Upload folder using huggingface_hub

ckpt/llavanext-google_siglip-so400m-patch14-384-Qwen_Qwen2-7B-Instruct-cut3r_lora_diff_lr_03_12/checkpoint-707/adapter_config.json

@@ -0,0 +1,215 @@
+{
+  "auto_mapping": null,
+  "base_model_name_or_path": "checkpoints/LLaVA-Video-7B-Qwen2",
+  "bias": "none",
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "lora_alpha": 256,
+  "lora_dropout": 0.05,
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "r": 128,
+  "revision": null,
+  "target_modules": [
+    "model.layers.20.mlp.down_proj",
+    "model.layers.23.mlp.up_proj",
+    "model.layers.13.self_attn.k_proj",
+    "model.layers.22.self_attn.q_proj",
+    "model.layers.23.mlp.down_proj",
+    "model.layers.13.self_attn.o_proj",
+    "model.layers.3.self_attn.o_proj",
+    "model.layers.4.self_attn.k_proj",
+    "model.layers.11.mlp.down_proj",
+    "model.layers.6.mlp.gate_proj",
+    "model.layers.17.mlp.up_proj",
+    "model.layers.25.self_attn.k_proj",
+    "model.layers.16.mlp.gate_proj",
+    "model.layers.18.self_attn.o_proj",
+    "model.layers.18.mlp.up_proj",
+    "model.layers.2.self_attn.k_proj",
+    "model.layers.1.self_attn.o_proj",
+    "model.layers.7.self_attn.o_proj",
+    "model.layers.3.mlp.up_proj",
+    "model.layers.2.self_attn.q_proj",
+    "model.layers.5.mlp.down_proj",
+    "model.layers.22.mlp.down_proj",
+    "model.layers.10.self_attn.q_proj",
+    "model.layers.11.self_attn.k_proj",
+    "model.layers.4.self_attn.q_proj",
+    "model.layers.18.mlp.down_proj",
+    "model.layers.23.mlp.gate_proj",
+    "model.layers.15.self_attn.v_proj",
+    "model.layers.4.mlp.up_proj",
+    "model.layers.22.mlp.gate_proj",
+    "model.layers.24.self_attn.q_proj",
+    "model.layers.17.self_attn.k_proj",
+    "model.layers.0.mlp.gate_proj",
+    "model.layers.17.mlp.down_proj",
+    "model.layers.25.mlp.up_proj",
+    "model.layers.3.self_attn.k_proj",
+    "model.layers.9.self_attn.k_proj",
+    "model.layers.12.mlp.down_proj",
+    "model.layers.16.mlp.down_proj",
+    "model.layers.7.self_attn.k_proj",
+    "model.layers.14.self_attn.o_proj",
+    "model.layers.25.self_attn.q_proj",
+    "model.layers.8.self_attn.k_proj",
+    "model.layers.24.self_attn.o_proj",
+    "model.layers.17.self_attn.o_proj",
+    "model.layers.2.self_attn.v_proj",
+    "model.layers.20.self_attn.o_proj",
+    "model.layers.3.mlp.gate_proj",
+    "model.layers.0.self_attn.k_proj",
+    "model.layers.10.self_attn.o_proj",
+    "model.layers.12.self_attn.v_proj",
+    "model.layers.15.mlp.gate_proj",
+    "model.layers.17.self_attn.v_proj",
+    "model.layers.20.mlp.gate_proj",
+    "model.layers.26.mlp.gate_proj",
+    "model.layers.27.self_attn.o_proj",
+    "model.layers.23.self_attn.v_proj",
+    "model.layers.11.self_attn.q_proj",
+    "model.layers.6.mlp.down_proj",
+    "model.layers.9.self_attn.o_proj",
+    "model.layers.12.mlp.up_proj",
+    "model.layers.8.mlp.gate_proj",
+    "model.layers.14.mlp.up_proj",
+    "model.layers.19.self_attn.v_proj",
+    "model.layers.23.self_attn.o_proj",
+    "model.layers.0.self_attn.q_proj",
+    "model.layers.19.self_attn.o_proj",
+    "model.layers.13.mlp.up_proj",
+    "model.layers.19.mlp.gate_proj",
+    "model.layers.10.mlp.up_proj",
+    "model.layers.9.self_attn.q_proj",
+    "model.layers.4.mlp.down_proj",
+    "model.layers.14.self_attn.v_proj",
+    "model.layers.1.self_attn.k_proj",
+    "model.layers.15.self_attn.q_proj",
+    "model.layers.20.self_attn.k_proj",
+    "model.layers.24.self_attn.k_proj",
+    "model.layers.16.mlp.up_proj",
+    "model.layers.19.self_attn.q_proj",
+    "model.layers.27.self_attn.q_proj",
+    "model.layers.4.mlp.gate_proj",
+    "model.layers.8.mlp.up_proj",
+    "model.layers.5.self_attn.k_proj",
+    "model.layers.0.mlp.up_proj",
+    "model.layers.5.mlp.up_proj",
+    "model.layers.24.mlp.down_proj",
+    "model.layers.21.mlp.up_proj",
+    "model.layers.27.self_attn.v_proj",
+    "model.layers.22.self_attn.k_proj",
+    "model.layers.16.self_attn.o_proj",
+    "model.layers.25.mlp.down_proj",
+    "model.layers.2.mlp.up_proj",
+    "model.layers.10.self_attn.k_proj",
+    "model.layers.16.self_attn.q_proj",
+    "model.layers.23.self_attn.k_proj",
+    "model.layers.1.mlp.up_proj",
+    "model.layers.26.self_attn.v_proj",
+    "model.layers.26.self_attn.q_proj",
+    "model.layers.11.self_attn.o_proj",
+    "model.layers.2.mlp.down_proj",
+    "model.layers.25.self_attn.v_proj",
+    "model.layers.4.self_attn.o_proj",
+    "model.layers.27.mlp.up_proj",
+    "model.layers.7.mlp.gate_proj",
+    "model.layers.7.self_attn.q_proj",
+    "model.layers.14.self_attn.k_proj",
+    "model.layers.1.mlp.down_proj",
+    "model.layers.26.mlp.down_proj",
+    "model.layers.0.self_attn.v_proj",
+    "model.layers.27.mlp.gate_proj",
+    "model.layers.19.self_attn.k_proj",
+    "model.layers.21.mlp.gate_proj",
+    "model.layers.11.mlp.up_proj",
+    "model.layers.24.self_attn.v_proj",
+    "model.layers.24.mlp.gate_proj",
+    "model.layers.18.mlp.gate_proj",
+    "model.layers.3.self_attn.q_proj",
+    "model.layers.6.self_attn.q_proj",
+    "model.layers.21.self_attn.q_proj",
+    "model.layers.14.mlp.gate_proj",
+    "model.layers.20.mlp.up_proj",
+    "model.layers.21.self_attn.v_proj",
+    "model.layers.6.self_attn.o_proj",
+    "model.layers.27.self_attn.k_proj",
+    "model.layers.21.mlp.down_proj",
+    "model.layers.12.self_attn.q_proj",
+    "model.layers.26.self_attn.o_proj",
+    "model.layers.1.mlp.gate_proj",
+    "model.layers.4.self_attn.v_proj",
+    "model.layers.5.mlp.gate_proj",
+    "model.layers.18.self_attn.v_proj",
+    "model.layers.8.self_attn.q_proj",
+    "model.layers.12.self_attn.k_proj",
+    "model.layers.18.self_attn.k_proj",
+    "model.layers.6.mlp.up_proj",
+    "model.layers.15.self_attn.o_proj",
+    "model.layers.5.self_attn.o_proj",
+    "model.layers.6.self_attn.k_proj",
+    "model.layers.25.self_attn.o_proj",
+    "model.layers.24.mlp.up_proj",
+    "model.layers.8.mlp.down_proj",
+    "model.layers.9.mlp.up_proj",
+    "model.layers.0.mlp.down_proj",
+    "model.layers.0.self_attn.o_proj",
+    "model.layers.22.self_attn.o_proj",
+    "model.layers.7.mlp.up_proj",
+    "model.layers.21.self_attn.k_proj",
+    "model.layers.13.mlp.down_proj",
+    "model.layers.2.self_attn.o_proj",
+    "model.layers.6.self_attn.v_proj",
+    "model.layers.15.self_attn.k_proj",
+    "model.layers.9.mlp.gate_proj",
+    "model.layers.13.self_attn.v_proj",
+    "model.layers.13.mlp.gate_proj",
+    "model.layers.8.self_attn.o_proj",
+    "model.layers.1.self_attn.v_proj",
+    "model.layers.11.self_attn.v_proj",
+    "model.layers.7.self_attn.v_proj",
+    "model.layers.3.mlp.down_proj",
+    "model.layers.15.mlp.down_proj",
+    "model.layers.14.self_attn.q_proj",
+    "model.layers.21.self_attn.o_proj",
+    "model.layers.23.self_attn.q_proj",
+    "model.layers.26.mlp.up_proj",
+    "model.layers.2.mlp.gate_proj",
+    "model.layers.20.self_attn.q_proj",
+    "model.layers.14.mlp.down_proj",
+    "model.layers.18.self_attn.q_proj",
+    "model.layers.1.self_attn.q_proj",
+    "model.layers.11.mlp.gate_proj",
+    "model.layers.7.mlp.down_proj",
+    "model.layers.8.self_attn.v_proj",
+    "model.layers.5.self_attn.v_proj",
+    "model.layers.10.mlp.down_proj",
+    "model.layers.12.self_attn.o_proj",
+    "model.layers.15.mlp.up_proj",
+    "model.layers.19.mlp.down_proj",
+    "model.layers.20.self_attn.v_proj",
+    "model.layers.9.mlp.down_proj",
+    "model.layers.10.self_attn.v_proj",
+    "model.layers.26.self_attn.k_proj",
+    "model.layers.27.mlp.down_proj",
+    "model.layers.10.mlp.gate_proj",
+    "model.layers.22.mlp.up_proj",
+    "model.layers.16.self_attn.k_proj",
+    "model.layers.12.mlp.gate_proj",
+    "model.layers.25.mlp.gate_proj",
+    "model.layers.5.self_attn.q_proj",
+    "model.layers.17.self_attn.q_proj",
+    "model.layers.22.self_attn.v_proj",
+    "model.layers.17.mlp.gate_proj",
+    "model.layers.3.self_attn.v_proj",
+    "model.layers.13.self_attn.q_proj",
+    "model.layers.16.self_attn.v_proj",
+    "model.layers.9.self_attn.v_proj",
+    "model.layers.19.mlp.up_proj"
+  ],
+  "task_type": "CAUSAL_LM"
+}

ckpt/llavanext-google_siglip-so400m-patch14-384-Qwen_Qwen2-7B-Instruct-cut3r_lora_diff_lr_03_12/checkpoint-707/adapter_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d530eb1037b1788250fe3a2ffdb4090867186534c151ea0feab938272eac1016
+size 646061802

ckpt/llavanext-google_siglip-so400m-patch14-384-Qwen_Qwen2-7B-Instruct-cut3r_lora_diff_lr_03_12/checkpoint-707/config.json

@@ -0,0 +1,232 @@
+{
+  "_name_or_path": "checkpoints/LLaVA-Video-7B-Qwen2",
+  "add_faster_video": false,
+  "add_time_instruction": true,
+  "architectures": [
+    "LlavaQwenForCausalLM"
+  ],
+  "attention_dropout": 0.0,
+  "bos_token_id": 151643,
+  "eos_token_id": 151645,
+  "faster_token_stride": 10,
+  "force_sample": true,
+  "freeze_mm_mlp_adapter": false,
+  "freeze_mm_vision_resampler": false,
+  "fusion_block": "cross_attention",
+  "fusion_block_lr": 0.002,
+  "hidden_act": "silu",
+  "hidden_size": 3584,
+  "ignore_index": -100,
+  "image_aspect_ratio": "anyres_max_9",
+  "image_crop_resolution": null,
+  "image_grid_pinpoints": [
+    [
+      384,
+      384
+    ],
+    [
+      384,
+      768
+    ],
+    [
+      384,
+      1152
+    ],
+    [
+      384,
+      1536
+    ],
+    [
+      384,
+      1920
+    ],
+    [
+      384,
+      2304
+    ],
+    [
+      768,
+      384
+    ],
+    [
+      768,
+      768
+    ],
+    [
+      768,
+      1152
+    ],
+    [
+      768,
+      1536
+    ],
+    [
+      768,
+      1920
+    ],
+    [
+      768,
+      2304
+    ],
+    [
+      1152,
+      384
+    ],
+    [
+      1152,
+      768
+    ],
+    [
+      1152,
+      1152
+    ],
+    [
+      1152,
+      1536
+    ],
+    [
+      1152,
+      1920
+    ],
+    [
+      1152,
+      2304
+    ],
+    [
+      1536,
+      384
+    ],
+    [
+      1536,
+      768
+    ],
+    [
+      1536,
+      1152
+    ],
+    [
+      1536,
+      1536
+    ],
+    [
+      1536,
+      1920
+    ],
+    [
+      1536,
+      2304
+    ],
+    [
+      1920,
+      384
+    ],
+    [
+      1920,
+      768
+    ],
+    [
+      1920,
+      1152
+    ],
+    [
+      1920,
+      1536
+    ],
+    [
+      1920,
+      1920
+    ],
+    [
+      1920,
+      2304
+    ],
+    [
+      2304,
+      384
+    ],
+    [
+      2304,
+      768
+    ],
+    [
+      2304,
+      1152
+    ],
+    [
+      2304,
+      1536
+    ],
+    [
+      2304,
+      1920
+    ],
+    [
+      2304,
+      2304
+    ]
+  ],
+  "image_split_resolution": null,
+  "image_token_index": 151646,
+  "initializer_range": 0.02,
+  "intermediate_size": 18944,
+  "max_position_embeddings": 32768,
+  "max_window_layers": 28,
+  "mm_hidden_size": 1152,
+  "mm_newline_position": "grid",
+  "mm_patch_merge_type": "spatial_unpad",
+  "mm_projector_lr": null,
+  "mm_projector_type": "mlp2x_gelu",
+  "mm_resampler_type": null,
+  "mm_spatial_pool_mode": "bilinear",
+  "mm_spatial_pool_stride": 2,
+  "mm_spatial_tower": "cut3r",
+  "mm_tunable_parts": "mm_vision_tower,mm_mlp_adapter,mm_language_model",
+  "mm_use_im_patch_token": false,
+  "mm_use_im_start_end": false,
+  "mm_vision_select_feature": "patch",
+  "mm_vision_select_layer": -2,
+  "mm_vision_tower": "google/siglip-so400m-patch14-384",
+  "mm_vision_tower_lr": null,
+  "model_type": "llava",
+  "num_attention_heads": 28,
+  "num_hidden_layers": 28,
+  "num_key_value_heads": 4,
+  "pos_skipping_range": 4096,
+  "projector_hidden_act": "gelu",
+  "rms_norm_eps": 1e-06,
+  "rope_scaling": null,
+  "rope_theta": 1000000.0,
+  "sliding_window": 131072,
+  "spatial_tower": "cut3r",
+  "text_config": {
+    "model_type": "llama"
+  },
+  "tie_word_embeddings": false,
+  "tokenizer_model_max_length": 32768,
+  "tokenizer_padding_side": "right",
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.40.0.dev0",
+  "tune_fusion_block": true,
+  "tune_mm_mlp_adapter": true,
+  "tune_mm_vision_resampler": false,
+  "tune_spatial_tower": false,
+  "unfreeze_mm_vision_tower": false,
+  "use_cache": false,
+  "use_mm_proj": true,
+  "use_pos_skipping": false,
+  "use_sliding_window": false,
+  "vision_config": {
+    "hidden_size": 1024,
+    "image_size": 336,
+    "intermediate_size": 4096,
+    "model_type": "clip_vision_model",
+    "num_attention_heads": 16,
+    "num_hidden_layers": 24,
+    "patch_size": 14,
+    "projection_dim": 768,
+    "vocab_size": 32000
+  },
+  "vision_feature_layer": -2,
+  "vision_feature_select_strategy": "default",
+  "vision_tower_pretrained": null
+}

ckpt/llavanext-google_siglip-so400m-patch14-384-Qwen_Qwen2-7B-Instruct-cut3r_lora_diff_lr_03_12/checkpoint-707/latest

@@ -0,0 +1 @@
+global_step707

ckpt/llavanext-google_siglip-so400m-patch14-384-Qwen_Qwen2-7B-Instruct-cut3r_lora_diff_lr_03_12/checkpoint-707/non_lora_trainables.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:31d0793d6b83f2efc66cc3945064e04053054cc900799a17caafbd9911ec2152
+size 53464972

ckpt/llavanext-google_siglip-so400m-patch14-384-Qwen_Qwen2-7B-Instruct-cut3r_lora_diff_lr_03_12/checkpoint-707/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d36d89e1330ee2b1d4b2aa755856a7d859c13ef0c42543c351373c89098503b1
+size 14256

ckpt/llavanext-google_siglip-so400m-patch14-384-Qwen_Qwen2-7B-Instruct-cut3r_lora_diff_lr_03_12/checkpoint-707/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3d9d4155bb73e7d3b425749b3461ae9c674f20c877d8bf36fdb0b2f244d5251f
+size 14256

ckpt/llavanext-google_siglip-so400m-patch14-384-Qwen_Qwen2-7B-Instruct-cut3r_lora_diff_lr_03_12/checkpoint-707/rng_state_10.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7b7f7348cd5d62e5d520828a7236e4ecb80da4166bfbb74c06c254976681c841
+size 14262

ckpt/llavanext-google_siglip-so400m-patch14-384-Qwen_Qwen2-7B-Instruct-cut3r_lora_diff_lr_03_12/checkpoint-707/rng_state_11.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:248d003676a4cc2cbf4a1284f94cbf1959b4b8c2dfb5dba38612fd5776b90d5c
+size 14262

ckpt/llavanext-google_siglip-so400m-patch14-384-Qwen_Qwen2-7B-Instruct-cut3r_lora_diff_lr_03_12/checkpoint-707/rng_state_12.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6377ebef2c642a5c4313426570bfc72bc38e4c01c8f12f43583524d7eea07e5a
+size 14262

ckpt/llavanext-google_siglip-so400m-patch14-384-Qwen_Qwen2-7B-Instruct-cut3r_lora_diff_lr_03_12/checkpoint-707/rng_state_13.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d58bb1be5f1c976405ecc9c0ab3cacb31e71cf17691ddf5c5ac557e470a52341
+size 14262

ckpt/llavanext-google_siglip-so400m-patch14-384-Qwen_Qwen2-7B-Instruct-cut3r_lora_diff_lr_03_12/checkpoint-707/rng_state_14.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:abb821e044c917764860abdf6f05f157569286b8bdf43b61ca0d6aef2a6e144e
+size 14262

ckpt/llavanext-google_siglip-so400m-patch14-384-Qwen_Qwen2-7B-Instruct-cut3r_lora_diff_lr_03_12/checkpoint-707/rng_state_15.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fd90db125656aff198f47d5858fc8bf4cdefd5612b47b16d000f8d6b0e00976e
+size 14262

ckpt/llavanext-google_siglip-so400m-patch14-384-Qwen_Qwen2-7B-Instruct-cut3r_lora_diff_lr_03_12/checkpoint-707/rng_state_16.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ececb857e2ee2506970a77d377048ca4f941add1f0e3929ac5a6e2c0122e855b
+size 14262

ckpt/llavanext-google_siglip-so400m-patch14-384-Qwen_Qwen2-7B-Instruct-cut3r_lora_diff_lr_03_12/checkpoint-707/rng_state_17.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:66aeca058fc1f82301425d028198191fc0bf6a230af18a3ad6c70712814d1d6f
+size 14262

ckpt/llavanext-google_siglip-so400m-patch14-384-Qwen_Qwen2-7B-Instruct-cut3r_lora_diff_lr_03_12/checkpoint-707/rng_state_18.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:186ab004ca5742adb866c67dbba48046cf9e81960ac4ac82f1a6de92e4d44fcb
+size 14262

ckpt/llavanext-google_siglip-so400m-patch14-384-Qwen_Qwen2-7B-Instruct-cut3r_lora_diff_lr_03_12/checkpoint-707/rng_state_19.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cad6e1901c262faf271ce70ba6d55b5279520bd1941b61f7db8c3f0871c98b20
+size 14262

ckpt/llavanext-google_siglip-so400m-patch14-384-Qwen_Qwen2-7B-Instruct-cut3r_lora_diff_lr_03_12/checkpoint-707/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fd2041c0d53e0d5b37113d2bf3e8a538aa3c03ef77719bdf0ab2ae01391d5e28
+size 14256

ckpt/llavanext-google_siglip-so400m-patch14-384-Qwen_Qwen2-7B-Instruct-cut3r_lora_diff_lr_03_12/checkpoint-707/rng_state_20.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:175ec70a0d1571c2a80dd460ee2fcdf8c1f5563cce0a2a12b06987df8b3e0e38
+size 14262

ckpt/llavanext-google_siglip-so400m-patch14-384-Qwen_Qwen2-7B-Instruct-cut3r_lora_diff_lr_03_12/checkpoint-707/rng_state_21.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9386d36206ecdf8448af832d0c65d3e012e9986ec443b5eac69a9b26c29fdd8f
+size 14262

ckpt/llavanext-google_siglip-so400m-patch14-384-Qwen_Qwen2-7B-Instruct-cut3r_lora_diff_lr_03_12/checkpoint-707/rng_state_22.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b1396d292133d3ebbb63e831407c11b0a393a2dc8daa81d8a74a7dbb2f13c2cd
+size 14262

ckpt/llavanext-google_siglip-so400m-patch14-384-Qwen_Qwen2-7B-Instruct-cut3r_lora_diff_lr_03_12/checkpoint-707/rng_state_23.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6ad90e2d89994b9c63768d2cb3a0952f90b24faec86e87223aedb61ecf172736
+size 14262

ckpt/llavanext-google_siglip-so400m-patch14-384-Qwen_Qwen2-7B-Instruct-cut3r_lora_diff_lr_03_12/checkpoint-707/rng_state_24.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4b2251a2b1bf61a38fe4329b814b57180ad6a667aaa4f9b50feb1a0b2081c380
+size 14262

ckpt/llavanext-google_siglip-so400m-patch14-384-Qwen_Qwen2-7B-Instruct-cut3r_lora_diff_lr_03_12/checkpoint-707/rng_state_25.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9a6e713729e0c6d6325e200ad2def22c1797ec42eb4fccc019f0a6812295d5c2
+size 14262

ckpt/llavanext-google_siglip-so400m-patch14-384-Qwen_Qwen2-7B-Instruct-cut3r_lora_diff_lr_03_12/checkpoint-707/rng_state_26.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d5ebca4191ac5173da35c591d732fa9c1bfeb14a22fc9ad20a2ad27d3d76991a
+size 14262

ckpt/llavanext-google_siglip-so400m-patch14-384-Qwen_Qwen2-7B-Instruct-cut3r_lora_diff_lr_03_12/checkpoint-707/rng_state_27.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:472e19a8203030ff6ca593402b5b7c42127fd289ed8fa06f9198985b24ca9aa0
+size 14262

ckpt/llavanext-google_siglip-so400m-patch14-384-Qwen_Qwen2-7B-Instruct-cut3r_lora_diff_lr_03_12/checkpoint-707/rng_state_28.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f862ce60ccaa72bd0a62b90264b811acf53f5a5059d9c40895ee11e42d6d5f57
+size 14262

ckpt/llavanext-google_siglip-so400m-patch14-384-Qwen_Qwen2-7B-Instruct-cut3r_lora_diff_lr_03_12/checkpoint-707/rng_state_29.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:662904a83ff8036bef9019a7de90ed7ec25614670e47135ba879f09344637f96
+size 14262

ckpt/llavanext-google_siglip-so400m-patch14-384-Qwen_Qwen2-7B-Instruct-cut3r_lora_diff_lr_03_12/checkpoint-707/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:402f9be1cb2dd2e180401e6799bc406cfaef270a7f52a1170449b0663e642246
+size 14256

ckpt/llavanext-google_siglip-so400m-patch14-384-Qwen_Qwen2-7B-Instruct-cut3r_lora_diff_lr_03_12/checkpoint-707/rng_state_30.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0e54e51065bd072c9d025808a8f30d73bcbc1beb924de05f326a03720c8f4359
+size 14262

ckpt/llavanext-google_siglip-so400m-patch14-384-Qwen_Qwen2-7B-Instruct-cut3r_lora_diff_lr_03_12/checkpoint-707/rng_state_31.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b4645d4846ed92a7d093af41d82ed2462b0501190835dc562ff5255b9032c85f
+size 14262

ckpt/llavanext-google_siglip-so400m-patch14-384-Qwen_Qwen2-7B-Instruct-cut3r_lora_diff_lr_03_12/checkpoint-707/rng_state_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b6615b629095828ec92dc57a8b81c53b78b97e10ffc56f2ad2bd5faf9ca62ec3
+size 14256

ckpt/llavanext-google_siglip-so400m-patch14-384-Qwen_Qwen2-7B-Instruct-cut3r_lora_diff_lr_03_12/checkpoint-707/rng_state_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fb02568dc0c7fa711c57a3a68b0cbcaef3c403a13c7b810b5652a317d2458edf
+size 14256

ckpt/llavanext-google_siglip-so400m-patch14-384-Qwen_Qwen2-7B-Instruct-cut3r_lora_diff_lr_03_12/checkpoint-707/rng_state_6.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1eb4d561674e03e6edfe08cede15fc4b4aae489d28fb28879d0c4a12a6eab957
+size 14256

ckpt/llavanext-google_siglip-so400m-patch14-384-Qwen_Qwen2-7B-Instruct-cut3r_lora_diff_lr_03_12/checkpoint-707/rng_state_7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f85a8efe0965179462561f6f0fb646c4399aa10c84ee3e9e03f1acb7b25afb06
+size 14256

ckpt/llavanext-google_siglip-so400m-patch14-384-Qwen_Qwen2-7B-Instruct-cut3r_lora_diff_lr_03_12/checkpoint-707/rng_state_8.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dd92a4fef47d7dc9e362cd5ac83f6dda1e594262cb7fdfb18745294d75fc2ecb
+size 14256

ckpt/llavanext-google_siglip-so400m-patch14-384-Qwen_Qwen2-7B-Instruct-cut3r_lora_diff_lr_03_12/checkpoint-707/rng_state_9.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b060ae0f8fed9f14d3ea29498d9679cfacd093b74b6946be3154c42f13f8627c
+size 14256

ckpt/llavanext-google_siglip-so400m-patch14-384-Qwen_Qwen2-7B-Instruct-cut3r_lora_diff_lr_03_12/checkpoint-707/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2f2d08b16af46d1a7a8883881ec6262cf4be1211815f5311ad397ae8a960c26c
+size 1064

ckpt/llavanext-google_siglip-so400m-patch14-384-Qwen_Qwen2-7B-Instruct-cut3r_lora_diff_lr_03_12/checkpoint-707/zero_to_fp32.py

@@ -0,0 +1,604 @@
+#!/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 . pytorch_model.bin
+
+import argparse
+import torch
+import glob
+import math
+import os
+import re
+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)
+
+        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 files:
+        state_dict = torch.load(f, map_location=device)
+        # 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 not ZERO_STAGE 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}")
+
+    if zero_stage <= 2:
+        fp32_flat_groups = [state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key] for i in range(len(state_dicts))]
+    elif zero_stage == 3:
+        # if there is more than one param group, there will be multiple flattened tensors - one
+        # flattened tensor per group - for simplicity merge them into a single tensor
+        #
+        # XXX: could make the script more memory efficient for when there are multiple groups - it
+        # will require matching the sub-lists of param_shapes for each param group flattened tensor
+
+        fp32_flat_groups = [
+            torch.cat(state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key], 0) 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")
+
+
+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 = fp32_flat_groups[0].numel() * 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
+    for name, shape in param_shapes.items():
+
+        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}"
+            )
+
+        # XXX: memory usage doubles here
+        state_dict[name] = torch.cat(
+            tuple(fp32_flat_groups[i].narrow(0, offset, partitioned_numel) for i in range(world_size)),
+            0).narrow(0, 0, unpartitioned_numel).view(shape)
+        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 get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag=None, exclude_frozen_parameters=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
+
+    Returns:
+        - pytorch ``state_dict``
+
+    Note: this approach may not work if your application doesn't have sufficient free CPU memory and
+    you may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
+    the checkpoint.
+
+    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.
+
+    """
+    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")
+
+    return _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters)
+
+
+def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir, output_file, 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_file``: path to the pytorch fp32 state_dict output file (e.g. path/pytorch_model.bin)
+        - ``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
+    """
+
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag, exclude_frozen_parameters)
+    print(f"Saving fp32 state dict to {output_file}")
+    torch.save(state_dict, output_file)
+
+
+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(f"Extracting fp32 weights")
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
+
+    logger.info(f"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_file",
+        type=str,
+        help="path to the pytorch fp32 state_dict output file (e.g. path/checkpoint-12/pytorch_model.bin)")
+    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_file,
+                                               tag=args.tag,
+                                               exclude_frozen_parameters=args.exclude_frozen_parameters)