Upload folder multi_stage2_run_stage1_both to stage_2/multi_stage2_run_stage1_both

stage_2/multi_stage2_run_stage1_both/stage2a/20250925_202658/vis_data/config.py

@@ -0,0 +1,261 @@
+SYSTEM = ''
+accumulative_counts = 64
+batch_size = 1
+betas = (
+    0.9,
+    0.999,
+)
+bnb = dict(
+    bnb_4bit_compute_dtype='torch.bfloat16',
+    bnb_4bit_quant_type='nf4',
+    bnb_4bit_use_double_quant=True,
+    llm_int8_has_fp16_weight=False,
+    llm_int8_threshold=6.0,
+    load_in_4bit=True,
+    load_in_8bit=False,
+    type='transformers.BitsAndBytesConfig')
+custom_hooks = [
+    dict(
+        tokenizer=dict(
+            padding_side='right',
+            pretrained_model_name_or_path='Qwen/Qwen2.5-7B-Instruct',
+            trust_remote_code=True,
+            type='transformers.AutoTokenizer.from_pretrained'),
+        type='xtuner.engine.hooks.DatasetInfoHook'),
+    dict(
+        evaluation_images=[
+            '/mnt/bn/xudong-va/meilong/datasets/Token_Compression/skcm_224x224_b20_t15/h5_files/TCGA-EB-A5UN-06Z-00-DX1.h5',
+            '/mnt/bn/xudong-va/meilong/datasets/Token_Compression/skcm_224x224_b20_t15/h5_files/TCGA-EE-A3AG-01Z-00-DX1.h5',
+            '/mnt/bn/xudong-va/meilong/datasets/Token_Compression/lusc_224x224_b20_t15/h5_files/TCGA-NC-A5HP-01Z-00-DX1.h5',
+        ],
+        evaluation_inputs=[
+            'Are the tumor cells organized in a lobulated pattern within the slide?',
+            'Craft a comprehensive outline capturing the key findings of the pathology report based on the whole slide image.',
+            'Based on the observed features, what do you think is the correct histological classification of the tumor? A) Poorly differentiated keratinizing squamous cell carcinoma B) Moderately differentiated squamous cell carcinoma C) Well-differentiated squamous cell carcinoma D) Adenocarcinoma',
+        ],
+        every_n_iters=512,
+        prompt_template='xtuner.utils.PROMPT_TEMPLATE.qwen_chat',
+        system='',
+        tokenizer=dict(
+            padding_side='right',
+            pretrained_model_name_or_path='Qwen/Qwen2.5-7B-Instruct',
+            trust_remote_code=True,
+            type='transformers.AutoTokenizer.from_pretrained'),
+        type='xtuner.engine.hooks.EvaluateChatHookResampler'),
+    dict(type='xtuner.engine.hooks.ThroughputHook'),
+]
+data_path = '/mnt/bn/yuxuanwang/meilong/code/projects/efficient_foundation_wsi_llava/merged_dataset/stage2_tasks_plus_report.json'
+dataloader_num_workers = 10
+default_hooks = dict(
+    checkpoint=dict(
+        by_epoch=False,
+        interval=4096,
+        max_keep_ckpts=8,
+        type='mmengine.hooks.CheckpointHook'),
+    logger=dict(
+        interval=10,
+        log_metric_by_epoch=False,
+        type='mmengine.hooks.LoggerHook'),
+    param_scheduler=dict(type='mmengine.hooks.ParamSchedulerHook'),
+    sampler_seed=dict(type='mmengine.hooks.DistSamplerSeedHook'),
+    timer=dict(type='mmengine.hooks.IterTimerHook'))
+env_cfg = dict(
+    cudnn_benchmark=False,
+    dist_cfg=dict(backend='nccl'),
+    mp_cfg=dict(mp_start_method='fork', opencv_num_threads=0))
+evaluation_freq = 512
+evaluation_images = [
+    '/mnt/bn/xudong-va/meilong/datasets/Token_Compression/skcm_224x224_b20_t15/h5_files/TCGA-EB-A5UN-06Z-00-DX1.h5',
+    '/mnt/bn/xudong-va/meilong/datasets/Token_Compression/skcm_224x224_b20_t15/h5_files/TCGA-EE-A3AG-01Z-00-DX1.h5',
+    '/mnt/bn/xudong-va/meilong/datasets/Token_Compression/lusc_224x224_b20_t15/h5_files/TCGA-NC-A5HP-01Z-00-DX1.h5',
+]
+evaluation_inputs = [
+    'Are the tumor cells organized in a lobulated pattern within the slide?',
+    'Craft a comprehensive outline capturing the key findings of the pathology report based on the whole slide image.',
+    'Based on the observed features, what do you think is the correct histological classification of the tumor? A) Poorly differentiated keratinizing squamous cell carcinoma B) Moderately differentiated squamous cell carcinoma C) Well-differentiated squamous cell carcinoma D) Adenocarcinoma',
+]
+image_path_list = None
+launcher = 'pytorch'
+llava_dataset = dict(
+    data_path=
+    '/mnt/bn/yuxuanwang/meilong/code/projects/efficient_foundation_wsi_llava/merged_dataset/stage2_tasks_plus_report.json',
+    dataset_map_fn='xtuner.dataset.map_fns.llava_map_fn',
+    identifier='_224x224_b20_t15',
+    image_feature_prefix='/mnt/bn/xudong-va/meilong/datasets/Token_Compression',
+    image_feature_suffix='.h5',
+    image_folder='',
+    image_path_list=None,
+    max_length=15836,
+    pad_image_to_square=False,
+    per_image_length=10240,
+    sample_num=10240,
+    sample_strategy='linspace',
+    template_map_fn=dict(
+        template='xtuner.utils.PROMPT_TEMPLATE.qwen_chat',
+        type='xtuner.dataset.map_fns.template_map_fn_factory'),
+    tokenizer=dict(
+        padding_side='right',
+        pretrained_model_name_or_path='Qwen/Qwen2.5-7B-Instruct',
+        trust_remote_code=True,
+        type='transformers.AutoTokenizer.from_pretrained'),
+    type='xtuner.dataset.LLaVADataset',
+    unwanted_prefix_csv=
+    '/mnt/bn/yuxuanwang/meilong/code/projects/efficient_foundation_wsi_llava/merged_dataset/missing_slides3.csv'
+)
+llm_lora = dict(
+    bias='none',
+    lora_alpha=256,
+    lora_dropout=0.05,
+    r=128,
+    task_type='CAUSAL_LM',
+    type='peft.LoraConfig')
+llm_name_or_path = 'Qwen/Qwen2.5-7B-Instruct'
+load_from = None
+log_level = 'INFO'
+log_processor = dict(
+    by_epoch=False,
+    mean_pattern='.*(loss|time|data_time|grad_norm|tflops).*',
+    window_size=1)
+lr = 5e-06
+max_epochs = 2
+max_length = 15836
+max_norm = 1
+model = dict(
+    enable_token_merge=True,
+    freeze_llm=True,
+    freeze_mm_in_stage2=False,
+    llm=dict(
+        attn_implementation='flash_attention_2',
+        pretrained_model_name_or_path='Qwen/Qwen2.5-7B-Instruct',
+        quantization_config=dict(
+            bnb_4bit_compute_dtype='torch.bfloat16',
+            bnb_4bit_quant_type='nf4',
+            bnb_4bit_use_double_quant=True,
+            llm_int8_has_fp16_weight=False,
+            llm_int8_threshold=6.0,
+            load_in_4bit=True,
+            load_in_8bit=False,
+            type='transformers.BitsAndBytesConfig'),
+        torch_dtype='torch.bfloat16',
+        trust_remote_code=True,
+        type='transformers.AutoModelForCausalLM.from_pretrained'),
+    llm_lora=dict(
+        bias='none',
+        lora_alpha=256,
+        lora_dropout=0.05,
+        r=128,
+        task_type='CAUSAL_LM',
+        type='peft.LoraConfig'),
+    max_position_embeddings=None,
+    projector_pth=
+    '/mnt/bn/yuxuanwang/meilong/code/projects/efficient_foundation_wsi_llava/checkpoints/stage_1/token_merge_plus_resampler/stage1_qwen25_both_hf/projector/projector.safetensors',
+    resampler_num_latents=100,
+    resampler_pth=
+    '/mnt/bn/yuxuanwang/meilong/code/projects/efficient_foundation_wsi_llava/checkpoints/stage_1/token_merge_plus_resampler/stage1_qwen25_both_hf/resampler/resampler.safetensors',
+    token_merge_pth=
+    '/mnt/bn/yuxuanwang/meilong/code/projects/efficient_foundation_wsi_llava/checkpoints/stage_1/token_merge_plus_resampler/stage1_qwen25_both_hf/token_merger/merger.safetensors',
+    train_stage='2',
+    type='xtuner.model.llava_no_longnet_simple_sampler.LLaVAModel',
+    use_resampler=True)
+optim_type = 'torch.optim.AdamW'
+optim_wrapper = dict(
+    optimizer=dict(
+        betas=(
+            0.9,
+            0.999,
+        ),
+        lr=2e-06,
+        type='torch.optim.AdamW',
+        weight_decay=0.01),
+    paramwise_cfg=dict(
+        bias_decay_mult=0.0,
+        norm_decay_mult=0.0,
+        paramwise_cfg=dict(
+            custom_keys=dict({'^projector\.': dict(lr_mult=1.0)}))),
+    type='DeepSpeedOptimWrapper')
+param_scheduler = [
+    dict(
+        begin=0,
+        by_epoch=True,
+        convert_to_iter_based=True,
+        end=0.1,
+        start_factor=0.01,
+        type='mmengine.optim.LinearLR'),
+    dict(
+        begin=0.1,
+        by_epoch=True,
+        convert_to_iter_based=True,
+        end=2,
+        eta_min=0.0,
+        type='mmengine.optim.CosineAnnealingLR'),
+]
+per_image_length = 10240
+prompt_template = 'xtuner.utils.PROMPT_TEMPLATE.qwen_chat'
+randomness = dict(deterministic=False, seed=None)
+resume = False
+runner_type = 'FlexibleRunner'
+sample_type = 'wsi'
+save_steps = 4096
+save_total_limit = 8
+seed = 42
+strategy = dict(
+    config=dict(
+        bf16=dict(enabled=True),
+        fp16=dict(enabled=False, initial_scale_power=16),
+        gradient_accumulation_steps='auto',
+        gradient_clipping='auto',
+        train_micro_batch_size_per_gpu='auto',
+        zero_allow_untested_optimizer=True,
+        zero_force_ds_cpu_optimizer=False,
+        zero_optimization=dict(overlap_comm=False, stage=2)),
+    exclude_frozen_parameters=True,
+    gradient_accumulation_steps=64,
+    gradient_clipping=1,
+    sequence_parallel_size=1,
+    train_micro_batch_size_per_gpu=1,
+    type='xtuner.engine.DeepSpeedStrategy')
+tokenizer = dict(
+    padding_side='right',
+    pretrained_model_name_or_path='Qwen/Qwen2.5-7B-Instruct',
+    trust_remote_code=True,
+    type='transformers.AutoTokenizer.from_pretrained')
+train_cfg = dict(max_epochs=1, type='xtuner.engine.runner.TrainLoop')
+train_dataloader = dict(
+    batch_size=1,
+    collate_fn=dict(type='xtuner.dataset.collate_fns.default_collate_fn'),
+    dataset=dict(
+        data_path=
+        '/mnt/bn/yuxuanwang/meilong/code/projects/efficient_foundation_wsi_llava/curriculum_training/merged_dataset_curriculum/stage2a_easy.json',
+        dataset_map_fn='xtuner.dataset.map_fns.llava_map_fn',
+        identifier='_224x224_b20_t15',
+        image_feature_prefix=
+        '/mnt/bn/xudong-va/meilong/datasets/Token_Compression',
+        image_feature_suffix='.h5',
+        image_folder='',
+        image_path_list=None,
+        max_length=15836,
+        pad_image_to_square=False,
+        per_image_length=10240,
+        sample_num=10240,
+        sample_strategy='linspace',
+        template_map_fn=dict(
+            template='xtuner.utils.PROMPT_TEMPLATE.qwen_chat',
+            type='xtuner.dataset.map_fns.template_map_fn_factory'),
+        tokenizer=dict(
+            padding_side='right',
+            pretrained_model_name_or_path='Qwen/Qwen2.5-7B-Instruct',
+            trust_remote_code=True,
+            type='transformers.AutoTokenizer.from_pretrained'),
+        type='xtuner.dataset.LLaVADataset',
+        unwanted_prefix_csv=
+        '/mnt/bn/yuxuanwang/meilong/code/projects/efficient_foundation_wsi_llava/merged_dataset/missing_slides3.csv'
+    ),
+    num_workers=10,
+    persistent_workers=True,
+    pin_memory=True,
+    prefetch_factor=4,
+    sampler=dict(shuffle=True, type='mmengine.dataset.DefaultSampler'))
+visualizer = None
+warmup_ratio = 0.05
+weight_decay = 0.01
+work_dir = '/mnt/bn/yuxuanwang/meilong/code/projects/efficient_foundation_wsi_llava/curriculum_training/models/outputs/multi_stage2_run_stage1_both/stage2a'

stage_2/multi_stage2_run_stage1_both/stage2a/20250925_202658/vis_data/eval_outputs_iter_4095.txt

@@ -0,0 +1,24 @@
+Eval output 1:
+<|im_start|>user
+<image>
+Are the tumor cells organized in a lobulated pattern within the slide?<|im_end|>
+<|im_start|>assistant
+Yes, the tumor cells are organized in a lobulated pattern. The lobules are well-defined and show a clear boundary. The lobules are arranged in a radial pattern, with the luminal surface facing towards the center of the lobule. The luminal surface of the tumor cells is smooth and regular, with no evidence of necrosis or hemorrhage. The tumor cells are well-differentiated and show a cohesive growth pattern.<|im_end|>
+
+
+Eval output 2:
+<|im_start|>user
+<image>
+Craft a comprehensive outline capturing the key findings of the pathology report based on the whole slide image.<|im_end|>
+<|im_start|>assistant
+The pathology report indicates that the tissue sample is composed of a mixture of benign and malignant cells. The malignant cells are characterized by their irregular nuclear morphology, hyperchromatic nuclei, and pleomorphic features. The benign cells exhibit a more uniform nuclear appearance. The report also notes the presence of necrosis, which is a significant finding as it suggests the potential for aggressive behavior. Additionally, the report mentions the presence of vascular invasion, indicating that the cancer has spread into the blood vessels, which is a concerning feature. The report concludes by stating that the overall assessment is of invasive ductal carcinoma, which is a type of breast cancer.<|im_end|>
+
+
+Eval output 3:
+<|im_start|>user
+<image>
+Based on the observed features, what do you think is the correct histological classification of the tumor? A) Poorly differentiated keratinizing squamous cell carcinoma B) Moderately differentiated squamous cell carcinoma C) Well-differentiated squamous cell carcinoma D) Adenocarcinoma<|im_end|>
+<|im_start|>assistant
+A) Poorly differentiated keratinizing squamous cell carcinoma<|im_end|>
+
+

stage_2/multi_stage2_run_stage1_both/stage2a/20250925_202658/vis_data/eval_outputs_iter_7451.txt

@@ -0,0 +1,24 @@
+Eval output 1:
+<|im_start|>user
+<image>
+Are the tumor cells organized in a lobulated pattern within the slide?<|im_end|>
+<|im_start|>assistant
+Yes, the tumor cells are organized in a lobulated pattern. The lobules are well-defined and separated by fibrous stroma.<|im_end|>
+
+
+Eval output 2:
+<|im_start|>user
+<image>
+Craft a comprehensive outline capturing the key findings of the pathology report based on the whole slide image.<|im_end|>
+<|im_start|>assistant
+The whole slide image reveals a well-demarcated area of neoplastic tissue, characterized by the presence of malignant cells. These cells exhibit marked pleomorphism, with significant variation in cell size and shape. The neoplastic cells are arranged in irregular clusters and cords, with areas of necrosis noted within the tumor. The tumor cells show marked atypia, with irregular nuclear contours, prominent nucleoli, and frequent mitotic figures. The tumor displays a high degree of cellular atypia, indicative of a poorly differentiated adenocarcinoma. The tumor cells infiltrate the surrounding stroma, with evidence of vascular invasion. The tumor also shows invasion into the adjacent lung parenchyma. The tumor cells express positivity for cytokeratin, indicating their epithelial origin. The absence of estrogen and progesterone receptors suggests that the tumor is hormone receptor negative. The Ki-67 proliferation index is elevated, indicating a high proliferative activity. The tumor cells show no evidence of neuroendocrine differentiation.<|im_end|>
+
+
+Eval output 3:
+<|im_start|>user
+<image>
+Based on the observed features, what do you think is the correct histological classification of the tumor? A) Poorly differentiated keratinizing squamous cell carcinoma B) Moderately differentiated squamous cell carcinoma C) Well-differentiated squamous cell carcinoma D) Adenocarcinoma<|im_end|>
+<|im_start|>assistant
+A) Poorly differentiated keratinizing squamous cell carcinoma<|im_end|>
+
+

stage_2/multi_stage2_run_stage1_both/stage2a/iter_4096.pth/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2d5842f4d284346e6db2ad5997038da4344f71e98951f885bc230ec8ef6b56b7
+size 612302570

stage_2/multi_stage2_run_stage1_both/stage2a/iter_4096.pth/bf16_zero_pp_rank_1_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:12be85db2d909747eabc50fc0351de056ffca5b6bce768099cff1509e55b3976
+size 612302570

stage_2/multi_stage2_run_stage1_both/stage2a/iter_4096.pth/bf16_zero_pp_rank_2_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ac680c0853d9ad64f02e09c742329b769c6089f5cf8a3da97859304c236f4f6a
+size 612302570

stage_2/multi_stage2_run_stage1_both/stage2a/iter_4096.pth/bf16_zero_pp_rank_3_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3aa8684a6a4155aef482387caf4a86fe66eb2388b98c6444c6f119e857d216a6
+size 612302570

stage_2/multi_stage2_run_stage1_both/stage2a/iter_4096.pth/bf16_zero_pp_rank_4_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fd3846b8336309c1fb952f3fdf3d435c202b6a8c62b24b54797faa35161e5b17
+size 612302570

stage_2/multi_stage2_run_stage1_both/stage2a/iter_4096.pth/bf16_zero_pp_rank_5_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cd26f8cda580dd803ee810b47caaf7ed39e4de12815ffaf2b9d397132ed12527
+size 612302570

stage_2/multi_stage2_run_stage1_both/stage2a/iter_4096.pth/bf16_zero_pp_rank_6_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:debad177fe141a327e4fa1fd11aea62f61b3a7d06c31aa808838473342882085
+size 612302570

stage_2/multi_stage2_run_stage1_both/stage2a/iter_4096.pth/bf16_zero_pp_rank_7_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:caf4d6100a49c554e31933d3243a5e5b2f1366d8314e8bfa718aae3471d4917d
+size 612302570

stage_2/multi_stage2_run_stage1_both/stage2a/iter_4096.pth/mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:98315f72ba79f7808402bc6aa4be55a8a80cc87b6515dc5714449dc6d0e53db5
+size 816343368

stage_2/multi_stage2_run_stage1_both/stage2a/iter_7452.pth/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1cfe182e40920254b6b0fe68dd569cd5fd34cb701a64ba9a8501667b12f7d878
+size 612302570

stage_2/multi_stage2_run_stage1_both/stage2a/iter_7452.pth/bf16_zero_pp_rank_1_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7870fa18f29473e40c786a3a02167012e510c579682ea0146c60d747599b46da
+size 612302570

stage_2/multi_stage2_run_stage1_both/stage2a/iter_7452.pth/bf16_zero_pp_rank_2_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:165d183b68447ad9d4b5d2c6536820b7de1a0b90e732760775cb29e235752755
+size 612302570

stage_2/multi_stage2_run_stage1_both/stage2a/iter_7452.pth/bf16_zero_pp_rank_3_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d7d52fcbfca606673d5e4e3ec3adaf5ddc8f113b03c09b1cb8c7146e0a09ea2f
+size 612302570

stage_2/multi_stage2_run_stage1_both/stage2a/iter_7452.pth/bf16_zero_pp_rank_4_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:39cd739e387326c6a2b6917b94469b128f65a1a651812c0b7a5360183179d4dd
+size 612302570

stage_2/multi_stage2_run_stage1_both/stage2a/iter_7452.pth/bf16_zero_pp_rank_5_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:14529d0a241e44b3b84ff31847263c61eeddb5363cbb832400c3dcfaeb217393
+size 612302570

stage_2/multi_stage2_run_stage1_both/stage2a/iter_7452.pth/bf16_zero_pp_rank_6_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b9e65c2783921f2b01719c81e0d0bb25179af58996475a7d1da0a49f86183095
+size 612302570

stage_2/multi_stage2_run_stage1_both/stage2a/iter_7452.pth/bf16_zero_pp_rank_7_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6ba09e83a6d8404a6044c6d20e583494fb34379aea9e07b42d408d3cb1706cec
+size 612302570

stage_2/multi_stage2_run_stage1_both/stage2a/iter_7452.pth/mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c50da5073632e92ab4e32249ac8ebb32249e31179d63f6cbe16cdf6664176001
+size 816784840

stage_2/multi_stage2_run_stage1_both/stage2a/last_checkpoint

@@ -0,0 +1 @@
+/mnt/bn/yuxuanwang/meilong/code/projects/efficient_foundation_wsi_llava/curriculum_training/models/outputs/multi_stage2_run_stage1_both/stage2a/iter_7452.pth

stage_2/multi_stage2_run_stage1_both/stage2a/temp_config_stage_2a.py

@@ -0,0 +1,261 @@
+SYSTEM = ''
+accumulative_counts = 64
+batch_size = 1
+betas = (
+    0.9,
+    0.999,
+)
+bnb = dict(
+    bnb_4bit_compute_dtype='torch.bfloat16',
+    bnb_4bit_quant_type='nf4',
+    bnb_4bit_use_double_quant=True,
+    llm_int8_has_fp16_weight=False,
+    llm_int8_threshold=6.0,
+    load_in_4bit=True,
+    load_in_8bit=False,
+    type='transformers.BitsAndBytesConfig')
+custom_hooks = [
+    dict(
+        tokenizer=dict(
+            padding_side='right',
+            pretrained_model_name_or_path='Qwen/Qwen2.5-7B-Instruct',
+            trust_remote_code=True,
+            type='transformers.AutoTokenizer.from_pretrained'),
+        type='xtuner.engine.hooks.DatasetInfoHook'),
+    dict(
+        evaluation_images=[
+            '/mnt/bn/xudong-va/meilong/datasets/Token_Compression/skcm_224x224_b20_t15/h5_files/TCGA-EB-A5UN-06Z-00-DX1.h5',
+            '/mnt/bn/xudong-va/meilong/datasets/Token_Compression/skcm_224x224_b20_t15/h5_files/TCGA-EE-A3AG-01Z-00-DX1.h5',
+            '/mnt/bn/xudong-va/meilong/datasets/Token_Compression/lusc_224x224_b20_t15/h5_files/TCGA-NC-A5HP-01Z-00-DX1.h5',
+        ],
+        evaluation_inputs=[
+            'Are the tumor cells organized in a lobulated pattern within the slide?',
+            'Craft a comprehensive outline capturing the key findings of the pathology report based on the whole slide image.',
+            'Based on the observed features, what do you think is the correct histological classification of the tumor? A) Poorly differentiated keratinizing squamous cell carcinoma B) Moderately differentiated squamous cell carcinoma C) Well-differentiated squamous cell carcinoma D) Adenocarcinoma',
+        ],
+        every_n_iters=512,
+        prompt_template='xtuner.utils.PROMPT_TEMPLATE.qwen_chat',
+        system='',
+        tokenizer=dict(
+            padding_side='right',
+            pretrained_model_name_or_path='Qwen/Qwen2.5-7B-Instruct',
+            trust_remote_code=True,
+            type='transformers.AutoTokenizer.from_pretrained'),
+        type='xtuner.engine.hooks.EvaluateChatHookResampler'),
+    dict(type='xtuner.engine.hooks.ThroughputHook'),
+]
+data_path = '/mnt/bn/yuxuanwang/meilong/code/projects/efficient_foundation_wsi_llava/merged_dataset/stage2_tasks_plus_report.json'
+dataloader_num_workers = 10
+default_hooks = dict(
+    checkpoint=dict(
+        by_epoch=False,
+        interval=4096,
+        max_keep_ckpts=8,
+        type='mmengine.hooks.CheckpointHook'),
+    logger=dict(
+        interval=10,
+        log_metric_by_epoch=False,
+        type='mmengine.hooks.LoggerHook'),
+    param_scheduler=dict(type='mmengine.hooks.ParamSchedulerHook'),
+    sampler_seed=dict(type='mmengine.hooks.DistSamplerSeedHook'),
+    timer=dict(type='mmengine.hooks.IterTimerHook'))
+env_cfg = dict(
+    cudnn_benchmark=False,
+    dist_cfg=dict(backend='nccl'),
+    mp_cfg=dict(mp_start_method='fork', opencv_num_threads=0))
+evaluation_freq = 512
+evaluation_images = [
+    '/mnt/bn/xudong-va/meilong/datasets/Token_Compression/skcm_224x224_b20_t15/h5_files/TCGA-EB-A5UN-06Z-00-DX1.h5',
+    '/mnt/bn/xudong-va/meilong/datasets/Token_Compression/skcm_224x224_b20_t15/h5_files/TCGA-EE-A3AG-01Z-00-DX1.h5',
+    '/mnt/bn/xudong-va/meilong/datasets/Token_Compression/lusc_224x224_b20_t15/h5_files/TCGA-NC-A5HP-01Z-00-DX1.h5',
+]
+evaluation_inputs = [
+    'Are the tumor cells organized in a lobulated pattern within the slide?',
+    'Craft a comprehensive outline capturing the key findings of the pathology report based on the whole slide image.',
+    'Based on the observed features, what do you think is the correct histological classification of the tumor? A) Poorly differentiated keratinizing squamous cell carcinoma B) Moderately differentiated squamous cell carcinoma C) Well-differentiated squamous cell carcinoma D) Adenocarcinoma',
+]
+image_path_list = None
+launcher = 'pytorch'
+llava_dataset = dict(
+    data_path=
+    '/mnt/bn/yuxuanwang/meilong/code/projects/efficient_foundation_wsi_llava/merged_dataset/stage2_tasks_plus_report.json',
+    dataset_map_fn='xtuner.dataset.map_fns.llava_map_fn',
+    identifier='_224x224_b20_t15',
+    image_feature_prefix='/mnt/bn/xudong-va/meilong/datasets/Token_Compression',
+    image_feature_suffix='.h5',
+    image_folder='',
+    image_path_list=None,
+    max_length=15836,
+    pad_image_to_square=False,
+    per_image_length=10240,
+    sample_num=10240,
+    sample_strategy='linspace',
+    template_map_fn=dict(
+        template='xtuner.utils.PROMPT_TEMPLATE.qwen_chat',
+        type='xtuner.dataset.map_fns.template_map_fn_factory'),
+    tokenizer=dict(
+        padding_side='right',
+        pretrained_model_name_or_path='Qwen/Qwen2.5-7B-Instruct',
+        trust_remote_code=True,
+        type='transformers.AutoTokenizer.from_pretrained'),
+    type='xtuner.dataset.LLaVADataset',
+    unwanted_prefix_csv=
+    '/mnt/bn/yuxuanwang/meilong/code/projects/efficient_foundation_wsi_llava/merged_dataset/missing_slides3.csv'
+)
+llm_lora = dict(
+    bias='none',
+    lora_alpha=256,
+    lora_dropout=0.05,
+    r=128,
+    task_type='CAUSAL_LM',
+    type='peft.LoraConfig')
+llm_name_or_path = 'Qwen/Qwen2.5-7B-Instruct'
+load_from = None
+log_level = 'INFO'
+log_processor = dict(
+    by_epoch=False,
+    mean_pattern='.*(loss|time|data_time|grad_norm|tflops).*',
+    window_size=1)
+lr = 5e-06
+max_epochs = 2
+max_length = 15836
+max_norm = 1
+model = dict(
+    enable_token_merge=True,
+    freeze_llm=True,
+    freeze_mm_in_stage2=False,
+    llm=dict(
+        attn_implementation='flash_attention_2',
+        pretrained_model_name_or_path='Qwen/Qwen2.5-7B-Instruct',
+        quantization_config=dict(
+            bnb_4bit_compute_dtype='torch.bfloat16',
+            bnb_4bit_quant_type='nf4',
+            bnb_4bit_use_double_quant=True,
+            llm_int8_has_fp16_weight=False,
+            llm_int8_threshold=6.0,
+            load_in_4bit=True,
+            load_in_8bit=False,
+            type='transformers.BitsAndBytesConfig'),
+        torch_dtype='torch.bfloat16',
+        trust_remote_code=True,
+        type='transformers.AutoModelForCausalLM.from_pretrained'),
+    llm_lora=dict(
+        bias='none',
+        lora_alpha=256,
+        lora_dropout=0.05,
+        r=128,
+        task_type='CAUSAL_LM',
+        type='peft.LoraConfig'),
+    max_position_embeddings=None,
+    projector_pth=
+    '/mnt/bn/yuxuanwang/meilong/code/projects/efficient_foundation_wsi_llava/checkpoints/stage_1/token_merge_plus_resampler/stage1_qwen25_both_hf/projector/projector.safetensors',
+    resampler_num_latents=100,
+    resampler_pth=
+    '/mnt/bn/yuxuanwang/meilong/code/projects/efficient_foundation_wsi_llava/checkpoints/stage_1/token_merge_plus_resampler/stage1_qwen25_both_hf/resampler/resampler.safetensors',
+    token_merge_pth=
+    '/mnt/bn/yuxuanwang/meilong/code/projects/efficient_foundation_wsi_llava/checkpoints/stage_1/token_merge_plus_resampler/stage1_qwen25_both_hf/token_merger/merger.safetensors',
+    train_stage='2',
+    type='xtuner.model.llava_no_longnet_simple_sampler.LLaVAModel',
+    use_resampler=True)
+optim_type = 'torch.optim.AdamW'
+optim_wrapper = dict(
+    optimizer=dict(
+        betas=(
+            0.9,
+            0.999,
+        ),
+        lr=2e-06,
+        type='torch.optim.AdamW',
+        weight_decay=0.01),
+    paramwise_cfg=dict(
+        bias_decay_mult=0.0,
+        norm_decay_mult=0.0,
+        paramwise_cfg=dict(
+            custom_keys=dict({'^projector\.': dict(lr_mult=1.0)}))),
+    type='DeepSpeedOptimWrapper')
+param_scheduler = [
+    dict(
+        begin=0,
+        by_epoch=True,
+        convert_to_iter_based=True,
+        end=0.1,
+        start_factor=0.01,
+        type='mmengine.optim.LinearLR'),
+    dict(
+        begin=0.1,
+        by_epoch=True,
+        convert_to_iter_based=True,
+        end=2,
+        eta_min=0.0,
+        type='mmengine.optim.CosineAnnealingLR'),
+]
+per_image_length = 10240
+prompt_template = 'xtuner.utils.PROMPT_TEMPLATE.qwen_chat'
+randomness = dict(deterministic=False, seed=None)
+resume = False
+runner_type = 'FlexibleRunner'
+sample_type = 'wsi'
+save_steps = 4096
+save_total_limit = 8
+seed = 42
+strategy = dict(
+    config=dict(
+        bf16=dict(enabled=True),
+        fp16=dict(enabled=False, initial_scale_power=16),
+        gradient_accumulation_steps='auto',
+        gradient_clipping='auto',
+        train_micro_batch_size_per_gpu='auto',
+        zero_allow_untested_optimizer=True,
+        zero_force_ds_cpu_optimizer=False,
+        zero_optimization=dict(overlap_comm=False, stage=2)),
+    exclude_frozen_parameters=True,
+    gradient_accumulation_steps=64,
+    gradient_clipping=1,
+    sequence_parallel_size=1,
+    train_micro_batch_size_per_gpu=1,
+    type='xtuner.engine.DeepSpeedStrategy')
+tokenizer = dict(
+    padding_side='right',
+    pretrained_model_name_or_path='Qwen/Qwen2.5-7B-Instruct',
+    trust_remote_code=True,
+    type='transformers.AutoTokenizer.from_pretrained')
+train_cfg = dict(max_epochs=1, type='xtuner.engine.runner.TrainLoop')
+train_dataloader = dict(
+    batch_size=1,
+    collate_fn=dict(type='xtuner.dataset.collate_fns.default_collate_fn'),
+    dataset=dict(
+        data_path=
+        '/mnt/bn/yuxuanwang/meilong/code/projects/efficient_foundation_wsi_llava/curriculum_training/merged_dataset_curriculum/stage2a_easy.json',
+        dataset_map_fn='xtuner.dataset.map_fns.llava_map_fn',
+        identifier='_224x224_b20_t15',
+        image_feature_prefix=
+        '/mnt/bn/xudong-va/meilong/datasets/Token_Compression',
+        image_feature_suffix='.h5',
+        image_folder='',
+        image_path_list=None,
+        max_length=15836,
+        pad_image_to_square=False,
+        per_image_length=10240,
+        sample_num=10240,
+        sample_strategy='linspace',
+        template_map_fn=dict(
+            template='xtuner.utils.PROMPT_TEMPLATE.qwen_chat',
+            type='xtuner.dataset.map_fns.template_map_fn_factory'),
+        tokenizer=dict(
+            padding_side='right',
+            pretrained_model_name_or_path='Qwen/Qwen2.5-7B-Instruct',
+            trust_remote_code=True,
+            type='transformers.AutoTokenizer.from_pretrained'),
+        type='xtuner.dataset.LLaVADataset',
+        unwanted_prefix_csv=
+        '/mnt/bn/yuxuanwang/meilong/code/projects/efficient_foundation_wsi_llava/merged_dataset/missing_slides3.csv'
+    ),
+    num_workers=10,
+    persistent_workers=True,
+    pin_memory=True,
+    prefetch_factor=4,
+    sampler=dict(shuffle=True, type='mmengine.dataset.DefaultSampler'))
+visualizer = None
+warmup_ratio = 0.05
+weight_decay = 0.01
+work_dir = '/mnt/bn/yuxuanwang/meilong/code/projects/efficient_foundation_wsi_llava/curriculum_training/models/outputs/multi_stage2_run_stage1_both/stage2a'

stage_2/multi_stage2_run_stage1_both/stage2a/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)

stage_2/multi_stage2_run_stage1_both/stage2b/20250925_230352/vis_data/config.py

@@ -0,0 +1,261 @@
+SYSTEM = ''
+accumulative_counts = 64
+batch_size = 1
+betas = (
+    0.9,
+    0.999,
+)
+bnb = dict(
+    bnb_4bit_compute_dtype='torch.bfloat16',
+    bnb_4bit_quant_type='nf4',
+    bnb_4bit_use_double_quant=True,
+    llm_int8_has_fp16_weight=False,
+    llm_int8_threshold=6.0,
+    load_in_4bit=True,
+    load_in_8bit=False,
+    type='transformers.BitsAndBytesConfig')
+custom_hooks = [
+    dict(
+        tokenizer=dict(
+            padding_side='right',
+            pretrained_model_name_or_path='Qwen/Qwen2.5-7B-Instruct',
+            trust_remote_code=True,
+            type='transformers.AutoTokenizer.from_pretrained'),
+        type='xtuner.engine.hooks.DatasetInfoHook'),
+    dict(
+        evaluation_images=[
+            '/mnt/bn/xudong-va/meilong/datasets/Token_Compression/skcm_224x224_b20_t15/h5_files/TCGA-EB-A5UN-06Z-00-DX1.h5',
+            '/mnt/bn/xudong-va/meilong/datasets/Token_Compression/skcm_224x224_b20_t15/h5_files/TCGA-EE-A3AG-01Z-00-DX1.h5',
+            '/mnt/bn/xudong-va/meilong/datasets/Token_Compression/lusc_224x224_b20_t15/h5_files/TCGA-NC-A5HP-01Z-00-DX1.h5',
+        ],
+        evaluation_inputs=[
+            'Are the tumor cells organized in a lobulated pattern within the slide?',
+            'Craft a comprehensive outline capturing the key findings of the pathology report based on the whole slide image.',
+            'Based on the observed features, what do you think is the correct histological classification of the tumor? A) Poorly differentiated keratinizing squamous cell carcinoma B) Moderately differentiated squamous cell carcinoma C) Well-differentiated squamous cell carcinoma D) Adenocarcinoma',
+        ],
+        every_n_iters=512,
+        prompt_template='xtuner.utils.PROMPT_TEMPLATE.qwen_chat',
+        system='',
+        tokenizer=dict(
+            padding_side='right',
+            pretrained_model_name_or_path='Qwen/Qwen2.5-7B-Instruct',
+            trust_remote_code=True,
+            type='transformers.AutoTokenizer.from_pretrained'),
+        type='xtuner.engine.hooks.EvaluateChatHookResampler'),
+    dict(type='xtuner.engine.hooks.ThroughputHook'),
+]
+data_path = '/mnt/bn/yuxuanwang/meilong/code/projects/efficient_foundation_wsi_llava/merged_dataset/stage2_tasks_plus_report.json'
+dataloader_num_workers = 10
+default_hooks = dict(
+    checkpoint=dict(
+        by_epoch=False,
+        interval=4096,
+        max_keep_ckpts=8,
+        type='mmengine.hooks.CheckpointHook'),
+    logger=dict(
+        interval=10,
+        log_metric_by_epoch=False,
+        type='mmengine.hooks.LoggerHook'),
+    param_scheduler=dict(type='mmengine.hooks.ParamSchedulerHook'),
+    sampler_seed=dict(type='mmengine.hooks.DistSamplerSeedHook'),
+    timer=dict(type='mmengine.hooks.IterTimerHook'))
+env_cfg = dict(
+    cudnn_benchmark=False,
+    dist_cfg=dict(backend='nccl'),
+    mp_cfg=dict(mp_start_method='fork', opencv_num_threads=0))
+evaluation_freq = 512
+evaluation_images = [
+    '/mnt/bn/xudong-va/meilong/datasets/Token_Compression/skcm_224x224_b20_t15/h5_files/TCGA-EB-A5UN-06Z-00-DX1.h5',
+    '/mnt/bn/xudong-va/meilong/datasets/Token_Compression/skcm_224x224_b20_t15/h5_files/TCGA-EE-A3AG-01Z-00-DX1.h5',
+    '/mnt/bn/xudong-va/meilong/datasets/Token_Compression/lusc_224x224_b20_t15/h5_files/TCGA-NC-A5HP-01Z-00-DX1.h5',
+]
+evaluation_inputs = [
+    'Are the tumor cells organized in a lobulated pattern within the slide?',
+    'Craft a comprehensive outline capturing the key findings of the pathology report based on the whole slide image.',
+    'Based on the observed features, what do you think is the correct histological classification of the tumor? A) Poorly differentiated keratinizing squamous cell carcinoma B) Moderately differentiated squamous cell carcinoma C) Well-differentiated squamous cell carcinoma D) Adenocarcinoma',
+]
+image_path_list = None
+launcher = 'pytorch'
+llava_dataset = dict(
+    data_path=
+    '/mnt/bn/yuxuanwang/meilong/code/projects/efficient_foundation_wsi_llava/merged_dataset/stage2_tasks_plus_report.json',
+    dataset_map_fn='xtuner.dataset.map_fns.llava_map_fn',
+    identifier='_224x224_b20_t15',
+    image_feature_prefix='/mnt/bn/xudong-va/meilong/datasets/Token_Compression',
+    image_feature_suffix='.h5',
+    image_folder='',
+    image_path_list=None,
+    max_length=15836,
+    pad_image_to_square=False,
+    per_image_length=10240,
+    sample_num=10240,
+    sample_strategy='linspace',
+    template_map_fn=dict(
+        template='xtuner.utils.PROMPT_TEMPLATE.qwen_chat',
+        type='xtuner.dataset.map_fns.template_map_fn_factory'),
+    tokenizer=dict(
+        padding_side='right',
+        pretrained_model_name_or_path='Qwen/Qwen2.5-7B-Instruct',
+        trust_remote_code=True,
+        type='transformers.AutoTokenizer.from_pretrained'),
+    type='xtuner.dataset.LLaVADataset',
+    unwanted_prefix_csv=
+    '/mnt/bn/yuxuanwang/meilong/code/projects/efficient_foundation_wsi_llava/merged_dataset/missing_slides3.csv'
+)
+llm_lora = dict(
+    bias='none',
+    lora_alpha=256,
+    lora_dropout=0.05,
+    r=128,
+    task_type='CAUSAL_LM',
+    type='peft.LoraConfig')
+llm_name_or_path = 'Qwen/Qwen2.5-7B-Instruct'
+load_from = '/mnt/bn/yuxuanwang/meilong/code/projects/efficient_foundation_wsi_llava/curriculum_training/models/outputs/multi_stage2_run_stage1_both/stage2a/iter_7452.pth'
+log_level = 'INFO'
+log_processor = dict(
+    by_epoch=False,
+    mean_pattern='.*(loss|time|data_time|grad_norm|tflops).*',
+    window_size=1)
+lr = 5e-06
+max_epochs = 2
+max_length = 15836
+max_norm = 1
+model = dict(
+    enable_token_merge=True,
+    freeze_llm=True,
+    freeze_mm_in_stage2=False,
+    llm=dict(
+        attn_implementation='flash_attention_2',
+        pretrained_model_name_or_path='Qwen/Qwen2.5-7B-Instruct',
+        quantization_config=dict(
+            bnb_4bit_compute_dtype='torch.bfloat16',
+            bnb_4bit_quant_type='nf4',
+            bnb_4bit_use_double_quant=True,
+            llm_int8_has_fp16_weight=False,
+            llm_int8_threshold=6.0,
+            load_in_4bit=True,
+            load_in_8bit=False,
+            type='transformers.BitsAndBytesConfig'),
+        torch_dtype='torch.bfloat16',
+        trust_remote_code=True,
+        type='transformers.AutoModelForCausalLM.from_pretrained'),
+    llm_lora=dict(
+        bias='none',
+        lora_alpha=256,
+        lora_dropout=0.05,
+        r=128,
+        task_type='CAUSAL_LM',
+        type='peft.LoraConfig'),
+    max_position_embeddings=None,
+    projector_pth=
+    '/mnt/bn/yuxuanwang/meilong/code/projects/efficient_foundation_wsi_llava/checkpoints/stage_1/token_merge_plus_resampler/stage1_qwen25_both_hf/projector/projector.safetensors',
+    resampler_num_latents=100,
+    resampler_pth=
+    '/mnt/bn/yuxuanwang/meilong/code/projects/efficient_foundation_wsi_llava/checkpoints/stage_1/token_merge_plus_resampler/stage1_qwen25_both_hf/resampler/resampler.safetensors',
+    token_merge_pth=
+    '/mnt/bn/yuxuanwang/meilong/code/projects/efficient_foundation_wsi_llava/checkpoints/stage_1/token_merge_plus_resampler/stage1_qwen25_both_hf/token_merger/merger.safetensors',
+    train_stage='2',
+    type='xtuner.model.llava_no_longnet_simple_sampler.LLaVAModel',
+    use_resampler=True)
+optim_type = 'torch.optim.AdamW'
+optim_wrapper = dict(
+    optimizer=dict(
+        betas=(
+            0.9,
+            0.999,
+        ),
+        lr=2e-05,
+        type='torch.optim.AdamW',
+        weight_decay=0.01),
+    paramwise_cfg=dict(
+        bias_decay_mult=0.0,
+        norm_decay_mult=0.0,
+        paramwise_cfg=dict(
+            custom_keys=dict({'^projector\.': dict(lr_mult=1.0)}))),
+    type='DeepSpeedOptimWrapper')
+param_scheduler = [
+    dict(
+        begin=0,
+        by_epoch=True,
+        convert_to_iter_based=True,
+        end=0.1,
+        start_factor=0.01,
+        type='mmengine.optim.LinearLR'),
+    dict(
+        begin=0.1,
+        by_epoch=True,
+        convert_to_iter_based=True,
+        end=2,
+        eta_min=0.0,
+        type='mmengine.optim.CosineAnnealingLR'),
+]
+per_image_length = 10240
+prompt_template = 'xtuner.utils.PROMPT_TEMPLATE.qwen_chat'
+randomness = dict(deterministic=False, seed=None)
+resume = False
+runner_type = 'FlexibleRunner'
+sample_type = 'wsi'
+save_steps = 4096
+save_total_limit = 8
+seed = 42
+strategy = dict(
+    config=dict(
+        bf16=dict(enabled=True),
+        fp16=dict(enabled=False, initial_scale_power=16),
+        gradient_accumulation_steps='auto',
+        gradient_clipping='auto',
+        train_micro_batch_size_per_gpu='auto',
+        zero_allow_untested_optimizer=True,
+        zero_force_ds_cpu_optimizer=False,
+        zero_optimization=dict(overlap_comm=False, stage=2)),
+    exclude_frozen_parameters=True,
+    gradient_accumulation_steps=64,
+    gradient_clipping=1,
+    sequence_parallel_size=1,
+    train_micro_batch_size_per_gpu=1,
+    type='xtuner.engine.DeepSpeedStrategy')
+tokenizer = dict(
+    padding_side='right',
+    pretrained_model_name_or_path='Qwen/Qwen2.5-7B-Instruct',
+    trust_remote_code=True,
+    type='transformers.AutoTokenizer.from_pretrained')
+train_cfg = dict(max_epochs=1, type='xtuner.engine.runner.TrainLoop')
+train_dataloader = dict(
+    batch_size=1,
+    collate_fn=dict(type='xtuner.dataset.collate_fns.default_collate_fn'),
+    dataset=dict(
+        data_path=
+        '/mnt/bn/yuxuanwang/meilong/code/projects/efficient_foundation_wsi_llava/curriculum_training/merged_dataset_curriculum/stage2b_medium.json',
+        dataset_map_fn='xtuner.dataset.map_fns.llava_map_fn',
+        identifier='_224x224_b20_t15',
+        image_feature_prefix=
+        '/mnt/bn/xudong-va/meilong/datasets/Token_Compression',
+        image_feature_suffix='.h5',
+        image_folder='',
+        image_path_list=None,
+        max_length=15836,
+        pad_image_to_square=False,
+        per_image_length=10240,
+        sample_num=10240,
+        sample_strategy='linspace',
+        template_map_fn=dict(
+            template='xtuner.utils.PROMPT_TEMPLATE.qwen_chat',
+            type='xtuner.dataset.map_fns.template_map_fn_factory'),
+        tokenizer=dict(
+            padding_side='right',
+            pretrained_model_name_or_path='Qwen/Qwen2.5-7B-Instruct',
+            trust_remote_code=True,
+            type='transformers.AutoTokenizer.from_pretrained'),
+        type='xtuner.dataset.LLaVADataset',
+        unwanted_prefix_csv=
+        '/mnt/bn/yuxuanwang/meilong/code/projects/efficient_foundation_wsi_llava/merged_dataset/missing_slides3.csv'
+    ),
+    num_workers=10,
+    persistent_workers=True,
+    pin_memory=True,
+    prefetch_factor=4,
+    sampler=dict(shuffle=True, type='mmengine.dataset.DefaultSampler'))
+visualizer = None
+warmup_ratio = 0.05
+weight_decay = 0.01
+work_dir = '/mnt/bn/yuxuanwang/meilong/code/projects/efficient_foundation_wsi_llava/curriculum_training/models/outputs/multi_stage2_run_stage1_both/stage2b'

stage_2/multi_stage2_run_stage1_both/stage2b/20250925_230352/vis_data/eval_outputs_iter_4095.txt

@@ -0,0 +1,24 @@
+Eval output 1:
+<|im_start|>user
+<image>
+Are the tumor cells organized in a lobulated pattern within the slide?<|im_end|>
+<|im_start|>assistant
+Yes, the tumor cells are organized in a lobulated pattern. This pattern is characterized by the formation of distinct lobules, each containing a cluster of tumor cells. The lobules are separated by fibrous stroma, contributing to the overall architectural organization of the tumor.<|im_end|>
+
+
+Eval output 2:
+<|im_start|>user
+<image>
+Craft a comprehensive outline capturing the key findings of the pathology report based on the whole slide image.<|im_end|>
+<|im_start|>assistant
+The pathology report indicates a poorly differentiated adenocarcinoma with significant architectural and nuclear atypia. The tumor exhibits a high degree of pleomorphism, with marked nuclear atypia and prominent nucleoli. The presence of microvascular invasion and lymphovascular invasion is noted, along with the involvement of the perineural space. The tumor is characterized by a high mitotic rate, with a high Ki-67 proliferation index, reflecting a high-grade malignancy. The absence of specific molecular markers such as EGFR, HER2, and p53 further supports the aggressive nature of the tumor.<|im_end|>
+
+
+Eval output 3:
+<|im_start|>user
+<image>
+Based on the observed features, what do you think is the correct histological classification of the tumor? A) Poorly differentiated keratinizing squamous cell carcinoma B) Moderately differentiated squamous cell carcinoma C) Well-differentiated squamous cell carcinoma D) Adenocarcinoma<|im_end|>
+<|im_start|>assistant
+C) Well-differentiated squamous cell carcinoma<|im_end|>
+
+

stage_2/multi_stage2_run_stage1_both/stage2b/20250925_230352/vis_data/eval_outputs_iter_4602.txt

@@ -0,0 +1,24 @@
+Eval output 1:
+<|im_start|>user
+<image>
+Are the tumor cells organized in a lobulated pattern within the slide?<|im_end|>
+<|im_start|>assistant
+Yes, the tumor cells are organized in a lobulated pattern, which is a characteristic feature of papillary thyroid carcinoma. This pattern is evident in the slide, reflecting the glandular architecture typical of this type of cancer.<|im_end|>
+
+
+Eval output 2:
+<|im_start|>user
+<image>
+Craft a comprehensive outline capturing the key findings of the pathology report based on the whole slide image.<|im_end|>
+<|im_start|>assistant
+The pathology report indicates a poorly differentiated adenocarcinoma with significant architectural and nuclear atypia. The tumor is characterized by glandular structures with irregular shapes and sizes, and there is marked nuclear pleomorphism. The presence of microvascular invasion and lymphovascular invasion is noted, suggesting aggressive tumor behavior. The tumor cells exhibit high mitotic activity, indicating a high proliferation rate. The absence of lymphovascular invasion and perineural invasion is a positive finding, but the presence of vascular invasion is concerning. The tumor is poorly differentiated, reflecting a high degree of malignancy. The absence of lymphovascular invasion and perineural invasion is a positive finding, but the presence of vascular invasion is concerning.<|im_end|>
+
+
+Eval output 3:
+<|im_start|>user
+<image>
+Based on the observed features, what do you think is the correct histological classification of the tumor? A) Poorly differentiated keratinizing squamous cell carcinoma B) Moderately differentiated squamous cell carcinoma C) Well-differentiated squamous cell carcinoma D) Adenocarcinoma<|im_end|>
+<|im_start|>assistant
+C) Well-differentiated squamous cell carcinoma<|im_end|>
+
+

stage_2/multi_stage2_run_stage1_both/stage2b/iter_4096.pth/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1a28678eb3dc6429dcabf65d7724b98cc43f6225d47b10e57d31cf5dce33a0ec
+size 612302570

stage_2/multi_stage2_run_stage1_both/stage2b/iter_4096.pth/bf16_zero_pp_rank_1_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4ec4d8aa103c7b7b58786013531cd939ddb49125e28a90ee791cff3c115dd4cb
+size 612302570

stage_2/multi_stage2_run_stage1_both/stage2b/iter_4096.pth/bf16_zero_pp_rank_2_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:62ff35fa2d294789df71fb685891df93ef00350afaabdd1ead1ff8b2550f2e1e
+size 612302570

stage_2/multi_stage2_run_stage1_both/stage2b/iter_4096.pth/bf16_zero_pp_rank_3_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:478c62f01e77c8cd856001abf0ffddb4b6ffed8b342c5e6780da31797b9b27dc
+size 612302570

stage_2/multi_stage2_run_stage1_both/stage2b/iter_4096.pth/bf16_zero_pp_rank_4_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d35ac2f6cce58392b20ca9d8ce85f6d2d7a9992851916ad543ee252c9632648a
+size 612302570

stage_2/multi_stage2_run_stage1_both/stage2b/iter_4096.pth/bf16_zero_pp_rank_5_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bbd90ff6ad347780f3295a3d20b8a7021d92a0c16f3297ee038fdb122fd98c3d
+size 612302570

stage_2/multi_stage2_run_stage1_both/stage2b/iter_4096.pth/bf16_zero_pp_rank_6_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:21da77a582762da9cf57e5de6421f48157dfcb190a44db580ace941aea1c7dc4
+size 612302570

stage_2/multi_stage2_run_stage1_both/stage2b/iter_4096.pth/bf16_zero_pp_rank_7_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ef6253fcfdf1047ca70c85c5c59ecc46f19971081742bc3d15727a7f18eb80d4
+size 612302570

stage_2/multi_stage2_run_stage1_both/stage2b/iter_4096.pth/mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9cec4603b1782bace539ad6dcfc1a5a70cf59177b193fa4b8bb850ed304003f7
+size 816345224

stage_2/multi_stage2_run_stage1_both/stage2b/iter_4603.pth/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:59445f3060bcf0b356c4db24c03801cdf3a495d7f8e7876e3ff19fb4eefad7df
+size 612302570

stage_2/multi_stage2_run_stage1_both/stage2b/iter_4603.pth/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:b11b09176a772f17783f976b80df5b65fc844be287a16049b4ddb16fb0070a68
+size 612302570

stage_2/multi_stage2_run_stage1_both/stage2b/iter_4603.pth/bf16_zero_pp_rank_2_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9360d83af5a87c9e82d2c5a7bec96675f95172b647e79d2c64caf1f7e3a2c632
+size 612302570

stage_2/multi_stage2_run_stage1_both/stage2b/iter_4603.pth/bf16_zero_pp_rank_3_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0b6decd45e95c679ff940c5ee7676e06e7fc313d589c7bd65af1395df1562be8
+size 612302570

stage_2/multi_stage2_run_stage1_both/stage2b/iter_4603.pth/bf16_zero_pp_rank_4_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e02610aa01d72c8d5966b55db7a809573d33195c8eee3ac93a67dfd0bda57b0a
+size 612302570

stage_2/multi_stage2_run_stage1_both/stage2b/iter_4603.pth/bf16_zero_pp_rank_5_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:75e651db681814b6e7beb541136a6a3273d60006d2ef984757aed1181831cf6a
+size 612302570

stage_2/multi_stage2_run_stage1_both/stage2b/iter_4603.pth/bf16_zero_pp_rank_6_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:28b1a6c920baaf3c8c98e30d594957c76f0437b461f2c181234e95f2aed5ead9
+size 612302570

stage_2/multi_stage2_run_stage1_both/stage2b/iter_4603.pth/bf16_zero_pp_rank_7_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:452055e852d1d0d831b6ddc97a19768823cd8a2f5037bc36d9850374ad4e9ad1
+size 612302570