Upload folder using huggingface_hub

README.md

@@ -1,4 +1,6 @@
 ---
 license: mit
-base_model: OpenGVLab/InternVL3-2B-Instruct
----
+base_model: OpenGVLab/InternVL2-2B
+---
+
+# InternVL3 2B Clarification SFT

adapter_config.json

@@ -0,0 +1,34 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "/home/.cache/modelscope/hub/models/OpenGVLab/InternVL3-2B-Instruct",
+  "bias": "none",
+  "corda_config": null,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 32,
+  "lora_bias": false,
+  "lora_dropout": 0.05,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": [],
+  "peft_type": "LORA",
+  "qalora_group_size": 16,
+  "r": 8,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": "^(language_model.*\\.(q_proj|k_proj|gate_proj|down_proj|o_proj|up_proj|v_proj))$",
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}

adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d9d7cf783a1b491f55e41b99ae8cc504a9f904f85b12ec491322b7924ebcd308
+size 18522336

additional_config.json

@@ -0,0 +1 @@
+{"lora_dtype": null, "lorap_lr_ratio": null, "lorap_emb_lr": 1e-06}

args.json

@@ -0,0 +1,381 @@
+{
+  "output_dir": "/home/clarify-ambiguity/output/internvl3_2b_clarification_sft/v4-20251224-184807",
+  "overwrite_output_dir": false,
+  "do_train": false,
+  "do_eval": false,
+  "do_predict": false,
+  "eval_strategy": "no",
+  "prediction_loss_only": false,
+  "per_device_train_batch_size": 1,
+  "per_device_eval_batch_size": 1,
+  "per_gpu_train_batch_size": null,
+  "per_gpu_eval_batch_size": null,
+  "gradient_accumulation_steps": 8,
+  "eval_accumulation_steps": null,
+  "eval_delay": 0,
+  "torch_empty_cache_steps": null,
+  "learning_rate": 0.0001,
+  "weight_decay": 0.1,
+  "adam_beta1": 0.9,
+  "adam_beta2": 0.95,
+  "adam_epsilon": 1e-08,
+  "max_grad_norm": 1.0,
+  "num_train_epochs": 3.0,
+  "max_steps": -1,
+  "lr_scheduler_type": "cosine",
+  "lr_scheduler_kwargs": null,
+  "warmup_ratio": 0.05,
+  "warmup_steps": 0,
+  "log_level": "passive",
+  "log_level_replica": "warning",
+  "log_on_each_node": true,
+  "logging_dir": "/home/clarify-ambiguity/output/internvl3_2b_clarification_sft/v4-20251224-184807/runs",
+  "logging_strategy": "steps",
+  "logging_first_step": true,
+  "logging_steps": 5,
+  "logging_nan_inf_filter": true,
+  "save_strategy": "steps",
+  "save_steps": 100.0,
+  "save_total_limit": 2,
+  "save_safetensors": true,
+  "save_on_each_node": false,
+  "save_only_model": false,
+  "restore_callback_states_from_checkpoint": false,
+  "no_cuda": false,
+  "use_cpu": false,
+  "use_mps_device": false,
+  "seed": 42,
+  "data_seed": 42,
+  "jit_mode_eval": false,
+  "use_ipex": false,
+  "bf16": true,
+  "fp16": false,
+  "fp16_opt_level": "O1",
+  "half_precision_backend": "auto",
+  "bf16_full_eval": false,
+  "fp16_full_eval": false,
+  "tf32": null,
+  "local_rank": 0,
+  "ddp_backend": null,
+  "tpu_num_cores": null,
+  "tpu_metrics_debug": false,
+  "debug": null,
+  "dataloader_drop_last": false,
+  "eval_steps": 100.0,
+  "dataloader_num_workers": 4,
+  "dataloader_prefetch_factor": null,
+  "past_index": -1,
+  "run_name": "/home/clarify-ambiguity/output/internvl3_2b_clarification_sft/v4-20251224-184807",
+  "disable_tqdm": null,
+  "remove_unused_columns": true,
+  "label_names": null,
+  "load_best_model_at_end": false,
+  "metric_for_best_model": "loss",
+  "greater_is_better": false,
+  "ignore_data_skip": false,
+  "fsdp": "",
+  "fsdp_min_num_params": 0,
+  "fsdp_config": null,
+  "tp_size": 0,
+  "fsdp_transformer_layer_cls_to_wrap": null,
+  "accelerator_config": {
+    "dispatch_batches": false
+  },
+  "deepspeed": {
+    "fp16": {
+      "enabled": "auto",
+      "loss_scale": 0,
+      "loss_scale_window": 1000,
+      "initial_scale_power": 16,
+      "hysteresis": 2,
+      "min_loss_scale": 1
+    },
+    "bf16": {
+      "enabled": "auto"
+    },
+    "zero_optimization": {
+      "stage": 2,
+      "offload_optimizer": {
+        "device": "none",
+        "pin_memory": true
+      },
+      "allgather_partitions": true,
+      "allgather_bucket_size": 200000000.0,
+      "overlap_comm": false,
+      "reduce_scatter": true,
+      "reduce_bucket_size": 200000000.0,
+      "contiguous_gradients": true
+    },
+    "gradient_accumulation_steps": "auto",
+    "gradient_clipping": "auto",
+    "steps_per_print": 2000,
+    "train_batch_size": "auto",
+    "train_micro_batch_size_per_gpu": "auto",
+    "wall_clock_breakdown": false
+  },
+  "label_smoothing_factor": 0.0,
+  "optim": "adamw_torch",
+  "optim_args": null,
+  "adafactor": false,
+  "group_by_length": false,
+  "length_column_name": "length",
+  "report_to": [
+    "tensorboard"
+  ],
+  "ddp_find_unused_parameters": null,
+  "ddp_bucket_cap_mb": null,
+  "ddp_broadcast_buffers": null,
+  "dataloader_pin_memory": true,
+  "dataloader_persistent_workers": false,
+  "skip_memory_metrics": true,
+  "use_legacy_prediction_loop": false,
+  "push_to_hub": false,
+  "resume_from_checkpoint": null,
+  "hub_model_id": null,
+  "hub_strategy": "every_save",
+  "hub_token": null,
+  "hub_private_repo": null,
+  "hub_always_push": false,
+  "gradient_checkpointing": true,
+  "gradient_checkpointing_kwargs": null,
+  "include_inputs_for_metrics": false,
+  "include_for_metrics": [],
+  "eval_do_concat_batches": true,
+  "fp16_backend": "auto",
+  "push_to_hub_model_id": null,
+  "push_to_hub_organization": null,
+  "push_to_hub_token": null,
+  "mp_parameters": "",
+  "auto_find_batch_size": false,
+  "full_determinism": false,
+  "torchdynamo": null,
+  "ray_scope": "last",
+  "ddp_timeout": 18000000,
+  "torch_compile": false,
+  "torch_compile_backend": null,
+  "torch_compile_mode": null,
+  "include_tokens_per_second": false,
+  "include_num_input_tokens_seen": false,
+  "neftune_noise_alpha": null,
+  "optim_target_modules": null,
+  "batch_eval_metrics": false,
+  "eval_on_start": false,
+  "use_liger_kernel": false,
+  "eval_use_gather_object": false,
+  "average_tokens_across_devices": false,
+  "sortish_sampler": false,
+  "predict_with_generate": false,
+  "generation_max_length": null,
+  "generation_num_beams": null,
+  "generation_config": null,
+  "tuner_backend": "peft",
+  "vit_gradient_checkpointing": null,
+  "router_aux_loss_coef": 0.0,
+  "enable_dft_loss": false,
+  "enable_channel_loss": false,
+  "check_model": true,
+  "acc_strategy": "token",
+  "train_dataloader_shuffle": true,
+  "max_epochs": null,
+  "aligner_lr": null,
+  "vit_lr": null,
+  "use_logits_to_keep": null,
+  "ds3_gather_for_generation": true,
+  "resume_only_model": false,
+  "optimizer": null,
+  "loss_type": null,
+  "metric": null,
+  "eval_use_evalscope": false,
+  "eval_dataset": [],
+  "eval_dataset_args": null,
+  "eval_limit": null,
+  "eval_generation_config": null,
+  "extra_eval_args": null,
+  "use_flash_ckpt": false,
+  "use_ray": false,
+  "ray_exp_name": null,
+  "device_groups": null,
+  "model": "OpenGVLab/InternVL3-2B-Instruct",
+  "model_type": "internvl3",
+  "model_revision": null,
+  "task_type": "causal_lm",
+  "torch_dtype": "bfloat16",
+  "attn_impl": null,
+  "new_special_tokens": [],
+  "num_labels": null,
+  "problem_type": null,
+  "rope_scaling": null,
+  "device_map": null,
+  "max_memory": {},
+  "max_model_len": null,
+  "local_repo_path": null,
+  "init_strategy": null,
+  "template": "internvl2_5",
+  "system": null,
+  "max_length": 4096,
+  "truncation_strategy": "delete",
+  "max_pixels": null,
+  "agent_template": null,
+  "norm_bbox": null,
+  "use_chat_template": true,
+  "padding_free": false,
+  "padding_side": "right",
+  "loss_scale": "default",
+  "sequence_parallel_size": 1,
+  "response_prefix": null,
+  "template_backend": "swift",
+  "dataset": [
+    "swift_dataset.jsonl"
+  ],
+  "val_dataset": [],
+  "cached_dataset": [],
+  "cached_val_dataset": [],
+  "split_dataset_ratio": 0.0,
+  "dataset_num_proc": 1,
+  "load_from_cache_file": false,
+  "dataset_shuffle": true,
+  "val_dataset_shuffle": false,
+  "streaming": false,
+  "interleave_prob": null,
+  "stopping_strategy": "first_exhausted",
+  "shuffle_buffer_size": 1000,
+  "download_mode": "reuse_dataset_if_exists",
+  "columns": {},
+  "strict": false,
+  "model_name": [
+    "swift-robot"
+  ],
+  "model_author": [
+    "swift"
+  ],
+  "custom_dataset_info": [],
+  "quant_method": null,
+  "quant_bits": null,
+  "hqq_axis": null,
+  "bnb_4bit_compute_dtype": "bfloat16",
+  "bnb_4bit_quant_type": "nf4",
+  "bnb_4bit_use_double_quant": true,
+  "bnb_4bit_quant_storage": null,
+  "max_new_tokens": 64,
+  "temperature": 0.0,
+  "top_k": null,
+  "top_p": null,
+  "repetition_penalty": null,
+  "num_beams": 1,
+  "stream": false,
+  "stop_words": [],
+  "logprobs": false,
+  "top_logprobs": null,
+  "ckpt_dir": null,
+  "lora_modules": [],
+  "train_type": "lora",
+  "adapters": [],
+  "external_plugins": [],
+  "model_kwargs": {},
+  "load_args": false,
+  "load_data_args": false,
+  "packing": false,
+  "packing_length": null,
+  "packing_num_proc": 1,
+  "lazy_tokenize": true,
+  "custom_register_path": [],
+  "use_hf": false,
+  "ignore_args_error": false,
+  "use_swift_lora": false,
+  "freeze_parameters": [],
+  "freeze_parameters_regex": null,
+  "freeze_parameters_ratio": 0.0,
+  "trainable_parameters": [],
+  "trainable_parameters_regex": null,
+  "freeze_llm": false,
+  "freeze_vit": true,
+  "freeze_aligner": true,
+  "target_modules": [
+    "all-linear"
+  ],
+  "target_regex": null,
+  "target_parameters": null,
+  "modules_to_save": [],
+  "lora_rank": 8,
+  "lora_alpha": 32,
+  "lora_dropout": 0.05,
+  "lora_bias": "none",
+  "lora_dtype": null,
+  "lorap_lr_ratio": null,
+  "use_rslora": false,
+  "use_dora": false,
+  "lora_ga_batch_size": 2,
+  "lora_ga_iters": 2,
+  "lora_ga_max_length": 1024,
+  "lora_ga_direction": "ArB2r",
+  "lora_ga_scale": "stable",
+  "lora_ga_stable_gamma": 16,
+  "init_weights": true,
+  "fourier_n_frequency": 2000,
+  "fourier_scaling": 300.0,
+  "boft_block_size": 4,
+  "boft_block_num": 0,
+  "boft_n_butterfly_factor": 1,
+  "boft_dropout": 0.0,
+  "vera_rank": 256,
+  "vera_projection_prng_key": 0,
+  "vera_dropout": 0.0,
+  "vera_d_initial": 0.1,
+  "adapter_act": "gelu",
+  "adapter_length": 128,
+  "use_galore": false,
+  "galore_target_modules": null,
+  "galore_rank": 128,
+  "galore_update_proj_gap": 50,
+  "galore_scale": 1.0,
+  "galore_proj_type": "std",
+  "galore_optim_per_parameter": false,
+  "galore_with_embedding": false,
+  "galore_quantization": false,
+  "galore_proj_quant": false,
+  "galore_proj_bits": 4,
+  "galore_proj_group_size": 256,
+  "galore_cos_threshold": 0.4,
+  "galore_gamma_proj": 2,
+  "galore_queue_size": 5,
+  "adalora_target_r": 8,
+  "adalora_init_r": 12,
+  "adalora_tinit": 0,
+  "adalora_tfinal": 0,
+  "adalora_deltaT": 1,
+  "adalora_beta1": 0.85,
+  "adalora_beta2": 0.85,
+  "adalora_orth_reg_weight": 0.5,
+  "llamapro_num_new_blocks": 4,
+  "llamapro_num_groups": null,
+  "lisa_activated_layers": 0,
+  "lisa_step_interval": 20,
+  "reft_layer_key": null,
+  "reft_layers": null,
+  "reft_rank": 4,
+  "reft_intervention_type": "LoreftIntervention",
+  "reft_args": null,
+  "swanlab_token": null,
+  "swanlab_project": null,
+  "swanlab_workspace": null,
+  "swanlab_exp_name": null,
+  "swanlab_lark_webhook_url": null,
+  "swanlab_lark_secret": null,
+  "swanlab_mode": "cloud",
+  "add_version": true,
+  "create_checkpoint_symlink": false,
+  "zero_hpz_partition_size": null,
+  "deepspeed_autotp_size": null,
+  "early_stop_interval": null,
+  "rank": 0,
+  "global_world_size": 2,
+  "local_world_size": 2,
+  "model_suffix": "InternVL3-2B-Instruct",
+  "model_info": "ModelInfo(model_type='internvl3', model_dir='/home/.cache/modelscope/hub/models/OpenGVLab/InternVL3-2B-Instruct', torch_dtype=torch.bfloat16, max_model_len=32768, quant_method=None, quant_bits=None, rope_scaling={'factor': 2.0, 'rope_type': 'dynamic', 'type': 'dynamic'}, is_moe_model=False, is_multimodal=True, config=None, task_type='causal_lm', num_labels=None)",
+  "model_meta": "ModelMeta(model_type='internvl3', model_groups=[ModelGroup(models=[Model(ms_model_id='OpenGVLab/InternVL3-1B-Pretrained', hf_model_id='OpenGVLab/InternVL3-1B-Pretrained', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='OpenGVLab/InternVL3-2B-Pretrained', hf_model_id='OpenGVLab/InternVL3-2B-Pretrained', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='OpenGVLab/InternVL3-8B-Pretrained', hf_model_id='OpenGVLab/InternVL3-8B-Pretrained', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='OpenGVLab/InternVL3-9B-Pretrained', hf_model_id='OpenGVLab/InternVL3-9B-Pretrained', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='OpenGVLab/InternVL3-14B-Pretrained', hf_model_id='OpenGVLab/InternVL3-14B-Pretrained', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='OpenGVLab/InternVL3-38B-Pretrained', hf_model_id='OpenGVLab/InternVL3-38B-Pretrained', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='OpenGVLab/InternVL3-78B-Pretrained', hf_model_id='OpenGVLab/InternVL3-78B-Pretrained', model_path=None, ms_revision=None, hf_revision=None)], ignore_patterns=None, requires=None, tags=[]), ModelGroup(models=[Model(ms_model_id='OpenGVLab/InternVL3-1B-Instruct', hf_model_id='OpenGVLab/InternVL3-1B-Instruct', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='OpenGVLab/InternVL3-2B-Instruct', hf_model_id='OpenGVLab/InternVL3-2B-Instruct', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='OpenGVLab/InternVL3-8B-Instruct', hf_model_id='OpenGVLab/InternVL3-8B-Instruct', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='OpenGVLab/InternVL3-9B-Instruct', hf_model_id='OpenGVLab/InternVL3-9B-Instruct', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='OpenGVLab/InternVL3-14B-Instruct', hf_model_id='OpenGVLab/InternVL3-14B-Instruct', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='OpenGVLab/InternVL3-38B-Instruct', hf_model_id='OpenGVLab/InternVL3-38B-Instruct', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='OpenGVLab/InternVL3-78B-Instruct', hf_model_id='OpenGVLab/InternVL3-78B-Instruct', model_path=None, ms_revision=None, hf_revision=None)], ignore_patterns=None, requires=None, tags=[]), ModelGroup(models=[Model(ms_model_id='OpenGVLab/InternVL3-1B', hf_model_id='OpenGVLab/InternVL3-1B', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='OpenGVLab/InternVL3-2B', hf_model_id='OpenGVLab/InternVL3-2B', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='OpenGVLab/InternVL3-8B', hf_model_id='OpenGVLab/InternVL3-8B', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='OpenGVLab/InternVL3-9B', hf_model_id='OpenGVLab/InternVL3-9B', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='OpenGVLab/InternVL3-14B', hf_model_id='OpenGVLab/InternVL3-14B', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='OpenGVLab/InternVL3-38B', hf_model_id='OpenGVLab/InternVL3-38B', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='OpenGVLab/InternVL3-78B', hf_model_id='OpenGVLab/InternVL3-78B', model_path=None, ms_revision=None, hf_revision=None)], ignore_patterns=None, requires=None, tags=[]), ModelGroup(models=[Model(ms_model_id='OpenGVLab/InternVL3-1B-AWQ', hf_model_id='OpenGVLab/InternVL3-1B-AWQ', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='OpenGVLab/InternVL3-2B-AWQ', hf_model_id='OpenGVLab/InternVL3-2B-AWQ', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='OpenGVLab/InternVL3-8B-AWQ', hf_model_id='OpenGVLab/InternVL3-8B-AWQ', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='OpenGVLab/InternVL3-9B-AWQ', hf_model_id='OpenGVLab/InternVL3-9B-AWQ', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='OpenGVLab/InternVL3-14B-AWQ', hf_model_id='OpenGVLab/InternVL3-14B-AWQ', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='OpenGVLab/InternVL3-38B-AWQ', hf_model_id='OpenGVLab/InternVL3-38B-AWQ', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='OpenGVLab/InternVL3-78B-AWQ', hf_model_id='OpenGVLab/InternVL3-78B-AWQ', model_path=None, ms_revision=None, hf_revision=None)], ignore_patterns=None, requires=None, tags=[]), ModelGroup(models=[Model(ms_model_id='SenseNova/SenseNova-SI-InternVL3-2B', hf_model_id='sensenova/SenseNova-SI-InternVL3-2B', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='SenseNova/SenseNova-SI-InternVL3-8B', hf_model_id='sensenova/SenseNova-SI-InternVL3-8B', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='SenseNova/SenseNova-SI-1.1-InternVL3-2B', hf_model_id='sensenova/SenseNova-SI-1.1-InternVL3-2B', model_path=None, ms_revision=None, hf_revision=None), Model(ms_model_id='SenseNova/SenseNova-SI-1.1-InternVL3-8B', hf_model_id='sensenova/SenseNova-SI-1.1-InternVL3-8B', model_path=None, ms_revision=None, hf_revision=None)], ignore_patterns=None, requires=None, tags=[])], template='internvl2_5', get_function=<function get_model_tokenizer_internvl at 0x7fe8373518a0>, model_arch=MultiModelKeys(arch_name='internvl', embedding=None, module_list=None, lm_head=None, q_proj=None, k_proj=None, v_proj=None, o_proj=None, attention=None, mlp=None, down_proj=None, qkv_proj=None, qk_proj=None, qa_proj=None, qb_proj=None, kv_proj=None, kva_proj=None, kvb_proj=None, language_model=['language_model'], aligner=['mlp1'], vision_tower=['vision_model'], generator=[]), architectures=['InternVLChatModel'], additional_saved_files=[], torch_dtype=None, is_multimodal=True, is_reward=False, is_reranker=False, task_type=None, ignore_patterns=None, requires=['transformers>=4.37.2', 'timm'], tags=['vision', 'video'])",
+  "model_dir": "/home/.cache/modelscope/hub/models/OpenGVLab/InternVL3-2B-Instruct",
+  "_val_dataset_exists": [],
+  "hub": "<class 'swift.hub.hub.MSHub'>",
+  "evaluation_strategy": "steps",
+  "training_args": "Seq2SeqTrainingArguments(output_dir='/home/clarify-ambiguity/output/internvl3_2b_clarification_sft/v4-20251224-184807', overwrite_output_dir=False, do_train=False, do_eval=False, do_predict=False, eval_strategy=<IntervalStrategy.NO: 'no'>, prediction_loss_only=False, per_device_train_batch_size=1, per_device_eval_batch_size=1, per_gpu_train_batch_size=None, per_gpu_eval_batch_size=None, gradient_accumulation_steps=8, eval_accumulation_steps=None, eval_delay=0, torch_empty_cache_steps=None, learning_rate=0.0001, weight_decay=0.1, adam_beta1=0.9, adam_beta2=0.95, adam_epsilon=1e-08, max_grad_norm=1.0, num_train_epochs=3.0, max_steps=-1, lr_scheduler_type=<SchedulerType.COSINE: 'cosine'>, lr_scheduler_kwargs=None, warmup_ratio=0.05, warmup_steps=0, log_level='passive', log_level_replica='warning', log_on_each_node=True, logging_dir='/home/clarify-ambiguity/output/internvl3_2b_clarification_sft/v4-20251224-184807/runs', logging_strategy=<IntervalStrategy.STEPS: 'steps'>, logging_first_step=True, logging_steps=5, logging_nan_inf_filter=True, save_strategy=<SaveStrategy.STEPS: 'steps'>, save_steps=100, save_total_limit=2, save_safetensors=True, save_on_each_node=False, save_only_model=False, restore_callback_states_from_checkpoint=False, no_cuda=False, use_cpu=False, use_mps_device=False, seed=42, data_seed=42, jit_mode_eval=False, use_ipex=False, bf16=True, fp16=False, fp16_opt_level='O1', half_precision_backend='auto', bf16_full_eval=False, fp16_full_eval=False, tf32=None, local_rank=0, ddp_backend=None, tpu_num_cores=None, tpu_metrics_debug=False, debug=[], dataloader_drop_last=False, eval_steps=100.0, dataloader_num_workers=4, dataloader_prefetch_factor=10, past_index=-1, run_name='/home/clarify-ambiguity/output/internvl3_2b_clarification_sft/v4-20251224-184807', disable_tqdm=False, remove_unused_columns=False, label_names=None, load_best_model_at_end=False, metric_for_best_model='loss', greater_is_better=False, ignore_data_skip=False, fsdp=[], fsdp_min_num_params=0, fsdp_config={'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}, tp_size=0, fsdp_transformer_layer_cls_to_wrap=None, accelerator_config=AcceleratorConfig(split_batches=False, dispatch_batches=False, even_batches=True, use_seedable_sampler=True, non_blocking=False, gradient_accumulation_kwargs=None, use_configured_state=False), deepspeed={'fp16': {'enabled': 'auto', 'loss_scale': 0, 'loss_scale_window': 1000, 'initial_scale_power': 16, 'hysteresis': 2, 'min_loss_scale': 1}, 'bf16': {'enabled': 'auto'}, 'zero_optimization': {'stage': 2, 'offload_optimizer': {'device': 'none', 'pin_memory': True}, 'allgather_partitions': True, 'allgather_bucket_size': 200000000.0, 'overlap_comm': False, 'reduce_scatter': True, 'reduce_bucket_size': 200000000.0, 'contiguous_gradients': True}, 'gradient_accumulation_steps': 'auto', 'gradient_clipping': 'auto', 'steps_per_print': 2000, 'train_batch_size': 'auto', 'train_micro_batch_size_per_gpu': 'auto', 'wall_clock_breakdown': False}, label_smoothing_factor=0.0, optim=<OptimizerNames.ADAMW_TORCH: 'adamw_torch'>, optim_args=None, adafactor=False, group_by_length=False, length_column_name='length', report_to=['tensorboard'], ddp_find_unused_parameters=None, ddp_bucket_cap_mb=None, ddp_broadcast_buffers=None, dataloader_pin_memory=True, dataloader_persistent_workers=False, skip_memory_metrics=True, use_legacy_prediction_loop=False, push_to_hub=False, resume_from_checkpoint=None, hub_model_id=None, hub_strategy=<HubStrategy.EVERY_SAVE: 'every_save'>, hub_token=None, hub_private_repo=None, hub_always_push=False, gradient_checkpointing=True, gradient_checkpointing_kwargs=None, include_inputs_for_metrics=False, include_for_metrics=[], eval_do_concat_batches=True, fp16_backend='auto', push_to_hub_model_id=None, push_to_hub_organization=None, push_to_hub_token=None, mp_parameters='', auto_find_batch_size=False, full_determinism=False, torchdynamo=None, ray_scope='last', ddp_timeout=18000000, torch_compile=False, torch_compile_backend=None, torch_compile_mode=None, include_tokens_per_second=None, include_num_input_tokens_seen=None, neftune_noise_alpha=None, optim_target_modules=None, batch_eval_metrics=False, eval_on_start=False, use_liger_kernel=False, eval_use_gather_object=False, average_tokens_across_devices=None, sortish_sampler=False, predict_with_generate=False, generation_max_length=None, generation_num_beams=None, generation_config=None, tuner_backend='peft', vit_gradient_checkpointing=True, router_aux_loss_coef=0.0, enable_dft_loss=False, enable_channel_loss=False, check_model=True, acc_strategy='token', train_dataloader_shuffle=True, max_epochs=None, aligner_lr=None, vit_lr=None, use_logits_to_keep=None, ds3_gather_for_generation=True, resume_only_model=False, optimizer=None, loss_type=None, metric=None, eval_use_evalscope=False, eval_dataset=[], eval_dataset_args=None, eval_limit=None, eval_generation_config=None, extra_eval_args=None, use_flash_ckpt=False, sft_alpha=0, chord_sft_dataset=[], chord_sft_per_device_train_batch_size=None, chord_enable_phi_function=False, chord_mu_warmup_steps=None, chord_mu_decay_steps=None, chord_mu_peak=None, chord_mu_valley=None, train_type='lora', local_repo_path=None, galore_config=None, padding_side='right', padding_free=False, task_type='causal_lm', problem_type=None)"
+}

global_step33/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:95cca669b21bccf7d8f4273fb4cd3bffbab33dfcb272a6e8abfe7fe4eab12431
+size 55422128

global_step33/bf16_zero_pp_rank_1_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e0b9bd528834877d0c87c9a3a94f30cc60629d623cc737fc491e123791ed937f
+size 55422256

global_step33/mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ae02381a0103524d061ff52dff030b3ad9be31857065ee2c0d1aab45b3865c87
+size 18842732

infer_result/20251225-072421_clari.jsonl

@@ -0,0 +1,42 @@
+{"response": "Is the rope you are holding on to moving forward or backward?", "labels": "Are you moving left or right?", "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/8f85517967.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>Am I moving forward or backward?", "loss": null}, {"role": "assistant", "content": "Is the rope you are holding on to moving forward or backward?"}]}
+{"response": "Is the train moving towards you or away from you?", "labels": "Is it approaching or moving away?", "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/c16a5320fa.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>Is the subway going in the same direction as I face?", "loss": null}, {"role": "assistant", "content": "Is the train moving towards you or away from you?"}]}
+{"response": "Is the tram moving towards the camera or away from the camera?", "labels": "Did the person cross the track before the tram reached them, or after?", "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/19ca14e7ea.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>Is the tram moving towards the photographer or away from the photographer?", "loss": null}, {"role": "assistant", "content": "Is the tram moving towards the camera or away from the camera?"}]}
+{"response": "Is this a group of zebras migrating northward, southward, or moving in a different direction?", "labels": "Is the camera facing east, west, south or north?", "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/a5bfc9e079.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>What is the direction of the zebra migration? East, west, south or north?", "loss": null}, {"role": "assistant", "content": "Is this a group of zebras migrating northward, southward, or moving in a different direction?"}]}
+{"response": "Is the road going to the left or to the right?", "labels": "Are the zebras moving east, west, south or north?", "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/a5771bce93.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>Which direction is the photographer facing? East, west, south or north?", "loss": null}, {"role": "assistant", "content": "Is the road going to the left or to the right?"}]}
+{"response": "Is the sun rising or setting in the image?", "labels": "Is the sun rising or setting?", "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/0640966322.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>Which direction are the horses heading, east or west?", "loss": null}, {"role": "assistant", "content": "Is the sun rising or setting in the image?"}]}
+{"response": "Is the zipper moving to the left or to the right?", "labels": "Are you pulling the zipper to the left or to the right?", "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/e4da3b7fbb.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>Am I opening or closing my bag?", "loss": null}, {"role": "assistant", "content": "Is the zipper moving to the left or to the right?"}]}
+{"response": "Is the red capsule of the ferris wheel going up or down?", "labels": "From the photographer’s perspective, is the Ferris wheel rotating clockwise or counterclockwise?", "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/6ea9ab1baa.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>Is the red capsule of the ferris wheel going up or down?", "loss": null}, {"role": "assistant", "content": "Is the red capsule of the ferris wheel going up or down?"}]}
+{"response": "Is the box being delivered to the customer or picked up from the customer?", "labels": "Is the courier picking it up or putting it down?", "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/65ded5353c.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>Is the courier moving the box into the car or out of the car?", "loss": null}, {"role": "assistant", "content": "Is the box being delivered to the customer or picked up from the customer?"}]}
+{"response": "Is the drawer moving left or right from the perspective of the photographer?", "labels": "Is she pulling the drawer or pushing it?", "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/9bf31c7ff0.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>Is the drawer moving left or right from the perspective of the photographer?", "loss": null}, {"role": "assistant", "content": "Is the drawer moving left or right from the perspective of the photographer?"}]}
+{"response": "Is the oven door open or closed?", "labels": "From your perspective, are your hands moving left or right?", "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/013d407166.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>Am I putting the plate into the oven or taking it out of the oven?", "loss": null}, {"role": "assistant", "content": "Is the oven door open or closed?"}]}
+{"response": "Is the player serving or returning?", "labels": "From whose perspective should the direction be determined?", "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/8216124282.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>Is this ball heading left or right?", "loss": null}, {"role": "assistant", "content": "Is the player serving or returning?"}]}
+{"response": "Is the box being delivered or picked up?", "labels": "Is the van being loaded or unloaded?", "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/6f4922f455.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>Is the man putting the box into the car or taking the box out of the car?", "loss": null}, {"role": "assistant", "content": "Is the box being delivered or picked up?"}]}
+{"response": "Is the bucket on the pallet or on the ground?", "labels": "Are they stocking items or retrieving items?", "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/9fc3d7152b.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>Is the man on the left in red hat picking up the bucket or putting down the bucket?", "loss": null}, {"role": "assistant", "content": "Is the bucket on the pallet or on the ground?"}]}
+{"response": "Is this a train station or a train depot?", "labels": "Is this photo from the morning or late at night?", "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/a3f390d88e.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>The clock is on 12-hour format. Please tell me what time it is on a 24-hour schedule.", "loss": null}, {"role": "assistant", "content": "Is this a train station or a train depot?"}]}
+{"response": "Is the door open or closed?", "labels": "To determine your direction, I need to know the purpose of your action. Are you opening or closing the door?", "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/1679091c5a.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>Am I going from left to right or from right to left?", "loss": null}, {"role": "assistant", "content": "Is the door open or closed?"}]}
+{"response": "Is the person in the image wearing the headset or taking it off?", "labels": "I can't tell the direction of movement from the photo. To figure it out, can you tell me what you are about to do?", "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/45c48cce2e.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>Am I putting on or taking off my headset?", "loss": null}, {"role": "assistant", "content": "Is the person in the image wearing the headset or taking it off?"}]}
+{"response": "Which country are you in?", "labels": "Is the sign already showing the side you want next, or does it need to be flipped?", "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/02e74f10e0.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>I am the owner of a shop. Should I flip the open sign or keep the position of the sign unchanged?", "loss": null}, {"role": "assistant", "content": "Which country are you in?"}]}
+{"response": "Is the box being loaded into the car or unloaded from the car?", "labels": "Are they delivering the packages to a recipient or collecting them from a sender?", "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/1afa34a7f9.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>Is the man in red hat picking up the box out of the car or putting down the box into the car?", "loss": null}, {"role": "assistant", "content": "Is the box being loaded into the car or unloaded from the car?"}]}
+{"response": "Is this a red light in the United States or in another country?", "labels": "The rules for turning right on a red light vary significantly depending on your location. In which city are you driving?", "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/ec8956637a.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>Can I make a right turn on the red traffic light?", "loss": null}, {"role": "assistant", "content": "Is this a red light in the United States or in another country?"}]}
+{"response": "Is this a country road or a highway?", "labels": "That depends on the country's traffic laws. In which country was this photo taken?", "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/f899139df5.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>I am driving on the left lane. Is it allowed?", "loss": null}, {"role": "assistant", "content": "Is this a country road or a highway?"}]}
+{"response": "Is your brother a minor?", "labels": "The legality depends on your brother's age and the specific location (country and state). How old is he and where was this photo taken?", "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/26657d5ff9.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>My brother is 20 years old. Is my brother's behavior legal?", "loss": null}, {"role": "assistant", "content": "Is your brother a minor?"}]}
+{"response": "Is the woman drinking alone or with others?", "labels": "What are the specific circumstances that lead you to believe her behavior might be illegal?", "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/ed3d2c2199.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>The women drinking alcohol is 20 years old. Is her behavior legal?", "loss": null}, {"role": "assistant", "content": "Is the woman drinking alone or with others?"}]}
+{"response": "Is he riding his bike on a road or a path?", "labels": "Where was this photo taken (country—and if relevant, state/city)? Rules differ by jurisdiction on whether a 10-year-old may ride in the street vs. a path, whether adult supervision is required.", "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/bd686fd640.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>This is the photo of my 10-year-old-son. Is his the behavior legal?", "loss": null}, {"role": "assistant", "content": "Is he riding his bike on a road or a path?"}]}
+{"response": "Is this a public place or a private one?", "labels": "In what country or region is this gesture being used?", "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/a597e50502.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>I want to buy two bottles of milk. Is my gesture appropriate?", "loss": null}, {"role": "assistant", "content": "Is this a public place or a private one?"}]}
+{"response": "Is this a one-way road?", "labels": "The image shows that you are in the right-hand lane. Whether that is correct or not depends on the country's traffic laws. In which country are you driving?", "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/38b3eff8ba.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>Am I driving in the correct lane?", "loss": null}, {"role": "assistant", "content": "Is this a one-way road?"}]}
+{"response": "Is the face of the cube that contains side B a front face or a back face?", "labels": "In your 3D perception of the image, how many cutes in this picture?", "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/202cb962ac.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>In the 3D perception of the images. Is side B part of the face of a cube?", "loss": null}, {"role": "assistant", "content": "Is the face of the cube that contains side B a front face or a back face?"}]}
+{"response": "Is this stock listed on a major stock exchange?", "labels": "That is impossible to say without knowing the market of origin, as the meaning of the colors differs by region. Is this ticker from Mainland China or from another market like Hong Kong or the US?\"", "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/7f6ffaa6bb.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>Based on the colors on the screen, is this stock performing well?", "loss": null}, {"role": "assistant", "content": "Is this stock listed on a major stock exchange?"}]}
+{"response": "Is he opening or closing his bag?", "labels": "In which direction is his hand moving the zipper?", "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/ec5decca5e.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>Is he opening or closing his bag?", "loss": null}, {"role": "assistant", "content": "Is he opening or closing his bag?"}]}
+{"response": "Is the time on the left clock 14:15 or 23:00?", "labels": "There are two clocks in the image, which clock are you referring to, the right one or the left one?", "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/aa00000001.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>Based on the image, what time is it now? One of the clock shows the correct time.", "loss": null}, {"role": "assistant", "content": "Is the time on the left clock 14:15 or 23:00?"}]}
+{"response": "Is the person in the image wearing a hat or a wig?", "labels": "This depends on which person you see. In your perception, is the person's ear visible?", "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/c8ffe9a587.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>Does the person in the image look old or young?", "loss": null}, {"role": "assistant", "content": "Is the person in the image wearing a hat or a wig?"}]}
+{"response": "Is the yellow dot on the left or the right?", "labels": "Which plane is at the front, the one with a red dot or the one with a blue dot?", "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/6f3ef77ac0.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>Which direction is the side with a yellow dot facing? Up, down, left or right?", "loss": null}, {"role": "assistant", "content": "Is the yellow dot on the left or the right?"}]}
+{"response": "Is the woman wearing a hat?", "labels": "Which woman are you referring to: the one looking away from the viewer or the one shown in profile?", "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/c8ffe9a587.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>Can we see the woman's ear in the picture?", "loss": null}, {"role": "assistant", "content": "Is the woman wearing a hat?"}]}
+{"response": "Is the baby in a bathtub or a basin?", "labels": "I can see the baby is playing with water. To know if there is a hidden danger, I need to know the temperature of the water being poured. Is it hot or cold?", "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/d09bf41544.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>The baby is play water when sitting in a basin. Is the baby in danger?", "loss": null}, {"role": "assistant", "content": "Is the baby in a bathtub or a basin?"}]}
+{"response": "Is the blue dot on the front plane or the red dot?", "labels": "From your perspective, are the stairs ascending or descending from left to right?", "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/6f3ef77ac0.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>Is blue dot on the front plane or the red dot?", "loss": null}, {"role": "assistant", "content": "Is the blue dot on the front plane or the red dot?"}]}
+{"response": "Is your mother walking with her family or with her friends?", "labels": "I cannot tell from the image alone. Can you tell me who the other person is in relation to your mother?", "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/7f39f8317f.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>My mother is walking with her families. Which one is my mother?", "loss": null}, {"role": "assistant", "content": "Is your mother walking with her family or with her friends?"}]}
+{"response": "Is the dog a pet or a working animal?", "labels": "I cannot tell from the image alone. Can you clarify the relationship between the two people in the photo?", "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/903ce9225f.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>Who is the owner of the dog?", "loss": null}, {"role": "assistant", "content": "Is the dog a pet or a working animal?"}]}
+{"response": "Is the box of water you received from your friend in the foreground or in the background?", "labels": "I cannot tell from the image alone. To identify your friend, can you describe your role in the action?", "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/1d7f7abc18.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>My friend handed me a box of water. Which one is my friend?", "loss": null}, {"role": "assistant", "content": "Is the box of water you received from your friend in the foreground or in the background?"}]}
+{"response": "I can't tell from the image alone. Can you tell me your role in the project?", "labels": "To identify your leader, can you tell me who is making a report or giving instructions?", "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/03afdbd66e.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>I and my leader are discussing a project. Who is my leader?", "loss": null}, {"role": "assistant", "content": "I can't tell from the image alone. Can you tell me your role in the project?"}]}
+{"response": "Is your daughter the one wearing the blue and white dress, or the one in the green dress?", "labels": "Are you referring to the young girl or the adult woman walking with the man?", "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/c0c7c76d30.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>My daughter is one of the two women. Who is my daughter?", "loss": null}, {"role": "assistant", "content": "Is your daughter the one wearing the blue and white dress, or the one in the green dress?"}]}
+{"response": "Is he giving a speech, or is he listening to someone else?", "labels": "I cannot tell who the principal is from the image alone. Can you describe what the principal is doing in the scene?", "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/9a1158154d.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>Our principle is one of the two men. What is the principle doing?", "loss": null}, {"role": "assistant", "content": "Is he giving a speech, or is he listening to someone else?"}]}
+{"response": "Is your sister the one who is cooking or the one holding the baby?", "labels": "I cannot tell from the image alone. Can you tell me what your sister is doing in the picture?", "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/072b030ba1.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>My sister is one of the two women. Who is my sister?", "loss": null}, {"role": "assistant", "content": "Is your sister the one who is cooking or the one holding the baby?"}]}

infer_result/20251225-194658.jsonl

@@ -0,0 +1,42 @@
+{"response": "Is the rope you are holding on to moving forward or backward?", "labels": null, "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/8f85517967.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>Am I moving forward or backward?", "loss": null}, {"role": "assistant", "content": "Is the rope you are holding on to moving forward or backward?"}]}
+{"response": "Is the train moving towards you or away from you?", "labels": null, "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/c16a5320fa.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>Is the subway going in the same direction as I face?", "loss": null}, {"role": "assistant", "content": "Is the train moving towards you or away from you?"}]}
+{"response": "Is the tram moving towards the camera or away from the camera?", "labels": null, "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/19ca14e7ea.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>Is the tram moving towards the photographer or away from the photographer?", "loss": null}, {"role": "assistant", "content": "Is the tram moving towards the camera or away from the camera?"}]}
+{"response": "Is this a group of zebras migrating northward, southward, or moving in a different direction?", "labels": null, "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/a5bfc9e079.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>What is the direction of the zebra migration? East, west, south or north?", "loss": null}, {"role": "assistant", "content": "Is this a group of zebras migrating northward, southward, or moving in a different direction?"}]}
+{"response": "Is the road going to the left or to the right?", "labels": null, "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/a5771bce93.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>Which direction is the photographer facing? East, west, south or north?", "loss": null}, {"role": "assistant", "content": "Is the road going to the left or to the right?"}]}
+{"response": "Is the sun rising or setting in the image?", "labels": null, "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/0640966322.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>Which direction are the horses heading, east or west?", "loss": null}, {"role": "assistant", "content": "Is the sun rising or setting in the image?"}]}
+{"response": "Is the zipper moving to the left or to the right?", "labels": null, "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/e4da3b7fbb.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>Am I opening or closing my bag?", "loss": null}, {"role": "assistant", "content": "Is the zipper moving to the left or to the right?"}]}
+{"response": "Is the red capsule of the ferris wheel going up or down?", "labels": null, "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/6ea9ab1baa.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>Is the red capsule of the ferris wheel going up or down?", "loss": null}, {"role": "assistant", "content": "Is the red capsule of the ferris wheel going up or down?"}]}
+{"response": "Is the box being delivered to the customer or picked up from the customer?", "labels": null, "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/65ded5353c.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>Is the courier moving the box into the car or out of the car?", "loss": null}, {"role": "assistant", "content": "Is the box being delivered to the customer or picked up from the customer?"}]}
+{"response": "Is the drawer moving left or right from the perspective of the photographer?", "labels": null, "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/9bf31c7ff0.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>Is the drawer moving left or right from the perspective of the photographer?", "loss": null}, {"role": "assistant", "content": "Is the drawer moving left or right from the perspective of the photographer?"}]}
+{"response": "Is the oven door open or closed?", "labels": null, "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/013d407166.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>Am I putting the plate into the oven or taking it out of the oven?", "loss": null}, {"role": "assistant", "content": "Is the oven door open or closed?"}]}
+{"response": "Is the player serving or returning?", "labels": null, "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/8216124282.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>Is this ball heading left or right?", "loss": null}, {"role": "assistant", "content": "Is the player serving or returning?"}]}
+{"response": "Is the box being delivered or picked up?", "labels": null, "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/6f4922f455.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>Is the man putting the box into the car or taking the box out of the car?", "loss": null}, {"role": "assistant", "content": "Is the box being delivered or picked up?"}]}
+{"response": "Is the bucket on the pallet or on the ground?", "labels": null, "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/9fc3d7152b.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>Is the man on the left in red hat picking up the bucket or putting down the bucket?", "loss": null}, {"role": "assistant", "content": "Is the bucket on the pallet or on the ground?"}]}
+{"response": "Is this a train station or a train depot?", "labels": null, "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/a3f390d88e.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>The clock is on 12-hour format. Please tell me what time it is on a 24-hour schedule.", "loss": null}, {"role": "assistant", "content": "Is this a train station or a train depot?"}]}
+{"response": "Is the door open or closed?", "labels": null, "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/1679091c5a.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>Am I going from left to right or from right to left?", "loss": null}, {"role": "assistant", "content": "Is the door open or closed?"}]}
+{"response": "Is the person in the image wearing the headset or taking it off?", "labels": null, "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/45c48cce2e.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>Am I putting on or taking off my headset?", "loss": null}, {"role": "assistant", "content": "Is the person in the image wearing the headset or taking it off?"}]}
+{"response": "Which country are you in?", "labels": null, "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/02e74f10e0.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>I am the owner of a shop. Should I flip the open sign or keep the position of the sign unchanged?", "loss": null}, {"role": "assistant", "content": "Which country are you in?"}]}
+{"response": "Is the box being loaded into the car or unloaded from the car?", "labels": null, "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/1afa34a7f9.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>Is the man in red hat picking up the box out of the car or putting down the box into the car?", "loss": null}, {"role": "assistant", "content": "Is the box being loaded into the car or unloaded from the car?"}]}
+{"response": "Is this a red light in the United States or in another country?", "labels": null, "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/ec8956637a.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>Can I make a right turn on the red traffic light?", "loss": null}, {"role": "assistant", "content": "Is this a red light in the United States or in another country?"}]}
+{"response": "Is this a country road or a highway?", "labels": null, "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/f899139df5.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>I am driving on the left lane. Is it allowed?", "loss": null}, {"role": "assistant", "content": "Is this a country road or a highway?"}]}
+{"response": "Is your brother a minor?", "labels": null, "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/26657d5ff9.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>My brother is 20 years old. Is my brother's behavior legal?", "loss": null}, {"role": "assistant", "content": "Is your brother a minor?"}]}
+{"response": "Is the woman drinking alone or with others?", "labels": null, "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/ed3d2c2199.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>The women drinking alcohol is 20 years old. Is her behavior legal?", "loss": null}, {"role": "assistant", "content": "Is the woman drinking alone or with others?"}]}
+{"response": "Is he riding his bike on a road or a path?", "labels": null, "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/bd686fd640.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>This is the photo of my 10-year-old-son. Is his the behavior legal?", "loss": null}, {"role": "assistant", "content": "Is he riding his bike on a road or a path?"}]}
+{"response": "Is this a public place or a private one?", "labels": null, "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/a597e50502.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>I want to buy two bottles of milk. Is my gesture appropriate?", "loss": null}, {"role": "assistant", "content": "Is this a public place or a private one?"}]}
+{"response": "Is this a one-way road?", "labels": null, "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/38b3eff8ba.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>Am I driving in the correct lane?", "loss": null}, {"role": "assistant", "content": "Is this a one-way road?"}]}
+{"response": "Is the face of the cube that contains side B a front face or a back face?", "labels": null, "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/202cb962ac.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>In the 3D perception of the images. Is side B part of the face of a cube?", "loss": null}, {"role": "assistant", "content": "Is the face of the cube that contains side B a front face or a back face?"}]}
+{"response": "Is this stock listed on a major stock exchange?", "labels": null, "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/7f6ffaa6bb.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>Based on the colors on the screen, is this stock performing well?", "loss": null}, {"role": "assistant", "content": "Is this stock listed on a major stock exchange?"}]}
+{"response": "Is he opening or closing his bag?", "labels": null, "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/ec5decca5e.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>Is he opening or closing his bag?", "loss": null}, {"role": "assistant", "content": "Is he opening or closing his bag?"}]}
+{"response": "Is the time on the left clock 14:15 or 23:00?", "labels": null, "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/aa00000001.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>Based on the image, what time is it now? One of the clock shows the correct time.", "loss": null}, {"role": "assistant", "content": "Is the time on the left clock 14:15 or 23:00?"}]}
+{"response": "Is the person in the image wearing a hat or a wig?", "labels": null, "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/c8ffe9a587.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>Does the person in the image look old or young?", "loss": null}, {"role": "assistant", "content": "Is the person in the image wearing a hat or a wig?"}]}
+{"response": "Is the yellow dot on the left or the right?", "labels": null, "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/6f3ef77ac0.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>Which direction is the side with a yellow dot facing? Up, down, left or right?", "loss": null}, {"role": "assistant", "content": "Is the yellow dot on the left or the right?"}]}
+{"response": "Is the woman wearing a hat?", "labels": null, "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/c8ffe9a587.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>Can we see the woman's ear in the picture?", "loss": null}, {"role": "assistant", "content": "Is the woman wearing a hat?"}]}
+{"response": "Is the baby in a bathtub or a basin?", "labels": null, "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/d09bf41544.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>The baby is play water when sitting in a basin. Is the baby in danger?", "loss": null}, {"role": "assistant", "content": "Is the baby in a bathtub or a basin?"}]}
+{"response": "Is the blue dot on the front plane or the red dot?", "labels": null, "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/6f3ef77ac0.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>Is blue dot on the front plane or the red dot?", "loss": null}, {"role": "assistant", "content": "Is the blue dot on the front plane or the red dot?"}]}
+{"response": "Is your mother walking with her family or with her friends?", "labels": null, "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/7f39f8317f.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>My mother is walking with her families. Which one is my mother?", "loss": null}, {"role": "assistant", "content": "Is your mother walking with her family or with her friends?"}]}
+{"response": "Is the dog a pet or a working animal?", "labels": null, "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/903ce9225f.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>Who is the owner of the dog?", "loss": null}, {"role": "assistant", "content": "Is the dog a pet or a working animal?"}]}
+{"response": "Is the box of water you received from your friend in the foreground or in the background?", "labels": null, "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/1d7f7abc18.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>My friend handed me a box of water. Which one is my friend?", "loss": null}, {"role": "assistant", "content": "Is the box of water you received from your friend in the foreground or in the background?"}]}
+{"response": "I can't tell from the image alone. Can you tell me your role in the project?", "labels": null, "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/03afdbd66e.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>I and my leader are discussing a project. Who is my leader?", "loss": null}, {"role": "assistant", "content": "I can't tell from the image alone. Can you tell me your role in the project?"}]}
+{"response": "Is your daughter the one wearing the blue and white dress, or the one in the green dress?", "labels": null, "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/c0c7c76d30.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>My daughter is one of the two women. Who is my daughter?", "loss": null}, {"role": "assistant", "content": "Is your daughter the one wearing the blue and white dress, or the one in the green dress?"}]}
+{"response": "Is he giving a speech, or is he listening to someone else?", "labels": null, "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/9a1158154d.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>Our principle is one of the two men. What is the principle doing?", "loss": null}, {"role": "assistant", "content": "Is he giving a speech, or is he listening to someone else?"}]}
+{"response": "Is your sister the one who is cooking or the one holding the baby?", "labels": null, "logprobs": null, "images": [{"bytes": null, "path": "/home/clarify-ambiguity/GRIT_data/clarification/images/072b030ba1.jpg"}], "messages": [{"role": "system", "content": "You write exactly one clarifying question. Do not answer the original question.\n\nGiven an image and an original question, ask for the single missing fact that would determine the answer.\n\nInput: (1) image, (2) original question.\n\nOutput: exactly one short clarification question without any prefix.\n\nNow generate the clarification question.", "loss": null}, {"role": "user", "content": "<image>My sister is one of the two women. Who is my sister?", "loss": null}, {"role": "assistant", "content": "Is your sister the one who is cooking or the one holding the baby?"}]}

latest

@@ -0,0 +1 @@
+global_step33

rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aeae17d90e4d6cb0d1ed3aa98e766042ffa1dc8abf62a1a6be5f8c2712d19356
+size 14512

rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1c1a61136fd0e8e7a644c56360ed64802e5b1e134b24c989b8bc9c9a6e2cbf6f
+size 14512

scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:027542fad38f771b478b22bb3927e76abad00a039aa3c38c656a3739dcd9ab5a
+size 1064

trainer_state.json

@@ -0,0 +1,90 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.9263157894736844,
+  "eval_steps": 100.0,
+  "global_step": 33,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.08421052631578947,
+      "grad_norm": 0.09988608956336975,
+      "learning_rate": 5e-05,
+      "loss": 2.3771400451660156,
+      "step": 1,
+      "token_acc": 0.4664310954063604
+    },
+    {
+      "epoch": 0.42105263157894735,
+      "grad_norm": 0.04726523533463478,
+      "learning_rate": 9.770696282000244e-05,
+      "loss": 2.1574275493621826,
+      "step": 5,
+      "token_acc": 0.519051651143099
+    },
+    {
+      "epoch": 0.8421052631578947,
+      "grad_norm": 0.03580109775066376,
+      "learning_rate": 8.444834595378434e-05,
+      "loss": 1.7195571899414062,
+      "step": 10,
+      "token_acc": 0.5904558404558404
+    },
+    {
+      "epoch": 1.3368421052631578,
+      "grad_norm": 0.03144058212637901,
+      "learning_rate": 6.253262661293604e-05,
+      "loss": 1.85762939453125,
+      "step": 15,
+      "token_acc": 0.6245161290322581
+    },
+    {
+      "epoch": 1.7578947368421054,
+      "grad_norm": 0.0300274845212698,
+      "learning_rate": 3.746737338706397e-05,
+      "loss": 1.3842062950134277,
+      "step": 20,
+      "token_acc": 0.653372008701958
+    },
+    {
+      "epoch": 2.2526315789473683,
+      "grad_norm": 0.02691994234919548,
+      "learning_rate": 1.555165404621567e-05,
+      "loss": 1.6822933197021483,
+      "step": 25,
+      "token_acc": 0.6409290646578782
+    },
+    {
+      "epoch": 2.6736842105263157,
+      "grad_norm": 0.030764643102884293,
+      "learning_rate": 2.2930371799975594e-06,
+      "loss": 1.3119495391845704,
+      "step": 30,
+      "token_acc": 0.6437125748502994
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 33,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 3,
+  "save_steps": 100,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.1367359130171867e+19,
+  "train_batch_size": 1,
+  "trial_name": null,
+  "trial_params": null
+}

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2baaf879875a908a92f4526f4132a93c1178dbfb106f007a21c0ad8d15e07699
+size 8632

zero_to_fp32.py

@@ -0,0 +1,760 @@
+#!/usr/bin/env python
+
+# Copyright (c) Microsoft Corporation.
+# SPDX-License-Identifier: Apache-2.0
+
+# DeepSpeed Team
+
+# This script extracts fp32 consolidated weights from a zero 1, 2 and 3 DeepSpeed checkpoints. It gets
+# copied into the top level checkpoint dir, so the user can easily do the conversion at any point in
+# the future. Once extracted, the weights don't require DeepSpeed and can be used in any
+# application.
+#
+# example:
+#   python zero_to_fp32.py . output_dir/
+#   or
+#   python zero_to_fp32.py . output_dir/ --safe_serialization
+
+import argparse
+import torch
+import glob
+import math
+import os
+import re
+import gc
+import json
+import numpy as np
+from tqdm import tqdm
+from collections import OrderedDict
+from dataclasses import dataclass
+
+# while this script doesn't use deepspeed to recover data, since the checkpoints are pickled with
+# DeepSpeed data structures it has to be available in the current python environment.
+from deepspeed.utils import logger
+from deepspeed.checkpoint.constants import (DS_VERSION, OPTIMIZER_STATE_DICT, SINGLE_PARTITION_OF_FP32_GROUPS,
+                                            FP32_FLAT_GROUPS, ZERO_STAGE, PARTITION_COUNT, PARAM_SHAPES, BUFFER_NAMES,
+                                            FROZEN_PARAM_SHAPES, FROZEN_PARAM_FRAGMENTS)
+
+
+@dataclass
+class zero_model_state:
+    buffers: dict()
+    param_shapes: dict()
+    shared_params: list
+    ds_version: int
+    frozen_param_shapes: dict()
+    frozen_param_fragments: dict()
+
+
+debug = 0
+
+# load to cpu
+device = torch.device('cpu')
+
+
+def atoi(text):
+    return int(text) if text.isdigit() else text
+
+
+def natural_keys(text):
+    '''
+    alist.sort(key=natural_keys) sorts in human order
+    http://nedbatchelder.com/blog/200712/human_sorting.html
+    (See Toothy's implementation in the comments)
+    '''
+    return [atoi(c) for c in re.split(r'(\d+)', text)]
+
+
+def get_model_state_file(checkpoint_dir, zero_stage):
+    if not os.path.isdir(checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{checkpoint_dir}' doesn't exist")
+
+    # there should be only one file
+    if zero_stage <= 2:
+        file = os.path.join(checkpoint_dir, "mp_rank_00_model_states.pt")
+    elif zero_stage == 3:
+        file = os.path.join(checkpoint_dir, "zero_pp_rank_0_mp_rank_00_model_states.pt")
+
+    if not os.path.exists(file):
+        raise FileNotFoundError(f"can't find model states file at '{file}'")
+
+    return file
+
+
+def get_checkpoint_files(checkpoint_dir, glob_pattern):
+    # XXX: need to test that this simple glob rule works for multi-node setup too
+    ckpt_files = sorted(glob.glob(os.path.join(checkpoint_dir, glob_pattern)), key=natural_keys)
+
+    if len(ckpt_files) == 0:
+        raise FileNotFoundError(f"can't find {glob_pattern} files in directory '{checkpoint_dir}'")
+
+    return ckpt_files
+
+
+def get_optim_files(checkpoint_dir):
+    return get_checkpoint_files(checkpoint_dir, "*_optim_states.pt")
+
+
+def get_model_state_files(checkpoint_dir):
+    return get_checkpoint_files(checkpoint_dir, "*_model_states.pt")
+
+
+def parse_model_states(files):
+    zero_model_states = []
+    for file in files:
+        state_dict = torch.load(file, map_location=device, weights_only=False)
+
+        if BUFFER_NAMES not in state_dict:
+            raise ValueError(f"{file} is not a model state checkpoint")
+        buffer_names = state_dict[BUFFER_NAMES]
+        if debug:
+            print("Found buffers:", buffer_names)
+
+        # recover just the buffers while restoring them to fp32 if they were saved in fp16
+        buffers = {k: v.float() for k, v in state_dict["module"].items() if k in buffer_names}
+        param_shapes = state_dict[PARAM_SHAPES]
+
+        # collect parameters that are included in param_shapes
+        param_names = []
+        for s in param_shapes:
+            for name in s.keys():
+                param_names.append(name)
+
+        # update with frozen parameters
+        frozen_param_shapes = state_dict.get(FROZEN_PARAM_SHAPES, None)
+        if frozen_param_shapes is not None:
+            if debug:
+                print(f"Found frozen_param_shapes: {frozen_param_shapes}")
+            param_names += list(frozen_param_shapes.keys())
+
+        # handle shared params
+        shared_params = [[k, v] for k, v in state_dict["shared_params"].items()]
+
+        ds_version = state_dict.get(DS_VERSION, None)
+
+        frozen_param_fragments = state_dict.get(FROZEN_PARAM_FRAGMENTS, None)
+
+        z_model_state = zero_model_state(buffers=buffers,
+                                         param_shapes=param_shapes,
+                                         shared_params=shared_params,
+                                         ds_version=ds_version,
+                                         frozen_param_shapes=frozen_param_shapes,
+                                         frozen_param_fragments=frozen_param_fragments)
+        zero_model_states.append(z_model_state)
+
+    return zero_model_states
+
+
+def parse_optim_states(files, ds_checkpoint_dir):
+    total_files = len(files)
+    state_dicts = []
+    for f in tqdm(files, desc='Loading checkpoint shards'):
+        state_dict = torch.load(f, map_location=device, mmap=True, weights_only=False)
+        # immediately discard the potentially huge 2 optimizer states as we only care for fp32 master weights
+        # and also handle the case where it was already removed by another helper script
+        state_dict["optimizer_state_dict"].pop("optimizer_state_dict", None)
+        state_dicts.append(state_dict)
+
+    if ZERO_STAGE not in state_dicts[0][OPTIMIZER_STATE_DICT]:
+        raise ValueError(f"{files[0]} is not a zero checkpoint")
+    zero_stage = state_dicts[0][OPTIMIZER_STATE_DICT][ZERO_STAGE]
+    world_size = state_dicts[0][OPTIMIZER_STATE_DICT][PARTITION_COUNT]
+
+    # For ZeRO-2 each param group can have different partition_count as data parallelism for expert
+    # parameters can be different from data parallelism for non-expert parameters. So we can just
+    # use the max of the partition_count to get the dp world_size.
+
+    if type(world_size) is list:
+        world_size = max(world_size)
+
+    if world_size != total_files:
+        raise ValueError(
+            f"Expected {world_size} of '*_optim_states.pt' under '{ds_checkpoint_dir}' but found {total_files} files. "
+            "Possibly due to an overwrite of an old checkpoint, or a checkpoint didn't get saved by one or more processes."
+        )
+
+    # the groups are named differently in each stage
+    if zero_stage <= 2:
+        fp32_groups_key = SINGLE_PARTITION_OF_FP32_GROUPS
+    elif zero_stage == 3:
+        fp32_groups_key = FP32_FLAT_GROUPS
+    else:
+        raise ValueError(f"unknown zero stage {zero_stage}")
+
+    fp32_flat_groups = [state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key] for i in range(len(state_dicts))]
+    return zero_stage, world_size, fp32_flat_groups
+
+
+def _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters):
+    """
+    Returns fp32 state_dict reconstructed from ds checkpoint
+
+    Args:
+        - ``ds_checkpoint_dir``: path to the deepspeed checkpoint folder (where the optimizer files are)
+
+    """
+    print(f"Processing zero checkpoint '{ds_checkpoint_dir}'")
+
+    optim_files = get_optim_files(ds_checkpoint_dir)
+    zero_stage, world_size, fp32_flat_groups = parse_optim_states(optim_files, ds_checkpoint_dir)
+    print(f"Detected checkpoint of type zero stage {zero_stage}, world_size: {world_size}")
+
+    model_files = get_model_state_files(ds_checkpoint_dir)
+
+    zero_model_states = parse_model_states(model_files)
+    print(f'Parsing checkpoint created by deepspeed=={zero_model_states[0].ds_version}')
+
+    if zero_stage <= 2:
+        return _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+    elif zero_stage == 3:
+        return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+
+
+def _zero2_merge_frozen_params(state_dict, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+    frozen_param_fragments = zero_model_states[0].frozen_param_fragments
+
+    if debug:
+        num_elem = sum(s.numel() for s in frozen_param_shapes.values())
+        print(f'rank 0: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in frozen_param_fragments.values()])
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        state_dict[name] = frozen_param_fragments[name]
+
+        if debug:
+            print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _has_callable(obj, fn):
+    attr = getattr(obj, fn, None)
+    return callable(attr)
+
+
+def _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+
+    # Reconstruction protocol:
+    #
+    # XXX: document this
+
+    if debug:
+        for i in range(world_size):
+            for j in range(len(fp32_flat_groups[0])):
+                print(f"{FP32_FLAT_GROUPS}[{i}][{j}].shape={fp32_flat_groups[i][j].shape}")
+
+    # XXX: memory usage doubles here (zero2)
+    num_param_groups = len(fp32_flat_groups[0])
+    merged_single_partition_of_fp32_groups = []
+    for i in range(num_param_groups):
+        merged_partitions = [sd[i] for sd in fp32_flat_groups]
+        full_single_fp32_vector = torch.cat(merged_partitions, 0)
+        merged_single_partition_of_fp32_groups.append(full_single_fp32_vector)
+    avail_numel = sum(
+        [full_single_fp32_vector.numel() for full_single_fp32_vector in merged_single_partition_of_fp32_groups])
+
+    if debug:
+        wanted_params = sum([len(shapes) for shapes in param_shapes])
+        wanted_numel = sum([sum(shape.numel() for shape in shapes.values()) for shapes in param_shapes])
+        # not asserting if there is a mismatch due to possible padding
+        print(f"Have {avail_numel} numels to process.")
+        print(f"Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    total_numel = 0
+    total_params = 0
+    for shapes, full_single_fp32_vector in zip(param_shapes, merged_single_partition_of_fp32_groups):
+        offset = 0
+        avail_numel = full_single_fp32_vector.numel()
+        for name, shape in shapes.items():
+
+            unpartitioned_numel = shape.numel() if _has_callable(shape, 'numel') else math.prod(shape)
+            total_numel += unpartitioned_numel
+            total_params += 1
+
+            if debug:
+                print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+            state_dict[name] = full_single_fp32_vector.narrow(0, offset, unpartitioned_numel).view(shape)
+            offset += unpartitioned_numel
+
+        # Z2 started to align to 2*world_size to improve nccl performance. Therefore both offset and
+        # avail_numel can differ by anywhere between 0..2*world_size. Due to two unrelated complex
+        # paddings performed in the code it's almost impossible to predict the exact numbers w/o the
+        # live optimizer object, so we are checking that the numbers are within the right range
+        align_to = 2 * world_size
+
+        def zero2_align(x):
+            return align_to * math.ceil(x / align_to)
+
+        if debug:
+            print(f"original offset={offset}, avail_numel={avail_numel}")
+
+        offset = zero2_align(offset)
+        avail_numel = zero2_align(avail_numel)
+
+        if debug:
+            print(f"aligned  offset={offset}, avail_numel={avail_numel}")
+
+        # Sanity check
+        if offset != avail_numel:
+            raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _zero2_merge_frozen_params(state_dict, zero_model_states)
+
+    _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def zero3_partitioned_param_info(unpartitioned_numel, world_size):
+    remainder = unpartitioned_numel % world_size
+    padding_numel = (world_size - remainder) if remainder else 0
+    partitioned_numel = math.ceil(unpartitioned_numel / world_size)
+    return partitioned_numel, padding_numel
+
+
+def _zero3_merge_frozen_params(state_dict, world_size, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    if debug:
+        for i in range(world_size):
+            num_elem = sum(s.numel() for s in zero_model_states[i].frozen_param_fragments.values())
+            print(f'rank {i}: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in zero_model_states[0].frozen_param_fragments.values()]) * world_size
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in zero_model_states[0].frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        param_frags = tuple(model_state.frozen_param_fragments[name] for model_state in zero_model_states)
+        state_dict[name] = torch.cat(param_frags, 0).narrow(0, 0, unpartitioned_numel).view(shape)
+
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Frozen params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+class GatheredTensor:
+    """
+    A pseudo tensor that collects partitioned weights.
+    It is more memory efficient when there are multiple groups.
+    """
+
+    def __init__(self, flat_groups, flat_groups_offset, offset, partitioned_numel, shape):
+        self.flat_groups = flat_groups
+        self.flat_groups_offset = flat_groups_offset
+        self.offset = offset
+        self.partitioned_numel = partitioned_numel
+        self.shape = shape
+        self.dtype = self.flat_groups[0][0].dtype
+
+    def contiguous(self):
+        """
+        Merge partitioned weights from flat_groups into a single tensor.
+        """
+        end_idx = self.offset + self.partitioned_numel
+        world_size = len(self.flat_groups)
+        pad_flat_param_chunks = []
+
+        for rank_i in range(world_size):
+            # for each rank, we need to collect weights from related group/groups
+            flat_groups_at_rank_i = self.flat_groups[rank_i]
+            start_group_id = None
+            end_group_id = None
+            for group_id in range(len(self.flat_groups_offset)):
+                if self.flat_groups_offset[group_id] <= self.offset < self.flat_groups_offset[group_id + 1]:
+                    start_group_id = group_id
+                if self.flat_groups_offset[group_id] < end_idx <= self.flat_groups_offset[group_id + 1]:
+                    end_group_id = group_id
+                    break
+            # collect weights from related group/groups
+            for group_id in range(start_group_id, end_group_id + 1):
+                flat_tensor = flat_groups_at_rank_i[group_id]
+                start_offset = self.offset - self.flat_groups_offset[group_id]
+                end_offset = min(end_idx, self.flat_groups_offset[group_id + 1]) - self.flat_groups_offset[group_id]
+                pad_flat_param_chunks.append(flat_tensor[start_offset:end_offset])
+
+        # collect weights from all ranks
+        pad_flat_param = torch.cat(pad_flat_param_chunks, dim=0)
+        param = pad_flat_param[:self.shape.numel()].view(self.shape).contiguous()
+        return param
+
+
+def _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+    avail_numel = sum([flat_group.numel() for flat_group in fp32_flat_groups[0]]) * world_size
+
+    # Reconstruction protocol: For zero3 we need to zip the partitions together at boundary of each
+    # param, re-consolidating each param, while dealing with padding if any
+
+    # merge list of dicts, preserving order
+    param_shapes = {k: v for d in param_shapes for k, v in d.items()}
+
+    if debug:
+        for i in range(world_size):
+            print(f"{FP32_FLAT_GROUPS}[{i}].shape={fp32_flat_groups[i].shape}")
+
+        wanted_params = len(param_shapes)
+        wanted_numel = sum(shape.numel() for shape in param_shapes.values())
+        # not asserting if there is a mismatch due to possible padding
+        avail_numel = fp32_flat_groups[0].numel() * world_size
+        print(f"Trainable params: Have {avail_numel} numels to process.")
+        print(f"Trainable params: Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    offset = 0
+    total_numel = 0
+    total_params = 0
+    flat_groups_offset = [0] + list(np.cumsum([flat_tensor.numel() for flat_tensor in fp32_flat_groups[0]]))
+    for name, shape in tqdm(param_shapes.items(), desc='Gathering sharded weights'):
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+        total_params += 1
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Trainable params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+        # memory efficient tensor
+        tensor = GatheredTensor(fp32_flat_groups, flat_groups_offset, offset, partitioned_numel, shape)
+        state_dict[name] = tensor
+        offset += partitioned_numel
+
+    offset *= world_size
+
+    # Sanity check
+    if offset != avail_numel:
+        raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed Trainable fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _zero3_merge_frozen_params(state_dict, world_size, zero_model_states)
+
+    _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def to_torch_tensor(state_dict, return_empty_tensor=False):
+    """
+    Convert state_dict of GatheredTensor to torch tensor
+    """
+    torch_state_dict = {}
+    converted_tensors = {}
+    for name, tensor in state_dict.items():
+        tensor_id = id(tensor)
+        if tensor_id in converted_tensors:  # shared tensors
+            shared_tensor = torch_state_dict[converted_tensors[tensor_id]]
+            torch_state_dict[name] = shared_tensor
+        else:
+            converted_tensors[tensor_id] = name
+            if return_empty_tensor:
+                torch_state_dict[name] = torch.empty(tensor.shape, dtype=tensor.dtype)
+            else:
+                torch_state_dict[name] = tensor.contiguous()
+    return torch_state_dict
+
+
+def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir,
+                                             tag=None,
+                                             exclude_frozen_parameters=False,
+                                             lazy_mode=False):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated state_dict that can be loaded with
+    ``load_state_dict()`` and used for training without DeepSpeed or shared with others, for example
+    via a model hub.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in 'latest' file. e.g., ``global_step14``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+        - ``lazy_mode``: get state_dict in lazy mode. It returns a dict of pesduo tensor instead of torch tensor, which is more memory efficient.
+          Convert the pesduo tensor to torch tensor by ``.contiguous()``
+
+    Returns:
+        - pytorch ``state_dict``
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
+        # do the training and checkpoint saving
+        state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir) # already on cpu
+        model = model.cpu() # move to cpu
+        model.load_state_dict(state_dict)
+        # submit to model hub or save the model to share with others
+
+    In this example the ``model`` will no longer be usable in the deepspeed context of the same
+    application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    If you want it all done for you, use ``load_state_dict_from_zero_checkpoint`` instead.
+
+    Note: the above usage may not work if your application doesn't have sufficient free CPU memory.
+    You may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
+    the checkpoint. Or you can load state_dict in lazy mode ::
+
+        from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
+        state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, lazy_mode=True) # not on cpu
+        for name, lazy_tensor in state_dict.item():
+            tensor = lazy_tensor.contiguous()  # to cpu
+            print(name, tensor)
+            # del tensor to release memory if it no longer in use
+    """
+    if tag is None:
+        latest_path = os.path.join(checkpoint_dir, 'latest')
+        if os.path.isfile(latest_path):
+            with open(latest_path, 'r') as fd:
+                tag = fd.read().strip()
+        else:
+            raise ValueError(f"Unable to find 'latest' file at {latest_path}")
+
+    ds_checkpoint_dir = os.path.join(checkpoint_dir, tag)
+
+    if not os.path.isdir(ds_checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{ds_checkpoint_dir}' doesn't exist")
+
+    state_dict = _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters)
+    if lazy_mode:
+        return state_dict
+    else:
+        return to_torch_tensor(state_dict)
+
+
+def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir,
+                                               output_dir,
+                                               max_shard_size="5GB",
+                                               safe_serialization=False,
+                                               tag=None,
+                                               exclude_frozen_parameters=False):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict`` file that can be
+    loaded with ``torch.load(file)`` + ``load_state_dict()`` and used for training without DeepSpeed.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``output_dir``: directory to the pytorch fp32 state_dict output files
+        - ``max_shard_size``: the maximum size for a checkpoint before being sharded, default value is 5GB
+        - ``safe_serialization``:  whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+    """
+
+    # Dependency pre-check
+    if safe_serialization:
+        try:
+            from safetensors.torch import save_file
+        except ImportError:
+            print('If you want to use `safe_serialization`, please `pip install safetensors`')
+            raise
+    if max_shard_size is not None:
+        try:
+            from huggingface_hub import split_torch_state_dict_into_shards
+        except ImportError:
+            print('If you want to use `max_shard_size`, please `pip install huggingface_hub`')
+            raise
+
+    # Convert zero checkpoint to state_dict
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir,
+                                                          tag,
+                                                          exclude_frozen_parameters,
+                                                          lazy_mode=True)
+
+    # Shard the model if it is too big.
+    weights_name = "model.safetensors" if safe_serialization else "pytorch_model.bin"
+    if max_shard_size is not None:
+        filename_pattern = weights_name.replace(".bin", "{suffix}.bin").replace(".safetensors", "{suffix}.safetensors")
+        # an memory-efficient approach for sharding
+        empty_state_dict = to_torch_tensor(state_dict, return_empty_tensor=True)
+        state_dict_split = split_torch_state_dict_into_shards(empty_state_dict,
+                                                              filename_pattern=filename_pattern,
+                                                              max_shard_size=max_shard_size)
+    else:
+        from collections import namedtuple
+        StateDictSplit = namedtuple("StateDictSplit", ["is_sharded", "filename_to_tensors"])
+        state_dict_split = StateDictSplit(is_sharded=False,
+                                          filename_to_tensors={weights_name: list(state_dict.keys())})
+
+    # Save the model by shard
+    os.makedirs(output_dir, exist_ok=True)
+    filename_to_tensors = state_dict_split.filename_to_tensors.items()
+    for shard_file, tensors in tqdm(filename_to_tensors, desc="Saving checkpoint shards"):
+        shard_state_dict = {tensor_name: state_dict[tensor_name] for tensor_name in tensors}
+        shard_state_dict = to_torch_tensor(shard_state_dict)
+        output_path = os.path.join(output_dir, shard_file)
+        if safe_serialization:
+            save_file(shard_state_dict, output_path, metadata={"format": "pt"})
+        else:
+            torch.save(shard_state_dict, output_path)
+        # release the memory of current shard
+        for tensor_name in list(shard_state_dict.keys()):
+            del state_dict[tensor_name]
+            del shard_state_dict[tensor_name]
+        del shard_state_dict
+        gc.collect()
+
+    # Save index if sharded
+    if state_dict_split.is_sharded:
+        index = {
+            "metadata": state_dict_split.metadata,
+            "weight_map": state_dict_split.tensor_to_filename,
+        }
+        save_index_file = "model.safetensors.index.json" if safe_serialization else "pytorch_model.bin.index.json"
+        save_index_file = os.path.join(output_dir, save_index_file)
+        with open(save_index_file, "w", encoding="utf-8") as f:
+            content = json.dumps(index, indent=2, sort_keys=True) + "\n"
+            f.write(content)
+
+
+def load_state_dict_from_zero_checkpoint(model, checkpoint_dir, tag=None):
+    """
+    1. Put the provided model to cpu
+    2. Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict``
+    3. Load it into the provided model
+
+    Args:
+        - ``model``: the model object to update
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
+
+    Returns:
+        - ``model`: modified model
+
+    Make sure you have plenty of CPU memory available before you call this function. If you don't
+    have enough use the ``zero_to_fp32.py`` utility to do the conversion. You will find it
+    conveniently placed for you in the checkpoint folder.
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import load_state_dict_from_zero_checkpoint
+        model = load_state_dict_from_zero_checkpoint(trainer.model, checkpoint_dir)
+        # submit to model hub or save the model to share with others
+
+    Note, that once this was run, the ``model`` will no longer be usable in the deepspeed context
+    of the same application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    """
+    logger.info("Extracting fp32 weights")
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
+
+    logger.info("Overwriting model with fp32 weights")
+    model = model.cpu()
+    model.load_state_dict(state_dict, strict=False)
+
+    return model
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    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 to the pytorch fp32 state_dict output files"
+                        "(e.g. path/checkpoint-12-output/)")
+    parser.add_argument(
+        "--max_shard_size",
+        type=str,
+        default="5GB",
+        help="The maximum size for a checkpoint before being sharded. Checkpoints shard will then be each of size"
+        "lower than this size. If expressed as a string, needs to be digits followed by a unit (like `5MB`"
+        "We default it to 5GB in order for models to be able to run easily on free-tier google colab instances"
+        "without CPU OOM issues.")
+    parser.add_argument(
+        "--safe_serialization",
+        default=False,
+        action='store_true',
+        help="Whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).")
+    parser.add_argument("-t",
+                        "--tag",
+                        type=str,
+                        default=None,
+                        help="checkpoint tag used as a unique identifier for checkpoint. 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")
+    args = parser.parse_args()
+
+    debug = args.debug
+
+    convert_zero_checkpoint_to_fp32_state_dict(args.checkpoint_dir,
+                                               args.output_dir,
+                                               max_shard_size=args.max_shard_size,
+                                               safe_serialization=args.safe_serialization,
+                                               tag=args.tag,
+                                               exclude_frozen_parameters=args.exclude_frozen_parameters)