{"cells":[{"cell_type":"markdown","metadata":{"id":"Q0ZGy2-ECTCY"},"source":["# Build SDNQ Joycaption"]},{"cell_type":"code","execution_count":null,"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"elapsed":2205,"status":"ok","timestamp":1777149303868,"user":{"displayName":"fukU Google","userId":"02763165356193834046"},"user_tz":-120},"id":"GlOXOGUCCX8v","outputId":"3bf6e9c9-c0a2-4e3d-9469-7fdf62c16110"},"outputs":[{"name":"stdout","output_type":"stream","text":["Drive already mounted at /content/drive; to attempt to forcibly remount, call drive.mount(\"/content/drive\", force_remount=True).\n","✅ Drive + HF login ready\n"]}],"source":["#@markdown ## 1 - Setup environment\n","\n","from google.colab import drive, userdata\n","\n","drive.mount('/content/drive')\n","\n","hf_token = userdata.get(\"HF_TOKEN\")\n","\n","\n","if hf_token:\n","    from huggingface_hub import login\n","    login(token=hf_token)\n","\n","print(\"✅ Drive + HF login ready\")"]},{"cell_type":"code","execution_count":null,"metadata":{"cellView":"form","colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"elapsed":16747,"status":"ok","timestamp":1777145643084,"user":{"displayName":"fukU Google","userId":"02763165356193834046"},"user_tz":-120},"id":"eSZIjx7AClZb","outputId":"c9870fd3-7aff-4d40-ab8d-0418be4c3755"},"outputs":[{"name":"stdout","output_type":"stream","text":["\u001b[?25l   \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m0.0/104.0 kB\u001b[0m \u001b[31m?\u001b[0m eta \u001b[36m-:--:--\u001b[0m\r\u001b[2K   \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m104.0/104.0 kB\u001b[0m \u001b[31m4.3 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n","\u001b[?25h✅ Dependencies installed\n"]}],"source":["#@markdown ## 2 - Install dependencies\n","\n","!pip install -U bitsandbytes>=0.46.1\n","\n","!pip install -q \\\n","    transformers \\\n","    tokenizers \\\n","    accelerate \\\n","    huggingface_hub \\\n","    safetensors \\\n","    sdnq\n","\n","print(\"✅ Dependencies installed\")"]},{"cell_type":"code","execution_count":null,"metadata":{"colab":{"base_uri":"https://localhost:8080/","height":197},"executionInfo":{"elapsed":99030,"status":"ok","timestamp":1777149404729,"user":{"displayName":"fukU Google","userId":"02763165356193834046"},"user_tz":-120},"id":"_S6JR1AuXSxF","outputId":"94b4d6ec-4329-4f4c-d5ba-0392ccde299a"},"outputs":[{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"dd60acbf14f8438bb587a4fdf11a2b76","version_major":2,"version_minor":0},"text/plain":["Downloading (incomplete total...): 0.00B [00:00, ?B/s]"]},"metadata":{},"output_type":"display_data"},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"36c6d11d657448ef9146bd3068f490d9","version_major":2,"version_minor":0},"text/plain":["Fetching 4 files:   0%|          | 0/4 [00:00<?, ?it/s]"]},"metadata":{},"output_type":"display_data"},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"a937be16926f407e88055f0ef5c86320","version_major":2,"version_minor":0},"text/plain":["Loading weights:   0%|          | 0/743 [00:00<?, ?it/s]"]},"metadata":{},"output_type":"display_data"},{"name":"stderr","output_type":"stream","text":["The image processor of type `SiglipImageProcessor` is now loaded as a fast processor by default, even if the model checkpoint was saved with a slow processor. This is a breaking change and may produce slightly different outputs. To continue using the slow processor, instantiate this class with `use_fast=False`. \n"]},{"name":"stdout","output_type":"stream","text":["✅ Model loaded in 4bit (bootstrap only)\n"]}],"source":["#@markdown ## 3 - Load 4bit model (temporary bootstrap – low VRAM)\n","import torch\n","from transformers import BitsAndBytesConfig, AutoProcessor, LlavaForConditionalGeneration\n","\n","MODEL_NAME = \"fancyfeast/llama-joycaption-beta-one-hf-llava\"\n","\n","bnb = BitsAndBytesConfig(\n","    load_in_4bit=True,\n","    bnb_4bit_compute_dtype=torch.float16,\n","    bnb_4bit_quant_type=\"nf4\"\n",")\n","\n","model = LlavaForConditionalGeneration.from_pretrained(\n","    MODEL_NAME,\n","    quantization_config=bnb,\n","    device_map=\"auto\",          # temporary, will be moved to CPU immediately\n","    trust_remote_code=True\n",")\n","\n","processor = AutoProcessor.from_pretrained(MODEL_NAME)\n","\n","print(\"✅ Model loaded in 4bit (bootstrap only)\")"]},{"cell_type":"code","execution_count":null,"metadata":{"cellView":"form","id":"EOZGbgp-p90C"},"outputs":[],"source":["#@markdown ## 4 - Low-RAM dequant + fast sharding (CPU-only + ROBUST resume)\n","\n","import os\n","import gc\n","import torch\n","import json\n","from safetensors.torch import save_file, load_file\n","import bitsandbytes.functional as F\n","\n","# ========================= CONFIG =========================\n","SHARD_DIR = \"/content/joycaption_dequant_shards\"\n","os.makedirs(SHARD_DIR, exist_ok=True)\n","\n","MAX_SHARD_SIZE_GB = 0.5\n","# =========================================================\n","\n","def _get_safetensors_keys(path):\n","    \"\"\"Ultra-reliable way to read keys from any safetensors file (no device=meta needed)\"\"\"\n","    try:\n","        with open(path, \"rb\") as f:\n","            length = int.from_bytes(f.read(8), \"little\")\n","            header = json.loads(f.read(length).decode(\"utf-8\"))\n","            # All keys except the internal __metadata__ block\n","            keys = [k for k in header.keys() if k != \"__metadata__\"]\n","            return keys\n","    except Exception as e:\n","        print(f\"   ❌ Failed to parse header of {os.path.basename(path)}: {e}\")\n","        return []\n","\n","def get_processed_keys(shard_dir=SHARD_DIR):\n","    \"\"\"Scan shards WITHOUT loading any tensors\"\"\"\n","    processed = set()\n","    shard_files = sorted([f for f in os.listdir(shard_dir) if f.endswith(\".safetensors\")])\n","\n","    print(f\"🔍 Scanning {len(shard_files)} existing shards for already-processed layers...\")\n","    if not shard_files:\n","        print(\"   → No shards found yet\")\n","        return processed\n","\n","    for shard_file in shard_files:\n","        path = os.path.join(shard_dir, shard_file)\n","        print(f\"   Checking {shard_file}...\", end=\" \")\n","        keys = _get_safetensors_keys(path)\n","        if keys:\n","            print(f\"✅ {len(keys)} keys found\")\n","            processed.update(keys)\n","        else:\n","            print(\"⚠️ no keys parsed\")\n","\n","    print(f\"✅ Total processed keys detected: {len(processed)}\\n\")\n","    return processed\n","\n","def load_dequant_shards(model, shard_dir=SHARD_DIR):\n","    \"\"\"Load shards ONLY at the very end – forces CPU\"\"\"\n","    shard_files = sorted([f for f in os.listdir(shard_dir) if f.endswith(\".safetensors\")])\n","    if not shard_files:\n","        print(\"No shards found – starting from scratch\")\n","        return model\n","\n","    print(f\"🔄 Loading {len(shard_files)} dequant shards (CPU only)...\")\n","    model = model.cpu()\n","    torch.cuda.empty_cache()\n","    gc.collect()\n","\n","    for shard_file in shard_files:\n","        path = os.path.join(shard_dir, shard_file)\n","        state = load_file(path, device=\"cpu\")\n","        for key, tensor in state.items():\n","            try:\n","                submodule_name = \".\".join(key.split(\".\")[:-1])\n","                submodule = model.get_submodule(submodule_name)\n","                if hasattr(submodule, \"weight\"):\n","                    submodule.weight = torch.nn.Parameter(tensor.to(torch.float16))\n","                    print(f\"   ✅ Restored {key}\")\n","            except Exception as e:\n","                print(f\"   ⚠️ Could not restore {key}: {e}\")\n","    print(\"✅ All shards loaded on CPU\")\n","    return model\n","\n","def dequantize_model_to_fp16(model, shard_dir=SHARD_DIR, max_shard_gb=0.5):\n","    model = model.cpu()\n","    torch.cuda.empty_cache()\n","    gc.collect()\n","\n","    processed_keys = get_processed_keys(shard_dir)\n","\n","    cumulative_bytes = 0\n","    current_shard_dict = {}\n","    shard_id = len([f for f in os.listdir(shard_dir) if f.endswith(\".safetensors\")])\n","\n","    print(\"🚀 Starting CPU-only dequantization with immediate offloading...\")\n","\n","    for name, module in list(model.named_modules()):\n","        if not (hasattr(module, \"weight\") and module.weight is not None):\n","            continue\n","        if not hasattr(module.weight, \"quant_state\"):\n","            continue\n","\n","        key = name + \".weight\"\n","        if key in processed_keys:\n","            print(f\"⏭️ Skipping already processed: {name}\")\n","            continue\n","\n","        print(f\"Dequantizing: {name}\")\n","\n","        try:\n","            weight = module.weight\n","            quant_state = weight.quant_state\n","\n","            weight_cpu = weight.to(\"cpu\")\n","            del module.weight\n","            del weight\n","            torch.cuda.empty_cache()\n","            gc.collect()\n","\n","            dequant = F.dequantize_4bit(weight_cpu, quant_state)\n","\n","            module.weight = torch.nn.Parameter(dequant.to(torch.float16))\n","\n","            current_shard_dict[key] = module.weight.data.clone()\n","            layer_bytes = module.weight.numel() * module.weight.element_size()\n","            cumulative_bytes += layer_bytes\n","\n","            print(f\"   → {layer_bytes / (1024**3):.3f} GB | cumulative {cumulative_bytes / (1024**3):.3f} GB\")\n","\n","            if cumulative_bytes / (1024**3) >= max_shard_gb:\n","                shard_path = os.path.join(shard_dir, f\"dequant_shard_{shard_id:05d}.safetensors\")\n","                save_file(current_shard_dict, shard_path)\n","                print(f\"💾 Saved shard → {shard_path} ({len(current_shard_dict)} keys)\")\n","\n","                # IMMEDIATE OFFLOAD\n","                for saved_key in list(current_shard_dict.keys()):\n","                    try:\n","                        sub_name = \".\".join(saved_key.split(\".\")[:-1])\n","                        sub = model.get_submodule(sub_name)\n","                        if hasattr(sub, \"weight\"):\n","                            del sub.weight\n","                            sub.weight = None\n","                    except:\n","                        pass\n","\n","                current_shard_dict = {}\n","                cumulative_bytes = 0\n","                shard_id += 1\n","\n","            del dequant\n","            del weight_cpu\n","            gc.collect()\n","\n","        except Exception as e:\n","            print(f\"❌ Skipped {name}: {e}\")\n","            gc.collect()\n","\n","    # final shard\n","    if current_shard_dict:\n","        shard_path = os.path.join(shard_dir, f\"dequant_shard_{shard_id:05d}.safetensors\")\n","        save_file(current_shard_dict, shard_path)\n","        print(f\"💾 Saved final shard → {shard_path}\")\n","\n","        for saved_key in list(current_shard_dict.keys()):\n","            try:\n","                sub_name = \".\".join(saved_key.split(\".\")[:-1])\n","                sub = model.get_submodule(sub_name)\n","                if hasattr(sub, \"weight\"):\n","                    del sub.weight\n","                    sub.weight = None\n","            except:\n","                pass\n","\n","    print(\"✅ Dequantization complete – all weights offloaded to disk\")\n","    return model"]},{"cell_type":"code","execution_count":null,"metadata":{"cellView":"form","colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"elapsed":19642,"status":"ok","timestamp":1777149424409,"user":{"displayName":"fukU Google","userId":"02763165356193834046"},"user_tz":-120},"id":"XeDK_bTjafNo","outputId":"3c8f74f8-fd41-4dfb-c732-86d4cf370a93"},"outputs":[{"name":"stdout","output_type":"stream","text":["🔍 Scanning 26 existing shards for already-processed layers...\n","   Checking dequant_shard_00000.safetensors... ✅ 107 keys found\n","   Checking dequant_shard_00001.safetensors... ✅ 65 keys found\n","   Checking dequant_shard_00002.safetensors... ✅ 8 keys found\n","   Checking dequant_shard_00003.safetensors... ✅ 8 keys found\n","   Checking dequant_shard_00004.safetensors... ✅ 12 keys found\n","   Checking dequant_shard_00005.safetensors... ✅ 8 keys found\n","   Checking dequant_shard_00006.safetensors... ✅ 8 keys found\n","   Checking dequant_shard_00007.safetensors... ✅ 12 keys found\n","   Checking dequant_shard_00008.safetensors... ✅ 8 keys found\n","   Checking dequant_shard_00009.safetensors... ✅ 8 keys found\n","   Checking dequant_shard_00010.safetensors... ✅ 12 keys found\n","   Checking dequant_shard_00011.safetensors... ✅ 8 keys found\n","   Checking dequant_shard_00012.safetensors... ✅ 8 keys found\n","   Checking dequant_shard_00013.safetensors... ✅ 12 keys found\n","   Checking dequant_shard_00014.safetensors... ✅ 8 keys found\n","   Checking dequant_shard_00015.safetensors... ✅ 8 keys found\n","   Checking dequant_shard_00016.safetensors... ✅ 12 keys found\n","   Checking dequant_shard_00017.safetensors... ✅ 8 keys found\n","   Checking dequant_shard_00018.safetensors... ✅ 8 keys found\n","   Checking dequant_shard_00019.safetensors... ✅ 12 keys found\n","   Checking dequant_shard_00020.safetensors... ✅ 8 keys found\n","   Checking dequant_shard_00021.safetensors... ✅ 8 keys found\n","   Checking dequant_shard_00022.safetensors... ✅ 12 keys found\n","   Checking dequant_shard_00023.safetensors... ✅ 8 keys found\n","   Checking dequant_shard_00024.safetensors... ✅ 8 keys found\n","   Checking dequant_shard_00025.safetensors... ✅ 7 keys found\n","✅ Total processed keys detected: 391\n","\n","🚀 Starting CPU-only dequantization with immediate offloading...\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.0.self_attn.k_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.0.self_attn.v_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.0.self_attn.q_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.0.self_attn.out_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.0.mlp.fc1\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.0.mlp.fc2\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.1.self_attn.k_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.1.self_attn.v_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.1.self_attn.q_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.1.self_attn.out_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.1.mlp.fc1\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.1.mlp.fc2\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.2.self_attn.k_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.2.self_attn.v_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.2.self_attn.q_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.2.self_attn.out_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.2.mlp.fc1\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.2.mlp.fc2\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.3.self_attn.k_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.3.self_attn.v_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.3.self_attn.q_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.3.self_attn.out_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.3.mlp.fc1\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.3.mlp.fc2\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.4.self_attn.k_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.4.self_attn.v_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.4.self_attn.q_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.4.self_attn.out_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.4.mlp.fc1\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.4.mlp.fc2\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.5.self_attn.k_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.5.self_attn.v_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.5.self_attn.q_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.5.self_attn.out_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.5.mlp.fc1\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.5.mlp.fc2\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.6.self_attn.k_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.6.self_attn.v_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.6.self_attn.q_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.6.self_attn.out_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.6.mlp.fc1\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.6.mlp.fc2\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.7.self_attn.k_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.7.self_attn.v_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.7.self_attn.q_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.7.self_attn.out_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.7.mlp.fc1\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.7.mlp.fc2\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.8.self_attn.k_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.8.self_attn.v_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.8.self_attn.q_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.8.self_attn.out_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.8.mlp.fc1\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.8.mlp.fc2\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.9.self_attn.k_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.9.self_attn.v_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.9.self_attn.q_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.9.self_attn.out_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.9.mlp.fc1\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.9.mlp.fc2\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.10.self_attn.k_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.10.self_attn.v_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.10.self_attn.q_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.10.self_attn.out_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.10.mlp.fc1\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.10.mlp.fc2\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.11.self_attn.k_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.11.self_attn.v_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.11.self_attn.q_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.11.self_attn.out_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.11.mlp.fc1\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.11.mlp.fc2\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.12.self_attn.k_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.12.self_attn.v_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.12.self_attn.q_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.12.self_attn.out_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.12.mlp.fc1\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.12.mlp.fc2\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.13.self_attn.k_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.13.self_attn.v_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.13.self_attn.q_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.13.self_attn.out_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.13.mlp.fc1\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.13.mlp.fc2\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.14.self_attn.k_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.14.self_attn.v_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.14.self_attn.q_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.14.self_attn.out_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.14.mlp.fc1\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.14.mlp.fc2\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.15.self_attn.k_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.15.self_attn.v_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.15.self_attn.q_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.15.self_attn.out_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.15.mlp.fc1\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.15.mlp.fc2\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.16.self_attn.k_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.16.self_attn.v_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.16.self_attn.q_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.16.self_attn.out_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.16.mlp.fc1\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.16.mlp.fc2\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.17.self_attn.k_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.17.self_attn.v_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.17.self_attn.q_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.17.self_attn.out_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.17.mlp.fc1\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.17.mlp.fc2\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.18.self_attn.k_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.18.self_attn.v_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.18.self_attn.q_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.18.self_attn.out_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.18.mlp.fc1\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.18.mlp.fc2\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.19.self_attn.k_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.19.self_attn.v_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.19.self_attn.q_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.19.self_attn.out_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.19.mlp.fc1\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.19.mlp.fc2\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.20.self_attn.k_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.20.self_attn.v_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.20.self_attn.q_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.20.self_attn.out_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.20.mlp.fc1\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.20.mlp.fc2\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.21.self_attn.k_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.21.self_attn.v_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.21.self_attn.q_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.21.self_attn.out_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.21.mlp.fc1\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.21.mlp.fc2\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.22.self_attn.k_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.22.self_attn.v_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.22.self_attn.q_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.22.self_attn.out_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.22.mlp.fc1\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.22.mlp.fc2\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.23.self_attn.k_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.23.self_attn.v_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.23.self_attn.q_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.23.self_attn.out_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.23.mlp.fc1\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.23.mlp.fc2\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.24.self_attn.k_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.24.self_attn.v_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.24.self_attn.q_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.24.self_attn.out_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.24.mlp.fc1\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.24.mlp.fc2\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.25.self_attn.k_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.25.self_attn.v_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.25.self_attn.q_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.25.self_attn.out_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.25.mlp.fc1\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.25.mlp.fc2\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.26.self_attn.k_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.26.self_attn.v_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.26.self_attn.q_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.26.self_attn.out_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.26.mlp.fc1\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.26.mlp.fc2\n","⏭️ Skipping already processed: model.vision_tower.vision_model.head.attention.out_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.head.mlp.fc1\n","⏭️ Skipping already processed: model.vision_tower.vision_model.head.mlp.fc2\n","⏭️ Skipping already processed: model.multi_modal_projector.linear_1\n","⏭️ Skipping already processed: model.multi_modal_projector.linear_2\n","⏭️ Skipping already processed: model.language_model.layers.0.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.0.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.0.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.0.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.0.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.0.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.0.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.1.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.1.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.1.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.1.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.1.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.1.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.1.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.2.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.2.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.2.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.2.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.2.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.2.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.2.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.3.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.3.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.3.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.3.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.3.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.3.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.3.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.4.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.4.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.4.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.4.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.4.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.4.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.4.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.5.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.5.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.5.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.5.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.5.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.5.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.5.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.6.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.6.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.6.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.6.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.6.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.6.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.6.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.7.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.7.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.7.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.7.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.7.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.7.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.7.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.8.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.8.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.8.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.8.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.8.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.8.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.8.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.9.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.9.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.9.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.9.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.9.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.9.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.9.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.10.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.10.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.10.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.10.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.10.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.10.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.10.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.11.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.11.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.11.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.11.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.11.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.11.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.11.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.12.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.12.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.12.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.12.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.12.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.12.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.12.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.13.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.13.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.13.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.13.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.13.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.13.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.13.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.14.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.14.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.14.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.14.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.14.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.14.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.14.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.15.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.15.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.15.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.15.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.15.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.15.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.15.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.16.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.16.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.16.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.16.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.16.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.16.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.16.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.17.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.17.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.17.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.17.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.17.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.17.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.17.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.18.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.18.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.18.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.18.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.18.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.18.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.18.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.19.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.19.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.19.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.19.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.19.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.19.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.19.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.20.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.20.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.20.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.20.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.20.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.20.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.20.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.21.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.21.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.21.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.21.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.21.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.21.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.21.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.22.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.22.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.22.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.22.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.22.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.22.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.22.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.23.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.23.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.23.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.23.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.23.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.23.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.23.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.24.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.24.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.24.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.24.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.24.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.24.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.24.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.25.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.25.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.25.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.25.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.25.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.25.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.25.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.26.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.26.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.26.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.26.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.26.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.26.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.26.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.27.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.27.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.27.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.27.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.27.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.27.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.27.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.28.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.28.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.28.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.28.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.28.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.28.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.28.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.29.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.29.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.29.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.29.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.29.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.29.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.29.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.30.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.30.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.30.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.30.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.30.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.30.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.30.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.31.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.31.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.31.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.31.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.31.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.31.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.31.mlp.down_proj\n","✅ Dequantization complete – all weights offloaded to disk\n","🔄 Loading 26 dequant shards (CPU only)...\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.0.mlp.fc1.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.0.mlp.fc2.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.0.self_attn.k_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.0.self_attn.out_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.0.self_attn.q_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.0.self_attn.v_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.1.mlp.fc1.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.1.mlp.fc2.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.1.self_attn.k_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.1.self_attn.out_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.1.self_attn.q_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.1.self_attn.v_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.10.mlp.fc1.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.10.mlp.fc2.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.10.self_attn.k_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.10.self_attn.out_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.10.self_attn.q_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.10.self_attn.v_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.11.mlp.fc1.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.11.mlp.fc2.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.11.self_attn.k_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.11.self_attn.out_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.11.self_attn.q_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.11.self_attn.v_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.12.mlp.fc1.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.12.mlp.fc2.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.12.self_attn.k_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.12.self_attn.out_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.12.self_attn.q_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.12.self_attn.v_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.13.mlp.fc1.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.13.mlp.fc2.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.13.self_attn.k_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.13.self_attn.out_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.13.self_attn.q_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.13.self_attn.v_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.14.mlp.fc1.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.14.mlp.fc2.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.14.self_attn.k_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.14.self_attn.out_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.14.self_attn.q_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.14.self_attn.v_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.15.mlp.fc1.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.15.mlp.fc2.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.15.self_attn.k_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.15.self_attn.out_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.15.self_attn.q_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.15.self_attn.v_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.16.mlp.fc1.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.16.mlp.fc2.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.16.self_attn.k_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.16.self_attn.out_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.16.self_attn.q_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.16.self_attn.v_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.17.mlp.fc1.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.17.self_attn.k_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.17.self_attn.out_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.17.self_attn.q_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.17.self_attn.v_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.2.mlp.fc1.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.2.mlp.fc2.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.2.self_attn.k_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.2.self_attn.out_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.2.self_attn.q_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.2.self_attn.v_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.3.mlp.fc1.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.3.mlp.fc2.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.3.self_attn.k_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.3.self_attn.out_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.3.self_attn.q_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.3.self_attn.v_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.4.mlp.fc1.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.4.mlp.fc2.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.4.self_attn.k_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.4.self_attn.out_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.4.self_attn.q_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.4.self_attn.v_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.5.mlp.fc1.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.5.mlp.fc2.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.5.self_attn.k_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.5.self_attn.out_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.5.self_attn.q_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.5.self_attn.v_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.6.mlp.fc1.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.6.mlp.fc2.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.6.self_attn.k_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.6.self_attn.out_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.6.self_attn.q_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.6.self_attn.v_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.7.mlp.fc1.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.7.mlp.fc2.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.7.self_attn.k_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.7.self_attn.out_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.7.self_attn.q_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.7.self_attn.v_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.8.mlp.fc1.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.8.mlp.fc2.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.8.self_attn.k_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.8.self_attn.out_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.8.self_attn.q_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.8.self_attn.v_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.9.mlp.fc1.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.9.mlp.fc2.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.9.self_attn.k_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.9.self_attn.out_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.9.self_attn.q_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.9.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.0.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.0.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.0.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.0.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.0.self_attn.v_proj.weight\n","   ✅ Restored model.multi_modal_projector.linear_1.weight\n","   ✅ Restored model.multi_modal_projector.linear_2.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.17.mlp.fc2.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.18.mlp.fc1.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.18.mlp.fc2.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.18.self_attn.k_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.18.self_attn.out_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.18.self_attn.q_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.18.self_attn.v_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.19.mlp.fc1.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.19.mlp.fc2.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.19.self_attn.k_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.19.self_attn.out_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.19.self_attn.q_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.19.self_attn.v_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.20.mlp.fc1.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.20.mlp.fc2.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.20.self_attn.k_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.20.self_attn.out_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.20.self_attn.q_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.20.self_attn.v_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.21.mlp.fc1.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.21.mlp.fc2.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.21.self_attn.k_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.21.self_attn.out_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.21.self_attn.q_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.21.self_attn.v_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.22.mlp.fc1.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.22.mlp.fc2.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.22.self_attn.k_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.22.self_attn.out_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.22.self_attn.q_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.22.self_attn.v_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.23.mlp.fc1.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.23.mlp.fc2.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.23.self_attn.k_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.23.self_attn.out_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.23.self_attn.q_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.23.self_attn.v_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.24.mlp.fc1.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.24.mlp.fc2.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.24.self_attn.k_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.24.self_attn.out_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.24.self_attn.q_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.24.self_attn.v_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.25.mlp.fc1.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.25.mlp.fc2.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.25.self_attn.k_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.25.self_attn.out_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.25.self_attn.q_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.25.self_attn.v_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.26.mlp.fc1.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.26.mlp.fc2.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.26.self_attn.k_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.26.self_attn.out_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.26.self_attn.q_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.26.self_attn.v_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.head.attention.out_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.head.mlp.fc1.weight\n","   ✅ Restored model.vision_tower.vision_model.head.mlp.fc2.weight\n","   ✅ Restored model.language_model.layers.0.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.0.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.1.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.1.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.1.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.1.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.1.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.1.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.1.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.2.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.2.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.2.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.2.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.2.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.2.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.2.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.3.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.3.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.3.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.3.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.3.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.3.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.3.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.4.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.4.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.4.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.4.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.4.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.4.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.4.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.5.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.5.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.5.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.5.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.5.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.5.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.5.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.6.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.6.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.6.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.6.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.6.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.6.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.6.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.7.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.7.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.7.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.7.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.7.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.7.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.7.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.8.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.8.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.8.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.8.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.8.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.8.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.8.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.9.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.9.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.9.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.9.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.9.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.9.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.10.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.10.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.10.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.10.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.10.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.10.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.10.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.9.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.11.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.11.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.11.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.11.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.11.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.11.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.11.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.12.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.12.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.12.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.12.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.12.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.12.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.12.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.13.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.13.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.13.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.13.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.13.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.13.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.13.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.14.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.14.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.14.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.14.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.14.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.14.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.14.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.15.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.15.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.15.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.15.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.15.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.15.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.15.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.16.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.16.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.16.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.16.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.16.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.16.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.16.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.17.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.17.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.17.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.17.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.17.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.17.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.17.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.18.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.18.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.18.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.18.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.18.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.18.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.18.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.19.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.19.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.19.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.19.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.19.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.19.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.19.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.20.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.20.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.20.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.20.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.20.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.20.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.20.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.21.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.21.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.21.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.21.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.21.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.21.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.21.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.22.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.22.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.22.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.22.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.22.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.22.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.22.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.23.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.23.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.23.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.23.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.23.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.23.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.23.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.24.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.24.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.24.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.24.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.24.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.24.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.24.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.25.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.25.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.25.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.25.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.25.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.25.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.25.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.26.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.26.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.26.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.26.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.26.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.26.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.26.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.27.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.27.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.27.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.27.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.27.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.27.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.27.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.28.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.28.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.28.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.28.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.28.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.28.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.28.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.29.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.29.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.29.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.29.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.29.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.29.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.29.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.30.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.30.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.30.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.30.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.30.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.30.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.30.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.31.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.31.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.31.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.31.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.31.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.31.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.31.self_attn.v_proj.weight\n","✅ All shards loaded on CPU\n","✅ Fully dequantized on CPU – ready for SDNQ\n"]}],"source":["#@markdown ## Run dequantization (Step 4)\n","# Resume + continue dequant (CPU only)\n","# ERROR! This function cant detect previously safed safetensors!\n","model = dequantize_model_to_fp16(model, SHARD_DIR, MAX_SHARD_SIZE_GB)\n","\n","# Now load everything back on CPU (this is where your previous VRAM OOM happened)\n","model = load_dequant_shards(model, SHARD_DIR)\n","\n","print(\"✅ Fully dequantized on CPU – ready for SDNQ\")"]},{"cell_type":"code","execution_count":null,"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"elapsed":691,"status":"ok","timestamp":1777149425101,"user":{"displayName":"fukU Google","userId":"02763165356193834046"},"user_tz":-120},"id":"7K1i4Bn8ItHl","outputId":"e2b4c7c5-ee65-42e9-af9e-fe8beb80078d"},"outputs":[{"data":{"text/plain":["0"]},"execution_count":5,"metadata":{},"output_type":"execute_result"}],"source":["#del model\n","import gc , torch\n","torch.cuda.empty_cache()\n","gc.collect()"]},{"cell_type":"code","execution_count":null,"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"elapsed":519440,"status":"ok","timestamp":1777149944552,"user":{"displayName":"fukU Google","userId":"02763165356193834046"},"user_tz":-120},"id":"qzTQnpEa_uk3","outputId":"ba57325d-3499-4b04-8595-9d3ec38b3959"},"outputs":[{"name":"stdout","output_type":"stream","text":["🚀 Found 26 FP16 shards. Starting REAL SDNQ uint4 quantization...\n","\n","[1/26] Processing dequant_shard_00000.safetensors ...\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.0.mlp.fc1.weight  shape=torch.Size([4304, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.0.mlp.fc2.weight  shape=torch.Size([1152, 4304])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.0.self_attn.k_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.0.self_attn.out_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.0.self_attn.q_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.0.self_attn.v_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.1.mlp.fc1.weight  shape=torch.Size([4304, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.1.mlp.fc2.weight  shape=torch.Size([1152, 4304])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.1.self_attn.k_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.1.self_attn.out_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.1.self_attn.q_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.1.self_attn.v_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.10.mlp.fc1.weight  shape=torch.Size([4304, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.10.mlp.fc2.weight  shape=torch.Size([1152, 4304])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.10.self_attn.k_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.10.self_attn.out_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.10.self_attn.q_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.10.self_attn.v_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.11.mlp.fc1.weight  shape=torch.Size([4304, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.11.mlp.fc2.weight  shape=torch.Size([1152, 4304])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.11.self_attn.k_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.11.self_attn.out_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.11.self_attn.q_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.11.self_attn.v_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.12.mlp.fc1.weight  shape=torch.Size([4304, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.12.mlp.fc2.weight  shape=torch.Size([1152, 4304])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.12.self_attn.k_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.12.self_attn.out_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.12.self_attn.q_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.12.self_attn.v_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.13.mlp.fc1.weight  shape=torch.Size([4304, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.13.mlp.fc2.weight  shape=torch.Size([1152, 4304])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.13.self_attn.k_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.13.self_attn.out_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.13.self_attn.q_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.13.self_attn.v_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.14.mlp.fc1.weight  shape=torch.Size([4304, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.14.mlp.fc2.weight  shape=torch.Size([1152, 4304])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.14.self_attn.k_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.14.self_attn.out_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.14.self_attn.q_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.14.self_attn.v_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.15.mlp.fc1.weight  shape=torch.Size([4304, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.15.mlp.fc2.weight  shape=torch.Size([1152, 4304])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.15.self_attn.k_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.15.self_attn.out_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.15.self_attn.q_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.15.self_attn.v_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.16.mlp.fc1.weight  shape=torch.Size([4304, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.16.mlp.fc2.weight  shape=torch.Size([1152, 4304])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.16.self_attn.k_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.16.self_attn.out_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.16.self_attn.q_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.16.self_attn.v_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.17.mlp.fc1.weight  shape=torch.Size([4304, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.17.self_attn.k_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.17.self_attn.out_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.17.self_attn.q_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.17.self_attn.v_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.2.mlp.fc1.weight  shape=torch.Size([4304, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.2.mlp.fc2.weight  shape=torch.Size([1152, 4304])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.2.self_attn.k_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.2.self_attn.out_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.2.self_attn.q_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.2.self_attn.v_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.3.mlp.fc1.weight  shape=torch.Size([4304, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.3.mlp.fc2.weight  shape=torch.Size([1152, 4304])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.3.self_attn.k_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.3.self_attn.out_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.3.self_attn.q_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.3.self_attn.v_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.4.mlp.fc1.weight  shape=torch.Size([4304, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.4.mlp.fc2.weight  shape=torch.Size([1152, 4304])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.4.self_attn.k_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.4.self_attn.out_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.4.self_attn.q_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.4.self_attn.v_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.5.mlp.fc1.weight  shape=torch.Size([4304, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.5.mlp.fc2.weight  shape=torch.Size([1152, 4304])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.5.self_attn.k_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.5.self_attn.out_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.5.self_attn.q_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.5.self_attn.v_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.6.mlp.fc1.weight  shape=torch.Size([4304, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.6.mlp.fc2.weight  shape=torch.Size([1152, 4304])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.6.self_attn.k_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.6.self_attn.out_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.6.self_attn.q_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.6.self_attn.v_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.7.mlp.fc1.weight  shape=torch.Size([4304, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.7.mlp.fc2.weight  shape=torch.Size([1152, 4304])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.7.self_attn.k_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.7.self_attn.out_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.7.self_attn.q_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.7.self_attn.v_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.8.mlp.fc1.weight  shape=torch.Size([4304, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.8.mlp.fc2.weight  shape=torch.Size([1152, 4304])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.8.self_attn.k_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.8.self_attn.out_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.8.self_attn.q_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.8.self_attn.v_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.9.mlp.fc1.weight  shape=torch.Size([4304, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.9.mlp.fc2.weight  shape=torch.Size([1152, 4304])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.9.self_attn.k_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.9.self_attn.out_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.9.self_attn.q_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.9.self_attn.v_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   💾 Saved quantized shard → /content/joycaption_sdnq_shards/dequant_shard_00000.safetensors\n","\n","[2/26] Processing dequant_shard_00001.safetensors ...\n","   → Quantizing model.language_model.layers.0.mlp.gate_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.0.self_attn.k_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.0.self_attn.o_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.0.self_attn.q_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.0.self_attn.v_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.multi_modal_projector.linear_1.weight  shape=torch.Size([4096, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.multi_modal_projector.linear_2.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.17.mlp.fc2.weight  shape=torch.Size([1152, 4304])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.18.mlp.fc1.weight  shape=torch.Size([4304, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.18.mlp.fc2.weight  shape=torch.Size([1152, 4304])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.18.self_attn.k_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.18.self_attn.out_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.18.self_attn.q_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.18.self_attn.v_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.19.mlp.fc1.weight  shape=torch.Size([4304, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.19.mlp.fc2.weight  shape=torch.Size([1152, 4304])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.19.self_attn.k_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.19.self_attn.out_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.19.self_attn.q_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.19.self_attn.v_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.20.mlp.fc1.weight  shape=torch.Size([4304, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.20.mlp.fc2.weight  shape=torch.Size([1152, 4304])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.20.self_attn.k_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.20.self_attn.out_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.20.self_attn.q_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.20.self_attn.v_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.21.mlp.fc1.weight  shape=torch.Size([4304, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.21.mlp.fc2.weight  shape=torch.Size([1152, 4304])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.21.self_attn.k_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.21.self_attn.out_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.21.self_attn.q_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.21.self_attn.v_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.22.mlp.fc1.weight  shape=torch.Size([4304, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.22.mlp.fc2.weight  shape=torch.Size([1152, 4304])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.22.self_attn.k_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.22.self_attn.out_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.22.self_attn.q_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.22.self_attn.v_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.23.mlp.fc1.weight  shape=torch.Size([4304, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.23.mlp.fc2.weight  shape=torch.Size([1152, 4304])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.23.self_attn.k_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.23.self_attn.out_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.23.self_attn.q_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.23.self_attn.v_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.24.mlp.fc1.weight  shape=torch.Size([4304, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.24.mlp.fc2.weight  shape=torch.Size([1152, 4304])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.24.self_attn.k_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.24.self_attn.out_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.24.self_attn.q_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.24.self_attn.v_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.25.mlp.fc1.weight  shape=torch.Size([4304, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.25.mlp.fc2.weight  shape=torch.Size([1152, 4304])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.25.self_attn.k_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.25.self_attn.out_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.25.self_attn.q_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.25.self_attn.v_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.26.mlp.fc1.weight  shape=torch.Size([4304, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.26.mlp.fc2.weight  shape=torch.Size([1152, 4304])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.26.self_attn.k_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.26.self_attn.out_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.26.self_attn.q_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.encoder.layers.26.self_attn.v_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.head.attention.out_proj.weight  shape=torch.Size([1152, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.head.mlp.fc1.weight  shape=torch.Size([4304, 1152])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.vision_tower.vision_model.head.mlp.fc2.weight  shape=torch.Size([1152, 4304])\n","   ✅ Success (uint4 packed)\n","   💾 Saved quantized shard → /content/joycaption_sdnq_shards/dequant_shard_00001.safetensors\n","\n","[3/26] Processing dequant_shard_00002.safetensors ...\n","   → Quantizing model.language_model.layers.0.mlp.down_proj.weight  shape=torch.Size([4096, 14336])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.0.mlp.up_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.1.mlp.gate_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.1.mlp.up_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.1.self_attn.k_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.1.self_attn.o_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.1.self_attn.q_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.1.self_attn.v_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   💾 Saved quantized shard → /content/joycaption_sdnq_shards/dequant_shard_00002.safetensors\n","\n","[4/26] Processing dequant_shard_00003.safetensors ...\n","   → Quantizing model.language_model.layers.1.mlp.down_proj.weight  shape=torch.Size([4096, 14336])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.2.mlp.down_proj.weight  shape=torch.Size([4096, 14336])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.2.mlp.gate_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.2.mlp.up_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.2.self_attn.k_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.2.self_attn.o_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.2.self_attn.q_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.2.self_attn.v_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   💾 Saved quantized shard → /content/joycaption_sdnq_shards/dequant_shard_00003.safetensors\n","\n","[5/26] Processing dequant_shard_00004.safetensors ...\n","   → Quantizing model.language_model.layers.3.mlp.down_proj.weight  shape=torch.Size([4096, 14336])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.3.mlp.gate_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.3.mlp.up_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.3.self_attn.k_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.3.self_attn.o_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.3.self_attn.q_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.3.self_attn.v_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.4.mlp.gate_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.4.self_attn.k_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.4.self_attn.o_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.4.self_attn.q_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.4.self_attn.v_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   💾 Saved quantized shard → /content/joycaption_sdnq_shards/dequant_shard_00004.safetensors\n","\n","[6/26] Processing dequant_shard_00005.safetensors ...\n","   → Quantizing model.language_model.layers.4.mlp.down_proj.weight  shape=torch.Size([4096, 14336])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.4.mlp.up_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.5.mlp.gate_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.5.mlp.up_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.5.self_attn.k_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.5.self_attn.o_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.5.self_attn.q_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.5.self_attn.v_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   💾 Saved quantized shard → /content/joycaption_sdnq_shards/dequant_shard_00005.safetensors\n","\n","[7/26] Processing dequant_shard_00006.safetensors ...\n","   → Quantizing model.language_model.layers.5.mlp.down_proj.weight  shape=torch.Size([4096, 14336])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.6.mlp.down_proj.weight  shape=torch.Size([4096, 14336])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.6.mlp.gate_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.6.mlp.up_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.6.self_attn.k_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.6.self_attn.o_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.6.self_attn.q_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.6.self_attn.v_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   💾 Saved quantized shard → /content/joycaption_sdnq_shards/dequant_shard_00006.safetensors\n","\n","[8/26] Processing dequant_shard_00007.safetensors ...\n","   → Quantizing model.language_model.layers.7.mlp.down_proj.weight  shape=torch.Size([4096, 14336])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.7.mlp.gate_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.7.mlp.up_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.7.self_attn.k_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.7.self_attn.o_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.7.self_attn.q_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.7.self_attn.v_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.8.mlp.gate_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.8.self_attn.k_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.8.self_attn.o_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.8.self_attn.q_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.8.self_attn.v_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   💾 Saved quantized shard → /content/joycaption_sdnq_shards/dequant_shard_00007.safetensors\n","\n","[9/26] Processing dequant_shard_00008.safetensors ...\n","   → Quantizing model.language_model.layers.8.mlp.down_proj.weight  shape=torch.Size([4096, 14336])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.8.mlp.up_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.9.mlp.gate_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.9.mlp.up_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.9.self_attn.k_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.9.self_attn.o_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.9.self_attn.q_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.9.self_attn.v_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   💾 Saved quantized shard → /content/joycaption_sdnq_shards/dequant_shard_00008.safetensors\n","\n","[10/26] Processing dequant_shard_00009.safetensors ...\n","   → Quantizing model.language_model.layers.10.mlp.down_proj.weight  shape=torch.Size([4096, 14336])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.10.mlp.gate_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.10.mlp.up_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.10.self_attn.k_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.10.self_attn.o_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.10.self_attn.q_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.10.self_attn.v_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.9.mlp.down_proj.weight  shape=torch.Size([4096, 14336])\n","   ✅ Success (uint4 packed)\n","   💾 Saved quantized shard → /content/joycaption_sdnq_shards/dequant_shard_00009.safetensors\n","\n","[11/26] Processing dequant_shard_00010.safetensors ...\n","   → Quantizing model.language_model.layers.11.mlp.down_proj.weight  shape=torch.Size([4096, 14336])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.11.mlp.gate_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.11.mlp.up_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.11.self_attn.k_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.11.self_attn.o_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.11.self_attn.q_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.11.self_attn.v_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.12.mlp.gate_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.12.self_attn.k_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.12.self_attn.o_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.12.self_attn.q_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.12.self_attn.v_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   💾 Saved quantized shard → /content/joycaption_sdnq_shards/dequant_shard_00010.safetensors\n","\n","[12/26] Processing dequant_shard_00011.safetensors ...\n","   → Quantizing model.language_model.layers.12.mlp.down_proj.weight  shape=torch.Size([4096, 14336])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.12.mlp.up_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.13.mlp.gate_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.13.mlp.up_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.13.self_attn.k_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.13.self_attn.o_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.13.self_attn.q_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.13.self_attn.v_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   💾 Saved quantized shard → /content/joycaption_sdnq_shards/dequant_shard_00011.safetensors\n","\n","[13/26] Processing dequant_shard_00012.safetensors ...\n","   → Quantizing model.language_model.layers.13.mlp.down_proj.weight  shape=torch.Size([4096, 14336])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.14.mlp.down_proj.weight  shape=torch.Size([4096, 14336])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.14.mlp.gate_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.14.mlp.up_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.14.self_attn.k_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.14.self_attn.o_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.14.self_attn.q_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.14.self_attn.v_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   💾 Saved quantized shard → /content/joycaption_sdnq_shards/dequant_shard_00012.safetensors\n","\n","[14/26] Processing dequant_shard_00013.safetensors ...\n","   → Quantizing model.language_model.layers.15.mlp.down_proj.weight  shape=torch.Size([4096, 14336])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.15.mlp.gate_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.15.mlp.up_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.15.self_attn.k_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.15.self_attn.o_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.15.self_attn.q_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.15.self_attn.v_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.16.mlp.gate_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.16.self_attn.k_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.16.self_attn.o_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.16.self_attn.q_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.16.self_attn.v_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   💾 Saved quantized shard → /content/joycaption_sdnq_shards/dequant_shard_00013.safetensors\n","\n","[15/26] Processing dequant_shard_00014.safetensors ...\n","   → Quantizing model.language_model.layers.16.mlp.down_proj.weight  shape=torch.Size([4096, 14336])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.16.mlp.up_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.17.mlp.gate_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.17.mlp.up_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.17.self_attn.k_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.17.self_attn.o_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.17.self_attn.q_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.17.self_attn.v_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   💾 Saved quantized shard → /content/joycaption_sdnq_shards/dequant_shard_00014.safetensors\n","\n","[16/26] Processing dequant_shard_00015.safetensors ...\n","   → Quantizing model.language_model.layers.17.mlp.down_proj.weight  shape=torch.Size([4096, 14336])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.18.mlp.down_proj.weight  shape=torch.Size([4096, 14336])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.18.mlp.gate_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.18.mlp.up_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.18.self_attn.k_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.18.self_attn.o_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.18.self_attn.q_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.18.self_attn.v_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   💾 Saved quantized shard → /content/joycaption_sdnq_shards/dequant_shard_00015.safetensors\n","\n","[17/26] Processing dequant_shard_00016.safetensors ...\n","   → Quantizing model.language_model.layers.19.mlp.down_proj.weight  shape=torch.Size([4096, 14336])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.19.mlp.gate_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.19.mlp.up_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.19.self_attn.k_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.19.self_attn.o_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.19.self_attn.q_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.19.self_attn.v_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.20.mlp.gate_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.20.self_attn.k_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.20.self_attn.o_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.20.self_attn.q_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.20.self_attn.v_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   💾 Saved quantized shard → /content/joycaption_sdnq_shards/dequant_shard_00016.safetensors\n","\n","[18/26] Processing dequant_shard_00017.safetensors ...\n","   → Quantizing model.language_model.layers.20.mlp.down_proj.weight  shape=torch.Size([4096, 14336])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.20.mlp.up_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.21.mlp.gate_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.21.mlp.up_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.21.self_attn.k_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.21.self_attn.o_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.21.self_attn.q_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.21.self_attn.v_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   💾 Saved quantized shard → /content/joycaption_sdnq_shards/dequant_shard_00017.safetensors\n","\n","[19/26] Processing dequant_shard_00018.safetensors ...\n","   → Quantizing model.language_model.layers.21.mlp.down_proj.weight  shape=torch.Size([4096, 14336])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.22.mlp.down_proj.weight  shape=torch.Size([4096, 14336])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.22.mlp.gate_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.22.mlp.up_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.22.self_attn.k_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.22.self_attn.o_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.22.self_attn.q_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.22.self_attn.v_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   💾 Saved quantized shard → /content/joycaption_sdnq_shards/dequant_shard_00018.safetensors\n","\n","[20/26] Processing dequant_shard_00019.safetensors ...\n","   → Quantizing model.language_model.layers.23.mlp.down_proj.weight  shape=torch.Size([4096, 14336])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.23.mlp.gate_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.23.mlp.up_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.23.self_attn.k_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.23.self_attn.o_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.23.self_attn.q_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.23.self_attn.v_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.24.mlp.gate_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.24.self_attn.k_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.24.self_attn.o_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.24.self_attn.q_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.24.self_attn.v_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   💾 Saved quantized shard → /content/joycaption_sdnq_shards/dequant_shard_00019.safetensors\n","\n","[21/26] Processing dequant_shard_00020.safetensors ...\n","   → Quantizing model.language_model.layers.24.mlp.down_proj.weight  shape=torch.Size([4096, 14336])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.24.mlp.up_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.25.mlp.gate_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.25.mlp.up_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.25.self_attn.k_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.25.self_attn.o_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.25.self_attn.q_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.25.self_attn.v_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   💾 Saved quantized shard → /content/joycaption_sdnq_shards/dequant_shard_00020.safetensors\n","\n","[22/26] Processing dequant_shard_00021.safetensors ...\n","   → Quantizing model.language_model.layers.25.mlp.down_proj.weight  shape=torch.Size([4096, 14336])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.26.mlp.down_proj.weight  shape=torch.Size([4096, 14336])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.26.mlp.gate_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.26.mlp.up_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.26.self_attn.k_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.26.self_attn.o_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.26.self_attn.q_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.26.self_attn.v_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   💾 Saved quantized shard → /content/joycaption_sdnq_shards/dequant_shard_00021.safetensors\n","\n","[23/26] Processing dequant_shard_00022.safetensors ...\n","   → Quantizing model.language_model.layers.27.mlp.down_proj.weight  shape=torch.Size([4096, 14336])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.27.mlp.gate_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.27.mlp.up_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.27.self_attn.k_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.27.self_attn.o_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.27.self_attn.q_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.27.self_attn.v_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.28.mlp.gate_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.28.self_attn.k_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.28.self_attn.o_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.28.self_attn.q_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.28.self_attn.v_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   💾 Saved quantized shard → /content/joycaption_sdnq_shards/dequant_shard_00022.safetensors\n","\n","[24/26] Processing dequant_shard_00023.safetensors ...\n","   → Quantizing model.language_model.layers.28.mlp.down_proj.weight  shape=torch.Size([4096, 14336])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.28.mlp.up_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.29.mlp.gate_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.29.mlp.up_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.29.self_attn.k_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.29.self_attn.o_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.29.self_attn.q_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.29.self_attn.v_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   💾 Saved quantized shard → /content/joycaption_sdnq_shards/dequant_shard_00023.safetensors\n","\n","[25/26] Processing dequant_shard_00024.safetensors ...\n","   → Quantizing model.language_model.layers.29.mlp.down_proj.weight  shape=torch.Size([4096, 14336])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.30.mlp.down_proj.weight  shape=torch.Size([4096, 14336])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.30.mlp.gate_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.30.mlp.up_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.30.self_attn.k_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.30.self_attn.o_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.30.self_attn.q_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.30.self_attn.v_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   💾 Saved quantized shard → /content/joycaption_sdnq_shards/dequant_shard_00024.safetensors\n","\n","[26/26] Processing dequant_shard_00025.safetensors ...\n","   → Quantizing model.language_model.layers.31.mlp.down_proj.weight  shape=torch.Size([4096, 14336])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.31.mlp.gate_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.31.mlp.up_proj.weight  shape=torch.Size([14336, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.31.self_attn.k_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.31.self_attn.o_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.31.self_attn.q_proj.weight  shape=torch.Size([4096, 4096])\n","   ✅ Success (uint4 packed)\n","   → Quantizing model.language_model.layers.31.self_attn.v_proj.weight  shape=torch.Size([1024, 4096])\n","   ✅ Success (uint4 packed)\n","   💾 Saved quantized shard → /content/joycaption_sdnq_shards/dequant_shard_00025.safetensors\n","\n","🎉 ALL shards processed and saved to /content/joycaption_sdnq_shards\n"]}],"source":["import os\n","import torch\n","import gc\n","from safetensors.torch import load_file, save_file\n","from sdnq.quantizer import sdnq_quantize_layer   # low-level layer quantizer\n","\n","# ========================= CONFIG =========================\n","FP16_SHARD_DIR = \"/content/joycaption_dequant_shards\"\n","SDNQ_SHARD_DIR = \"/content/joycaption_sdnq_shards\"\n","os.makedirs(SDNQ_SHARD_DIR, exist_ok=True)\n","\n","# ONLY parameters that sdnq_quantize_layer actually accepts\n","QUANT_CONFIG = {\n","    \"weights_dtype\": \"uint4\",           # ← this is what you want\n","    \"group_size\": 128,\n","    \"quantization_device\": \"cpu\",\n","    \"return_device\": \"cpu\",\n","    \"use_svd\": False,\n","    \"quant_conv\": False,\n","    \"quant_embedding\": False,\n","    \"use_quantized_matmul\": False,\n","    \"use_quantized_matmul_conv\": False,\n","    \"use_dynamic_quantization\": False,\n","    \"dequantize_fp32\": True,\n","    \"non_blocking\": False,\n","}\n","# =========================================================\n","\n","shard_files = sorted([f for f in os.listdir(FP16_SHARD_DIR) if f.endswith(\".safetensors\")])\n","\n","print(f\"🚀 Found {len(shard_files)} FP16 shards. Starting REAL SDNQ uint4 quantization...\\n\")\n","\n","for i, shard_file in enumerate(shard_files):\n","    shard_path = os.path.join(FP16_SHARD_DIR, shard_file)\n","    print(f\"[{i+1}/{len(shard_files)}] Processing {shard_file} ...\")\n","\n","    state = load_file(shard_path, device=\"cpu\")\n","    new_state = {}\n","\n","    for key, tensor in state.items():\n","        if not key.endswith(\".weight\") or tensor.dtype != torch.float16:\n","            new_state[key] = tensor\n","            continue\n","\n","        print(f\"   → Quantizing {key}  shape={tensor.shape}\")\n","\n","        try:\n","            # Tiny dummy Linear layer for this single weight\n","            dummy = torch.nn.Linear(\n","                in_features=tensor.shape[1],\n","                out_features=tensor.shape[0],\n","                bias=False,\n","                dtype=torch.float16\n","            )\n","            dummy.weight = torch.nn.Parameter(tensor.clone())\n","\n","            # REAL SDNQ layer quantization (now with correct kwargs)\n","            quantized_dummy, _, _ = sdnq_quantize_layer(dummy, **QUANT_CONFIG)\n","\n","            # Extract packed weight + metadata\n","            new_state[key] = quantized_dummy.weight.data\n","\n","            for attr in [\"scale\", \"zero_point\", \"svd_up\", \"svd_down\"]:\n","                if hasattr(quantized_dummy, attr) and getattr(quantized_dummy, attr) is not None:\n","                    meta_key = key.replace(\".weight\", f\".{attr}\")\n","                    new_state[meta_key] = getattr(quantized_dummy, attr)\n","\n","            print(f\"   ✅ Success (uint4 packed)\")\n","\n","        except Exception as e:\n","            print(f\"   ⚠️ Failed {key}: {e}\")\n","            new_state[key] = tensor  # fallback to original\n","\n","        # Immediate cleanup\n","        del dummy, quantized_dummy, tensor\n","        gc.collect()\n","\n","    # Save the new quantized shard (same filename)\n","    new_shard_path = os.path.join(SDNQ_SHARD_DIR, shard_file)\n","    save_file(new_state, new_shard_path)\n","    print(f\"   💾 Saved quantized shard → {new_shard_path}\\n\")\n","\n","    del state, new_state\n","    gc.collect()\n","\n","\n","#This code works. Success!\n","#✅ Success (uint4 packed)\n","#File size of /content/joycaption_sdnq_shards/dequant_shard_00000.safetensors:\n","#-rw-r--r-- 1 root root 142M Apr 25 18:56 /content/joycaption_sdnq_shards/dequant_shard_00000.safetensors\n","\n","#💾 Saved quantized shard → /content/joycaption_sdnq_shards/dequant_shard_00025.safetensors\n","#🎉 ALL shards processed and saved to /content/joycaption_sdnq_shards\n","print(\"🎉 ALL shards processed and saved to\", SDNQ_SHARD_DIR)"]},{"cell_type":"code","execution_count":null,"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"elapsed":222,"status":"ok","timestamp":1777149944773,"user":{"displayName":"fukU Google","userId":"02763165356193834046"},"user_tz":-120},"id":"FrdD6iBIR0O-","outputId":"ce826a6d-0fdf-40f8-edde-446b169eb489"},"outputs":[{"data":{"text/plain":["0"]},"execution_count":7,"metadata":{},"output_type":"execute_result"}],"source":["import torch , gc\n","torch.cuda.empty_cache()\n","gc.collect()"]},{"cell_type":"code","execution_count":null,"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"elapsed":127496,"status":"ok","timestamp":1777150072273,"user":{"displayName":"fukU Google","userId":"02763165356193834046"},"user_tz":-120},"id":"oTuEt1K_Qgr1","outputId":"a8c27096-6257-47d7-98af-e802290143ba"},"outputs":[{"name":"stdout","output_type":"stream","text":["🚀 In-place SDNQ uint4 quantization on the existing model in RAM...\n","Quantizing model.vision_tower.vision_model.encoder.layers.0.self_attn.k_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.0.self_attn.v_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.0.self_attn.q_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.0.self_attn.out_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.0.mlp.fc1  shape=torch.Size([4304, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.0.mlp.fc2  shape=torch.Size([1152, 4304])\n","Quantizing model.vision_tower.vision_model.encoder.layers.1.self_attn.k_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.1.self_attn.v_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.1.self_attn.q_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.1.self_attn.out_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.1.mlp.fc1  shape=torch.Size([4304, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.1.mlp.fc2  shape=torch.Size([1152, 4304])\n","Quantizing model.vision_tower.vision_model.encoder.layers.2.self_attn.k_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.2.self_attn.v_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.2.self_attn.q_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.2.self_attn.out_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.2.mlp.fc1  shape=torch.Size([4304, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.2.mlp.fc2  shape=torch.Size([1152, 4304])\n","Quantizing model.vision_tower.vision_model.encoder.layers.3.self_attn.k_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.3.self_attn.v_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.3.self_attn.q_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.3.self_attn.out_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.3.mlp.fc1  shape=torch.Size([4304, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.3.mlp.fc2  shape=torch.Size([1152, 4304])\n","Quantizing model.vision_tower.vision_model.encoder.layers.4.self_attn.k_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.4.self_attn.v_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.4.self_attn.q_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.4.self_attn.out_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.4.mlp.fc1  shape=torch.Size([4304, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.4.mlp.fc2  shape=torch.Size([1152, 4304])\n","Quantizing model.vision_tower.vision_model.encoder.layers.5.self_attn.k_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.5.self_attn.v_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.5.self_attn.q_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.5.self_attn.out_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.5.mlp.fc1  shape=torch.Size([4304, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.5.mlp.fc2  shape=torch.Size([1152, 4304])\n","Quantizing model.vision_tower.vision_model.encoder.layers.6.self_attn.k_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.6.self_attn.v_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.6.self_attn.q_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.6.self_attn.out_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.6.mlp.fc1  shape=torch.Size([4304, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.6.mlp.fc2  shape=torch.Size([1152, 4304])\n","Quantizing model.vision_tower.vision_model.encoder.layers.7.self_attn.k_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.7.self_attn.v_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.7.self_attn.q_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.7.self_attn.out_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.7.mlp.fc1  shape=torch.Size([4304, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.7.mlp.fc2  shape=torch.Size([1152, 4304])\n","Quantizing model.vision_tower.vision_model.encoder.layers.8.self_attn.k_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.8.self_attn.v_proj  shape=torch.Size([1152, 1152])\n","   → 50 layers done\n","Quantizing model.vision_tower.vision_model.encoder.layers.8.self_attn.q_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.8.self_attn.out_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.8.mlp.fc1  shape=torch.Size([4304, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.8.mlp.fc2  shape=torch.Size([1152, 4304])\n","Quantizing model.vision_tower.vision_model.encoder.layers.9.self_attn.k_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.9.self_attn.v_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.9.self_attn.q_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.9.self_attn.out_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.9.mlp.fc1  shape=torch.Size([4304, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.9.mlp.fc2  shape=torch.Size([1152, 4304])\n","Quantizing model.vision_tower.vision_model.encoder.layers.10.self_attn.k_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.10.self_attn.v_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.10.self_attn.q_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.10.self_attn.out_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.10.mlp.fc1  shape=torch.Size([4304, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.10.mlp.fc2  shape=torch.Size([1152, 4304])\n","Quantizing model.vision_tower.vision_model.encoder.layers.11.self_attn.k_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.11.self_attn.v_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.11.self_attn.q_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.11.self_attn.out_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.11.mlp.fc1  shape=torch.Size([4304, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.11.mlp.fc2  shape=torch.Size([1152, 4304])\n","Quantizing model.vision_tower.vision_model.encoder.layers.12.self_attn.k_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.12.self_attn.v_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.12.self_attn.q_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.12.self_attn.out_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.12.mlp.fc1  shape=torch.Size([4304, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.12.mlp.fc2  shape=torch.Size([1152, 4304])\n","Quantizing model.vision_tower.vision_model.encoder.layers.13.self_attn.k_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.13.self_attn.v_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.13.self_attn.q_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.13.self_attn.out_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.13.mlp.fc1  shape=torch.Size([4304, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.13.mlp.fc2  shape=torch.Size([1152, 4304])\n","Quantizing model.vision_tower.vision_model.encoder.layers.14.self_attn.k_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.14.self_attn.v_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.14.self_attn.q_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.14.self_attn.out_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.14.mlp.fc1  shape=torch.Size([4304, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.14.mlp.fc2  shape=torch.Size([1152, 4304])\n","Quantizing model.vision_tower.vision_model.encoder.layers.15.self_attn.k_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.15.self_attn.v_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.15.self_attn.q_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.15.self_attn.out_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.15.mlp.fc1  shape=torch.Size([4304, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.15.mlp.fc2  shape=torch.Size([1152, 4304])\n","Quantizing model.vision_tower.vision_model.encoder.layers.16.self_attn.k_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.16.self_attn.v_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.16.self_attn.q_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.16.self_attn.out_proj  shape=torch.Size([1152, 1152])\n","   → 100 layers done\n","Quantizing model.vision_tower.vision_model.encoder.layers.16.mlp.fc1  shape=torch.Size([4304, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.16.mlp.fc2  shape=torch.Size([1152, 4304])\n","Quantizing model.vision_tower.vision_model.encoder.layers.17.self_attn.k_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.17.self_attn.v_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.17.self_attn.q_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.17.self_attn.out_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.17.mlp.fc1  shape=torch.Size([4304, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.17.mlp.fc2  shape=torch.Size([1152, 4304])\n","Quantizing model.vision_tower.vision_model.encoder.layers.18.self_attn.k_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.18.self_attn.v_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.18.self_attn.q_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.18.self_attn.out_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.18.mlp.fc1  shape=torch.Size([4304, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.18.mlp.fc2  shape=torch.Size([1152, 4304])\n","Quantizing model.vision_tower.vision_model.encoder.layers.19.self_attn.k_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.19.self_attn.v_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.19.self_attn.q_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.19.self_attn.out_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.19.mlp.fc1  shape=torch.Size([4304, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.19.mlp.fc2  shape=torch.Size([1152, 4304])\n","Quantizing model.vision_tower.vision_model.encoder.layers.20.self_attn.k_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.20.self_attn.v_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.20.self_attn.q_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.20.self_attn.out_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.20.mlp.fc1  shape=torch.Size([4304, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.20.mlp.fc2  shape=torch.Size([1152, 4304])\n","Quantizing model.vision_tower.vision_model.encoder.layers.21.self_attn.k_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.21.self_attn.v_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.21.self_attn.q_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.21.self_attn.out_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.21.mlp.fc1  shape=torch.Size([4304, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.21.mlp.fc2  shape=torch.Size([1152, 4304])\n","Quantizing model.vision_tower.vision_model.encoder.layers.22.self_attn.k_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.22.self_attn.v_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.22.self_attn.q_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.22.self_attn.out_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.22.mlp.fc1  shape=torch.Size([4304, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.22.mlp.fc2  shape=torch.Size([1152, 4304])\n","Quantizing model.vision_tower.vision_model.encoder.layers.23.self_attn.k_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.23.self_attn.v_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.23.self_attn.q_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.23.self_attn.out_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.23.mlp.fc1  shape=torch.Size([4304, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.23.mlp.fc2  shape=torch.Size([1152, 4304])\n","Quantizing model.vision_tower.vision_model.encoder.layers.24.self_attn.k_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.24.self_attn.v_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.24.self_attn.q_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.24.self_attn.out_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.24.mlp.fc1  shape=torch.Size([4304, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.24.mlp.fc2  shape=torch.Size([1152, 4304])\n","   → 150 layers done\n","Quantizing model.vision_tower.vision_model.encoder.layers.25.self_attn.k_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.25.self_attn.v_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.25.self_attn.q_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.25.self_attn.out_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.25.mlp.fc1  shape=torch.Size([4304, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.25.mlp.fc2  shape=torch.Size([1152, 4304])\n","Quantizing model.vision_tower.vision_model.encoder.layers.26.self_attn.k_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.26.self_attn.v_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.26.self_attn.q_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.26.self_attn.out_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.26.mlp.fc1  shape=torch.Size([4304, 1152])\n","Quantizing model.vision_tower.vision_model.encoder.layers.26.mlp.fc2  shape=torch.Size([1152, 4304])\n","Quantizing model.vision_tower.vision_model.head.attention.out_proj  shape=torch.Size([1152, 1152])\n","Quantizing model.vision_tower.vision_model.head.mlp.fc1  shape=torch.Size([4304, 1152])\n","Quantizing model.vision_tower.vision_model.head.mlp.fc2  shape=torch.Size([1152, 4304])\n","Quantizing model.multi_modal_projector.linear_1  shape=torch.Size([4096, 1152])\n","Quantizing model.multi_modal_projector.linear_2  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.0.self_attn.q_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.0.self_attn.k_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.0.self_attn.v_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.0.self_attn.o_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.0.mlp.gate_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.0.mlp.up_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.0.mlp.down_proj  shape=torch.Size([4096, 14336])\n","Quantizing model.language_model.layers.1.self_attn.q_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.1.self_attn.k_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.1.self_attn.v_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.1.self_attn.o_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.1.mlp.gate_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.1.mlp.up_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.1.mlp.down_proj  shape=torch.Size([4096, 14336])\n","Quantizing model.language_model.layers.2.self_attn.q_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.2.self_attn.k_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.2.self_attn.v_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.2.self_attn.o_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.2.mlp.gate_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.2.mlp.up_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.2.mlp.down_proj  shape=torch.Size([4096, 14336])\n","Quantizing model.language_model.layers.3.self_attn.q_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.3.self_attn.k_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.3.self_attn.v_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.3.self_attn.o_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.3.mlp.gate_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.3.mlp.up_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.3.mlp.down_proj  shape=torch.Size([4096, 14336])\n","Quantizing model.language_model.layers.4.self_attn.q_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.4.self_attn.k_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.4.self_attn.v_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.4.self_attn.o_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.4.mlp.gate_proj  shape=torch.Size([14336, 4096])\n","   → 200 layers done\n","Quantizing model.language_model.layers.4.mlp.up_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.4.mlp.down_proj  shape=torch.Size([4096, 14336])\n","Quantizing model.language_model.layers.5.self_attn.q_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.5.self_attn.k_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.5.self_attn.v_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.5.self_attn.o_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.5.mlp.gate_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.5.mlp.up_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.5.mlp.down_proj  shape=torch.Size([4096, 14336])\n","Quantizing model.language_model.layers.6.self_attn.q_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.6.self_attn.k_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.6.self_attn.v_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.6.self_attn.o_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.6.mlp.gate_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.6.mlp.up_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.6.mlp.down_proj  shape=torch.Size([4096, 14336])\n","Quantizing model.language_model.layers.7.self_attn.q_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.7.self_attn.k_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.7.self_attn.v_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.7.self_attn.o_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.7.mlp.gate_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.7.mlp.up_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.7.mlp.down_proj  shape=torch.Size([4096, 14336])\n","Quantizing model.language_model.layers.8.self_attn.q_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.8.self_attn.k_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.8.self_attn.v_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.8.self_attn.o_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.8.mlp.gate_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.8.mlp.up_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.8.mlp.down_proj  shape=torch.Size([4096, 14336])\n","Quantizing model.language_model.layers.9.self_attn.q_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.9.self_attn.k_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.9.self_attn.v_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.9.self_attn.o_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.9.mlp.gate_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.9.mlp.up_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.9.mlp.down_proj  shape=torch.Size([4096, 14336])\n","Quantizing model.language_model.layers.10.self_attn.q_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.10.self_attn.k_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.10.self_attn.v_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.10.self_attn.o_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.10.mlp.gate_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.10.mlp.up_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.10.mlp.down_proj  shape=torch.Size([4096, 14336])\n","Quantizing model.language_model.layers.11.self_attn.q_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.11.self_attn.k_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.11.self_attn.v_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.11.self_attn.o_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.11.mlp.gate_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.11.mlp.up_proj  shape=torch.Size([14336, 4096])\n","   → 250 layers done\n","Quantizing model.language_model.layers.11.mlp.down_proj  shape=torch.Size([4096, 14336])\n","Quantizing model.language_model.layers.12.self_attn.q_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.12.self_attn.k_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.12.self_attn.v_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.12.self_attn.o_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.12.mlp.gate_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.12.mlp.up_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.12.mlp.down_proj  shape=torch.Size([4096, 14336])\n","Quantizing model.language_model.layers.13.self_attn.q_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.13.self_attn.k_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.13.self_attn.v_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.13.self_attn.o_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.13.mlp.gate_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.13.mlp.up_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.13.mlp.down_proj  shape=torch.Size([4096, 14336])\n","Quantizing model.language_model.layers.14.self_attn.q_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.14.self_attn.k_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.14.self_attn.v_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.14.self_attn.o_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.14.mlp.gate_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.14.mlp.up_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.14.mlp.down_proj  shape=torch.Size([4096, 14336])\n","Quantizing model.language_model.layers.15.self_attn.q_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.15.self_attn.k_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.15.self_attn.v_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.15.self_attn.o_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.15.mlp.gate_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.15.mlp.up_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.15.mlp.down_proj  shape=torch.Size([4096, 14336])\n","Quantizing model.language_model.layers.16.self_attn.q_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.16.self_attn.k_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.16.self_attn.v_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.16.self_attn.o_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.16.mlp.gate_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.16.mlp.up_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.16.mlp.down_proj  shape=torch.Size([4096, 14336])\n","Quantizing model.language_model.layers.17.self_attn.q_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.17.self_attn.k_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.17.self_attn.v_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.17.self_attn.o_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.17.mlp.gate_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.17.mlp.up_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.17.mlp.down_proj  shape=torch.Size([4096, 14336])\n","Quantizing model.language_model.layers.18.self_attn.q_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.18.self_attn.k_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.18.self_attn.v_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.18.self_attn.o_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.18.mlp.gate_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.18.mlp.up_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.18.mlp.down_proj  shape=torch.Size([4096, 14336])\n","   → 300 layers done\n","Quantizing model.language_model.layers.19.self_attn.q_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.19.self_attn.k_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.19.self_attn.v_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.19.self_attn.o_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.19.mlp.gate_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.19.mlp.up_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.19.mlp.down_proj  shape=torch.Size([4096, 14336])\n","Quantizing model.language_model.layers.20.self_attn.q_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.20.self_attn.k_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.20.self_attn.v_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.20.self_attn.o_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.20.mlp.gate_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.20.mlp.up_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.20.mlp.down_proj  shape=torch.Size([4096, 14336])\n","Quantizing model.language_model.layers.21.self_attn.q_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.21.self_attn.k_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.21.self_attn.v_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.21.self_attn.o_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.21.mlp.gate_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.21.mlp.up_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.21.mlp.down_proj  shape=torch.Size([4096, 14336])\n","Quantizing model.language_model.layers.22.self_attn.q_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.22.self_attn.k_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.22.self_attn.v_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.22.self_attn.o_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.22.mlp.gate_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.22.mlp.up_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.22.mlp.down_proj  shape=torch.Size([4096, 14336])\n","Quantizing model.language_model.layers.23.self_attn.q_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.23.self_attn.k_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.23.self_attn.v_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.23.self_attn.o_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.23.mlp.gate_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.23.mlp.up_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.23.mlp.down_proj  shape=torch.Size([4096, 14336])\n","Quantizing model.language_model.layers.24.self_attn.q_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.24.self_attn.k_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.24.self_attn.v_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.24.self_attn.o_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.24.mlp.gate_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.24.mlp.up_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.24.mlp.down_proj  shape=torch.Size([4096, 14336])\n","Quantizing model.language_model.layers.25.self_attn.q_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.25.self_attn.k_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.25.self_attn.v_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.25.self_attn.o_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.25.mlp.gate_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.25.mlp.up_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.25.mlp.down_proj  shape=torch.Size([4096, 14336])\n","Quantizing model.language_model.layers.26.self_attn.q_proj  shape=torch.Size([4096, 4096])\n","   → 350 layers done\n","Quantizing model.language_model.layers.26.self_attn.k_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.26.self_attn.v_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.26.self_attn.o_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.26.mlp.gate_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.26.mlp.up_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.26.mlp.down_proj  shape=torch.Size([4096, 14336])\n","Quantizing model.language_model.layers.27.self_attn.q_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.27.self_attn.k_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.27.self_attn.v_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.27.self_attn.o_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.27.mlp.gate_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.27.mlp.up_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.27.mlp.down_proj  shape=torch.Size([4096, 14336])\n","Quantizing model.language_model.layers.28.self_attn.q_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.28.self_attn.k_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.28.self_attn.v_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.28.self_attn.o_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.28.mlp.gate_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.28.mlp.up_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.28.mlp.down_proj  shape=torch.Size([4096, 14336])\n","Quantizing model.language_model.layers.29.self_attn.q_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.29.self_attn.k_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.29.self_attn.v_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.29.self_attn.o_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.29.mlp.gate_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.29.mlp.up_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.29.mlp.down_proj  shape=torch.Size([4096, 14336])\n","Quantizing model.language_model.layers.30.self_attn.q_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.30.self_attn.k_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.30.self_attn.v_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.30.self_attn.o_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.30.mlp.gate_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.30.mlp.up_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.30.mlp.down_proj  shape=torch.Size([4096, 14336])\n","Quantizing model.language_model.layers.31.self_attn.q_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.31.self_attn.k_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.31.self_attn.v_proj  shape=torch.Size([1024, 4096])\n","Quantizing model.language_model.layers.31.self_attn.o_proj  shape=torch.Size([4096, 4096])\n","Quantizing model.language_model.layers.31.mlp.gate_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.31.mlp.up_proj  shape=torch.Size([14336, 4096])\n","Quantizing model.language_model.layers.31.mlp.down_proj  shape=torch.Size([4096, 14336])\n","✅ In-place quantization finished! Quantized 391 layers\n","🔍 Model memory after uint4: 5.85 GB  ← should be ~5–6 GB\n","   Sample: model.vision_tower.vision_model.encoder.layers.0.self_attn.q_proj.weight | dtype=torch.uint8 | shape=torch.Size([663552]) | type=Parameter\n"]}],"source":["import torch\n","import gc\n","from sdnq.quantizer import sdnq_quantize_layer\n","\n","print(\"🚀 In-place SDNQ uint4 quantization on the existing model in RAM...\")\n","\n","# Same config that worked for your shards\n","QUANT_CONFIG = {\n","    \"weights_dtype\": \"uint4\",\n","    \"group_size\": 128,\n","    \"quantization_device\": \"cpu\",\n","    \"return_device\": \"cpu\",\n","    \"use_svd\": False,\n","    \"quant_conv\": False,\n","    \"quant_embedding\": False,\n","    \"use_quantized_matmul\": False,\n","    \"use_quantized_matmul_conv\": False,\n","    \"use_dynamic_quantization\": False,\n","    \"dequantize_fp32\": True,\n","    \"non_blocking\": False,\n","}\n","\n","quantized_count = 0\n","\n","for name, module in list(model.named_modules()):\n","    if not hasattr(module, \"weight\") or module.weight is None:\n","        continue\n","    if module.weight.dtype != torch.float16:   # only the dequantized FP16 weights\n","        continue\n","\n","    print(f\"Quantizing {name}  shape={module.weight.shape}\")\n","\n","    try:\n","        # In-place quantization on the real module\n","        module = sdnq_quantize_layer(module, **QUANT_CONFIG)\n","        quantized_count += 1\n","\n","        if quantized_count % 50 == 0:\n","            print(f\"   → {quantized_count} layers done\")\n","\n","    except Exception as e:\n","        print(f\"   ⚠️ Skipped {name}: {e}\")\n","\n","print(f\"✅ In-place quantization finished! Quantized {quantized_count} layers\")\n","\n","# Quick memory check\n","def get_model_memory(m):\n","    mem = 0\n","    for p in m.parameters():\n","        if p is not None:\n","            mem += p.numel() * p.element_size()\n","    for b in m.buffers():\n","        mem += b.numel() * b.element_size()\n","    return mem / (1024**3)\n","\n","print(f\"🔍 Model memory after uint4: {get_model_memory(model):.2f} GB  ← should be ~5–6 GB\")\n","\n","# Sample check (should now be uint8 packed)\n","for name, param in model.named_parameters():\n","    if any(x in name for x in [\"q_proj.weight\", \"gate_proj.weight\"]):\n","        print(f\"   Sample: {name} | dtype={param.dtype} | shape={param.shape} | type={type(param).__name__}\")\n","        break\n","\n","\n","#---#\n","#Cell ran successfully\n","#Quantizing model.language_model.layers.31.mlp.down_proj  shape=torch.Size([4096, 14336])\n","#✅ In-place quantization finished! Quantized 391 layers\n","#🔍 Model memory after uint4: 5.85 GB  ← should be ~5–6 GB\n","#  Sample: model.vision_tower.vision_model.encoder.layers.0.self_attn.q_proj.weight | dtype=torch.uint8 | shape=torch.Size([663552]) | type=Parameter\n"]},{"cell_type":"code","execution_count":null,"metadata":{"id":"9-Q1-GPfSaAC"},"outputs":[],"source":["model=model.to(\"cpu\")"]},{"cell_type":"code","execution_count":null,"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"elapsed":4,"status":"ok","timestamp":1777150074807,"user":{"displayName":"fukU Google","userId":"02763165356193834046"},"user_tz":-120},"id":"GExaYBHxYiJM","outputId":"ff415fa8-2a5f-46e9-e00c-38444e5127b5"},"outputs":[{"data":{"text/plain":["torch.autograd.grad_mode.set_grad_enabled(mode=False)"]},"execution_count":11,"metadata":{},"output_type":"execute_result"}],"source":["for p in model.parameters():\n","    p.requires_grad_(False)\n","\n","torch.set_grad_enabled(False)"]},{"cell_type":"code","execution_count":null,"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"elapsed":7,"status":"ok","timestamp":1777150148228,"user":{"displayName":"fukU Google","userId":"02763165356193834046"},"user_tz":-120},"id":"Y-3mg72oRUIA","outputId":"d83715e1-4892-43ca-cf07-5b99ee175c26"},"outputs":[{"data":{"text/plain":["0"]},"execution_count":13,"metadata":{},"output_type":"execute_result"}],"source":["import torch , gc\n","torch.cuda.empty_cache()\n","gc.collect()"]},{"cell_type":"code","execution_count":null,"metadata":{"colab":{"base_uri":"https://localhost:8080/","height":105},"executionInfo":{"elapsed":329091,"status":"ok","timestamp":1777150768250,"user":{"displayName":"fukU Google","userId":"02763165356193834046"},"user_tz":-120},"id":"Ks5EAQkbZwyA","outputId":"b8929aa6-3500-4c07-a59c-733c007ea84f"},"outputs":[{"name":"stdout","output_type":"stream","text":["💾 Saving with low-RAM streaming (safe_serialization + max_shard_size)...\n"]},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"0ec8085fecec4947877614e55ded6293","version_major":2,"version_minor":0},"text/plain":["Writing model shards:   0%|          | 0/17 [00:00<?, ?it/s]"]},"metadata":{},"output_type":"display_data"},{"name":"stdout","output_type":"stream","text":["✅ Model saved to /content/llava_sdnq_single (single large safetensors file)\n"]}],"source":["import os\n","import shutil\n","import json\n","import gc\n","from huggingface_hub import HfApi, snapshot_download\n","\n","SAVE_DIR = \"/content/llava_sdnq_single\"\n","os.makedirs(SAVE_DIR, exist_ok=True)\n","\n","print(\"💾 Saving with low-RAM streaming (safe_serialization + max_shard_size)...\")\n","model = model.to(\"cpu\")   # make sure everything is on CPU\n","torch.cuda.empty_cache()\n","gc.collect()\n","\n","# This is the memory-efficient way — no full state_dict copy\n","model.save_pretrained(\n","    SAVE_DIR,\n","    safe_serialization=True,\n","    max_shard_size=\"0.3GB\",# forces basically a single file (~5.85 GB)\n","    state_dict = None,\n",")\n","\n","print(f\"✅ Model saved to {SAVE_DIR} (single large safetensors file)\")\n"]},{"cell_type":"code","execution_count":null,"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"elapsed":2375,"status":"ok","timestamp":1777150776119,"user":{"displayName":"fukU Google","userId":"02763165356193834046"},"user_tz":-120},"id":"hzcE066KanBA","outputId":"02a1cd78-b524-4a60-9fd5-cd3cb4441a6c"},"outputs":[{"data":{"text/plain":["261"]},"execution_count":16,"metadata":{},"output_type":"execute_result"}],"source":["import torch , gc\n","del model\n","torch.cuda.empty_cache()\n","gc.collect()"]},{"cell_type":"code","execution_count":null,"metadata":{"colab":{"base_uri":"https://localhost:8080/","height":1000},"executionInfo":{"elapsed":96228,"status":"ok","timestamp":1777150876670,"user":{"displayName":"fukU Google","userId":"02763165356193834046"},"user_tz":-120},"id":"qdClU0GBaach","outputId":"8635b233-4c83-4201-9264-a1fabe5f0a6c"},"outputs":[{"name":"stdout","output_type":"stream","text":["✅ Processor saved\n","📥 Copying config files, README, etc....\n"]},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"903b057e7c8848bbbfb441374132da80","version_major":2,"version_minor":0},"text/plain":["Downloading (incomplete total...): 0.00B [00:00, ?B/s]"]},"metadata":{},"output_type":"display_data"},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"78e997c8e9ae4dab8a1bd991c35f0107","version_major":2,"version_minor":0},"text/plain":["Fetching 13 files:   0%|          | 0/13 [00:00<?, ?it/s]"]},"metadata":{},"output_type":"display_data"},{"name":"stdout","output_type":"stream","text":["   Copied LLAMA_LICENSE\n","   Copied LLAMA_USE_POLICY.md\n","   Copied special_tokens_map.json\n","   Copied README.md\n","   Copied tokenizer.json\n","   Copied chat_template.json\n","   Copied processor_config.json\n","   Copied config.json\n","   Copied model.safetensors.index.json\n","   Copied tokenizer_config.json\n","   Copied .gitattributes\n","   Copied generation_config.json\n","   Copied preprocessor_config.json\n","✅ SDNQ quantization_config added to config.json\n","🌐 Pushing to https://huggingface.co/codeShare/joycaption_beta_one_SDNQ ...\n"]},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"e1b48944baec4287864b7343162cf515","version_major":2,"version_minor":0},"text/plain":["Processing Files (0 / 0)      : |          |  0.00B /  0.00B            "]},"metadata":{},"output_type":"display_data"},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"b3cc2dce1d544f9582204f413b57988d","version_major":2,"version_minor":0},"text/plain":["New Data Upload               : |          |  0.00B /  0.00B            "]},"metadata":{},"output_type":"display_data"},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"44bdca3e7b924f7189bc6153e1d12198","version_major":2,"version_minor":0},"text/plain":["  ...0014-of-00017.safetensors:   4%|4         | 13.4MB /  298MB            "]},"metadata":{},"output_type":"display_data"},{"name":"stdout","output_type":"stream","text":["   Uploaded model-00014-of-00017.safetensors\n"]},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"1a887acd9097454a85bcbcd4568e67d7","version_major":2,"version_minor":0},"text/plain":["Processing Files (0 / 0)      : |          |  0.00B /  0.00B            "]},"metadata":{},"output_type":"display_data"},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"9c004cd2b75b4db1a1a7abf054637863","version_major":2,"version_minor":0},"text/plain":["New Data Upload               : |          |  0.00B /  0.00B            "]},"metadata":{},"output_type":"display_data"},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"6984b88442b145cc91d6fa1178ff4a70","version_major":2,"version_minor":0},"text/plain":["  ...0006-of-00017.safetensors:  28%|##8       | 79.8MB /  283MB            "]},"metadata":{},"output_type":"display_data"},{"name":"stdout","output_type":"stream","text":["   Uploaded model-00006-of-00017.safetensors\n"]},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"ba9a927dc653401fa45f2ea14b11f5fd","version_major":2,"version_minor":0},"text/plain":["Processing Files (0 / 0)      : |          |  0.00B /  0.00B            "]},"metadata":{},"output_type":"display_data"},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"21d92c4de3a74165b0dcbba302078fef","version_major":2,"version_minor":0},"text/plain":["New Data Upload               : |          |  0.00B /  0.00B            "]},"metadata":{},"output_type":"display_data"},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"1f0b033c1764418d82bb4afb91b96e8c","version_major":2,"version_minor":0},"text/plain":["  ...0004-of-00017.safetensors:  43%|####2     |  119MB /  278MB            "]},"metadata":{},"output_type":"display_data"},{"name":"stdout","output_type":"stream","text":["   Uploaded model-00004-of-00017.safetensors\n"]},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"9b266633912b4edab4e749c286dcf191","version_major":2,"version_minor":0},"text/plain":["Processing Files (0 / 0)      : |          |  0.00B /  0.00B            "]},"metadata":{},"output_type":"display_data"},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"686ddf6c0327477ca43adf99dc2ad722","version_major":2,"version_minor":0},"text/plain":["New Data Upload               : |          |  0.00B /  0.00B            "]},"metadata":{},"output_type":"display_data"},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"490d9c4a19bb4e408282ac370066740b","version_major":2,"version_minor":0},"text/plain":["  ...0011-of-00017.safetensors:  43%|####2     |  128MB /  298MB            "]},"metadata":{},"output_type":"display_data"},{"name":"stdout","output_type":"stream","text":["   Uploaded model-00011-of-00017.safetensors\n"]},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"5a129cb0757d48a7bf67402423ce4075","version_major":2,"version_minor":0},"text/plain":["Processing Files (0 / 0)      : |          |  0.00B /  0.00B            "]},"metadata":{},"output_type":"display_data"},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"676f55cda4fe454b95c19c7ebee7aaeb","version_major":2,"version_minor":0},"text/plain":["New Data Upload               : |          |  0.00B /  0.00B            "]},"metadata":{},"output_type":"display_data"},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"9d938fff31784895a0315977c329675b","version_major":2,"version_minor":0},"text/plain":["  ...0016-of-00017.safetensors:  41%|####1     |  124MB /  299MB            "]},"metadata":{},"output_type":"display_data"},{"name":"stderr","output_type":"stream","text":["No files have been modified since last commit. Skipping to prevent empty commit.\n","WARNING:huggingface_hub.hf_api:No files have been modified since last commit. Skipping to prevent empty commit.\n"]},{"name":"stdout","output_type":"stream","text":["   Uploaded model-00016-of-00017.safetensors\n","   Uploaded LLAMA_LICENSE\n"]},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"69cf29529f694253929fc3cbdad55547","version_major":2,"version_minor":0},"text/plain":["Processing Files (0 / 0)      : |          |  0.00B /  0.00B            "]},"metadata":{},"output_type":"display_data"},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"10b83557c2214b79825fb5af8eec801b","version_major":2,"version_minor":0},"text/plain":["New Data Upload               : |          |  0.00B /  0.00B            "]},"metadata":{},"output_type":"display_data"},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"1bd0717fe9d4426796e99b2dfa988e65","version_major":2,"version_minor":0},"text/plain":["  ...0005-of-00017.safetensors:  32%|###2      | 95.8MB /  298MB            "]},"metadata":{},"output_type":"display_data"},{"name":"stdout","output_type":"stream","text":["   Uploaded model-00005-of-00017.safetensors\n"]},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"3d6475ba0b10481ba6fd29e055116dd5","version_major":2,"version_minor":0},"text/plain":["Processing Files (0 / 0)      : |          |  0.00B /  0.00B            "]},"metadata":{},"output_type":"display_data"},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"74def50d3b0b497f941e35b347e8e85c","version_major":2,"version_minor":0},"text/plain":["New Data Upload               : |          |  0.00B /  0.00B            "]},"metadata":{},"output_type":"display_data"},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"2bb0783775ad4ea38e6bbdb2de5400de","version_major":2,"version_minor":0},"text/plain":["  ...0001-of-00017.safetensors:   8%|7         | 79.9MB / 1.05GB            "]},"metadata":{},"output_type":"display_data"},{"name":"stdout","output_type":"stream","text":["   Uploaded model-00001-of-00017.safetensors\n"]},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"8c0c47729e8e45d7a70d0cc9d78568d4","version_major":2,"version_minor":0},"text/plain":["Processing Files (0 / 0)      : |          |  0.00B /  0.00B            "]},"metadata":{},"output_type":"display_data"},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"26bb81f027dd485c99ec3247b082662e","version_major":2,"version_minor":0},"text/plain":["New Data Upload               : |          |  0.00B /  0.00B            "]},"metadata":{},"output_type":"display_data"},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"0b728eec20ad451c943dbc97361b9398","version_major":2,"version_minor":0},"text/plain":["  ...0007-of-00017.safetensors:  31%|###1      | 87.6MB /  278MB            "]},"metadata":{},"output_type":"display_data"},{"name":"stderr","output_type":"stream","text":["No files have been modified since last commit. Skipping to prevent empty commit.\n","WARNING:huggingface_hub.hf_api:No files have been modified since last commit. Skipping to prevent empty commit.\n","No files have been modified since last commit. Skipping to prevent empty commit.\n","WARNING:huggingface_hub.hf_api:No files have been modified since last commit. Skipping to prevent empty commit.\n"]},{"name":"stdout","output_type":"stream","text":["   Uploaded model-00007-of-00017.safetensors\n","   Uploaded LLAMA_USE_POLICY.md\n","   Uploaded special_tokens_map.json\n"]},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"6bbbd18bbda24729b3436fc048d98662","version_major":2,"version_minor":0},"text/plain":["Processing Files (0 / 0)      : |          |  0.00B /  0.00B            "]},"metadata":{},"output_type":"display_data"},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"26e46af705324c59a41bec9abf706c1b","version_major":2,"version_minor":0},"text/plain":["New Data Upload               : |          |  0.00B /  0.00B            "]},"metadata":{},"output_type":"display_data"},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"81e95de8f5e749af9bbe07176ac52804","version_major":2,"version_minor":0},"text/plain":["  ...0003-of-00017.safetensors:  48%|####8     |  135MB /  280MB            "]},"metadata":{},"output_type":"display_data"},{"name":"stdout","output_type":"stream","text":["   Uploaded model-00003-of-00017.safetensors\n"]},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"82510e2fcfb24dd898b5e9d9d70b8ab8","version_major":2,"version_minor":0},"text/plain":["Processing Files (0 / 0)      : |          |  0.00B /  0.00B            "]},"metadata":{},"output_type":"display_data"},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"e414f83d54424ff993737cc2f0906d0a","version_major":2,"version_minor":0},"text/plain":["New Data Upload               : |          |  0.00B /  0.00B            "]},"metadata":{},"output_type":"display_data"},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"a92d4feb3d8e43c0b4b2877e3a5b3293","version_major":2,"version_minor":0},"text/plain":["  ...0008-of-00017.safetensors:  40%|####      |  120MB /  298MB            "]},"metadata":{},"output_type":"display_data"},{"name":"stderr","output_type":"stream","text":["No files have been modified since last commit. Skipping to prevent empty commit.\n","WARNING:huggingface_hub.hf_api:No files have been modified since last commit. Skipping to prevent empty commit.\n"]},{"name":"stdout","output_type":"stream","text":["   Uploaded model-00008-of-00017.safetensors\n","   Uploaded README.md\n"]},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"f13cccabb4ee43408e341026fd67a9ec","version_major":2,"version_minor":0},"text/plain":["Processing Files (0 / 0)      : |          |  0.00B /  0.00B            "]},"metadata":{},"output_type":"display_data"},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"d333c9505c3048a39d0964d7f4b9b206","version_major":2,"version_minor":0},"text/plain":["New Data Upload               : |          |  0.00B /  0.00B            "]},"metadata":{},"output_type":"display_data"},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"44c38d11ed2a410799a0ba009f45010f","version_major":2,"version_minor":0},"text/plain":["  ...0010-of-00017.safetensors:  52%|#####1    |  144MB /  278MB            "]},"metadata":{},"output_type":"display_data"},{"name":"stdout","output_type":"stream","text":["   Uploaded model-00010-of-00017.safetensors\n"]},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"7b87e27233e54c0d841f3a9e0433f015","version_major":2,"version_minor":0},"text/plain":["Processing Files (0 / 0)      : |          |  0.00B /  0.00B            "]},"metadata":{},"output_type":"display_data"},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"e370d9ab14fd4fc8a8f8d640bd43b230","version_major":2,"version_minor":0},"text/plain":["New Data Upload               : |          |  0.00B /  0.00B            "]},"metadata":{},"output_type":"display_data"},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"5672fb01d2074d9b865ba294c87339e8","version_major":2,"version_minor":0},"text/plain":["  ...0013-of-00017.safetensors:  29%|##8       | 79.9MB /  278MB            "]},"metadata":{},"output_type":"display_data"},{"name":"stdout","output_type":"stream","text":["   Uploaded model-00013-of-00017.safetensors\n"]},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"9fe8350e210a474fbea2f359d16cfd2f","version_major":2,"version_minor":0},"text/plain":["Processing Files (0 / 0)      : |          |  0.00B /  0.00B            "]},"metadata":{},"output_type":"display_data"},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"130eb63f508a497cb3fb5eab41d22e3b","version_major":2,"version_minor":0},"text/plain":["New Data Upload               : |          |  0.00B /  0.00B            "]},"metadata":{},"output_type":"display_data"},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"5e2b1ed0049a4eb5a4cb224192e49143","version_major":2,"version_minor":0},"text/plain":["  ...dnq_single/tokenizer.json: 100%|##########| 17.2MB / 17.2MB            "]},"metadata":{},"output_type":"display_data"},{"name":"stderr","output_type":"stream","text":["No files have been modified since last commit. Skipping to prevent empty commit.\n","WARNING:huggingface_hub.hf_api:No files have been modified since last commit. Skipping to prevent empty commit.\n"]},{"name":"stdout","output_type":"stream","text":["   Uploaded tokenizer.json\n"]},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"f634b78ae7c2422db85ca1cf9e35b9a4","version_major":2,"version_minor":0},"text/plain":["Processing Files (0 / 0)      : |          |  0.00B /  0.00B            "]},"metadata":{},"output_type":"display_data"},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"2dc4c80d102640cbab425ddfac9c46a0","version_major":2,"version_minor":0},"text/plain":["New Data Upload               : |          |  0.00B /  0.00B            "]},"metadata":{},"output_type":"display_data"},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"d60ddbeb32d44d11adb061be3fdb834e","version_major":2,"version_minor":0},"text/plain":["  ...0015-of-00017.safetensors:  45%|####5     |  128MB /  283MB            "]},"metadata":{},"output_type":"display_data"},{"name":"stderr","output_type":"stream","text":["No files have been modified since last commit. Skipping to prevent empty commit.\n","WARNING:huggingface_hub.hf_api:No files have been modified since last commit. Skipping to prevent empty commit.\n"]},{"name":"stdout","output_type":"stream","text":["   Uploaded model-00015-of-00017.safetensors\n","   Uploaded chat_template.json\n"]},{"name":"stderr","output_type":"stream","text":["No files have been modified since last commit. Skipping to prevent empty commit.\n","WARNING:huggingface_hub.hf_api:No files have been modified since last commit. Skipping to prevent empty commit.\n"]},{"name":"stdout","output_type":"stream","text":["   Uploaded processor_config.json\n"]},{"name":"stderr","output_type":"stream","text":["No files have been modified since last commit. Skipping to prevent empty commit.\n","WARNING:huggingface_hub.hf_api:No files have been modified since last commit. Skipping to prevent empty commit.\n","No files have been modified since last commit. Skipping to prevent empty commit.\n","WARNING:huggingface_hub.hf_api:No files have been modified since last commit. Skipping to prevent empty commit.\n"]},{"name":"stdout","output_type":"stream","text":["   Uploaded config.json\n","   Uploaded chat_template.jinja\n"]},{"name":"stderr","output_type":"stream","text":["No files have been modified since last commit. Skipping to prevent empty commit.\n","WARNING:huggingface_hub.hf_api:No files have been modified since last commit. Skipping to prevent empty commit.\n"]},{"name":"stdout","output_type":"stream","text":["   Uploaded model.safetensors.index.json\n","   Uploaded tokenizer_config.json\n"]},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"49f64565d39b46dcad12a55a2cd94e48","version_major":2,"version_minor":0},"text/plain":["Processing Files (0 / 0)      : |          |  0.00B /  0.00B            "]},"metadata":{},"output_type":"display_data"},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"442965d9c5b54da2ae7d2a26dfbbd59b","version_major":2,"version_minor":0},"text/plain":["New Data Upload               : |          |  0.00B /  0.00B            "]},"metadata":{},"output_type":"display_data"},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"5130eda354fb46f9a4d819feda8f9c09","version_major":2,"version_minor":0},"text/plain":["  ...0009-of-00017.safetensors:  45%|####5     |  128MB /  283MB            "]},"metadata":{},"output_type":"display_data"},{"name":"stderr","output_type":"stream","text":["No files have been modified since last commit. Skipping to prevent empty commit.\n","WARNING:huggingface_hub.hf_api:No files have been modified since last commit. Skipping to prevent empty commit.\n"]},{"name":"stdout","output_type":"stream","text":["   Uploaded model-00009-of-00017.safetensors\n","   Uploaded .gitattributes\n"]},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"88abce217fc04e29b948c135c3f607dc","version_major":2,"version_minor":0},"text/plain":["Processing Files (0 / 0)      : |          |  0.00B /  0.00B            "]},"metadata":{},"output_type":"display_data"},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"3f3b7273931b4200b4c687f8e1b7357c","version_major":2,"version_minor":0},"text/plain":["New Data Upload               : |          |  0.00B /  0.00B            "]},"metadata":{},"output_type":"display_data"},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"02b78a176e42434eb841d3008eb0704b","version_major":2,"version_minor":0},"text/plain":["  ...0017-of-00017.safetensors:  38%|###7      | 63.1MB /  168MB            "]},"metadata":{},"output_type":"display_data"},{"name":"stdout","output_type":"stream","text":["   Uploaded model-00017-of-00017.safetensors\n"]},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"c6c7b09c4df44bc691bf0c2bbefc2ac9","version_major":2,"version_minor":0},"text/plain":["Processing Files (0 / 0)      : |          |  0.00B /  0.00B            "]},"metadata":{},"output_type":"display_data"},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"2ca85086f2624dca9dbd6334577d4d02","version_major":2,"version_minor":0},"text/plain":["New Data Upload               : |          |  0.00B /  0.00B            "]},"metadata":{},"output_type":"display_data"},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"b00ef24545c84ce8bd396dad9752693e","version_major":2,"version_minor":0},"text/plain":["  ...0002-of-00017.safetensors:   1%|          | 9.31MB / 1.05GB            "]},"metadata":{},"output_type":"display_data"},{"name":"stderr","output_type":"stream","text":["No files have been modified since last commit. Skipping to prevent empty commit.\n","WARNING:huggingface_hub.hf_api:No files have been modified since last commit. Skipping to prevent empty commit.\n"]},{"name":"stdout","output_type":"stream","text":["   Uploaded model-00002-of-00017.safetensors\n","   Uploaded generation_config.json\n"]},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"181c6a503c7342099914b8b84c9ce5ca","version_major":2,"version_minor":0},"text/plain":["Processing Files (0 / 0)      : |          |  0.00B /  0.00B            "]},"metadata":{},"output_type":"display_data"},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"13ff2fc6de0641ba96120a4de297705e","version_major":2,"version_minor":0},"text/plain":["New Data Upload               : |          |  0.00B /  0.00B            "]},"metadata":{},"output_type":"display_data"},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"198c24d13a3140b5942ea451eb2b9905","version_major":2,"version_minor":0},"text/plain":["  ...0012-of-00017.safetensors:  51%|#####     |  144MB /  283MB            "]},"metadata":{},"output_type":"display_data"},{"name":"stderr","output_type":"stream","text":["No files have been modified since last commit. Skipping to prevent empty commit.\n","WARNING:huggingface_hub.hf_api:No files have been modified since last commit. Skipping to prevent empty commit.\n"]},{"name":"stdout","output_type":"stream","text":["   Uploaded model-00012-of-00017.safetensors\n","   Uploaded preprocessor_config.json\n","\n","🎉 DONE! Model is now live at https://huggingface.co/codeShare/joycaption_beta_one_SDNQ\n"]}],"source":["# Save processor\n","processor.save_pretrained(SAVE_DIR)\n","print(\"✅ Processor saved\")\n","\n","# Copy all small files from the original repo\n","print(\"📥 Copying config files, README, etc....\")\n","original_small = snapshot_download(\n","    repo_id=MODEL_NAME,\n","    ignore_patterns=[\"*.safetensors\", \"*.bin\", \"*.pt\"],\n","    cache_dir=\"/tmp/original_joycaption\"\n",")\n","\n","for item in os.listdir(original_small):\n","    src = os.path.join(original_small, item)\n","    dst = os.path.join(SAVE_DIR, item)\n","    if os.path.isfile(src):\n","        shutil.copy2(src, dst)\n","        print(f\"   Copied {item}\")\n","\n","# Add proper SDNQ quantization_config to config.json\n","config_path = os.path.join(SAVE_DIR, \"config.json\")\n","with open(config_path, \"r\") as f:\n","    config = json.load(f)\n","\n","config[\"quantization_config\"] = {\n","    \"quant_method\": \"sdnq\",\n","    \"weights_dtype\": \"uint4\",\n","    \"group_size\": 128,\n","    \"quantization_device\": \"cuda\",\n","    \"return_device\": \"cuda\",\n","    \"use_svd\": False,\n","    \"quant_conv\": False,\n","    \"quant_embedding\": False,\n","    \"use_quantized_matmul\": False,\n","    \"dequantize_fp32\": True\n","}\n","\n","with open(config_path, \"w\") as f:\n","    json.dump(config, f, indent=2)\n","print(\"✅ SDNQ quantization_config added to config.json\")\n","\n","# Push to Hugging Face\n","print(f\"🌐 Pushing to https://huggingface.co/codeShare/joycaption_beta_one_SDNQ ...\")\n","api = HfApi(token=hf_token)\n","repo_id = \"codeShare/joycaption_beta_one_SDNQ\"\n","api.create_repo(repo_id, exist_ok=True)\n","\n","for filename in os.listdir(SAVE_DIR):\n","    file_path = os.path.join(SAVE_DIR, filename)\n","    if os.path.isfile(file_path):\n","        api.upload_file(\n","            path_or_fileobj=file_path,\n","            path_in_repo=filename,\n","            repo_id=repo_id,\n","            commit_message=\"SDNQ uint4 JoyCaption — complete single-file safetensors (in-place CPU quant, 5.85 GB)\"\n","        )\n","        print(f\"   Uploaded {filename}\")\n","\n","print(\"\\n🎉 DONE! Model is now live at https://huggingface.co/codeShare/joycaption_beta_one_SDNQ\")\n","gc.collect()\n","torch.cuda.empty_cache()"]},{"cell_type":"markdown","metadata":{"id":"p1V0L8u_FGaK"},"source":["# Fixing index"]},{"cell_type":"code","execution_count":null,"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"elapsed":27807,"status":"ok","timestamp":1777151772256,"user":{"displayName":"gfuku63","userId":"10213382542081034600"},"user_tz":-120},"id":"lFjv4wrPFlKA","outputId":"41fe6d98-b4b0-4eea-8d0a-c1a07140e828"},"outputs":[{"name":"stdout","output_type":"stream","text":["Mounted at /content/drive\n","✅ Drive + HF login ready\n"]}],"source":["#@markdown ## 1 - Setup environment\n","\n","from google.colab import drive, userdata\n","\n","drive.mount('/content/drive')\n","\n","hf_token = userdata.get(\"HF_TOKEN\")\n","\n","\n","if hf_token:\n","    from huggingface_hub import login\n","    login(token=hf_token)\n","\n","print(\"✅ Drive + HF login ready\")"]},{"cell_type":"code","execution_count":null,"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"elapsed":8724,"status":"ok","timestamp":1777151780981,"user":{"displayName":"gfuku63","userId":"10213382542081034600"},"user_tz":-120},"id":"YyHworE2FuaK","outputId":"ad12e16f-c255-4af5-823f-a1f75f49d2b5"},"outputs":[{"name":"stdout","output_type":"stream","text":["\u001b[?25l   \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m0.0/104.0 kB\u001b[0m \u001b[31m?\u001b[0m eta \u001b[36m-:--:--\u001b[0m\r\u001b[2K   \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m104.0/104.0 kB\u001b[0m \u001b[31m10.5 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n","\u001b[?25h✅ Dependencies installed\n"]}],"source":["#@markdown ## 2 - Install dependencies\n","\n","#!pip install -U bitsandbytes>=0.46.1\n","\n","!pip install -q \\\n","    transformers \\\n","    tokenizers \\\n","    accelerate \\\n","    huggingface_hub \\\n","    safetensors \\\n","    sdnq\n","\n","print(\"✅ Dependencies installed\")"]},{"cell_type":"code","execution_count":null,"metadata":{"colab":{"base_uri":"https://localhost:8080/","height":173},"executionInfo":{"elapsed":55245,"status":"ok","timestamp":1777151836228,"user":{"displayName":"gfuku63","userId":"10213382542081034600"},"user_tz":-120},"id":"wjBL_qJUf47C","outputId":"18c0da46-ab4b-4651-830a-22ab1fb1679d"},"outputs":[{"name":"stderr","output_type":"stream","text":["/usr/local/lib/python3.12/dist-packages/huggingface_hub/utils/_validators.py:189: UserWarning: The `resume_download` argument is deprecated and ignored in `snapshot_download`. Downloads always resume whenever possible.\n","  warnings.warn(\n"]},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"79f8f84fc9b44e6a86b19a08696e071a","version_major":2,"version_minor":0},"text/plain":["Downloading (incomplete total...): 0.00B [00:00, ?B/s]"]},"metadata":{},"output_type":"display_data"},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"c8d38343e4fc486abd54c4ab5862ff8c","version_major":2,"version_minor":0},"text/plain":["Fetching 18 files:   0%|          | 0/18 [00:00<?, ?it/s]"]},"metadata":{},"output_type":"display_data"},{"name":"stdout","output_type":"stream","text":["✅ Model snapshot downloaded to: /content/joycaption_beta_one_SDNQ\n","   Files in folder: 19\n"]}],"source":["#@markdown ## 3 - Snapshot Download the Model (codeshare/joycaption_beta_one_SDNQ)\n","\n","from huggingface_hub import snapshot_download\n","import os\n","\n","# Repo ID as provided\n","repo_id = \"codeShare/joycaption_beta_one_SDNQ\"\n","\n","# Download to Colab runtime disk (fast).\n","# Change to \"/content/drive/MyDrive/joycaption_beta_one_SDNQ\" if you want to persist on Drive.\n","local_dir = \"/content/joycaption_beta_one_SDNQ\"\n","\n","# Download ONLY the safetensors shards + index (saves time/bandwidth)\n","snapshot_download(\n","    repo_id=repo_id,\n","    local_dir=local_dir,\n","    allow_patterns=[\"*.safetensors\", \"model.safetensors.index.json\"],\n","    resume_download=True,\n","    # token is already logged in via HF_TOKEN\n",")\n","\n","print(f\"✅ Model snapshot downloaded to: {local_dir}\")\n","print(f\"   Files in folder: {len(os.listdir(local_dir))}\")"]},{"cell_type":"code","execution_count":null,"metadata":{"colab":{"base_uri":"https://localhost:8080/","height":613},"executionInfo":{"elapsed":4270,"status":"ok","timestamp":1777151998620,"user":{"displayName":"gfuku63","userId":"10213382542081034600"},"user_tz":-120},"id":"sKoazPjEg0K4","outputId":"8370f586-022a-4fcc-e719-632bfb202b60"},"outputs":[{"name":"stdout","output_type":"stream","text":["Original total_size in index: 16,959,981,696 bytes\n","\n","Found 17 safetensor shards:\n","   • model-00001-of-00017.safetensors\n","   • model-00002-of-00017.safetensors\n","   • model-00003-of-00017.safetensors\n","   • model-00004-of-00017.safetensors\n","   • model-00005-of-00017.safetensors\n","   • model-00006-of-00017.safetensors\n","   • model-00007-of-00017.safetensors\n","   • model-00008-of-00017.safetensors\n","   • model-00009-of-00017.safetensors\n","   • model-00010-of-00017.safetensors\n","   • model-00011-of-00017.safetensors\n","   • model-00012-of-00017.safetensors\n","   • model-00013-of-00017.safetensors\n","   • model-00014-of-00017.safetensors\n","   • model-00015-of-00017.safetensors\n","   • model-00016-of-00017.safetensors\n","   • model-00017-of-00017.safetensors\n","\n","Computed new total_size: 6,284,681,856 bytes\n","\n","✅ SUCCESS! model.safetensors.index.json is now fully updated\n","   • Shards: 17 (was 4)\n","   • Weights remapped: 1,525\n","   • total_size updated: 16,959,981,696 → 6,284,681,856\n","   • File saved: /content/joycaption_beta_one_SDNQ/model.safetensors.index.json\n","\n","You can now load the model safely:\n","   from transformers import AutoModelForCausalLM\n","   model = AutoModelForCausalLM.from_pretrained('/content/joycaption_beta_one_SDNQ', device_map='auto', torch_dtype='auto')\n"]}],"source":["#@markdown ## 4 - Update model.safetensors.index.json (4 shards → 17 shards + correct total_size)\n","\n","import json\n","import os\n","from safetensors import safe_open\n","\n","index_path = os.path.join(local_dir, \"model.safetensors.index.json\")\n","\n","# 1. Load the existing index\n","with open(index_path, \"r\", encoding=\"utf-8\") as f:\n","    index_data = json.load(f)\n","\n","old_total_size = index_data.get(\"metadata\", {}).get(\"total_size\", 0)\n","print(f\"Original total_size in index: {old_total_size:,} bytes\")\n","\n","# 2. Find ALL current safetensor shards (17 files)\n","shard_files = sorted([\n","    f for f in os.listdir(local_dir)\n","    if f.startswith(\"model-\") and f.endswith(\".safetensors\")\n","])\n","\n","print(f\"\\nFound {len(shard_files)} safetensor shards:\")\n","for shard in shard_files:\n","    print(\"   •\", shard)\n","\n","# 3. Rebuild weight_map by inspecting every shard (each key lives in exactly one shard)\n","weight_map = {}\n","\n","for shard_file in shard_files:\n","    shard_path = os.path.join(local_dir, shard_file)\n","    with safe_open(shard_path, framework=\"pt\") as f:\n","        for key in f.keys():\n","            if key in weight_map:\n","                print(f\"⚠️  Duplicate key detected: {key} (in {shard_file})\")\n","            weight_map[key] = shard_file\n","\n","# 4. Compute the TRUE total_size from the actual tensor data in the 17 shards\n","#    (we parse the safetensors header — zero RAM usage, no tensors loaded)\n","total_size = 0\n","\n","for shard_file in shard_files:\n","    shard_path = os.path.join(local_dir, shard_file)\n","    with open(shard_path, \"rb\") as f:\n","        # First 8 bytes = header size (little-endian uint64)\n","        header_size = int.from_bytes(f.read(8), \"little\")\n","        # Read the JSON header\n","        header_json = f.read(header_size)\n","        header = json.loads(header_json.decode(\"utf-8\"))\n","\n","        # Sum data size of every tensor (data_offsets[1] - data_offsets[0])\n","        for tensor_name, info in header.items():\n","            if tensor_name == \"__metadata__\":\n","                continue\n","            offsets = info[\"data_offsets\"]\n","            total_size += offsets[1] - offsets[0]\n","\n","print(f\"\\nComputed new total_size: {total_size:,} bytes\")\n","\n","# 5. Update both the weight_map AND the total_size\n","if \"metadata\" not in index_data:\n","    index_data[\"metadata\"] = {}\n","index_data[\"metadata\"][\"total_size\"] = total_size\n","index_data[\"weight_map\"] = weight_map\n","\n","# 6. Save the fixed index back to disk\n","with open(index_path, \"w\", encoding=\"utf-8\") as f:\n","    json.dump(index_data, f, indent=2)\n","\n","print(\"\\n✅ SUCCESS! model.safetensors.index.json is now fully updated\")\n","print(f\"   • Shards: {len(shard_files)} (was 4)\")\n","print(f\"   • Weights remapped: {len(weight_map):,}\")\n","print(f\"   • total_size updated: {old_total_size:,} → {total_size:,}\")\n","print(f\"   • File saved: {index_path}\")\n","\n","print(\"\\nYou can now load the model safely:\")\n","print(\"   from transformers import AutoModelForCausalLM\")\n","print(f\"   model = AutoModelForCausalLM.from_pretrained('{local_dir}', device_map='auto', torch_dtype='auto')\")"]},{"cell_type":"code","execution_count":null,"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"elapsed":2361,"status":"ok","timestamp":1777152223184,"user":{"displayName":"gfuku63","userId":"10213382542081034600"},"user_tz":-120},"id":"rrERw9Uuhqqc","outputId":"8d543d0b-7db0-49ed-d8d2-d1b37bb325c6"},"outputs":[{"name":"stdout","output_type":"stream","text":["🔍 Loading old index.json (the 4-shard version from the repo)...\n","   • Old index contains 1,525 weights\n","   • Old total_size: 6,284,681,856 bytes\n","\n","🔍 Found 17 safetensor shards:\n","   • model-00001-of-00017.safetensors\n","   • model-00002-of-00017.safetensors\n","   • model-00003-of-00017.safetensors\n","   • model-00004-of-00017.safetensors\n","   • model-00005-of-00017.safetensors\n","   • model-00006-of-00017.safetensors\n","   • model-00007-of-00017.safetensors\n","   • model-00008-of-00017.safetensors\n","   • model-00009-of-00017.safetensors\n","   • model-00010-of-00017.safetensors\n","   ... and 7 more\n","\n","🔍 Extracting keys from all 17 shards...\n","\n","✅ VERIFICATION RESULTS\n","   • Keys in 17 shards : 1,525\n","   • Keys in old index : 1,525\n","   ✅ PERFECT MATCH — All keys are identical!\n","\n","✅ Verification PASSED — Proceeding with update & push...\n","\n","📏 Re-calculating true total_size from safetensors headers...\n","   Old total_size : 6,284,681,856 bytes\n","   New total_size : 6,284,681,856 bytes\n","\n","💾 Updated model.safetensors.index.json saved locally (156,178 bytes)\n","\n","🚀 Pushing updated index to Hugging Face repo...\n","🎉 SUCCESS! Index pushed to repo\n","   Commit URL: https://huggingface.co/codeShare/joycaption_beta_one_SDNQ/commit/7c895223e92fd352e1dcc2880d73c9ecc0fa9209\n","   View repo: https://huggingface.co/codeShare/joycaption_beta_one_SDNQ\n","\n","You can now safely use this model from the repo with 17 shards!\n"]}],"source":["#@markdown ## 5 - Verify keys (17 shards vs old index) + Push updated index to HF repo\n","\n","import json\n","import os\n","from safetensors import safe_open\n","from huggingface_hub import HfApi\n","\n","# --- Paths (from previous cells) ---\n","index_path = os.path.join(local_dir, \"model.safetensors.index.json\")\n","repo_id = \"codeShare/joycaption_beta_one_SDNQ\"\n","\n","print(\"🔍 Loading old index.json (the 4-shard version from the repo)...\")\n","with open(index_path, \"r\", encoding=\"utf-8\") as f:\n","    old_index = json.load(f)\n","\n","old_weight_map = old_index.get(\"weight_map\", {})\n","old_keys = set(old_weight_map.keys())\n","old_total = old_index.get(\"metadata\", {}).get(\"total_size\", 0)\n","\n","print(f\"   • Old index contains {len(old_keys):,} weights\")\n","print(f\"   • Old total_size: {old_total:,} bytes\\n\")\n","\n","# --- Find all 17 current shards ---\n","shard_files = sorted([\n","    f for f in os.listdir(local_dir)\n","    if f.startswith(\"model-\") and f.endswith(\".safetensors\")\n","])\n","\n","print(f\"🔍 Found {len(shard_files)} safetensor shards:\")\n","for shard in shard_files[:10]:          # show first 10 only\n","    print(\"   •\", shard)\n","if len(shard_files) > 10:\n","    print(f\"   ... and {len(shard_files)-10} more\")\n","\n","# --- Extract EVERY key from the 17 shards (no full tensors loaded) ---\n","print(\"\\n🔍 Extracting keys from all 17 shards...\")\n","new_weight_map = {}\n","for shard_file in shard_files:\n","    shard_path = os.path.join(local_dir, shard_file)\n","    with safe_open(shard_path, framework=\"pt\") as f:\n","        for key in f.keys():\n","            if key in new_weight_map:\n","                print(f\"⚠️  Duplicate key {key} found in {shard_file}\")\n","            new_weight_map[key] = shard_file\n","\n","new_keys = set(new_weight_map.keys())\n","\n","# --- VERIFICATION ---\n","print(\"\\n✅ VERIFICATION RESULTS\")\n","print(f\"   • Keys in 17 shards : {len(new_keys):,}\")\n","print(f\"   • Keys in old index : {len(old_keys):,}\")\n","\n","all_keys_exist_in_old = new_keys.issubset(old_keys)\n","all_old_keys_covered = old_keys.issubset(new_keys)\n","\n","if all_keys_exist_in_old and all_old_keys_covered:\n","    print(\"   ✅ PERFECT MATCH — All keys are identical!\")\n","elif all_keys_exist_in_old:\n","    print(\"   ⚠️  All new keys exist in old index, but some old keys are missing in shards\")\n","else:\n","    missing_in_old = new_keys - old_keys\n","    print(f\"   ❌ {len(missing_in_old):,} keys in the 17 shards do NOT exist in the old index!\")\n","    print(\"      First 10 missing:\", list(missing_in_old)[:10])\n","\n","if not all_keys_exist_in_old:\n","    print(\"\\n🚫 Verification FAILED — Aborting update & push.\")\n","    raise SystemExit(\"Key mismatch detected. Do not push until resolved.\")\n","\n","print(\"\\n✅ Verification PASSED — Proceeding with update & push...\\n\")\n","\n","# --- Recompute correct total_size from the real 17 shards ---\n","print(\"📏 Re-calculating true total_size from safetensors headers...\")\n","total_size = 0\n","for shard_file in shard_files:\n","    shard_path = os.path.join(local_dir, shard_file)\n","    with open(shard_path, \"rb\") as f:\n","        header_size = int.from_bytes(f.read(8), \"little\")\n","        header = json.loads(f.read(header_size).decode(\"utf-8\"))\n","        for tensor_name, info in header.items():\n","            if tensor_name == \"__metadata__\":\n","                continue\n","            offsets = info.get(\"data_offsets\", [0, 0])\n","            total_size += offsets[1] - offsets[0]\n","\n","print(f\"   Old total_size : {old_total:,} bytes\")\n","print(f\"   New total_size : {total_size:,} bytes\\n\")\n","\n","# --- Update the index in memory ---\n","if \"metadata\" not in old_index:\n","    old_index[\"metadata\"] = {}\n","old_index[\"metadata\"][\"total_size\"] = total_size\n","old_index[\"weight_map\"] = new_weight_map\n","\n","# --- Save updated index locally ---\n","with open(index_path, \"w\", encoding=\"utf-8\") as f:\n","    json.dump(old_index, f, indent=2)\n","\n","print(f\"💾 Updated model.safetensors.index.json saved locally ({os.path.getsize(index_path):,} bytes)\")\n","\n","# --- Push ONLY the fixed index.json back to the repo ---\n","print(\"\\n🚀 Pushing updated index to Hugging Face repo...\")\n","api = HfApi()\n","\n","commit = api.upload_file(\n","    path_or_fileobj=index_path,\n","    path_in_repo=\"model.safetensors.index.json\",\n","    repo_id=repo_id,\n","    repo_type=\"model\",\n","    commit_message=\"fix: update model.safetensors.index.json for 17 shards + correct total_size\",\n","    commit_description=\"Rebuilt weight_map from actual 17 safetensors shards and recomputed total_size. Keys verified 100% match.\",\n",")\n","\n","print(\"🎉 SUCCESS! Index pushed to repo\")\n","print(f\"   Commit URL: {commit.commit_url}\")\n","print(f\"   View repo: https://huggingface.co/{repo_id}\")\n","print(\"\\nYou can now safely use this model from the repo with 17 shards!\")"]},{"cell_type":"markdown","metadata":{"id":"XTTQShHIh56j"},"source":[]},{"cell_type":"markdown","metadata":{"id":"92MEeRpOh7eX"},"source":["# load model from repo"]},{"cell_type":"code","execution_count":null,"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"elapsed":28323,"status":"ok","timestamp":1777152369564,"user":{"displayName":"gfuku63","userId":"10213382542081034600"},"user_tz":-120},"id":"iJL37jOlh7eY","outputId":"e136b929-44ac-4f37-bc36-9516cccb1320"},"outputs":[{"name":"stdout","output_type":"stream","text":["Mounted at /content/drive\n","✅ Drive + HF login ready\n"]}],"source":["#@markdown ## 1 - Setup environment\n","\n","from google.colab import drive, userdata\n","\n","drive.mount('/content/drive')\n","\n","hf_token = userdata.get(\"HF_TOKEN\")\n","\n","\n","if hf_token:\n","    from huggingface_hub import login\n","    login(token=hf_token)\n","\n","print(\"✅ Drive + HF login ready\")"]},{"cell_type":"code","execution_count":null,"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"elapsed":7515,"status":"ok","timestamp":1777152377081,"user":{"displayName":"gfuku63","userId":"10213382542081034600"},"user_tz":-120},"id":"0ZWtubaRh7eZ","outputId":"9ab595c3-48bd-4a15-e3eb-20c6016d682f"},"outputs":[{"name":"stdout","output_type":"stream","text":["\u001b[?25l   \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m0.0/104.0 kB\u001b[0m \u001b[31m?\u001b[0m eta \u001b[36m-:--:--\u001b[0m\r\u001b[2K   \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m104.0/104.0 kB\u001b[0m \u001b[31m9.2 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n","\u001b[?25h✅ Dependencies installed\n"]}],"source":["#@markdown ## 2 - Install dependencies\n","\n","#!pip install -U bitsandbytes>=0.46.1\n","\n","!pip install -q \\\n","    transformers \\\n","    tokenizers \\\n","    accelerate \\\n","    huggingface_hub \\\n","    safetensors \\\n","    sdnq\n","\n","print(\"✅ Dependencies installed\")"]},{"cell_type":"code","execution_count":null,"metadata":{"colab":{"base_uri":"https://localhost:8080/","height":893},"executionInfo":{"elapsed":17866,"status":"error","timestamp":1777154435454,"user":{"displayName":"gfuku63","userId":"10213382542081034600"},"user_tz":-120},"id":"yIHQ4-21qDrJ","outputId":"674c8e7e-4d3d-4dc8-f97a-c3016e772da0"},"outputs":[{"name":"stdout","output_type":"stream","text":["Drive already mounted at /content/drive; to attempt to forcibly remount, call drive.mount(\"/content/drive\", force_remount=True).\n","✅ Drive + HF login ready\n","✅ SDNQ registered + idempotent torch init patch applied\n","🚀 Loading SDNQ uint4 JoyCaption in FP32...\n"]},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"01e259ecafe144378007551979a90cec","version_major":2,"version_minor":0},"text/plain":["Downloading (incomplete total...): 0.00B [00:00, ?B/s]"]},"metadata":{},"output_type":"display_data"},{"data":{"application/vnd.jupyter.widget-view+json":{"model_id":"3241c4c30d8a4a64943f9f4c3e89835f","version_major":2,"version_minor":0},"text/plain":["Fetching 17 files:   0%|          | 0/17 [00:00<?, ?it/s]"]},"metadata":{},"output_type":"display_data"},{"ename":"RecursionError","evalue":"maximum recursion depth exceeded","output_type":"error","traceback":["\u001b[0;31m---------------------------------------------------------------------------\u001b[0m","\u001b[0;31mRecursionError\u001b[0m                            Traceback (most recent call last)","\u001b[0;32m/tmp/ipykernel_20780/3426480698.py\u001b[0m in \u001b[0;36m<cell line: 0>\u001b[0;34m()\u001b[0m\n\u001b[1;32m     77\u001b[0m \u001b[0mprint\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m\"🚀 Loading SDNQ uint4 JoyCaption in FP32...\"\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     78\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 79\u001b[0;31m model = LlavaForConditionalGeneration.from_pretrained(\n\u001b[0m\u001b[1;32m     80\u001b[0m     \u001b[0;34m\"codeShare/joycaption_beta_one_SDNQ\"\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     81\u001b[0m     \u001b[0mdevice_map\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;34m\"auto\"\u001b[0m\u001b[0;34m,\u001b[0m              \u001b[0;31m# works on Tesla T4\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;32m/usr/local/lib/python3.12/dist-packages/transformers/modeling_utils.py\u001b[0m in \u001b[0;36mfrom_pretrained\u001b[0;34m(cls, pretrained_model_name_or_path, config, cache_dir, ignore_mismatched_sizes, force_download, local_files_only, token, revision, use_safetensors, weights_only, fusion_config, disable_mmap, *model_args, **kwargs)\u001b[0m\n\u001b[1;32m   4164\u001b[0m         \u001b[0mconfig\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mcopy\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mdeepcopy\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mconfig\u001b[0m\u001b[0;34m)\u001b[0m  \u001b[0;31m# We do not want to modify the config inplace in from_pretrained.\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   4165\u001b[0m         \u001b[0;32mwith\u001b[0m \u001b[0mContextManagers\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mmodel_init_context\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 4166\u001b[0;31m             \u001b[0mmodel\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mcls\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mconfig\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m*\u001b[0m\u001b[0mmodel_args\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mmodel_kwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   4167\u001b[0m             \u001b[0mpatch_output_recorders\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mmodel\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   4168\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;32m/usr/local/lib/python3.12/dist-packages/transformers/models/llava/modeling_llava.py\u001b[0m in \u001b[0;36m__init__\u001b[0;34m(self, config)\u001b[0m\n\u001b[1;32m    281\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0m__init__\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mconfig\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0mLlavaConfig\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    282\u001b[0m         \u001b[0msuper\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m__init__\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mconfig\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 283\u001b[0;31m         \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mmodel\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mLlavaModel\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mconfig\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    284\u001b[0m         \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlm_head\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mnn\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mLinear\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mconfig\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtext_config\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mhidden_size\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mconfig\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtext_config\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mvocab_size\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mbias\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mFalse\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    285\u001b[0m         \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mpost_init\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;32m/usr/local/lib/python3.12/dist-packages/transformers/models/llava/modeling_llava.py\u001b[0m in \u001b[0;36m__init__\u001b[0;34m(self, config)\u001b[0m\n\u001b[1;32m    131\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0m__init__\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mconfig\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0mLlavaConfig\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    132\u001b[0m         \u001b[0msuper\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m__init__\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mconfig\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 133\u001b[0;31m         \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mvision_tower\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mAutoModel\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mfrom_config\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mconfig\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mvision_config\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    134\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    135\u001b[0m         \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mmulti_modal_projector\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mLlavaMultiModalProjector\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mconfig\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;32m/usr/local/lib/python3.12/dist-packages/transformers/models/auto/auto_factory.py\u001b[0m in \u001b[0;36mfrom_config\u001b[0;34m(cls, config, **kwargs)\u001b[0m\n\u001b[1;32m    239\u001b[0m         \u001b[0;32melif\u001b[0m \u001b[0mhas_local_code\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    240\u001b[0m             \u001b[0mmodel_class\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0m_get_model_class\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mconfig\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mcls\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_model_mapping\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 241\u001b[0;31m             \u001b[0;32mreturn\u001b[0m \u001b[0mmodel_class\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_from_config\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mconfig\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    242\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    243\u001b[0m         raise ValueError(\n","\u001b[0;32m/usr/local/lib/python3.12/dist-packages/transformers/modeling_utils.py\u001b[0m in \u001b[0;36m_from_config\u001b[0;34m(cls, config, **kwargs)\u001b[0m\n\u001b[1;32m   1540\u001b[0m         \u001b[0;31m# Instantiate the model\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1541\u001b[0m         \u001b[0;32mwith\u001b[0m \u001b[0mContextManagers\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0minit_contexts\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1542\u001b[0;31m             \u001b[0mmodel\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mcls\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mconfig\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   1543\u001b[0m             \u001b[0mpatch_output_recorders\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mmodel\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1544\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;32m/usr/local/lib/python3.12/dist-packages/transformers/models/siglip/modeling_siglip.py\u001b[0m in \u001b[0;36m__init__\u001b[0;34m(self, config)\u001b[0m\n\u001b[1;32m    567\u001b[0m         \u001b[0membed_dim\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mconfig\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mhidden_size\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    568\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 569\u001b[0;31m         \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0membeddings\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mSiglipVisionEmbeddings\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mconfig\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    570\u001b[0m         \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mencoder\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mSiglipEncoder\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mconfig\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    571\u001b[0m         \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mpost_layernorm\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mnn\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mLayerNorm\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0membed_dim\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0meps\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mconfig\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlayer_norm_eps\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;32m/usr/local/lib/python3.12/dist-packages/transformers/models/siglip/modeling_siglip.py\u001b[0m in \u001b[0;36m__init__\u001b[0;34m(self, config)\u001b[0m\n\u001b[1;32m    122\u001b[0m         \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mpatch_size\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mconfig\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mpatch_size\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    123\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 124\u001b[0;31m         self.patch_embedding = nn.Conv2d(\n\u001b[0m\u001b[1;32m    125\u001b[0m             \u001b[0min_channels\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mconfig\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mnum_channels\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    126\u001b[0m             \u001b[0mout_channels\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0membed_dim\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;32m/usr/local/lib/python3.12/dist-packages/torch/nn/modules/conv.py\u001b[0m in \u001b[0;36m__init__\u001b[0;34m(self, in_channels, out_channels, kernel_size, stride, padding, dilation, groups, bias, padding_mode, device, dtype)\u001b[0m\n\u001b[1;32m    517\u001b[0m         \u001b[0mpadding_\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mpadding\u001b[0m \u001b[0;32mif\u001b[0m \u001b[0misinstance\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mpadding\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mstr\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;32melse\u001b[0m \u001b[0m_pair\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mpadding\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    518\u001b[0m         \u001b[0mdilation_\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0m_pair\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mdilation\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 519\u001b[0;31m         super().__init__(\n\u001b[0m\u001b[1;32m    520\u001b[0m             \u001b[0min_channels\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    521\u001b[0m             \u001b[0mout_channels\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;32m/usr/local/lib/python3.12/dist-packages/torch/nn/modules/conv.py\u001b[0m in \u001b[0;36m__init__\u001b[0;34m(self, in_channels, out_channels, kernel_size, stride, padding, dilation, transposed, output_padding, groups, bias, padding_mode, device, dtype)\u001b[0m\n\u001b[1;32m    178\u001b[0m             \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mregister_parameter\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m\"bias\"\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    179\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 180\u001b[0;31m         \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mreset_parameters\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    181\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    182\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0mreset_parameters\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;34m->\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;32m/usr/local/lib/python3.12/dist-packages/torch/nn/modules/conv.py\u001b[0m in \u001b[0;36mreset_parameters\u001b[0;34m(self)\u001b[0m\n\u001b[1;32m    184\u001b[0m         \u001b[0;31m# uniform(-1/sqrt(k), 1/sqrt(k)), where k = weight.size(1) * prod(*kernel_size)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    185\u001b[0m         \u001b[0;31m# For more details see: https://github.com/pytorch/pytorch/issues/15314#issuecomment-477448573\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 186\u001b[0;31m         \u001b[0minit\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mkaiming_uniform_\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mweight\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0ma\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mmath\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msqrt\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;36m5\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    187\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mbias\u001b[0m \u001b[0;32mis\u001b[0m \u001b[0;32mnot\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    188\u001b[0m             \u001b[0mfan_in\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0m_\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0minit\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_calculate_fan_in_and_fan_out\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mweight\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;32m/usr/local/lib/python3.12/dist-packages/torch/nn/init.py\u001b[0m in \u001b[0;36mkaiming_uniform_\u001b[0;34m(tensor, a, mode, nonlinearity, generator)\u001b[0m\n\u001b[1;32m    555\u001b[0m     \"\"\"\n\u001b[1;32m    556\u001b[0m     \u001b[0;32mif\u001b[0m \u001b[0mtorch\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0moverrides\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mhas_torch_function_variadic\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mtensor\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 557\u001b[0;31m         return torch.overrides.handle_torch_function(\n\u001b[0m\u001b[1;32m    558\u001b[0m             \u001b[0mkaiming_uniform_\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    559\u001b[0m             \u001b[0;34m(\u001b[0m\u001b[0mtensor\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;32m/usr/local/lib/python3.12/dist-packages/torch/overrides.py\u001b[0m in \u001b[0;36mhandle_torch_function\u001b[0;34m(public_api, relevant_args, *args, **kwargs)\u001b[0m\n\u001b[1;32m   1738\u001b[0m         \u001b[0;31m# this unsets it and calls directly into TorchFunctionStackMode's torch function\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1739\u001b[0m         \u001b[0;32mwith\u001b[0m \u001b[0m_pop_mode_temporarily\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;32mas\u001b[0m \u001b[0mmode\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1740\u001b[0;31m             \u001b[0mresult\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mmode\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m__torch_function__\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mpublic_api\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtypes\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   1741\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0mresult\u001b[0m \u001b[0;32mis\u001b[0m \u001b[0;32mnot\u001b[0m \u001b[0mNotImplemented\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1742\u001b[0m             \u001b[0;32mreturn\u001b[0m \u001b[0mresult\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;32m/usr/local/lib/python3.12/dist-packages/torch/utils/_device.py\u001b[0m in \u001b[0;36m__torch_function__\u001b[0;34m(self, func, types, args, kwargs)\u001b[0m\n\u001b[1;32m    107\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0mfunc\u001b[0m \u001b[0;32min\u001b[0m \u001b[0m_device_constructors\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;32mand\u001b[0m \u001b[0mkwargs\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mget\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m\"device\"\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;32mis\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    108\u001b[0m             \u001b[0mkwargs\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m\"device\"\u001b[0m\u001b[0;34m]\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mdevice\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 109\u001b[0;31m         \u001b[0;32mreturn\u001b[0m \u001b[0mfunc\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    110\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    111\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;32m/usr/local/lib/python3.12/dist-packages/torch/nn/init.py\u001b[0m in \u001b[0;36mkaiming_uniform_\u001b[0;34m(tensor, a, mode, nonlinearity, generator)\u001b[0m\n\u001b[1;32m    568\u001b[0m         \u001b[0mwarnings\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mwarn\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m\"Initializing zero-element tensors is a no-op\"\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mstacklevel\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;36m2\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    569\u001b[0m         \u001b[0;32mreturn\u001b[0m \u001b[0mtensor\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 570\u001b[0;31m     \u001b[0mfan\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0m_calculate_correct_fan\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mtensor\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mmode\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    571\u001b[0m     \u001b[0mgain\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mcalculate_gain\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mnonlinearity\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0ma\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    572\u001b[0m     \u001b[0mstd\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mgain\u001b[0m \u001b[0;34m/\u001b[0m \u001b[0mmath\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msqrt\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mfan\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;32m/usr/local/lib/python3.12/dist-packages/torch/nn/init.py\u001b[0m in \u001b[0;36m_calculate_correct_fan\u001b[0;34m(tensor, mode)\u001b[0m\n\u001b[1;32m    507\u001b[0m         \u001b[0;32mraise\u001b[0m \u001b[0mValueError\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34mf\"Mode {mode} not supported, please use one of {valid_modes}\"\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    508\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 509\u001b[0;31m     \u001b[0mfan_in\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfan_out\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0m_calculate_fan_in_and_fan_out\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mtensor\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    510\u001b[0m     \u001b[0;32mreturn\u001b[0m \u001b[0mfan_in\u001b[0m \u001b[0;32mif\u001b[0m \u001b[0mmode\u001b[0m \u001b[0;34m==\u001b[0m \u001b[0;34m\"fan_in\"\u001b[0m \u001b[0;32melse\u001b[0m \u001b[0mfan_out\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    511\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;32m/tmp/ipykernel_20780/3426480698.py\u001b[0m in \u001b[0;36msafe_fan_in_and_fan_out\u001b[0;34m(tensor)\u001b[0m\n\u001b[1;32m     65\u001b[0m     \u001b[0;32mif\u001b[0m \u001b[0mtensor\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mdim\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;34m<\u001b[0m \u001b[0;36m2\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     66\u001b[0m         \u001b[0;32mreturn\u001b[0m \u001b[0;36m1\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;36m1\u001b[0m          \u001b[0;31m# dummy values – we don't want to re-init\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 67\u001b[0;31m     \u001b[0;32mreturn\u001b[0m \u001b[0moriginal_fan\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mtensor\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     68\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     69\u001b[0m \u001b[0mtorch_init\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_calculate_fan_in_and_fan_out\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0msafe_fan_in_and_fan_out\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;32m/tmp/ipykernel_20780/592907678.py\u001b[0m in \u001b[0;36msafe_fan_in_and_fan_out\u001b[0;34m(tensor)\u001b[0m\n\u001b[1;32m     21\u001b[0m     \u001b[0;32mif\u001b[0m \u001b[0mtensor\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mdim\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;34m<\u001b[0m \u001b[0;36m2\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     22\u001b[0m         \u001b[0;32mreturn\u001b[0m \u001b[0;36m1\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;36m1\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 23\u001b[0;31m     \u001b[0;32mreturn\u001b[0m \u001b[0moriginal_fan\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mtensor\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     24\u001b[0m \u001b[0mtorch_init\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_calculate_fan_in_and_fan_out\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0msafe_fan_in_and_fan_out\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     25\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n","... last 1 frames repeated, from the frame below ...\n","\u001b[0;32m/tmp/ipykernel_20780/592907678.py\u001b[0m in \u001b[0;36msafe_fan_in_and_fan_out\u001b[0;34m(tensor)\u001b[0m\n\u001b[1;32m     21\u001b[0m     \u001b[0;32mif\u001b[0m \u001b[0mtensor\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mdim\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;34m<\u001b[0m \u001b[0;36m2\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     22\u001b[0m         \u001b[0;32mreturn\u001b[0m \u001b[0;36m1\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;36m1\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 23\u001b[0;31m     \u001b[0;32mreturn\u001b[0m \u001b[0moriginal_fan\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mtensor\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     24\u001b[0m \u001b[0mtorch_init\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_calculate_fan_in_and_fan_out\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0msafe_fan_in_and_fan_out\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     25\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;31mRecursionError\u001b[0m: maximum recursion depth exceeded"]}],"source":["#@markdown ## FIXED LOADER CELL (RecursionError + dtype crash solved)\n","\n","# ===================================================================\n","# BUILD ISSUES TO FIX WHEN WE REBUILD THE SDNQ MODEL (save this list!)\n","# ===================================================================\n","# 1. model.safetensors.index.json was not updated before first commit\n","#    → caused incomplete shard references (you fixed it manually later)\n","#\n","# 2. lm_head.scale and lm_head.zero_point were MISSING after load\n","#    → because we quantized the lm_head but never saved its meta tensors\n","#    → solution: either skip quantizing lm_head OR explicitly add its\n","#      scale/zero_point as buffers before save_pretrained()\n","#\n","# 3. vision_tower.head.attention.out_proj.scale / zero_point were UNEXPECTED\n","#    → some vision tower layers got partially quantized (only weight packed,\n","#      but no proper SDNQ wrapper / buffers registered)\n","#    → happens because sdnq_quantize_layer was called on nn.Linear but\n","#      the module was not fully replaced in the model tree\n","#\n","# 4. bfloat16 vs Byte dtype crash during inference\n","#    → model loaded in bfloat16 by default, but SDNQ dequantize_fp32=True\n","#      produces float32 → mismatch in matmul\n","#    → solution: force torch_dtype=torch.float32 + model.to(torch.float32)\n","#      at load time (or fix in build: save with torch_dtype=float32)\n","#\n","# 5. torch init recursion (current error)\n","#    → our monkey-patch was being re-applied on cell re-run → original_fan\n","#      became the safe version → infinite recursion\n","#    → fixed below with idempotent patch (only patch once)\n","#\n","# 6. Future improvements for a perfect build:\n","#    - Use sdnq's official save flow instead of in-place + save_pretrained\n","#    - Explicitly register every quantized layer with SDNQConfig\n","#    - Skip quantization on lm_head, vision_tower.head, and any 1D/embedding layers\n","#    - Add proper quantization_config to config.json with ALL keys\n","#    - Test load with device_map=\"auto\" + torch_dtype=float32 BEFORE pushing\n","# ===================================================================\n","\n","from google.colab import drive, userdata\n","drive.mount('/content/drive')\n","\n","hf_token = userdata.get(\"HF_TOKEN\")\n","if hf_token:\n","    from huggingface_hub import login\n","    login(token=hf_token)\n","\n","print(\"✅ Drive + HF login ready\")\n","\n","#@markdown ## 2 - Install + register SDNQ + idempotent safety patch\n","!pip install -q --upgrade sdnq transformers\n","\n","from sdnq import SDNQConfig\n","\n","# === IDEMPOTENT PATCH (fixes recursion on re-run) ===\n","import torch.nn.init as torch_init\n","\n","# Store the TRUE original function only once\n","if not hasattr(torch_init, \"_original_calculate_fan_in_and_fan_out\"):\n","    torch_init._original_calculate_fan_in_and_fan_out = torch_init._calculate_fan_in_and_fan_out\n","\n","original_fan = torch_init._original_calculate_fan_in_and_fan_out\n","\n","def safe_fan_in_and_fan_out(tensor):\n","    \"\"\"Skip fan calculation for 1D packed SDNQ uint4 weights\"\"\"\n","    if tensor.dim() < 2:\n","        return 1, 1          # dummy values – we don't want to re-init\n","    return original_fan(tensor)\n","\n","torch_init._calculate_fan_in_and_fan_out = safe_fan_in_and_fan_out\n","\n","print(\"✅ SDNQ registered + idempotent torch init patch applied\")\n","\n","#@markdown ## 3 - Load the model (FP32 = fixes dtype crash)\n","import torch\n","from transformers import LlavaForConditionalGeneration, AutoProcessor\n","\n","print(\"🚀 Loading SDNQ uint4 JoyCaption in FP32...\")\n","\n","model = LlavaForConditionalGeneration.from_pretrained(\n","    \"codeShare/joycaption_beta_one_SDNQ\",\n","    device_map=\"auto\",              # works on Tesla T4\n","    trust_remote_code=True,\n","    low_cpu_mem_usage=True,\n","    torch_dtype=torch.float32,      # CRITICAL for SDNQ dequantize_fp32=True\n",")\n","\n","processor = AutoProcessor.from_pretrained(\"codeShare/joycaption_beta_one_SDNQ\")\n","\n","# Extra safety\n","model = model.to(torch.float32)\n","model.eval()\n","torch.set_grad_enabled(False)\n","\n","print(\"✅ SDNQ JoyCaption loaded successfully in FP32!\")\n","print(f\"   Model memory usage: {sum(p.numel() * p.element_size() for p in model.parameters()) / (1024**3):.2f} GB\")"]},{"cell_type":"markdown","metadata":{"id":"etI08AHHq7eQ"},"source":["# Latest build code"]},{"cell_type":"code","execution_count":1,"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"elapsed":3558,"status":"ok","timestamp":1777158517876,"user":{"displayName":"gfuku63","userId":"10213382542081034600"},"user_tz":-120},"id":"L1QyiNUss5Fl","outputId":"8a88eabb-505f-406d-93c1-22199ea4c963","cellView":"form"},"outputs":[{"output_type":"stream","name":"stdout","text":["Drive already mounted at /content/drive; to attempt to forcibly remount, call drive.mount(\"/content/drive\", force_remount=True).\n","✅ Drive + HF login ready\n"]}],"source":["#@markdown ## 1 - Setup environment\n","\n","from google.colab import drive, userdata\n","\n","drive.mount('/content/drive')\n","\n","hf_token = userdata.get(\"HF_TOKEN\")\n","\n","if hf_token:\n","    from huggingface_hub import login\n","    login(token=hf_token)\n","\n","print(\"✅ Drive + HF login ready\")"]},{"cell_type":"code","execution_count":null,"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"elapsed":15010,"status":"ok","timestamp":1777155232953,"user":{"displayName":"gfuku63","userId":"10213382542081034600"},"user_tz":-120},"id":"c-ERF9zms75a","outputId":"b948b8ac-6b7e-4713-e805-c3083d82e2e3","cellView":"form"},"outputs":[{"name":"stdout","output_type":"stream","text":["\u001b[?25l   \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m0.0/104.0 kB\u001b[0m \u001b[31m?\u001b[0m eta \u001b[36m-:--:--\u001b[0m\r\u001b[2K   \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m104.0/104.0 kB\u001b[0m \u001b[31m3.4 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n","\u001b[?25h✅ Dependencies installed\n"]}],"source":["#@markdown ## 2 - Install dependencies\n","\n","!pip install -U bitsandbytes>=0.46.1\n","\n","!pip install -q \\\n","    transformers \\\n","    tokenizers \\\n","    accelerate \\\n","    huggingface_hub \\\n","    safetensors \\\n","    sdnq\n","\n","print(\"✅ Dependencies installed\")"]},{"cell_type":"code","execution_count":2,"metadata":{"colab":{"base_uri":"https://localhost:8080/","height":197,"referenced_widgets":["e0a1b32fbfb3476997088ef6cf8bd1a4","735f3bc73aa54a3fab17436188c31577","5264574b812840f5b379c5704f123b64","989634a219a44642a6f2e0748960655f","7bd8f165ed8c49f192d3f42739ac963a","dcb82126aae448f0a37c1c5661011795","2eabc79a4afd4b07a6ce7cca897e4450","7542c6ce9a544a81aad3cbf3669186ce","0b38405f4e864526b85af2fbc1267e03","d4905952a5c84841abc842cafc514366","45c11fc828d94b779e7345ea114179b0","f352233154ef4c12aa4c87f83ffa1af6","b6a0c2c218de4b9dba80f6bb39fe4dff","93176536d88041269f3e9d4c692e6557","4e02ede43b9045c2aff16a102b4c212a","681933b0634645b98be2c3949cdb6149","7fa9e8ffa7d2414ea4e7cefde801b958","44b0c67e2d044bfdb564dc1aea15cfd2","cf350b2c8159406ab09cc8ec12126ec4","828a3b771f664bcca4205ccd6ca14315","62bda97392ba43069274532eaaa56c1e","41ac9dc3ccc14c35a0ee71c229e815c7","9406b00f72b74c718ac970a04a3d81a2","a7923196bd3841a49231d5116da5c251","9d88f227be7f408f8b4414e3fdf7d203","c548dbc891004175830fde769445eb7a","3bbe588c03e947f885725681e4153101","89aaf170bce7437986064d241c8ee4fe","6f98b457b7a84522ae3b73c2d8e5d3cb","6255316ee1ae47c586a843beb99df90d","feba69d737494890b84ea889efc465c4","14055003c4244e02884877195cc24781","47b4f7649218455eb45bdbe9db89a265"]},"id":"LYaI_zLMtDJZ","cellView":"form","executionInfo":{"status":"ok","timestamp":1777158616983,"user_tz":-120,"elapsed":92132,"user":{"displayName":"gfuku63","userId":"10213382542081034600"}},"outputId":"68c1eddd-c40e-40e0-ba3d-d7b22fa4ffcc"},"outputs":[{"output_type":"display_data","data":{"text/plain":["Downloading (incomplete total...): 0.00B [00:00, ?B/s]"],"application/vnd.jupyter.widget-view+json":{"version_major":2,"version_minor":0,"model_id":"e0a1b32fbfb3476997088ef6cf8bd1a4"}},"metadata":{}},{"output_type":"display_data","data":{"text/plain":["Fetching 4 files:   0%|          | 0/4 [00:00<?, ?it/s]"],"application/vnd.jupyter.widget-view+json":{"version_major":2,"version_minor":0,"model_id":"f352233154ef4c12aa4c87f83ffa1af6"}},"metadata":{}},{"output_type":"display_data","data":{"text/plain":["Loading weights:   0%|          | 0/743 [00:00<?, ?it/s]"],"application/vnd.jupyter.widget-view+json":{"version_major":2,"version_minor":0,"model_id":"9406b00f72b74c718ac970a04a3d81a2"}},"metadata":{}},{"output_type":"stream","name":"stderr","text":["The image processor of type `SiglipImageProcessor` is now loaded as a fast processor by default, even if the model checkpoint was saved with a slow processor. This is a breaking change and may produce slightly different outputs. To continue using the slow processor, instantiate this class with `use_fast=False`. \n"]},{"output_type":"stream","name":"stdout","text":["✅ Model loaded in 4bit (bootstrap only)\n"]}],"source":["#@markdown ## 3 - Load 4bit model (temporary bootstrap – low VRAM)\n","\n","import torch\n","from transformers import BitsAndBytesConfig, AutoProcessor, LlavaForConditionalGeneration\n","\n","MODEL_NAME = \"fancyfeast/llama-joycaption-beta-one-hf-llava\"\n","\n","bnb = BitsAndBytesConfig(\n","    load_in_4bit=True,\n","    bnb_4bit_compute_dtype=torch.float16,\n","    bnb_4bit_quant_type=\"nf4\"\n",")\n","\n","model = LlavaForConditionalGeneration.from_pretrained(\n","    MODEL_NAME,\n","    quantization_config=bnb,\n","    device_map=\"auto\",\n","    trust_remote_code=True\n",")\n","\n","processor = AutoProcessor.from_pretrained(MODEL_NAME)\n","\n","print(\"✅ Model loaded in 4bit (bootstrap only)\")"]},{"cell_type":"code","execution_count":3,"metadata":{"id":"u_QVlRCGtJ34","colab":{"base_uri":"https://localhost:8080/"},"cellView":"form","executionInfo":{"status":"ok","timestamp":1777158628603,"user_tz":-120,"elapsed":11619,"user":{"displayName":"gfuku63","userId":"10213382542081034600"}},"outputId":"7ac41d1a-3e4c-42cf-f3a4-24b7c8de6ba4"},"outputs":[{"output_type":"stream","name":"stdout","text":["🔍 Scanning 26 existing shards...\n","   Checking dequant_shard_00000.safetensors... ✅ 107 keys\n","   Checking dequant_shard_00001.safetensors... ✅ 65 keys\n","   Checking dequant_shard_00002.safetensors... ✅ 8 keys\n","   Checking dequant_shard_00003.safetensors... ✅ 8 keys\n","   Checking dequant_shard_00004.safetensors... ✅ 12 keys\n","   Checking dequant_shard_00005.safetensors... ✅ 8 keys\n","   Checking dequant_shard_00006.safetensors... ✅ 8 keys\n","   Checking dequant_shard_00007.safetensors... ✅ 12 keys\n","   Checking dequant_shard_00008.safetensors... ✅ 8 keys\n","   Checking dequant_shard_00009.safetensors... ✅ 8 keys\n","   Checking dequant_shard_00010.safetensors... ✅ 12 keys\n","   Checking dequant_shard_00011.safetensors... ✅ 8 keys\n","   Checking dequant_shard_00012.safetensors... ✅ 8 keys\n","   Checking dequant_shard_00013.safetensors... ✅ 12 keys\n","   Checking dequant_shard_00014.safetensors... ✅ 8 keys\n","   Checking dequant_shard_00015.safetensors... ✅ 8 keys\n","   Checking dequant_shard_00016.safetensors... ✅ 12 keys\n","   Checking dequant_shard_00017.safetensors... ✅ 8 keys\n","   Checking dequant_shard_00018.safetensors... ✅ 8 keys\n","   Checking dequant_shard_00019.safetensors... ✅ 12 keys\n","   Checking dequant_shard_00020.safetensors... ✅ 8 keys\n","   Checking dequant_shard_00021.safetensors... ✅ 8 keys\n","   Checking dequant_shard_00022.safetensors... ✅ 12 keys\n","   Checking dequant_shard_00023.safetensors... ✅ 8 keys\n","   Checking dequant_shard_00024.safetensors... ✅ 8 keys\n","   Checking dequant_shard_00025.safetensors... ✅ 7 keys\n","✅ Total processed keys: 391\n","\n","🚀 Starting CPU-only dequantization...\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.0.self_attn.k_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.0.self_attn.v_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.0.self_attn.q_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.0.self_attn.out_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.0.mlp.fc1\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.0.mlp.fc2\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.1.self_attn.k_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.1.self_attn.v_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.1.self_attn.q_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.1.self_attn.out_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.1.mlp.fc1\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.1.mlp.fc2\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.2.self_attn.k_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.2.self_attn.v_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.2.self_attn.q_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.2.self_attn.out_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.2.mlp.fc1\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.2.mlp.fc2\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.3.self_attn.k_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.3.self_attn.v_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.3.self_attn.q_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.3.self_attn.out_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.3.mlp.fc1\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.3.mlp.fc2\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.4.self_attn.k_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.4.self_attn.v_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.4.self_attn.q_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.4.self_attn.out_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.4.mlp.fc1\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.4.mlp.fc2\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.5.self_attn.k_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.5.self_attn.v_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.5.self_attn.q_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.5.self_attn.out_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.5.mlp.fc1\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.5.mlp.fc2\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.6.self_attn.k_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.6.self_attn.v_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.6.self_attn.q_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.6.self_attn.out_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.6.mlp.fc1\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.6.mlp.fc2\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.7.self_attn.k_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.7.self_attn.v_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.7.self_attn.q_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.7.self_attn.out_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.7.mlp.fc1\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.7.mlp.fc2\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.8.self_attn.k_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.8.self_attn.v_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.8.self_attn.q_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.8.self_attn.out_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.8.mlp.fc1\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.8.mlp.fc2\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.9.self_attn.k_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.9.self_attn.v_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.9.self_attn.q_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.9.self_attn.out_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.9.mlp.fc1\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.9.mlp.fc2\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.10.self_attn.k_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.10.self_attn.v_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.10.self_attn.q_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.10.self_attn.out_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.10.mlp.fc1\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.10.mlp.fc2\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.11.self_attn.k_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.11.self_attn.v_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.11.self_attn.q_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.11.self_attn.out_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.11.mlp.fc1\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.11.mlp.fc2\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.12.self_attn.k_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.12.self_attn.v_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.12.self_attn.q_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.12.self_attn.out_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.12.mlp.fc1\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.12.mlp.fc2\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.13.self_attn.k_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.13.self_attn.v_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.13.self_attn.q_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.13.self_attn.out_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.13.mlp.fc1\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.13.mlp.fc2\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.14.self_attn.k_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.14.self_attn.v_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.14.self_attn.q_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.14.self_attn.out_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.14.mlp.fc1\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.14.mlp.fc2\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.15.self_attn.k_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.15.self_attn.v_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.15.self_attn.q_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.15.self_attn.out_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.15.mlp.fc1\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.15.mlp.fc2\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.16.self_attn.k_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.16.self_attn.v_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.16.self_attn.q_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.16.self_attn.out_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.16.mlp.fc1\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.16.mlp.fc2\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.17.self_attn.k_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.17.self_attn.v_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.17.self_attn.q_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.17.self_attn.out_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.17.mlp.fc1\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.17.mlp.fc2\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.18.self_attn.k_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.18.self_attn.v_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.18.self_attn.q_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.18.self_attn.out_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.18.mlp.fc1\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.18.mlp.fc2\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.19.self_attn.k_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.19.self_attn.v_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.19.self_attn.q_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.19.self_attn.out_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.19.mlp.fc1\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.19.mlp.fc2\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.20.self_attn.k_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.20.self_attn.v_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.20.self_attn.q_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.20.self_attn.out_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.20.mlp.fc1\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.20.mlp.fc2\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.21.self_attn.k_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.21.self_attn.v_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.21.self_attn.q_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.21.self_attn.out_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.21.mlp.fc1\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.21.mlp.fc2\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.22.self_attn.k_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.22.self_attn.v_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.22.self_attn.q_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.22.self_attn.out_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.22.mlp.fc1\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.22.mlp.fc2\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.23.self_attn.k_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.23.self_attn.v_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.23.self_attn.q_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.23.self_attn.out_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.23.mlp.fc1\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.23.mlp.fc2\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.24.self_attn.k_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.24.self_attn.v_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.24.self_attn.q_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.24.self_attn.out_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.24.mlp.fc1\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.24.mlp.fc2\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.25.self_attn.k_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.25.self_attn.v_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.25.self_attn.q_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.25.self_attn.out_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.25.mlp.fc1\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.25.mlp.fc2\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.26.self_attn.k_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.26.self_attn.v_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.26.self_attn.q_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.26.self_attn.out_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.26.mlp.fc1\n","⏭️ Skipping already processed: model.vision_tower.vision_model.encoder.layers.26.mlp.fc2\n","⏭️ Skipping already processed: model.vision_tower.vision_model.head.attention.out_proj\n","⏭️ Skipping already processed: model.vision_tower.vision_model.head.mlp.fc1\n","⏭️ Skipping already processed: model.vision_tower.vision_model.head.mlp.fc2\n","⏭️ Skipping already processed: model.multi_modal_projector.linear_1\n","⏭️ Skipping already processed: model.multi_modal_projector.linear_2\n","⏭️ Skipping already processed: model.language_model.layers.0.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.0.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.0.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.0.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.0.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.0.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.0.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.1.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.1.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.1.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.1.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.1.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.1.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.1.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.2.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.2.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.2.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.2.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.2.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.2.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.2.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.3.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.3.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.3.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.3.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.3.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.3.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.3.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.4.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.4.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.4.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.4.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.4.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.4.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.4.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.5.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.5.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.5.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.5.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.5.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.5.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.5.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.6.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.6.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.6.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.6.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.6.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.6.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.6.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.7.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.7.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.7.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.7.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.7.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.7.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.7.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.8.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.8.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.8.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.8.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.8.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.8.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.8.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.9.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.9.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.9.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.9.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.9.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.9.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.9.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.10.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.10.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.10.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.10.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.10.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.10.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.10.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.11.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.11.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.11.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.11.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.11.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.11.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.11.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.12.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.12.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.12.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.12.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.12.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.12.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.12.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.13.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.13.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.13.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.13.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.13.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.13.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.13.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.14.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.14.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.14.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.14.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.14.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.14.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.14.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.15.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.15.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.15.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.15.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.15.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.15.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.15.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.16.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.16.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.16.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.16.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.16.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.16.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.16.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.17.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.17.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.17.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.17.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.17.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.17.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.17.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.18.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.18.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.18.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.18.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.18.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.18.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.18.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.19.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.19.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.19.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.19.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.19.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.19.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.19.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.20.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.20.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.20.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.20.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.20.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.20.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.20.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.21.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.21.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.21.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.21.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.21.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.21.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.21.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.22.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.22.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.22.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.22.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.22.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.22.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.22.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.23.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.23.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.23.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.23.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.23.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.23.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.23.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.24.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.24.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.24.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.24.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.24.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.24.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.24.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.25.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.25.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.25.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.25.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.25.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.25.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.25.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.26.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.26.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.26.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.26.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.26.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.26.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.26.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.27.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.27.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.27.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.27.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.27.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.27.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.27.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.28.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.28.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.28.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.28.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.28.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.28.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.28.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.29.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.29.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.29.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.29.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.29.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.29.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.29.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.30.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.30.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.30.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.30.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.30.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.30.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.30.mlp.down_proj\n","⏭️ Skipping already processed: model.language_model.layers.31.self_attn.q_proj\n","⏭️ Skipping already processed: model.language_model.layers.31.self_attn.k_proj\n","⏭️ Skipping already processed: model.language_model.layers.31.self_attn.v_proj\n","⏭️ Skipping already processed: model.language_model.layers.31.self_attn.o_proj\n","⏭️ Skipping already processed: model.language_model.layers.31.mlp.gate_proj\n","⏭️ Skipping already processed: model.language_model.layers.31.mlp.up_proj\n","⏭️ Skipping already processed: model.language_model.layers.31.mlp.down_proj\n","✅ Dequantization complete – all weights offloaded to disk\n","🔄 Loading 26 dequant shards (CPU only)...\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.0.mlp.fc1.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.0.mlp.fc2.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.0.self_attn.k_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.0.self_attn.out_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.0.self_attn.q_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.0.self_attn.v_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.1.mlp.fc1.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.1.mlp.fc2.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.1.self_attn.k_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.1.self_attn.out_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.1.self_attn.q_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.1.self_attn.v_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.10.mlp.fc1.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.10.mlp.fc2.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.10.self_attn.k_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.10.self_attn.out_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.10.self_attn.q_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.10.self_attn.v_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.11.mlp.fc1.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.11.mlp.fc2.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.11.self_attn.k_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.11.self_attn.out_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.11.self_attn.q_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.11.self_attn.v_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.12.mlp.fc1.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.12.mlp.fc2.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.12.self_attn.k_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.12.self_attn.out_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.12.self_attn.q_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.12.self_attn.v_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.13.mlp.fc1.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.13.mlp.fc2.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.13.self_attn.k_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.13.self_attn.out_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.13.self_attn.q_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.13.self_attn.v_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.14.mlp.fc1.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.14.mlp.fc2.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.14.self_attn.k_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.14.self_attn.out_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.14.self_attn.q_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.14.self_attn.v_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.15.mlp.fc1.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.15.mlp.fc2.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.15.self_attn.k_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.15.self_attn.out_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.15.self_attn.q_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.15.self_attn.v_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.16.mlp.fc1.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.16.mlp.fc2.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.16.self_attn.k_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.16.self_attn.out_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.16.self_attn.q_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.16.self_attn.v_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.17.mlp.fc1.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.17.self_attn.k_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.17.self_attn.out_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.17.self_attn.q_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.17.self_attn.v_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.2.mlp.fc1.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.2.mlp.fc2.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.2.self_attn.k_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.2.self_attn.out_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.2.self_attn.q_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.2.self_attn.v_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.3.mlp.fc1.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.3.mlp.fc2.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.3.self_attn.k_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.3.self_attn.out_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.3.self_attn.q_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.3.self_attn.v_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.4.mlp.fc1.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.4.mlp.fc2.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.4.self_attn.k_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.4.self_attn.out_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.4.self_attn.q_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.4.self_attn.v_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.5.mlp.fc1.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.5.mlp.fc2.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.5.self_attn.k_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.5.self_attn.out_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.5.self_attn.q_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.5.self_attn.v_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.6.mlp.fc1.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.6.mlp.fc2.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.6.self_attn.k_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.6.self_attn.out_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.6.self_attn.q_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.6.self_attn.v_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.7.mlp.fc1.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.7.mlp.fc2.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.7.self_attn.k_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.7.self_attn.out_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.7.self_attn.q_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.7.self_attn.v_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.8.mlp.fc1.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.8.mlp.fc2.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.8.self_attn.k_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.8.self_attn.out_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.8.self_attn.q_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.8.self_attn.v_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.9.mlp.fc1.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.9.mlp.fc2.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.9.self_attn.k_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.9.self_attn.out_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.9.self_attn.q_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.9.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.0.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.0.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.0.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.0.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.0.self_attn.v_proj.weight\n","   ✅ Restored model.multi_modal_projector.linear_1.weight\n","   ✅ Restored model.multi_modal_projector.linear_2.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.17.mlp.fc2.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.18.mlp.fc1.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.18.mlp.fc2.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.18.self_attn.k_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.18.self_attn.out_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.18.self_attn.q_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.18.self_attn.v_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.19.mlp.fc1.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.19.mlp.fc2.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.19.self_attn.k_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.19.self_attn.out_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.19.self_attn.q_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.19.self_attn.v_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.20.mlp.fc1.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.20.mlp.fc2.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.20.self_attn.k_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.20.self_attn.out_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.20.self_attn.q_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.20.self_attn.v_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.21.mlp.fc1.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.21.mlp.fc2.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.21.self_attn.k_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.21.self_attn.out_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.21.self_attn.q_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.21.self_attn.v_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.22.mlp.fc1.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.22.mlp.fc2.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.22.self_attn.k_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.22.self_attn.out_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.22.self_attn.q_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.22.self_attn.v_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.23.mlp.fc1.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.23.mlp.fc2.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.23.self_attn.k_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.23.self_attn.out_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.23.self_attn.q_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.23.self_attn.v_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.24.mlp.fc1.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.24.mlp.fc2.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.24.self_attn.k_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.24.self_attn.out_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.24.self_attn.q_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.24.self_attn.v_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.25.mlp.fc1.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.25.mlp.fc2.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.25.self_attn.k_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.25.self_attn.out_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.25.self_attn.q_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.25.self_attn.v_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.26.mlp.fc1.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.26.mlp.fc2.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.26.self_attn.k_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.26.self_attn.out_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.26.self_attn.q_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.encoder.layers.26.self_attn.v_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.head.attention.out_proj.weight\n","   ✅ Restored model.vision_tower.vision_model.head.mlp.fc1.weight\n","   ✅ Restored model.vision_tower.vision_model.head.mlp.fc2.weight\n","   ✅ Restored model.language_model.layers.0.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.0.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.1.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.1.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.1.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.1.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.1.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.1.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.1.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.2.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.2.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.2.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.2.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.2.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.2.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.2.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.3.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.3.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.3.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.3.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.3.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.3.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.3.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.4.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.4.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.4.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.4.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.4.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.4.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.4.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.5.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.5.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.5.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.5.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.5.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.5.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.5.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.6.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.6.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.6.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.6.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.6.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.6.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.6.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.7.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.7.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.7.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.7.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.7.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.7.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.7.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.8.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.8.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.8.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.8.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.8.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.8.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.8.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.9.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.9.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.9.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.9.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.9.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.9.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.10.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.10.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.10.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.10.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.10.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.10.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.10.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.9.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.11.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.11.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.11.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.11.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.11.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.11.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.11.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.12.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.12.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.12.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.12.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.12.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.12.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.12.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.13.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.13.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.13.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.13.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.13.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.13.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.13.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.14.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.14.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.14.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.14.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.14.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.14.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.14.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.15.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.15.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.15.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.15.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.15.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.15.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.15.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.16.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.16.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.16.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.16.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.16.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.16.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.16.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.17.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.17.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.17.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.17.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.17.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.17.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.17.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.18.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.18.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.18.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.18.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.18.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.18.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.18.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.19.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.19.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.19.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.19.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.19.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.19.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.19.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.20.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.20.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.20.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.20.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.20.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.20.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.20.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.21.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.21.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.21.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.21.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.21.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.21.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.21.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.22.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.22.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.22.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.22.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.22.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.22.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.22.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.23.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.23.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.23.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.23.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.23.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.23.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.23.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.24.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.24.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.24.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.24.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.24.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.24.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.24.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.25.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.25.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.25.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.25.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.25.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.25.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.25.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.26.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.26.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.26.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.26.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.26.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.26.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.26.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.27.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.27.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.27.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.27.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.27.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.27.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.27.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.28.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.28.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.28.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.28.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.28.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.28.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.28.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.29.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.29.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.29.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.29.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.29.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.29.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.29.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.30.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.30.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.30.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.30.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.30.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.30.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.30.self_attn.v_proj.weight\n","   ✅ Restored model.language_model.layers.31.mlp.down_proj.weight\n","   ✅ Restored model.language_model.layers.31.mlp.gate_proj.weight\n","   ✅ Restored model.language_model.layers.31.mlp.up_proj.weight\n","   ✅ Restored model.language_model.layers.31.self_attn.k_proj.weight\n","   ✅ Restored model.language_model.layers.31.self_attn.o_proj.weight\n","   ✅ Restored model.language_model.layers.31.self_attn.q_proj.weight\n","   ✅ Restored model.language_model.layers.31.self_attn.v_proj.weight\n","✅ All shards loaded on CPU\n","✅ Fully dequantized on CPU – ready for SDNQ\n"]}],"source":["#@markdown ## 4 - Low-RAM dequant + fast sharding (CPU-only + ROBUST resume)\n","\n","import os\n","import gc\n","import torch\n","import json\n","from safetensors.torch import save_file, load_file\n","import bitsandbytes.functional as F\n","\n","# ========================= CONFIG =========================\n","SHARD_DIR = \"/content/joycaption_dequant_shards\"\n","os.makedirs(SHARD_DIR, exist_ok=True)\n","MAX_SHARD_SIZE_GB = 0.5\n","# =========================================================\n","\n","def _get_safetensors_keys(path):\n","    \"\"\"Ultra-reliable way to read keys from any safetensors file\"\"\"\n","    try:\n","        with open(path, \"rb\") as f:\n","            length = int.from_bytes(f.read(8), \"little\")\n","            header = json.loads(f.read(length).decode(\"utf-8\"))\n","            keys = [k for k in header.keys() if k != \"__metadata__\"]\n","            return keys\n","    except Exception as e:\n","        print(f\"   ❌ Failed to parse header of {os.path.basename(path)}: {e}\")\n","        return []\n","\n","def get_processed_keys(shard_dir=SHARD_DIR):\n","    processed = set()\n","    shard_files = sorted([f for f in os.listdir(shard_dir) if f.endswith(\".safetensors\")])\n","    print(f\"🔍 Scanning {len(shard_files)} existing shards...\")\n","    if not shard_files:\n","        print(\"   → No shards found yet\")\n","        return processed\n","    for shard_file in shard_files:\n","        path = os.path.join(shard_dir, shard_file)\n","        print(f\"   Checking {shard_file}...\", end=\" \")\n","        keys = _get_safetensors_keys(path)\n","        if keys:\n","            print(f\"✅ {len(keys)} keys\")\n","            processed.update(keys)\n","        else:\n","            print(\"⚠️ no keys parsed\")\n","    print(f\"✅ Total processed keys: {len(processed)}\\n\")\n","    return processed\n","\n","def load_dequant_shards(model, shard_dir=SHARD_DIR):\n","    shard_files = sorted([f for f in os.listdir(shard_dir) if f.endswith(\".safetensors\")])\n","    if not shard_files:\n","        print(\"No shards found – starting from scratch\")\n","        return model\n","    print(f\"🔄 Loading {len(shard_files)} dequant shards (CPU only)...\")\n","    model = model.cpu()\n","    torch.cuda.empty_cache()\n","    gc.collect()\n","    for shard_file in shard_files:\n","        path = os.path.join(shard_dir, shard_file)\n","        state = load_file(path, device=\"cpu\")\n","        for key, tensor in state.items():\n","            try:\n","                submodule_name = \".\".join(key.split(\".\")[:-1])\n","                submodule = model.get_submodule(submodule_name)\n","                if hasattr(submodule, \"weight\"):\n","                    submodule.weight = torch.nn.Parameter(tensor.to(torch.float16))\n","                    print(f\"   ✅ Restored {key}\")\n","            except Exception as e:\n","                print(f\"   ⚠️ Could not restore {key}: {e}\")\n","    print(\"✅ All shards loaded on CPU\")\n","    return model\n","\n","def dequantize_model_to_fp16(model, shard_dir=SHARD_DIR, max_shard_gb=0.5):\n","    model = model.cpu()\n","    torch.cuda.empty_cache()\n","    gc.collect()\n","    processed_keys = get_processed_keys(shard_dir)\n","    cumulative_bytes = 0\n","    current_shard_dict = {}\n","    shard_id = len([f for f in os.listdir(shard_dir) if f.endswith(\".safetensors\")])\n","    print(\"🚀 Starting CPU-only dequantization...\")\n","    for name, module in list(model.named_modules()):\n","        if not (hasattr(module, \"weight\") and module.weight is not None):\n","            continue\n","        if not hasattr(module.weight, \"quant_state\"):\n","            continue\n","        key = name + \".weight\"\n","        if key in processed_keys:\n","            print(f\"⏭️ Skipping already processed: {name}\")\n","            continue\n","        print(f\"Dequantizing: {name}\")\n","        try:\n","            weight = module.weight\n","            quant_state = weight.quant_state\n","            weight_cpu = weight.to(\"cpu\")\n","            del module.weight\n","            del weight\n","            torch.cuda.empty_cache()\n","            gc.collect()\n","            dequant = F.dequantize_4bit(weight_cpu, quant_state)\n","            module.weight = torch.nn.Parameter(dequant.to(torch.float16))\n","            current_shard_dict[key] = module.weight.data.clone()\n","            layer_bytes = module.weight.numel() * module.weight.element_size()\n","            cumulative_bytes += layer_bytes\n","            print(f\"   → {layer_bytes / (1024**3):.3f} GB | cumulative {cumulative_bytes / (1024**3):.3f} GB\")\n","            if cumulative_bytes / (1024**3) >= max_shard_gb:\n","                shard_path = os.path.join(shard_dir, f\"dequant_shard_{shard_id:05d}.safetensors\")\n","                save_file(current_shard_dict, shard_path)\n","                print(f\"💾 Saved shard → {shard_path} ({len(current_shard_dict)} keys)\")\n","                for saved_key in list(current_shard_dict.keys()):\n","                    try:\n","                        sub_name = \".\".join(saved_key.split(\".\")[:-1])\n","                        sub = model.get_submodule(sub_name)\n","                        if hasattr(sub, \"weight\"):\n","                            del sub.weight\n","                            sub.weight = None\n","                    except:\n","                        pass\n","                current_shard_dict = {}\n","                cumulative_bytes = 0\n","                shard_id += 1\n","            del dequant\n","            del weight_cpu\n","            gc.collect()\n","        except Exception as e:\n","            print(f\"❌ Skipped {name}: {e}\")\n","            gc.collect()\n","    if current_shard_dict:\n","        shard_path = os.path.join(shard_dir, f\"dequant_shard_{shard_id:05d}.safetensors\")\n","        save_file(current_shard_dict, shard_path)\n","        print(f\"💾 Saved final shard → {shard_path}\")\n","        for saved_key in list(current_shard_dict.keys()):\n","            try:\n","                sub_name = \".\".join(saved_key.split(\".\")[:-1])\n","                sub = model.get_submodule(sub_name)\n","                if hasattr(sub, \"weight\"):\n","                    del sub.weight\n","                    sub.weight = None\n","            except:\n","                pass\n","    print(\"✅ Dequantization complete – all weights offloaded to disk\")\n","    return model\n","\n","#@markdown ## Run dequantization (Step 4)\n","model = dequantize_model_to_fp16(model, SHARD_DIR, MAX_SHARD_SIZE_GB)\n","model = load_dequant_shards(model, SHARD_DIR)\n","print(\"✅ Fully dequantized on CPU – ready for SDNQ\")"]},{"cell_type":"code","source":["#@markdown ## 5 - SDNQ uint4 quantization (official high-level API + RESUME)\n","\n","import os\n","import torch\n","import gc\n","from safetensors.torch import load_file, save_file\n","from sdnq import sdnq_post_load_quant   # ← official post-load function\n","\n","FP16_SHARD_DIR = \"/content/joycaption_dequant_shards\"\n","SDNQ_SHARD_DIR = \"/content/joycaption_sdnq_shards\"\n","os.makedirs(SDNQ_SHARD_DIR, exist_ok=True)\n","\n","# ====================== SDNQ CONFIG (current official) ======================\n","SDNQ_CONFIG = {\n","    \"weights_dtype\": \"uint4\",           # supported by SDNQ\n","    \"group_size\": 128,\n","    \"quantization_device\": \"cpu\",\n","    \"return_device\": \"cpu\",\n","    \"use_svd\": False,\n","    \"quant_conv\": False,\n","    \"quant_embedding\": False,\n","    \"use_quantized_matmul\": False,\n","    \"use_quantized_matmul_conv\": False,\n","    \"use_dynamic_quantization\": False,\n","    \"dequantize_fp32\": True,\n","    \"non_blocking\": False,\n","    \"add_skip_keys\": True,\n","    \"modules_to_not_convert\": [\"lm_head\", \"embedding_projection\"],\n","}\n","\n","# ====================== RESUME LOGIC ======================\n","fp16_shard_files = sorted([f for f in os.listdir(FP16_SHARD_DIR) if f.endswith(\".safetensors\")])\n","already_quantized = {f for f in os.listdir(SDNQ_SHARD_DIR) if f.endswith(\".safetensors\")}\n","\n","remaining_shards = [f for f in fp16_shard_files if f not in already_quantized]\n","\n","print(f\"🚀 Found {len(fp16_shard_files)} FP16 shards.\")\n","print(f\"   → {len(already_quantized)} already quantized (resume mode)\")\n","print(f\"   → {len(remaining_shards)} shards need processing...\\n\")\n","\n","if not remaining_shards:\n","    print(\"✅ ALL shards already quantized! Skipping quantization.\")\n","else:\n","    print(\"🔄 Applying SDNQ uint4 quantization to full model (official API)...\")\n","\n","    # The model is already fully loaded in FP16 on CPU from Cell 4\n","    model = sdnq_post_load_quant(model, **SDNQ_CONFIG)\n","\n","    print(\"✅ SDNQ uint4 quantization complete!\")\n","\n","    # Optional: re-save the quantized weights as shards (for resume / safety)\n","    print(\"💾 Saving quantized weights as shards...\")\n","    shard_id = len(already_quantized)\n","    state = model.state_dict()\n","    current_shard = {}\n","    cumulative_bytes = 0\n","    MAX_SHARD_GB = 0.5\n","\n","    for key, tensor in state.items():\n","        current_shard[key] = tensor\n","        layer_bytes = tensor.numel() * tensor.element_size()\n","        cumulative_bytes += layer_bytes\n","\n","        if cumulative_bytes / (1024**3) >= MAX_SHARD_GB:\n","            shard_path = os.path.join(SDNQ_SHARD_DIR, f\"sdnq_shard_{shard_id:05d}.safetensors\")\n","            save_file(current_shard, shard_path)\n","            print(f\"   Saved {shard_path} ({len(current_shard)} keys)\")\n","            current_shard = {}\n","            cumulative_bytes = 0\n","            shard_id += 1\n","\n","    if current_shard:\n","        shard_path = os.path.join(SDNQ_SHARD_DIR, f\"sdnq_shard_{shard_id:05d}.safetensors\")\n","        save_file(current_shard, shard_path)\n","        print(f\"   Saved final shard {shard_path}\")\n","\n","print(\"🎉 SDNQ quantization finished – model is now uint4 on CPU\")\n","gc.collect()\n","torch.cuda.empty_cache()"],"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"cellView":"form","id":"0o6J6b9q8cSc","executionInfo":{"status":"ok","timestamp":1777159282456,"user_tz":-120,"elapsed":18134,"user":{"displayName":"gfuku63","userId":"10213382542081034600"}},"outputId":"a0ded726-0b2b-4edc-e465-69ef91fbf314"},"execution_count":1,"outputs":[{"output_type":"stream","name":"stdout","text":["🚀 Found 26 FP16 shards.\n","   → 26 already quantized (resume mode)\n","   → 0 shards need processing...\n","\n","✅ ALL shards already quantized! Skipping quantization.\n","🎉 SDNQ quantization finished – model is now uint4 on CPU\n"]}]},{"cell_type":"code","source":["#@markdown ## 6 - Save clean single-file SDNQ model + push to HF\n","\n","import os\n","import shutil\n","import json\n","import gc\n","from huggingface_hub import HfApi, snapshot_download\n","\n","SAVE_DIR = \"/content/llava_sdnq_single\"\n","os.makedirs(SAVE_DIR, exist_ok=True)\n","\n","print(\"💾 Saving clean SDNQ model (single large safetensors file)...\")\n","model = model.to(\"cpu\")\n","torch.cuda.empty_cache()\n","gc.collect()\n","\n","model.save_pretrained(\n","    SAVE_DIR,\n","    safe_serialization=True,\n","    max_shard_size=\"1GB\",   # forces basically one file\n",")\n","\n","processor.save_pretrained(SAVE_DIR)\n","print(f\"✅ Model + processor saved to {SAVE_DIR}\")\n","\n","# Copy original non-weight files (README, generation_config, etc.)\n","print(\"📥 Copying config files from original repo...\")\n","original_small = snapshot_download(\n","    repo_id=MODEL_NAME,\n","    ignore_patterns=[\"*.safetensors\", \"*.bin\", \"*.pt\"],\n","    cache_dir=\"/tmp/original_joycaption\"\n",")\n","\n","for item in os.listdir(original_small):\n","    src = os.path.join(original_small, item)\n","    dst = os.path.join(SAVE_DIR, item)\n","    if os.path.isfile(src) and not os.path.exists(dst):\n","        shutil.copy2(src, dst)\n","        print(f\"   Copied {item}\")\n","\n","# Ensure quantization_config is in config.json\n","config_path = os.path.join(SAVE_DIR, \"config.json\")\n","with open(config_path, \"r\") as f:\n","    config = json.load(f)\n","\n","config[\"quantization_config\"] = {\n","    \"quant_method\": \"sdnq\",\n","    \"weights_dtype\": \"uint4\",\n","    \"group_size\": 128,\n","    \"quantization_device\": \"cuda\",\n","    \"return_device\": \"cuda\",\n","    \"use_svd\": False,\n","    \"quant_conv\": False,\n","    \"quant_embedding\": False,\n","    \"use_quantized_matmul\": False,\n","    \"dequantize_fp32\": True,\n","    \"add_skip_keys\": True\n","}\n","\n","with open(config_path, \"w\") as f:\n","    json.dump(config, f, indent=2)\n","print(\"✅ quantization_config written to config.json\")\n","\n","# ====================== PUSH TO HF ======================\n","print(f\"🌐 Pushing to https://huggingface.co/codeShare/joycaption_beta_one_SDNQ ...\")\n","api = HfApi(token=hf_token)\n","repo_id = \"codeShare/joycaption_beta_one_SDNQ\"\n","api.create_repo(repo_id, exist_ok=True)\n","\n","for filename in sorted(os.listdir(SAVE_DIR)):\n","    file_path = os.path.join(SAVE_DIR, filename)\n","    if os.path.isfile(file_path):\n","        api.upload_file(\n","            path_or_fileobj=file_path,\n","            path_in_repo=filename,\n","            repo_id=repo_id,\n","            commit_message=\"SDNQ uint4 JoyCaption – clean single-file build (official SDNQ API)\"\n","        )\n","        print(f\"   Uploaded {filename}\")\n","\n","print(\"\\n🎉 DONE! Model is now live at https://huggingface.co/codeShare/joycaption_beta_one_SDNQ\")\n","print(\"   → Load it with your normal FP32 / SDNQ loader – it should now report clean uint4 weights.\")\n","gc.collect()\n","torch.cuda.empty_cache()"],"metadata":{"cellView":"form","id":"Ir-mrADZ8hru"},"execution_count":null,"outputs":[]}],"metadata":{"accelerator":"GPU","colab":{"collapsed_sections":["Q0ZGy2-ECTCY","p1V0L8u_FGaK","92MEeRpOh7eX"],"gpuType":"T4","provenance":[{"file_id":"https://huggingface.co/datasets/codeShare/lora-training-data/blob/main/build joycap sdnq.ipynb","timestamp":1777110475920}]},"kernelspec":{"display_name":"Python 3","name":"python3"},"language_info":{"name":"python"},"widgets":{"application/vnd.jupyter.widget-state+json":{"e0a1b32fbfb3476997088ef6cf8bd1a4":{"model_module":"@jupyter-widgets/controls","model_name":"HBoxModel","model_module_version":"1.5.0","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"HBoxModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"HBoxView","box_style":"","children":["IPY_MODEL_735f3bc73aa54a3fab17436188c31577","IPY_MODEL_5264574b812840f5b379c5704f123b64","IPY_MODEL_989634a219a44642a6f2e0748960655f"],"layout":"IPY_MODEL_7bd8f165ed8c49f192d3f42739ac963a"}},"735f3bc73aa54a3fab17436188c31577":{"model_module":"@jupyter-widgets/controls","model_name":"HTMLModel","model_module_version":"1.5.0","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"HTMLModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"HTMLView","description":"","description_tooltip":null,"layout":"IPY_MODEL_dcb82126aae448f0a37c1c5661011795","placeholder":"​","style":"IPY_MODEL_2eabc79a4afd4b07a6ce7cca897e4450","value":"Download complete: "}},"5264574b812840f5b379c5704f123b64":{"model_module":"@jupyter-widgets/controls","model_name":"FloatProgressModel","model_module_version":"1.5.0","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"FloatProgressModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"ProgressView","bar_style":"success","description":"","description_tooltip":null,"layout":"IPY_MODEL_7542c6ce9a544a81aad3cbf3669186ce","max":1,"min":0,"orientation":"horizontal","style":"IPY_MODEL_0b38405f4e864526b85af2fbc1267e03","value":0}},"989634a219a44642a6f2e0748960655f":{"model_module":"@jupyter-widgets/controls","model_name":"HTMLModel","model_module_version":"1.5.0","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"HTMLModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"HTMLView","description":"","description_tooltip":null,"layout":"IPY_MODEL_d4905952a5c84841abc842cafc514366","placeholder":"​","style":"IPY_MODEL_45c11fc828d94b779e7345ea114179b0","value":" 0.00/0.00 [00:00&lt;?, ?B/s]"}},"7bd8f165ed8c49f192d3f42739ac963a":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"dcb82126aae448f0a37c1c5661011795":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"2eabc79a4afd4b07a6ce7cca897e4450":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"DescriptionStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"StyleView","description_width":""}},"7542c6ce9a544a81aad3cbf3669186ce":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":"20px"}},"0b38405f4e864526b85af2fbc1267e03":{"model_module":"@jupyter-widgets/controls","model_name":"ProgressStyleModel","model_module_version":"1.5.0","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"ProgressStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"StyleView","bar_color":null,"description_width":""}},"d4905952a5c84841abc842cafc514366":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"45c11fc828d94b779e7345ea114179b0":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"DescriptionStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"StyleView","description_width":""}},"f352233154ef4c12aa4c87f83ffa1af6":{"model_module":"@jupyter-widgets/controls","model_name":"HBoxModel","model_module_version":"1.5.0","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"HBoxModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"HBoxView","box_style":"","children":["IPY_MODEL_b6a0c2c218de4b9dba80f6bb39fe4dff","IPY_MODEL_93176536d88041269f3e9d4c692e6557","IPY_MODEL_4e02ede43b9045c2aff16a102b4c212a"],"layout":"IPY_MODEL_681933b0634645b98be2c3949cdb6149"}},"b6a0c2c218de4b9dba80f6bb39fe4dff":{"model_module":"@jupyter-widgets/controls","model_name":"HTMLModel","model_module_version":"1.5.0","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"HTMLModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"HTMLView","description":"","description_tooltip":null,"layout":"IPY_MODEL_7fa9e8ffa7d2414ea4e7cefde801b958","placeholder":"​","style":"IPY_MODEL_44b0c67e2d044bfdb564dc1aea15cfd2","value":"Fetching 4 files: 100%"}},"93176536d88041269f3e9d4c692e6557":{"model_module":"@jupyter-widgets/controls","model_name":"FloatProgressModel","model_module_version":"1.5.0","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"FloatProgressModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"ProgressView","bar_style":"success","description":"","description_tooltip":null,"layout":"IPY_MODEL_cf350b2c8159406ab09cc8ec12126ec4","max":4,"min":0,"orientation":"horizontal","style":"IPY_MODEL_828a3b771f664bcca4205ccd6ca14315","value":4}},"4e02ede43b9045c2aff16a102b4c212a":{"model_module":"@jupyter-widgets/controls","model_name":"HTMLModel","model_module_version":"1.5.0","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"HTMLModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"HTMLView","description":"","description_tooltip":null,"layout":"IPY_MODEL_62bda97392ba43069274532eaaa56c1e","placeholder":"​","style":"IPY_MODEL_41ac9dc3ccc14c35a0ee71c229e815c7","value":" 4/4 [00:00&lt;00:00, 204.38it/s]"}},"681933b0634645b98be2c3949cdb6149":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"7fa9e8ffa7d2414ea4e7cefde801b958":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"44b0c67e2d044bfdb564dc1aea15cfd2":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"DescriptionStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"StyleView","description_width":""}},"cf350b2c8159406ab09cc8ec12126ec4":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"828a3b771f664bcca4205ccd6ca14315":{"model_module":"@jupyter-widgets/controls","model_name":"ProgressStyleModel","model_module_version":"1.5.0","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"ProgressStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"StyleView","bar_color":null,"description_width":""}},"62bda97392ba43069274532eaaa56c1e":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"41ac9dc3ccc14c35a0ee71c229e815c7":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"DescriptionStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"StyleView","description_width":""}},"9406b00f72b74c718ac970a04a3d81a2":{"model_module":"@jupyter-widgets/controls","model_name":"HBoxModel","model_module_version":"1.5.0","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"HBoxModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"HBoxView","box_style":"","children":["IPY_MODEL_a7923196bd3841a49231d5116da5c251","IPY_MODEL_9d88f227be7f408f8b4414e3fdf7d203","IPY_MODEL_c548dbc891004175830fde769445eb7a"],"layout":"IPY_MODEL_3bbe588c03e947f885725681e4153101"}},"a7923196bd3841a49231d5116da5c251":{"model_module":"@jupyter-widgets/controls","model_name":"HTMLModel","model_module_version":"1.5.0","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"HTMLModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"HTMLView","description":"","description_tooltip":null,"layout":"IPY_MODEL_89aaf170bce7437986064d241c8ee4fe","placeholder":"​","style":"IPY_MODEL_6f98b457b7a84522ae3b73c2d8e5d3cb","value":"Loading weights: 100%"}},"9d88f227be7f408f8b4414e3fdf7d203":{"model_module":"@jupyter-widgets/controls","model_name":"FloatProgressModel","model_module_version":"1.5.0","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"FloatProgressModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"ProgressView","bar_style":"success","description":"","description_tooltip":null,"layout":"IPY_MODEL_6255316ee1ae47c586a843beb99df90d","max":743,"min":0,"orientation":"horizontal","style":"IPY_MODEL_feba69d737494890b84ea889efc465c4","value":743}},"c548dbc891004175830fde769445eb7a":{"model_module":"@jupyter-widgets/controls","model_name":"HTMLModel","model_module_version":"1.5.0","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"HTMLModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"HTMLView","description":"","description_tooltip":null,"layout":"IPY_MODEL_14055003c4244e02884877195cc24781","placeholder":"​","style":"IPY_MODEL_47b4f7649218455eb45bdbe9db89a265","value":" 743/743 [01:10&lt;00:00, 586.88it/s, Materializing param=model.vision_tower.vision_model.post_layernorm.weight]"}},"3bbe588c03e947f885725681e4153101":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"89aaf170bce7437986064d241c8ee4fe":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"6f98b457b7a84522ae3b73c2d8e5d3cb":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"DescriptionStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"StyleView","description_width":""}},"6255316ee1ae47c586a843beb99df90d":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"feba69d737494890b84ea889efc465c4":{"model_module":"@jupyter-widgets/controls","model_name":"ProgressStyleModel","model_module_version":"1.5.0","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"ProgressStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"StyleView","bar_color":null,"description_width":""}},"14055003c4244e02884877195cc24781":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"47b4f7649218455eb45bdbe9db89a265":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"DescriptionStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"StyleView","description_width":""}}}}},"nbformat":4,"nbformat_minor":0}