, in answer_challenge
    message = connection.recv_bytes(256)         # reject large message
              ^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/multiprocessing/connection.py", line 216, in recv_bytes
    buf = self._recv_bytes(maxlength)
          ^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/multiprocessing/connection.py", line 430, in _recv_bytes
    buf = self._recv(4)
          ^^^^^^^^^^^^^
  File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/multiprocessing/connection.py", line 395, in _recv
    chunk = read(handle, remaining)
            ^^^^^^^^^^^^^^^^^^^^^^^
ConnectionResetError: [Errno 104] Connection reset by peer
[rank3]: Traceback (most recent call last):
[rank3]:   File "/home/dungpq6/Project/SpatialStack/src/qwen_vl/train/train_qwen.py", line 414, in <module>
[rank3]:     train(attn_implementation="flash_attention_2")
[rank3]:   File "/home/dungpq6/Project/SpatialStack/src/qwen_vl/train/train_qwen.py", line 378, in train
[rank3]:     trainer.train()
[rank3]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/transformers/trainer.py", line 1424, in train
[rank3]:     return inner_training_loop(
[rank3]:            ^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/transformers/trainer.py", line 1506, in _inner_training_loop
[rank3]:     self._run_epoch(
[rank3]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/transformers/trainer.py", line 1734, in _run_epoch
[rank3]:     tr_loss_step = self.training_step(model, inputs, num_items_in_batch)
[rank3]:                    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/transformers/trainer.py", line 1906, in training_step
[rank3]:     loss = self.compute_loss(model, inputs, num_items_in_batch=num_items_in_batch)
[rank3]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/transformers/trainer.py", line 1978, in compute_loss
[rank3]:     outputs = model(**inputs)
[rank3]:               ^^^^^^^^^^^^^^^
[rank3]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl
[rank3]:     return self._call_impl(*args, **kwargs)
[rank3]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl
[rank3]:     return forward_call(*args, **kwargs)
[rank3]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/deepspeed/utils/nvtx.py", line 18, in wrapped_fn
[rank3]:     ret_val = func(*args, **kwargs)
[rank3]:               ^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/deepspeed/runtime/engine.py", line 1987, in forward
[rank3]:     loss = self.module(*inputs, **kwargs)
[rank3]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl
[rank3]:     return self._call_impl(*args, **kwargs)
[rank3]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl
[rank3]:     return forward_call(*args, **kwargs)
[rank3]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/home/dungpq6/Project/SpatialStack/src/qwen_vl/model/modeling_qwen3_5.py", line 1077, in forward
[rank3]:     outputs = self.model(
[rank3]:               ^^^^^^^^^^^
[rank3]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl
[rank3]:     return self._call_impl(*args, **kwargs)
[rank3]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl
[rank3]:     return forward_call(*args, **kwargs)
[rank3]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/home/dungpq6/Project/SpatialStack/src/qwen_vl/model/modeling_qwen3_5.py", line 966, in forward
[rank3]:     geometry_layer_features = self._collect_geometry_layer_features(geometry_encoder_inputs)
[rank3]:                               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/home/dungpq6/Project/SpatialStack/src/qwen_vl/model/modeling_qwen3_5.py", line 762, in _collect_geometry_layer_features
[rank3]:     sample_features = self.geometry_encoder.encode_layers_with_mode(
[rank3]:                       ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/home/dungpq6/Project/SpatialStack/src/qwen_vl/model/geometry_encoders/vggt_encoder.py", line 647, in encode_layers_with_mode
[rank3]:     return self.encode_layers_streaming(
[rank3]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/home/dungpq6/Project/SpatialStack/src/qwen_vl/model/geometry_encoders/vggt_encoder.py", line 340, in encode_layers_streaming
[rank3]:     output = self.vggt.aggregator(
[rank3]:              ^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl
[rank3]:     return self._call_impl(*args, **kwargs)
[rank3]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl
[rank3]:     return forward_call(*args, **kwargs)
[rank3]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/home/dungpq6/Project/SpatialStack/src/qwen_vl/model/vggt/models/aggregator.py", line 229, in forward
[rank3]:     patch_tokens = self.patch_embed(images)
[rank3]:                    ^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl
[rank3]:     return self._call_impl(*args, **kwargs)
[rank3]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl
[rank3]:     return forward_call(*args, **kwargs)
[rank3]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/home/dungpq6/Project/SpatialStack/src/qwen_vl/model/vggt/layers/vision_transformer.py", line 326, in forward
[rank3]:     ret = self.forward_features(*args, **kwargs)
[rank3]:           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/home/dungpq6/Project/SpatialStack/src/qwen_vl/model/vggt/layers/vision_transformer.py", line 262, in forward_features
[rank3]:     x = blk(x)
[rank3]:         ^^^^^^
[rank3]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl
[rank3]:     return self._call_impl(*args, **kwargs)
[rank3]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl
[rank3]:     return forward_call(*args, **kwargs)
[rank3]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/home/dungpq6/Project/SpatialStack/src/qwen_vl/model/vggt/layers/block.py", line 257, in forward
[rank3]:     return super().forward(x_or_x_list)
[rank3]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/home/dungpq6/Project/SpatialStack/src/qwen_vl/model/vggt/layers/block.py", line 109, in forward
[rank3]:     x = x + attn_residual_func(x, pos=pos, attn_mask=attn_mask)
[rank3]:             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/home/dungpq6/Project/SpatialStack/src/qwen_vl/model/vggt/layers/block.py", line 82, in attn_residual_func
[rank3]:     return self.ls1(self.attn(self.norm1(x), pos=pos))
[rank3]:                               ^^^^^^^^^^^^^
[rank3]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl
[rank3]:     return self._call_impl(*args, **kwargs)
[rank3]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1756, in _call_impl
[rank3]:     forward_call = (self._slow_forward if torch._C._get_tracing_state() else self.forward)
[rank3]:                                           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/utils/data/_utils/signal_handling.py", line 73, in handler
[rank3]:     _error_if_any_worker_fails()
[rank3]: RuntimeError: DataLoader worker (pid 2237174) is killed by signal: Bus error. It is possible that dataloader's workers are out of shared memory. Please try to raise your shared memory limit.
[rank1]: Traceback (most recent call last):
[rank1]:   File "/home/dungpq6/Project/SpatialStack/src/qwen_vl/train/train_qwen.py", line 414, in <module>
[rank1]:     train(attn_implementation="flash_attention_2")
[rank1]:   File "/home/dungpq6/Project/SpatialStack/src/qwen_vl/train/train_qwen.py", line 378, in train
[rank1]:     trainer.train()
[rank1]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/transformers/trainer.py", line 1424, in train
[rank1]:     return inner_training_loop(
[rank1]:            ^^^^^^^^^^^^^^^^^^^^
[rank1]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/transformers/trainer.py", line 1506, in _inner_training_loop
[rank1]:     self._run_epoch(
[rank1]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/transformers/trainer.py", line 1734, in _run_epoch
[rank1]:     tr_loss_step = self.training_step(model, inputs, num_items_in_batch)
[rank1]:                    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank1]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/transformers/trainer.py", line 1906, in training_step
[rank1]:     loss = self.compute_loss(model, inputs, num_items_in_batch=num_items_in_batch)
[rank1]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank1]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/transformers/trainer.py", line 1978, in compute_loss
[rank1]:     outputs = model(**inputs)
[rank1]:               ^^^^^^^^^^^^^^^
[rank1]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl
[rank1]:     return self._call_impl(*args, **kwargs)
[rank1]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank1]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl
[rank1]:     return forward_call(*args, **kwargs)
[rank1]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank1]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/deepspeed/utils/nvtx.py", line 18, in wrapped_fn
[rank1]:     ret_val = func(*args, **kwargs)
[rank1]:               ^^^^^^^^^^^^^^^^^^^^^
[rank1]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/deepspeed/runtime/engine.py", line 1987, in forward
[rank1]:     loss = self.module(*inputs, **kwargs)
[rank1]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank1]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl
[rank1]:     return self._call_impl(*args, **kwargs)
[rank1]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank1]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl
[rank1]:     return forward_call(*args, **kwargs)
[rank1]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank1]:   File "/home/dungpq6/Project/SpatialStack/src/qwen_vl/model/modeling_qwen3_5.py", line 1077, in forward
[rank1]:     outputs = self.model(
[rank1]:               ^^^^^^^^^^^
[rank1]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl
[rank1]:     return self._call_impl(*args, **kwargs)
[rank1]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank1]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl
[rank1]:     return forward_call(*args, **kwargs)
[rank1]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank1]:   File "/home/dungpq6/Project/SpatialStack/src/qwen_vl/model/modeling_qwen3_5.py", line 966, in forward
[rank1]:     geometry_layer_features = self._collect_geometry_layer_features(geometry_encoder_inputs)
[rank1]:                               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank1]:   File "/home/dungpq6/Project/SpatialStack/src/qwen_vl/model/modeling_qwen3_5.py", line 762, in _collect_geometry_layer_features
[rank1]:     sample_features = self.geometry_encoder.encode_layers_with_mode(
[rank1]:                       ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank1]:   File "/home/dungpq6/Project/SpatialStack/src/qwen_vl/model/geometry_encoders/vggt_encoder.py", line 647, in encode_layers_with_mode
[rank1]:     return self.encode_layers_streaming(
[rank1]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank1]:   File "/home/dungpq6/Project/SpatialStack/src/qwen_vl/model/geometry_encoders/vggt_encoder.py", line 340, in encode_layers_streaming
[rank1]:     output = self.vggt.aggregator(
[rank1]:              ^^^^^^^^^^^^^^^^^^^^^
[rank1]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl
[rank1]:     return self._call_impl(*args, **kwargs)
[rank1]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank1]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl
[rank1]:     return forward_call(*args, **kwargs)
[rank1]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank1]:   File "/home/dungpq6/Project/SpatialStack/src/qwen_vl/model/vggt/models/aggregator.py", line 279, in forward
[rank1]:     tokens, frame_idx, frame_intermediates = self._process_frame_attention(
[rank1]:                                              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank1]:   File "/home/dungpq6/Project/SpatialStack/src/qwen_vl/model/vggt/models/aggregator.py", line 332, in _process_frame_attention
[rank1]:     tokens = self.frame_blocks[frame_idx](tokens, pos=pos)
[rank1]:              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank1]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl
[rank1]:     return self._call_impl(*args, **kwargs)
[rank1]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank1]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl
[rank1]:     return forward_call(*args, **kwargs)
[rank1]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank1]:   File "/home/dungpq6/Project/SpatialStack/src/qwen_vl/model/vggt/layers/block.py", line 109, in forward
[rank1]:     x = x + attn_residual_func(x, pos=pos, attn_mask=attn_mask)
[rank1]:             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank1]:   File "/home/dungpq6/Project/SpatialStack/src/qwen_vl/model/vggt/layers/block.py", line 82, in attn_residual_func
[rank1]:     return self.ls1(self.attn(self.norm1(x), pos=pos))
[rank1]:                     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank1]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl
[rank1]:     return self._call_impl(*args, **kwargs)
[rank1]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank1]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl
[rank1]:     return forward_call(*args, **kwargs)
[rank1]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank1]:   File "/home/dungpq6/Project/SpatialStack/src/qwen_vl/model/vggt/layers/attention.py", line 78, in forward
[rank1]:     k = self.rope(k, pos_k)
[rank1]:         ^^^^^^^^^^^^^^^^^^^
[rank1]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl
[rank1]:     return self._call_impl(*args, **kwargs)
[rank1]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank1]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl
[rank1]:     return forward_call(*args, **kwargs)
[rank1]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank1]:   File "/home/dungpq6/Project/SpatialStack/src/qwen_vl/model/vggt/layers/rope.py", line 184, in forward
[rank1]:     vertical_features = self._apply_1d_rope(vertical_features, positions[..., 0], cos_comp, sin_comp)
[rank1]:                         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank1]:   File "/home/dungpq6/Project/SpatialStack/src/qwen_vl/model/vggt/layers/rope.py", line 152, in _apply_1d_rope
[rank1]:     return (tokens * cos) + (self._rotate_features(tokens) * sin)
[rank1]:                              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank1]:   File "/home/dungpq6/Project/SpatialStack/src/qwen_vl/model/vggt/layers/rope.py", line 119, in _rotate_features
[rank1]:     @staticmethod

[rank1]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/utils/data/_utils/signal_handling.py", line 73, in handler
[rank1]:     _error_if_any_worker_fails()
[rank1]: RuntimeError: DataLoader worker (pid 2237177) is killed by signal: Bus error. It is possible that dataloader's workers are out of shared memory. Please try to raise your shared memory limit.
[rank2]: Traceback (most recent call last):
[rank2]:   File "/home/dungpq6/Project/SpatialStack/src/qwen_vl/train/train_qwen.py", line 414, in <module>
[rank2]:     train(attn_implementation="flash_attention_2")
[rank2]:   File "/home/dungpq6/Project/SpatialStack/src/qwen_vl/train/train_qwen.py", line 378, in train
[rank2]:     trainer.train()
[rank2]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/transformers/trainer.py", line 1424, in train
[rank2]:     return inner_training_loop(
[rank2]:            ^^^^^^^^^^^^^^^^^^^^
[rank2]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/transformers/trainer.py", line 1506, in _inner_training_loop
[rank2]:     self._run_epoch(
[rank2]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/transformers/trainer.py", line 1734, in _run_epoch
[rank2]:     tr_loss_step = self.training_step(model, inputs, num_items_in_batch)
[rank2]:                    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank2]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/transformers/trainer.py", line 1906, in training_step
[rank2]:     loss = self.compute_loss(model, inputs, num_items_in_batch=num_items_in_batch)
[rank2]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank2]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/transformers/trainer.py", line 1978, in compute_loss
[rank2]:     outputs = model(**inputs)
[rank2]:               ^^^^^^^^^^^^^^^
[rank2]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl
[rank2]:     return self._call_impl(*args, **kwargs)
[rank2]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank2]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl
[rank2]:     return forward_call(*args, **kwargs)
[rank2]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank2]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/deepspeed/utils/nvtx.py", line 18, in wrapped_fn
[rank2]:     ret_val = func(*args, **kwargs)
[rank2]:               ^^^^^^^^^^^^^^^^^^^^^
[rank2]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/deepspeed/runtime/engine.py", line 1987, in forward
[rank2]:     loss = self.module(*inputs, **kwargs)
[rank2]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank2]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl
[rank2]:     return self._call_impl(*args, **kwargs)
[rank2]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank2]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl
[rank2]:     return forward_call(*args, **kwargs)
[rank2]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank2]:   File "/home/dungpq6/Project/SpatialStack/src/qwen_vl/model/modeling_qwen3_5.py", line 1077, in forward
[rank2]:     outputs = self.model(
[rank2]:               ^^^^^^^^^^^
[rank2]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl
[rank2]:     return self._call_impl(*args, **kwargs)
[rank2]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank2]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl
[rank2]:     return forward_call(*args, **kwargs)
[rank2]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank2]:   File "/home/dungpq6/Project/SpatialStack/src/qwen_vl/model/modeling_qwen3_5.py", line 966, in forward
[rank2]:     geometry_layer_features = self._collect_geometry_layer_features(geometry_encoder_inputs)
[rank2]:                               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank2]:   File "/home/dungpq6/Project/SpatialStack/src/qwen_vl/model/modeling_qwen3_5.py", line 762, in _collect_geometry_layer_features
[rank2]:     sample_features = self.geometry_encoder.encode_layers_with_mode(
[rank2]:                       ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank2]:   File "/home/dungpq6/Project/SpatialStack/src/qwen_vl/model/geometry_encoders/vggt_encoder.py", line 647, in encode_layers_with_mode
[rank2]:     return self.encode_layers_streaming(
[rank2]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank2]:   File "/home/dungpq6/Project/SpatialStack/src/qwen_vl/model/geometry_encoders/vggt_encoder.py", line 340, in encode_layers_streaming
[rank2]:     output = self.vggt.aggregator(
[rank2]:              ^^^^^^^^^^^^^^^^^^^^^
[rank2]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl
[rank2]:     return self._call_impl(*args, **kwargs)
[rank2]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank2]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl
[rank2]:     return forward_call(*args, **kwargs)
[rank2]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank2]:   File "/home/dungpq6/Project/SpatialStack/src/qwen_vl/model/vggt/models/aggregator.py", line 279, in forward
[rank2]:     tokens, frame_idx, frame_intermediates = self._process_frame_attention(
[rank2]:                                              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank2]:   File "/home/dungpq6/Project/SpatialStack/src/qwen_vl/model/vggt/models/aggregator.py", line 332, in _process_frame_attention
[rank2]:     tokens = self.frame_blocks[frame_idx](tokens, pos=pos)
[rank2]:              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank2]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl
[rank2]:     return self._call_impl(*args, **kwargs)
[rank2]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank2]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl
[rank2]:     return forward_call(*args, **kwargs)
[rank2]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank2]:   File "/home/dungpq6/Project/SpatialStack/src/qwen_vl/model/vggt/layers/block.py", line 110, in forward
[rank2]:     x = x + ffn_residual_func(x)
[rank2]:             ^^^^^^^^^^^^^^^^^^^^
[rank2]:   File "/home/dungpq6/Project/SpatialStack/src/qwen_vl/model/vggt/layers/block.py", line 84, in ffn_residual_func
[rank2]:     return self.ls2(self.mlp(self.norm2(x)))
[rank2]:                     ^^^^^^^^^^^^^^^^^^^^^^^
[rank2]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl
[rank2]:     return self._call_impl(*args, **kwargs)
[rank2]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank2]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl
[rank2]:     return forward_call(*args, **kwargs)
[rank2]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank2]:   File "/home/dungpq6/Project/SpatialStack/src/qwen_vl/model/vggt/layers/mlp.py", line 38, in forward
[rank2]:     x = self.fc2(x)
[rank2]:         ^^^^^^^^^^^
[rank2]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1751, in _wrapped_call_impl
[rank2]:     return self._call_impl(*args, **kwargs)
[rank2]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank2]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1762, in _call_impl
[rank2]:     return forward_call(*args, **kwargs)
[rank2]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank2]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/nn/modules/linear.py", line 125, in forward
[rank2]:     return F.linear(input, self.weight, self.bias)
[rank2]:            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank2]:   File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/utils/data/_utils/signal_handling.py", line 73, in handler
[rank2]:     _error_if_any_worker_fails()
[rank2]: RuntimeError: DataLoader worker (pid 2237181) is killed by signal: Bus error. It is possible that dataloader's workers are out of shared memory. Please try to raise your shared memory limit.
Exception in thread Thread-2 (_pin_memory_loop):
Traceback (most recent call last):
  File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/threading.py", line 1075, in _bootstrap_inner
    self.run()
  File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/threading.py", line 1012, in run
    self._target(*self._args, **self._kwargs)
  File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/utils/data/_utils/pin_memory.py", line 61, in _pin_memory_loop
    do_one_step()
  File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/utils/data/_utils/pin_memory.py", line 37, in do_one_step
    r = in_queue.get(timeout=MP_STATUS_CHECK_INTERVAL)
        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/multiprocessing/queues.py", line 122, in get
    return _ForkingPickler.loads(res)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/multiprocessing/reductions.py", line 541, in rebuild_storage_fd
    fd = df.detach()
         ^^^^^^^^^^^
  File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/multiprocessing/resource_sharer.py", line 57, in detach
    with _resource_sharer.get_connection(self._id) as conn:
         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/multiprocessing/resource_sharer.py", line 86, in get_connection
    c = Client(address, authkey=process.current_process().authkey)
        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/multiprocessing/connection.py", line 519, in Client
    c = SocketClient(address)
        ^^^^^^^^^^^^^^^^^^^^^
  File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/multiprocessing/connection.py", line 647, in SocketClient
    s.connect(address)
ConnectionRefusedError: [Errno 111] Connection refused
Exception in thread Thread-2 (_pin_memory_loop):
Traceback (most recent call last):
  File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/threading.py", line 1075, in _bootstrap_inner
    self.run()
  File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/threading.py", line 1012, in run
    self._target(*self._args, **self._kwargs)
  File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/utils/data/_utils/pin_memory.py", line 61, in _pin_memory_loop
    do_one_step()
  File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/utils/data/_utils/pin_memory.py", line 37, in do_one_step
    r = in_queue.get(timeout=MP_STATUS_CHECK_INTERVAL)
        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/multiprocessing/queues.py", line 122, in get
    return _ForkingPickler.loads(res)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/multiprocessing/reductions.py", line 541, in rebuild_storage_fd
    fd = df.detach()
         ^^^^^^^^^^^
  File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/multiprocessing/resource_sharer.py", line 57, in detach
    with _resource_sharer.get_connection(self._id) as conn:
         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/multiprocessing/resource_sharer.py", line 86, in get_connection
    c = Client(address, authkey=process.current_process().authkey)
        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/multiprocessing/connection.py", line 519, in Client
    c = SocketClient(address)
        ^^^^^^^^^^^^^^^^^^^^^
  File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/multiprocessing/connection.py", line 647, in SocketClient
    s.connect(address)
ConnectionRefusedError: [Errno 111] Connection refused

  0%|          | 0/50 [00:32<?, ?it/s]
[rank0]:[W918 03:41:46.838099179 ProcessGroupNCCL.cpp:1476] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator())
W0918 03:41:49.020000 2236402 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2236406 closing signal SIGTERM
W0918 03:41:49.021000 2236402 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2236407 closing signal SIGTERM
W0918 03:41:49.023000 2236402 site-packages/torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2236408 closing signal SIGTERM
E0918 03:41:55.404000 2236402 site-packages/torch/distributed/elastic/multiprocessing/api.py:874] failed (exitcode: 1) local_rank: 0 (pid: 2236405) of binary: /home/dungpq6/anaconda3/envs/spatialstack-qwen35/bin/python3.12
Traceback (most recent call last):
  File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/bin/torchrun", line 6, in <module>
    sys.exit(main())
             ^^^^^^
  File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/distributed/elastic/multiprocessing/errors/__init__.py", line 355, in wrapper
    return f(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^
  File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/distributed/run.py", line 892, in main
    run(args)
  File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/distributed/run.py", line 883, in run
    elastic_launch(
  File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 139, in __call__
    return launch_agent(self._config, self._entrypoint, list(args))
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/dungpq6/anaconda3/envs/spatialstack-qwen35/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 270, in launch_agent
    raise ChildFailedError(
torch.distributed.elastic.multiprocessing.errors.ChildFailedError: 
============================================================
src/qwen_vl/train/train_qwen.py FAILED
------------------------------------------------------------
Failures:
  <NO_OTHER_FAILURES>
------------------------------------------------------------
Root Cause (first observed failure):
[0]:
  time      : 2026-09-18_03:41:49
  host      : worker-1.fpt-hpc-worker-svc.fpt-hpc.svc.cluster.local
  rank      : 0 (local_rank: 0)
  exitcode  : 1 (pid: 2236405)
  error_file: <N/A>
  traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html
============================================================