{ "cells": [ { "cell_type": "code", "execution_count": 1, "metadata": {}, "outputs": [], "source": [ "import numpy as np\n", "import torch\n", "import os\n", "\n" ] }, { "cell_type": "code", "execution_count": 4, "metadata": {}, "outputs": [], "source": [ "from PIL import Image\n" ] }, { "cell_type": "code", "execution_count": 3, "metadata": {}, "outputs": [], "source": [ "import matplotlib.pyplot as plt" ] }, { "cell_type": "code", "execution_count": 5, "metadata": {}, "outputs": [], "source": [ "im_path = \"/media/ubuntu/New Volume/jay/fundus_images/archive/REFUGE/val/Masks/V0001.png\"\n", "img = torch.as_tensor(np.array(Image.open(im_path).convert(\"RGB\")))\n" ] }, { "cell_type": "code", "execution_count": 8, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "array([0, 1, 2], dtype=uint8)" ] }, "execution_count": 8, "metadata": {}, "output_type": "execute_result" } ], "source": [ "np.unique(img[:,:,0])" ] }, { "cell_type": "code", "execution_count": 3, "metadata": {}, "outputs": [], "source": [ "model_dict_path = \"./eval/cholec8k/svd_shiftscale_cholec_tal_focal075_1e-4.pth\"\n", "model_dict = torch.load(model_dict_path, map_location='cpu')" ] }, { "cell_type": "code", "execution_count": 4, "metadata": {}, "outputs": [], "source": [ "keys = list(model_dict.keys())" ] }, { "cell_type": "code", "execution_count": 5, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "['sam_encoder.pos_embed',\n", " 'sam_encoder.patch_embed.proj.weight',\n", " 'sam_encoder.patch_embed.proj.bias',\n", " 'sam_encoder.blocks.0.norm1.weight',\n", " 'sam_encoder.blocks.0.norm1.bias',\n", " 'sam_encoder.blocks.0.attn.qkv.weight',\n", " 'sam_encoder.blocks.0.attn.qkv.bias',\n", " 'sam_encoder.blocks.0.attn.qkv.trainable_scale',\n", " 'sam_encoder.blocks.0.attn.qkv.trainable_shift',\n", " 'sam_encoder.blocks.0.attn.proj.weight',\n", " 'sam_encoder.blocks.0.attn.proj.bias',\n", " 'sam_encoder.blocks.0.attn.proj.trainable_scale',\n", " 'sam_encoder.blocks.0.attn.proj.trainable_shift',\n", " 'sam_encoder.blocks.0.norm2.weight',\n", " 'sam_encoder.blocks.0.norm2.bias',\n", " 'sam_encoder.blocks.0.mlp.lin1.weight',\n", " 'sam_encoder.blocks.0.mlp.lin1.bias',\n", " 'sam_encoder.blocks.0.mlp.lin1.trainable_scale',\n", " 'sam_encoder.blocks.0.mlp.lin1.trainable_shift',\n", " 'sam_encoder.blocks.0.mlp.lin2.weight',\n", " 'sam_encoder.blocks.0.mlp.lin2.bias',\n", " 'sam_encoder.blocks.0.mlp.lin2.trainable_scale',\n", " 'sam_encoder.blocks.0.mlp.lin2.trainable_shift',\n", " 'sam_encoder.blocks.1.norm1.weight',\n", " 'sam_encoder.blocks.1.norm1.bias',\n", " 'sam_encoder.blocks.1.attn.qkv.weight',\n", " 'sam_encoder.blocks.1.attn.qkv.bias',\n", " 'sam_encoder.blocks.1.attn.qkv.trainable_scale',\n", " 'sam_encoder.blocks.1.attn.qkv.trainable_shift',\n", " 'sam_encoder.blocks.1.attn.proj.weight',\n", " 'sam_encoder.blocks.1.attn.proj.bias',\n", " 'sam_encoder.blocks.1.attn.proj.trainable_scale',\n", " 'sam_encoder.blocks.1.attn.proj.trainable_shift',\n", " 'sam_encoder.blocks.1.norm2.weight',\n", " 'sam_encoder.blocks.1.norm2.bias',\n", " 'sam_encoder.blocks.1.mlp.lin1.weight',\n", " 'sam_encoder.blocks.1.mlp.lin1.bias',\n", " 'sam_encoder.blocks.1.mlp.lin1.trainable_scale',\n", " 'sam_encoder.blocks.1.mlp.lin1.trainable_shift',\n", " 'sam_encoder.blocks.1.mlp.lin2.weight',\n", " 'sam_encoder.blocks.1.mlp.lin2.bias',\n", " 'sam_encoder.blocks.1.mlp.lin2.trainable_scale',\n", " 'sam_encoder.blocks.1.mlp.lin2.trainable_shift',\n", " 'sam_encoder.blocks.2.norm1.weight',\n", " 'sam_encoder.blocks.2.norm1.bias',\n", " 'sam_encoder.blocks.2.attn.qkv.weight',\n", " 'sam_encoder.blocks.2.attn.qkv.bias',\n", " 'sam_encoder.blocks.2.attn.qkv.trainable_scale',\n", " 'sam_encoder.blocks.2.attn.qkv.trainable_shift',\n", " 'sam_encoder.blocks.2.attn.proj.weight',\n", " 'sam_encoder.blocks.2.attn.proj.bias',\n", " 'sam_encoder.blocks.2.attn.proj.trainable_scale',\n", " 'sam_encoder.blocks.2.attn.proj.trainable_shift',\n", " 'sam_encoder.blocks.2.norm2.weight',\n", " 'sam_encoder.blocks.2.norm2.bias',\n", " 'sam_encoder.blocks.2.mlp.lin1.weight',\n", " 'sam_encoder.blocks.2.mlp.lin1.bias',\n", " 'sam_encoder.blocks.2.mlp.lin1.trainable_scale',\n", " 'sam_encoder.blocks.2.mlp.lin1.trainable_shift',\n", " 'sam_encoder.blocks.2.mlp.lin2.weight',\n", " 'sam_encoder.blocks.2.mlp.lin2.bias',\n", " 'sam_encoder.blocks.2.mlp.lin2.trainable_scale',\n", " 'sam_encoder.blocks.2.mlp.lin2.trainable_shift',\n", " 'sam_encoder.blocks.3.norm1.weight',\n", " 'sam_encoder.blocks.3.norm1.bias',\n", " 'sam_encoder.blocks.3.attn.qkv.weight',\n", " 'sam_encoder.blocks.3.attn.qkv.bias',\n", " 'sam_encoder.blocks.3.attn.qkv.trainable_scale',\n", " 'sam_encoder.blocks.3.attn.qkv.trainable_shift',\n", " 'sam_encoder.blocks.3.attn.proj.weight',\n", " 'sam_encoder.blocks.3.attn.proj.bias',\n", " 'sam_encoder.blocks.3.attn.proj.trainable_scale',\n", " 'sam_encoder.blocks.3.attn.proj.trainable_shift',\n", " 'sam_encoder.blocks.3.norm2.weight',\n", " 'sam_encoder.blocks.3.norm2.bias',\n", " 'sam_encoder.blocks.3.mlp.lin1.weight',\n", " 'sam_encoder.blocks.3.mlp.lin1.bias',\n", " 'sam_encoder.blocks.3.mlp.lin1.trainable_scale',\n", " 'sam_encoder.blocks.3.mlp.lin1.trainable_shift',\n", " 'sam_encoder.blocks.3.mlp.lin2.weight',\n", " 'sam_encoder.blocks.3.mlp.lin2.bias',\n", " 'sam_encoder.blocks.3.mlp.lin2.trainable_scale',\n", " 'sam_encoder.blocks.3.mlp.lin2.trainable_shift',\n", " 'sam_encoder.blocks.4.norm1.weight',\n", " 'sam_encoder.blocks.4.norm1.bias',\n", " 'sam_encoder.blocks.4.attn.qkv.weight',\n", " 'sam_encoder.blocks.4.attn.qkv.bias',\n", " 'sam_encoder.blocks.4.attn.qkv.trainable_scale',\n", " 'sam_encoder.blocks.4.attn.qkv.trainable_shift',\n", " 'sam_encoder.blocks.4.attn.proj.weight',\n", " 'sam_encoder.blocks.4.attn.proj.bias',\n", " 'sam_encoder.blocks.4.attn.proj.trainable_scale',\n", " 'sam_encoder.blocks.4.attn.proj.trainable_shift',\n", " 'sam_encoder.blocks.4.norm2.weight',\n", " 'sam_encoder.blocks.4.norm2.bias',\n", " 'sam_encoder.blocks.4.mlp.lin1.weight',\n", " 'sam_encoder.blocks.4.mlp.lin1.bias',\n", " 'sam_encoder.blocks.4.mlp.lin1.trainable_scale',\n", " 'sam_encoder.blocks.4.mlp.lin1.trainable_shift',\n", " 'sam_encoder.blocks.4.mlp.lin2.weight',\n", " 'sam_encoder.blocks.4.mlp.lin2.bias',\n", " 'sam_encoder.blocks.4.mlp.lin2.trainable_scale',\n", " 'sam_encoder.blocks.4.mlp.lin2.trainable_shift',\n", " 'sam_encoder.blocks.5.norm1.weight',\n", " 'sam_encoder.blocks.5.norm1.bias',\n", " 'sam_encoder.blocks.5.attn.qkv.weight',\n", " 'sam_encoder.blocks.5.attn.qkv.bias',\n", " 'sam_encoder.blocks.5.attn.qkv.trainable_scale',\n", " 'sam_encoder.blocks.5.attn.qkv.trainable_shift',\n", " 'sam_encoder.blocks.5.attn.proj.weight',\n", " 'sam_encoder.blocks.5.attn.proj.bias',\n", " 'sam_encoder.blocks.5.attn.proj.trainable_scale',\n", " 'sam_encoder.blocks.5.attn.proj.trainable_shift',\n", " 'sam_encoder.blocks.5.norm2.weight',\n", " 'sam_encoder.blocks.5.norm2.bias',\n", " 'sam_encoder.blocks.5.mlp.lin1.weight',\n", " 'sam_encoder.blocks.5.mlp.lin1.bias',\n", " 'sam_encoder.blocks.5.mlp.lin1.trainable_scale',\n", " 'sam_encoder.blocks.5.mlp.lin1.trainable_shift',\n", " 'sam_encoder.blocks.5.mlp.lin2.weight',\n", " 'sam_encoder.blocks.5.mlp.lin2.bias',\n", " 'sam_encoder.blocks.5.mlp.lin2.trainable_scale',\n", " 'sam_encoder.blocks.5.mlp.lin2.trainable_shift',\n", " 'sam_encoder.blocks.6.norm1.weight',\n", " 'sam_encoder.blocks.6.norm1.bias',\n", " 'sam_encoder.blocks.6.attn.qkv.weight',\n", " 'sam_encoder.blocks.6.attn.qkv.bias',\n", " 'sam_encoder.blocks.6.attn.qkv.trainable_scale',\n", " 'sam_encoder.blocks.6.attn.qkv.trainable_shift',\n", " 'sam_encoder.blocks.6.attn.proj.weight',\n", " 'sam_encoder.blocks.6.attn.proj.bias',\n", " 'sam_encoder.blocks.6.attn.proj.trainable_scale',\n", " 'sam_encoder.blocks.6.attn.proj.trainable_shift',\n", " 'sam_encoder.blocks.6.norm2.weight',\n", " 'sam_encoder.blocks.6.norm2.bias',\n", " 'sam_encoder.blocks.6.mlp.lin1.weight',\n", " 'sam_encoder.blocks.6.mlp.lin1.bias',\n", " 'sam_encoder.blocks.6.mlp.lin1.trainable_scale',\n", " 'sam_encoder.blocks.6.mlp.lin1.trainable_shift',\n", " 'sam_encoder.blocks.6.mlp.lin2.weight',\n", " 'sam_encoder.blocks.6.mlp.lin2.bias',\n", " 'sam_encoder.blocks.6.mlp.lin2.trainable_scale',\n", " 'sam_encoder.blocks.6.mlp.lin2.trainable_shift',\n", " 'sam_encoder.blocks.7.norm1.weight',\n", " 'sam_encoder.blocks.7.norm1.bias',\n", " 'sam_encoder.blocks.7.attn.qkv.weight',\n", " 'sam_encoder.blocks.7.attn.qkv.bias',\n", " 'sam_encoder.blocks.7.attn.qkv.trainable_scale',\n", " 'sam_encoder.blocks.7.attn.qkv.trainable_shift',\n", " 'sam_encoder.blocks.7.attn.proj.weight',\n", " 'sam_encoder.blocks.7.attn.proj.bias',\n", " 'sam_encoder.blocks.7.attn.proj.trainable_scale',\n", " 'sam_encoder.blocks.7.attn.proj.trainable_shift',\n", " 'sam_encoder.blocks.7.norm2.weight',\n", " 'sam_encoder.blocks.7.norm2.bias',\n", " 'sam_encoder.blocks.7.mlp.lin1.weight',\n", " 'sam_encoder.blocks.7.mlp.lin1.bias',\n", " 'sam_encoder.blocks.7.mlp.lin1.trainable_scale',\n", " 'sam_encoder.blocks.7.mlp.lin1.trainable_shift',\n", " 'sam_encoder.blocks.7.mlp.lin2.weight',\n", " 'sam_encoder.blocks.7.mlp.lin2.bias',\n", " 'sam_encoder.blocks.7.mlp.lin2.trainable_scale',\n", " 'sam_encoder.blocks.7.mlp.lin2.trainable_shift',\n", " 'sam_encoder.blocks.8.norm1.weight',\n", " 'sam_encoder.blocks.8.norm1.bias',\n", " 'sam_encoder.blocks.8.attn.qkv.weight',\n", " 'sam_encoder.blocks.8.attn.qkv.bias',\n", " 'sam_encoder.blocks.8.attn.qkv.trainable_scale',\n", " 'sam_encoder.blocks.8.attn.qkv.trainable_shift',\n", " 'sam_encoder.blocks.8.attn.proj.weight',\n", " 'sam_encoder.blocks.8.attn.proj.bias',\n", " 'sam_encoder.blocks.8.attn.proj.trainable_scale',\n", " 'sam_encoder.blocks.8.attn.proj.trainable_shift',\n", " 'sam_encoder.blocks.8.norm2.weight',\n", " 'sam_encoder.blocks.8.norm2.bias',\n", " 'sam_encoder.blocks.8.mlp.lin1.weight',\n", " 'sam_encoder.blocks.8.mlp.lin1.bias',\n", " 'sam_encoder.blocks.8.mlp.lin1.trainable_scale',\n", " 'sam_encoder.blocks.8.mlp.lin1.trainable_shift',\n", " 'sam_encoder.blocks.8.mlp.lin2.weight',\n", " 'sam_encoder.blocks.8.mlp.lin2.bias',\n", " 'sam_encoder.blocks.8.mlp.lin2.trainable_scale',\n", " 'sam_encoder.blocks.8.mlp.lin2.trainable_shift',\n", " 'sam_encoder.blocks.9.norm1.weight',\n", " 'sam_encoder.blocks.9.norm1.bias',\n", " 'sam_encoder.blocks.9.attn.qkv.weight',\n", " 'sam_encoder.blocks.9.attn.qkv.bias',\n", " 'sam_encoder.blocks.9.attn.qkv.trainable_scale',\n", " 'sam_encoder.blocks.9.attn.qkv.trainable_shift',\n", " 'sam_encoder.blocks.9.attn.proj.weight',\n", " 'sam_encoder.blocks.9.attn.proj.bias',\n", " 'sam_encoder.blocks.9.attn.proj.trainable_scale',\n", " 'sam_encoder.blocks.9.attn.proj.trainable_shift',\n", " 'sam_encoder.blocks.9.norm2.weight',\n", " 'sam_encoder.blocks.9.norm2.bias',\n", " 'sam_encoder.blocks.9.mlp.lin1.weight',\n", " 'sam_encoder.blocks.9.mlp.lin1.bias',\n", " 'sam_encoder.blocks.9.mlp.lin1.trainable_scale',\n", " 'sam_encoder.blocks.9.mlp.lin1.trainable_shift',\n", " 'sam_encoder.blocks.9.mlp.lin2.weight',\n", " 'sam_encoder.blocks.9.mlp.lin2.bias',\n", " 'sam_encoder.blocks.9.mlp.lin2.trainable_scale',\n", " 'sam_encoder.blocks.9.mlp.lin2.trainable_shift',\n", " 'sam_encoder.blocks.10.norm1.weight',\n", " 'sam_encoder.blocks.10.norm1.bias',\n", " 'sam_encoder.blocks.10.attn.qkv.weight',\n", " 'sam_encoder.blocks.10.attn.qkv.bias',\n", " 'sam_encoder.blocks.10.attn.qkv.trainable_scale',\n", " 'sam_encoder.blocks.10.attn.qkv.trainable_shift',\n", " 'sam_encoder.blocks.10.attn.proj.weight',\n", " 'sam_encoder.blocks.10.attn.proj.bias',\n", " 'sam_encoder.blocks.10.attn.proj.trainable_scale',\n", " 'sam_encoder.blocks.10.attn.proj.trainable_shift',\n", " 'sam_encoder.blocks.10.norm2.weight',\n", " 'sam_encoder.blocks.10.norm2.bias',\n", " 'sam_encoder.blocks.10.mlp.lin1.weight',\n", " 'sam_encoder.blocks.10.mlp.lin1.bias',\n", " 'sam_encoder.blocks.10.mlp.lin1.trainable_scale',\n", " 'sam_encoder.blocks.10.mlp.lin1.trainable_shift',\n", " 'sam_encoder.blocks.10.mlp.lin2.weight',\n", " 'sam_encoder.blocks.10.mlp.lin2.bias',\n", " 'sam_encoder.blocks.10.mlp.lin2.trainable_scale',\n", " 'sam_encoder.blocks.10.mlp.lin2.trainable_shift',\n", " 'sam_encoder.blocks.11.norm1.weight',\n", " 'sam_encoder.blocks.11.norm1.bias',\n", " 'sam_encoder.blocks.11.attn.qkv.weight',\n", " 'sam_encoder.blocks.11.attn.qkv.bias',\n", " 'sam_encoder.blocks.11.attn.qkv.trainable_scale',\n", " 'sam_encoder.blocks.11.attn.qkv.trainable_shift',\n", " 'sam_encoder.blocks.11.attn.proj.weight',\n", " 'sam_encoder.blocks.11.attn.proj.bias',\n", " 'sam_encoder.blocks.11.attn.proj.trainable_scale',\n", " 'sam_encoder.blocks.11.attn.proj.trainable_shift',\n", " 'sam_encoder.blocks.11.norm2.weight',\n", " 'sam_encoder.blocks.11.norm2.bias',\n", " 'sam_encoder.blocks.11.mlp.lin1.weight',\n", " 'sam_encoder.blocks.11.mlp.lin1.bias',\n", " 'sam_encoder.blocks.11.mlp.lin1.trainable_scale',\n", " 'sam_encoder.blocks.11.mlp.lin1.trainable_shift',\n", " 'sam_encoder.blocks.11.mlp.lin2.weight',\n", " 'sam_encoder.blocks.11.mlp.lin2.bias',\n", " 'sam_encoder.blocks.11.mlp.lin2.trainable_scale',\n", " 'sam_encoder.blocks.11.mlp.lin2.trainable_shift',\n", " 'sam_encoder.neck.0.weight',\n", " 'sam_encoder.neck.0.trainable_scale',\n", " 'sam_encoder.neck.0.trainable_shift',\n", " 'sam_encoder.neck.1.weight',\n", " 'sam_encoder.neck.1.bias',\n", " 'sam_encoder.neck.2.weight',\n", " 'sam_encoder.neck.2.trainable_scale',\n", " 'sam_encoder.neck.2.trainable_shift',\n", " 'sam_encoder.neck.3.weight',\n", " 'sam_encoder.neck.3.bias',\n", " 'clip_model.positional_embedding',\n", " 'clip_model.text_projection',\n", " 'clip_model.logit_scale',\n", " 'clip_model.visual.class_embedding',\n", " 'clip_model.visual.positional_embedding',\n", " 'clip_model.visual.proj',\n", " 'clip_model.visual.conv1.weight',\n", " 'clip_model.visual.ln_pre.weight',\n", " 'clip_model.visual.ln_pre.bias',\n", " 'clip_model.visual.transformer.resblocks.0.attn.in_proj_weight',\n", " 'clip_model.visual.transformer.resblocks.0.attn.in_proj_bias',\n", " 'clip_model.visual.transformer.resblocks.0.attn.out_proj.weight',\n", " 'clip_model.visual.transformer.resblocks.0.attn.out_proj.bias',\n", " 'clip_model.visual.transformer.resblocks.0.ln_1.weight',\n", " 'clip_model.visual.transformer.resblocks.0.ln_1.bias',\n", " 'clip_model.visual.transformer.resblocks.0.mlp.c_fc.weight',\n", " 'clip_model.visual.transformer.resblocks.0.mlp.c_fc.bias',\n", " 'clip_model.visual.transformer.resblocks.0.mlp.c_proj.weight',\n", " 'clip_model.visual.transformer.resblocks.0.mlp.c_proj.bias',\n", " 'clip_model.visual.transformer.resblocks.0.ln_2.weight',\n", " 'clip_model.visual.transformer.resblocks.0.ln_2.bias',\n", " 'clip_model.visual.transformer.resblocks.1.attn.in_proj_weight',\n", " 'clip_model.visual.transformer.resblocks.1.attn.in_proj_bias',\n", " 'clip_model.visual.transformer.resblocks.1.attn.out_proj.weight',\n", " 'clip_model.visual.transformer.resblocks.1.attn.out_proj.bias',\n", " 'clip_model.visual.transformer.resblocks.1.ln_1.weight',\n", " 'clip_model.visual.transformer.resblocks.1.ln_1.bias',\n", " 'clip_model.visual.transformer.resblocks.1.mlp.c_fc.weight',\n", " 'clip_model.visual.transformer.resblocks.1.mlp.c_fc.bias',\n", " 'clip_model.visual.transformer.resblocks.1.mlp.c_proj.weight',\n", " 'clip_model.visual.transformer.resblocks.1.mlp.c_proj.bias',\n", " 'clip_model.visual.transformer.resblocks.1.ln_2.weight',\n", " 'clip_model.visual.transformer.resblocks.1.ln_2.bias',\n", " 'clip_model.visual.transformer.resblocks.2.attn.in_proj_weight',\n", " 'clip_model.visual.transformer.resblocks.2.attn.in_proj_bias',\n", " 'clip_model.visual.transformer.resblocks.2.attn.out_proj.weight',\n", " 'clip_model.visual.transformer.resblocks.2.attn.out_proj.bias',\n", " 'clip_model.visual.transformer.resblocks.2.ln_1.weight',\n", " 'clip_model.visual.transformer.resblocks.2.ln_1.bias',\n", " 'clip_model.visual.transformer.resblocks.2.mlp.c_fc.weight',\n", " 'clip_model.visual.transformer.resblocks.2.mlp.c_fc.bias',\n", " 'clip_model.visual.transformer.resblocks.2.mlp.c_proj.weight',\n", " 'clip_model.visual.transformer.resblocks.2.mlp.c_proj.bias',\n", " 'clip_model.visual.transformer.resblocks.2.ln_2.weight',\n", " 'clip_model.visual.transformer.resblocks.2.ln_2.bias',\n", " 'clip_model.visual.transformer.resblocks.3.attn.in_proj_weight',\n", " 'clip_model.visual.transformer.resblocks.3.attn.in_proj_bias',\n", " 'clip_model.visual.transformer.resblocks.3.attn.out_proj.weight',\n", " 'clip_model.visual.transformer.resblocks.3.attn.out_proj.bias',\n", " 'clip_model.visual.transformer.resblocks.3.ln_1.weight',\n", " 'clip_model.visual.transformer.resblocks.3.ln_1.bias',\n", " 'clip_model.visual.transformer.resblocks.3.mlp.c_fc.weight',\n", " 'clip_model.visual.transformer.resblocks.3.mlp.c_fc.bias',\n", " 'clip_model.visual.transformer.resblocks.3.mlp.c_proj.weight',\n", " 'clip_model.visual.transformer.resblocks.3.mlp.c_proj.bias',\n", " 'clip_model.visual.transformer.resblocks.3.ln_2.weight',\n", " 'clip_model.visual.transformer.resblocks.3.ln_2.bias',\n", " 'clip_model.visual.transformer.resblocks.4.attn.in_proj_weight',\n", " 'clip_model.visual.transformer.resblocks.4.attn.in_proj_bias',\n", " 'clip_model.visual.transformer.resblocks.4.attn.out_proj.weight',\n", " 'clip_model.visual.transformer.resblocks.4.attn.out_proj.bias',\n", " 'clip_model.visual.transformer.resblocks.4.ln_1.weight',\n", " 'clip_model.visual.transformer.resblocks.4.ln_1.bias',\n", " 'clip_model.visual.transformer.resblocks.4.mlp.c_fc.weight',\n", " 'clip_model.visual.transformer.resblocks.4.mlp.c_fc.bias',\n", " 'clip_model.visual.transformer.resblocks.4.mlp.c_proj.weight',\n", " 'clip_model.visual.transformer.resblocks.4.mlp.c_proj.bias',\n", " 'clip_model.visual.transformer.resblocks.4.ln_2.weight',\n", " 'clip_model.visual.transformer.resblocks.4.ln_2.bias',\n", " 'clip_model.visual.transformer.resblocks.5.attn.in_proj_weight',\n", " 'clip_model.visual.transformer.resblocks.5.attn.in_proj_bias',\n", " 'clip_model.visual.transformer.resblocks.5.attn.out_proj.weight',\n", " 'clip_model.visual.transformer.resblocks.5.attn.out_proj.bias',\n", " 'clip_model.visual.transformer.resblocks.5.ln_1.weight',\n", " 'clip_model.visual.transformer.resblocks.5.ln_1.bias',\n", " 'clip_model.visual.transformer.resblocks.5.mlp.c_fc.weight',\n", " 'clip_model.visual.transformer.resblocks.5.mlp.c_fc.bias',\n", " 'clip_model.visual.transformer.resblocks.5.mlp.c_proj.weight',\n", " 'clip_model.visual.transformer.resblocks.5.mlp.c_proj.bias',\n", " 'clip_model.visual.transformer.resblocks.5.ln_2.weight',\n", " 'clip_model.visual.transformer.resblocks.5.ln_2.bias',\n", " 'clip_model.visual.transformer.resblocks.6.attn.in_proj_weight',\n", " 'clip_model.visual.transformer.resblocks.6.attn.in_proj_bias',\n", " 'clip_model.visual.transformer.resblocks.6.attn.out_proj.weight',\n", " 'clip_model.visual.transformer.resblocks.6.attn.out_proj.bias',\n", " 'clip_model.visual.transformer.resblocks.6.ln_1.weight',\n", " 'clip_model.visual.transformer.resblocks.6.ln_1.bias',\n", " 'clip_model.visual.transformer.resblocks.6.mlp.c_fc.weight',\n", " 'clip_model.visual.transformer.resblocks.6.mlp.c_fc.bias',\n", " 'clip_model.visual.transformer.resblocks.6.mlp.c_proj.weight',\n", " 'clip_model.visual.transformer.resblocks.6.mlp.c_proj.bias',\n", " 'clip_model.visual.transformer.resblocks.6.ln_2.weight',\n", " 'clip_model.visual.transformer.resblocks.6.ln_2.bias',\n", " 'clip_model.visual.transformer.resblocks.7.attn.in_proj_weight',\n", " 'clip_model.visual.transformer.resblocks.7.attn.in_proj_bias',\n", " 'clip_model.visual.transformer.resblocks.7.attn.out_proj.weight',\n", " 'clip_model.visual.transformer.resblocks.7.attn.out_proj.bias',\n", " 'clip_model.visual.transformer.resblocks.7.ln_1.weight',\n", " 'clip_model.visual.transformer.resblocks.7.ln_1.bias',\n", " 'clip_model.visual.transformer.resblocks.7.mlp.c_fc.weight',\n", " 'clip_model.visual.transformer.resblocks.7.mlp.c_fc.bias',\n", " 'clip_model.visual.transformer.resblocks.7.mlp.c_proj.weight',\n", " 'clip_model.visual.transformer.resblocks.7.mlp.c_proj.bias',\n", " 'clip_model.visual.transformer.resblocks.7.ln_2.weight',\n", " 'clip_model.visual.transformer.resblocks.7.ln_2.bias',\n", " 'clip_model.visual.transformer.resblocks.8.attn.in_proj_weight',\n", " 'clip_model.visual.transformer.resblocks.8.attn.in_proj_bias',\n", " 'clip_model.visual.transformer.resblocks.8.attn.out_proj.weight',\n", " 'clip_model.visual.transformer.resblocks.8.attn.out_proj.bias',\n", " 'clip_model.visual.transformer.resblocks.8.ln_1.weight',\n", " 'clip_model.visual.transformer.resblocks.8.ln_1.bias',\n", " 'clip_model.visual.transformer.resblocks.8.mlp.c_fc.weight',\n", " 'clip_model.visual.transformer.resblocks.8.mlp.c_fc.bias',\n", " 'clip_model.visual.transformer.resblocks.8.mlp.c_proj.weight',\n", " 'clip_model.visual.transformer.resblocks.8.mlp.c_proj.bias',\n", " 'clip_model.visual.transformer.resblocks.8.ln_2.weight',\n", " 'clip_model.visual.transformer.resblocks.8.ln_2.bias',\n", " 'clip_model.visual.transformer.resblocks.9.attn.in_proj_weight',\n", " 'clip_model.visual.transformer.resblocks.9.attn.in_proj_bias',\n", " 'clip_model.visual.transformer.resblocks.9.attn.out_proj.weight',\n", " 'clip_model.visual.transformer.resblocks.9.attn.out_proj.bias',\n", " 'clip_model.visual.transformer.resblocks.9.ln_1.weight',\n", " 'clip_model.visual.transformer.resblocks.9.ln_1.bias',\n", " 'clip_model.visual.transformer.resblocks.9.mlp.c_fc.weight',\n", " 'clip_model.visual.transformer.resblocks.9.mlp.c_fc.bias',\n", " 'clip_model.visual.transformer.resblocks.9.mlp.c_proj.weight',\n", " 'clip_model.visual.transformer.resblocks.9.mlp.c_proj.bias',\n", " 'clip_model.visual.transformer.resblocks.9.ln_2.weight',\n", " 'clip_model.visual.transformer.resblocks.9.ln_2.bias',\n", " 'clip_model.visual.transformer.resblocks.10.attn.in_proj_weight',\n", " 'clip_model.visual.transformer.resblocks.10.attn.in_proj_bias',\n", " 'clip_model.visual.transformer.resblocks.10.attn.out_proj.weight',\n", " 'clip_model.visual.transformer.resblocks.10.attn.out_proj.bias',\n", " 'clip_model.visual.transformer.resblocks.10.ln_1.weight',\n", " 'clip_model.visual.transformer.resblocks.10.ln_1.bias',\n", " 'clip_model.visual.transformer.resblocks.10.mlp.c_fc.weight',\n", " 'clip_model.visual.transformer.resblocks.10.mlp.c_fc.bias',\n", " 'clip_model.visual.transformer.resblocks.10.mlp.c_proj.weight',\n", " 'clip_model.visual.transformer.resblocks.10.mlp.c_proj.bias',\n", " 'clip_model.visual.transformer.resblocks.10.ln_2.weight',\n", " 'clip_model.visual.transformer.resblocks.10.ln_2.bias',\n", " 'clip_model.visual.transformer.resblocks.11.attn.in_proj_weight',\n", " 'clip_model.visual.transformer.resblocks.11.attn.in_proj_bias',\n", " 'clip_model.visual.transformer.resblocks.11.attn.out_proj.weight',\n", " 'clip_model.visual.transformer.resblocks.11.attn.out_proj.bias',\n", " 'clip_model.visual.transformer.resblocks.11.ln_1.weight',\n", " 'clip_model.visual.transformer.resblocks.11.ln_1.bias',\n", " 'clip_model.visual.transformer.resblocks.11.mlp.c_fc.weight',\n", " 'clip_model.visual.transformer.resblocks.11.mlp.c_fc.bias',\n", " 'clip_model.visual.transformer.resblocks.11.mlp.c_proj.weight',\n", " 'clip_model.visual.transformer.resblocks.11.mlp.c_proj.bias',\n", " 'clip_model.visual.transformer.resblocks.11.ln_2.weight',\n", " 'clip_model.visual.transformer.resblocks.11.ln_2.bias',\n", " 'clip_model.visual.ln_post.weight',\n", " 'clip_model.visual.ln_post.bias',\n", " 'clip_model.transformer.resblocks.0.attn.in_proj_weight',\n", " 'clip_model.transformer.resblocks.0.attn.in_proj_bias',\n", " 'clip_model.transformer.resblocks.0.attn.out_proj.weight',\n", " 'clip_model.transformer.resblocks.0.attn.out_proj.bias',\n", " 'clip_model.transformer.resblocks.0.ln_1.weight',\n", " 'clip_model.transformer.resblocks.0.ln_1.bias',\n", " 'clip_model.transformer.resblocks.0.mlp.c_fc.weight',\n", " 'clip_model.transformer.resblocks.0.mlp.c_fc.bias',\n", " 'clip_model.transformer.resblocks.0.mlp.c_proj.weight',\n", " 'clip_model.transformer.resblocks.0.mlp.c_proj.bias',\n", " 'clip_model.transformer.resblocks.0.ln_2.weight',\n", " 'clip_model.transformer.resblocks.0.ln_2.bias',\n", " 'clip_model.transformer.resblocks.1.attn.in_proj_weight',\n", " 'clip_model.transformer.resblocks.1.attn.in_proj_bias',\n", " 'clip_model.transformer.resblocks.1.attn.out_proj.weight',\n", " 'clip_model.transformer.resblocks.1.attn.out_proj.bias',\n", " 'clip_model.transformer.resblocks.1.ln_1.weight',\n", " 'clip_model.transformer.resblocks.1.ln_1.bias',\n", " 'clip_model.transformer.resblocks.1.mlp.c_fc.weight',\n", " 'clip_model.transformer.resblocks.1.mlp.c_fc.bias',\n", " 'clip_model.transformer.resblocks.1.mlp.c_proj.weight',\n", " 'clip_model.transformer.resblocks.1.mlp.c_proj.bias',\n", " 'clip_model.transformer.resblocks.1.ln_2.weight',\n", " 'clip_model.transformer.resblocks.1.ln_2.bias',\n", " 'clip_model.transformer.resblocks.2.attn.in_proj_weight',\n", " 'clip_model.transformer.resblocks.2.attn.in_proj_bias',\n", " 'clip_model.transformer.resblocks.2.attn.out_proj.weight',\n", " 'clip_model.transformer.resblocks.2.attn.out_proj.bias',\n", " 'clip_model.transformer.resblocks.2.ln_1.weight',\n", " 'clip_model.transformer.resblocks.2.ln_1.bias',\n", " 'clip_model.transformer.resblocks.2.mlp.c_fc.weight',\n", " 'clip_model.transformer.resblocks.2.mlp.c_fc.bias',\n", " 'clip_model.transformer.resblocks.2.mlp.c_proj.weight',\n", " 'clip_model.transformer.resblocks.2.mlp.c_proj.bias',\n", " 'clip_model.transformer.resblocks.2.ln_2.weight',\n", " 'clip_model.transformer.resblocks.2.ln_2.bias',\n", " 'clip_model.transformer.resblocks.3.attn.in_proj_weight',\n", " 'clip_model.transformer.resblocks.3.attn.in_proj_bias',\n", " 'clip_model.transformer.resblocks.3.attn.out_proj.weight',\n", " 'clip_model.transformer.resblocks.3.attn.out_proj.bias',\n", " 'clip_model.transformer.resblocks.3.ln_1.weight',\n", " 'clip_model.transformer.resblocks.3.ln_1.bias',\n", " 'clip_model.transformer.resblocks.3.mlp.c_fc.weight',\n", " 'clip_model.transformer.resblocks.3.mlp.c_fc.bias',\n", " 'clip_model.transformer.resblocks.3.mlp.c_proj.weight',\n", " 'clip_model.transformer.resblocks.3.mlp.c_proj.bias',\n", " 'clip_model.transformer.resblocks.3.ln_2.weight',\n", " 'clip_model.transformer.resblocks.3.ln_2.bias',\n", " 'clip_model.transformer.resblocks.4.attn.in_proj_weight',\n", " 'clip_model.transformer.resblocks.4.attn.in_proj_bias',\n", " 'clip_model.transformer.resblocks.4.attn.out_proj.weight',\n", " 'clip_model.transformer.resblocks.4.attn.out_proj.bias',\n", " 'clip_model.transformer.resblocks.4.ln_1.weight',\n", " 'clip_model.transformer.resblocks.4.ln_1.bias',\n", " 'clip_model.transformer.resblocks.4.mlp.c_fc.weight',\n", " 'clip_model.transformer.resblocks.4.mlp.c_fc.bias',\n", " 'clip_model.transformer.resblocks.4.mlp.c_proj.weight',\n", " 'clip_model.transformer.resblocks.4.mlp.c_proj.bias',\n", " 'clip_model.transformer.resblocks.4.ln_2.weight',\n", " 'clip_model.transformer.resblocks.4.ln_2.bias',\n", " 'clip_model.transformer.resblocks.5.attn.in_proj_weight',\n", " 'clip_model.transformer.resblocks.5.attn.in_proj_bias',\n", " 'clip_model.transformer.resblocks.5.attn.out_proj.weight',\n", " 'clip_model.transformer.resblocks.5.attn.out_proj.bias',\n", " 'clip_model.transformer.resblocks.5.ln_1.weight',\n", " 'clip_model.transformer.resblocks.5.ln_1.bias',\n", " 'clip_model.transformer.resblocks.5.mlp.c_fc.weight',\n", " 'clip_model.transformer.resblocks.5.mlp.c_fc.bias',\n", " 'clip_model.transformer.resblocks.5.mlp.c_proj.weight',\n", " 'clip_model.transformer.resblocks.5.mlp.c_proj.bias',\n", " 'clip_model.transformer.resblocks.5.ln_2.weight',\n", " 'clip_model.transformer.resblocks.5.ln_2.bias',\n", " 'clip_model.transformer.resblocks.6.attn.in_proj_weight',\n", " 'clip_model.transformer.resblocks.6.attn.in_proj_bias',\n", " 'clip_model.transformer.resblocks.6.attn.out_proj.weight',\n", " 'clip_model.transformer.resblocks.6.attn.out_proj.bias',\n", " 'clip_model.transformer.resblocks.6.ln_1.weight',\n", " 'clip_model.transformer.resblocks.6.ln_1.bias',\n", " 'clip_model.transformer.resblocks.6.mlp.c_fc.weight',\n", " 'clip_model.transformer.resblocks.6.mlp.c_fc.bias',\n", " 'clip_model.transformer.resblocks.6.mlp.c_proj.weight',\n", " 'clip_model.transformer.resblocks.6.mlp.c_proj.bias',\n", " 'clip_model.transformer.resblocks.6.ln_2.weight',\n", " 'clip_model.transformer.resblocks.6.ln_2.bias',\n", " 'clip_model.transformer.resblocks.7.attn.in_proj_weight',\n", " 'clip_model.transformer.resblocks.7.attn.in_proj_bias',\n", " 'clip_model.transformer.resblocks.7.attn.out_proj.weight',\n", " 'clip_model.transformer.resblocks.7.attn.out_proj.bias',\n", " 'clip_model.transformer.resblocks.7.ln_1.weight',\n", " 'clip_model.transformer.resblocks.7.ln_1.bias',\n", " 'clip_model.transformer.resblocks.7.mlp.c_fc.weight',\n", " 'clip_model.transformer.resblocks.7.mlp.c_fc.bias',\n", " 'clip_model.transformer.resblocks.7.mlp.c_proj.weight',\n", " 'clip_model.transformer.resblocks.7.mlp.c_proj.bias',\n", " 'clip_model.transformer.resblocks.7.ln_2.weight',\n", " 'clip_model.transformer.resblocks.7.ln_2.bias',\n", " 'clip_model.transformer.resblocks.8.attn.in_proj_weight',\n", " 'clip_model.transformer.resblocks.8.attn.in_proj_bias',\n", " 'clip_model.transformer.resblocks.8.attn.out_proj.weight',\n", " 'clip_model.transformer.resblocks.8.attn.out_proj.bias',\n", " 'clip_model.transformer.resblocks.8.ln_1.weight',\n", " 'clip_model.transformer.resblocks.8.ln_1.bias',\n", " 'clip_model.transformer.resblocks.8.mlp.c_fc.weight',\n", " 'clip_model.transformer.resblocks.8.mlp.c_fc.bias',\n", " 'clip_model.transformer.resblocks.8.mlp.c_proj.weight',\n", " 'clip_model.transformer.resblocks.8.mlp.c_proj.bias',\n", " 'clip_model.transformer.resblocks.8.ln_2.weight',\n", " 'clip_model.transformer.resblocks.8.ln_2.bias',\n", " 'clip_model.transformer.resblocks.9.attn.in_proj_weight',\n", " 'clip_model.transformer.resblocks.9.attn.in_proj_bias',\n", " 'clip_model.transformer.resblocks.9.attn.out_proj.weight',\n", " 'clip_model.transformer.resblocks.9.attn.out_proj.bias',\n", " 'clip_model.transformer.resblocks.9.ln_1.weight',\n", " 'clip_model.transformer.resblocks.9.ln_1.bias',\n", " 'clip_model.transformer.resblocks.9.mlp.c_fc.weight',\n", " 'clip_model.transformer.resblocks.9.mlp.c_fc.bias',\n", " 'clip_model.transformer.resblocks.9.mlp.c_proj.weight',\n", " 'clip_model.transformer.resblocks.9.mlp.c_proj.bias',\n", " 'clip_model.transformer.resblocks.9.ln_2.weight',\n", " 'clip_model.transformer.resblocks.9.ln_2.bias',\n", " 'clip_model.transformer.resblocks.10.attn.in_proj_weight',\n", " 'clip_model.transformer.resblocks.10.attn.in_proj_bias',\n", " 'clip_model.transformer.resblocks.10.attn.out_proj.weight',\n", " 'clip_model.transformer.resblocks.10.attn.out_proj.bias',\n", " 'clip_model.transformer.resblocks.10.ln_1.weight',\n", " 'clip_model.transformer.resblocks.10.ln_1.bias',\n", " 'clip_model.transformer.resblocks.10.mlp.c_fc.weight',\n", " 'clip_model.transformer.resblocks.10.mlp.c_fc.bias',\n", " 'clip_model.transformer.resblocks.10.mlp.c_proj.weight',\n", " 'clip_model.transformer.resblocks.10.mlp.c_proj.bias',\n", " 'clip_model.transformer.resblocks.10.ln_2.weight',\n", " 'clip_model.transformer.resblocks.10.ln_2.bias',\n", " 'clip_model.transformer.resblocks.11.attn.in_proj_weight',\n", " 'clip_model.transformer.resblocks.11.attn.in_proj_bias',\n", " 'clip_model.transformer.resblocks.11.attn.out_proj.weight',\n", " 'clip_model.transformer.resblocks.11.attn.out_proj.bias',\n", " 'clip_model.transformer.resblocks.11.ln_1.weight',\n", " 'clip_model.transformer.resblocks.11.ln_1.bias',\n", " 'clip_model.transformer.resblocks.11.mlp.c_fc.weight',\n", " 'clip_model.transformer.resblocks.11.mlp.c_fc.bias',\n", " 'clip_model.transformer.resblocks.11.mlp.c_proj.weight',\n", " 'clip_model.transformer.resblocks.11.mlp.c_proj.bias',\n", " 'clip_model.transformer.resblocks.11.ln_2.weight',\n", " 'clip_model.transformer.resblocks.11.ln_2.bias',\n", " 'clip_model.token_embedding.weight',\n", " 'clip_model.ln_final.weight',\n", " 'clip_model.ln_final.bias',\n", " 'prompt_encoder.pe_layer.positional_encoding_gaussian_matrix',\n", " 'prompt_encoder.point_embeddings.0.weight',\n", " 'prompt_encoder.point_embeddings.1.weight',\n", " 'prompt_encoder.point_embeddings.2.weight',\n", " 'prompt_encoder.point_embeddings.3.weight',\n", " 'prompt_encoder.not_a_point_embed.weight',\n", " 'prompt_encoder.mask_downscaling.0.weight',\n", " 'prompt_encoder.mask_downscaling.0.bias',\n", " 'prompt_encoder.mask_downscaling.1.weight',\n", " 'prompt_encoder.mask_downscaling.1.bias',\n", " 'prompt_encoder.mask_downscaling.3.weight',\n", " 'prompt_encoder.mask_downscaling.3.bias',\n", " 'prompt_encoder.mask_downscaling.4.weight',\n", " 'prompt_encoder.mask_downscaling.4.bias',\n", " 'prompt_encoder.mask_downscaling.6.weight',\n", " 'prompt_encoder.mask_downscaling.6.bias',\n", " 'prompt_encoder.no_mask_embed.weight',\n", " 'mask_decoder.transformer.layers.0.self_attn.q_proj.weight',\n", " 'mask_decoder.transformer.layers.0.self_attn.q_proj.bias',\n", " 'mask_decoder.transformer.layers.0.self_attn.k_proj.weight',\n", " 'mask_decoder.transformer.layers.0.self_attn.k_proj.bias',\n", " 'mask_decoder.transformer.layers.0.self_attn.v_proj.weight',\n", " 'mask_decoder.transformer.layers.0.self_attn.v_proj.bias',\n", " 'mask_decoder.transformer.layers.0.self_attn.out_proj.weight',\n", " 'mask_decoder.transformer.layers.0.self_attn.out_proj.bias',\n", " 'mask_decoder.transformer.layers.0.norm1.weight',\n", " 'mask_decoder.transformer.layers.0.norm1.bias',\n", " 'mask_decoder.transformer.layers.0.cross_attn_token_to_image.q_proj.weight',\n", " 'mask_decoder.transformer.layers.0.cross_attn_token_to_image.q_proj.bias',\n", " 'mask_decoder.transformer.layers.0.cross_attn_token_to_image.k_proj.weight',\n", " 'mask_decoder.transformer.layers.0.cross_attn_token_to_image.k_proj.bias',\n", " 'mask_decoder.transformer.layers.0.cross_attn_token_to_image.v_proj.weight',\n", " 'mask_decoder.transformer.layers.0.cross_attn_token_to_image.v_proj.bias',\n", " 'mask_decoder.transformer.layers.0.cross_attn_token_to_image.out_proj.weight',\n", " 'mask_decoder.transformer.layers.0.cross_attn_token_to_image.out_proj.bias',\n", " 'mask_decoder.transformer.layers.0.norm2.weight',\n", " 'mask_decoder.transformer.layers.0.norm2.bias',\n", " 'mask_decoder.transformer.layers.0.mlp.lin1.weight',\n", " 'mask_decoder.transformer.layers.0.mlp.lin1.bias',\n", " 'mask_decoder.transformer.layers.0.mlp.lin1.trainable_scale',\n", " 'mask_decoder.transformer.layers.0.mlp.lin1.trainable_shift',\n", " 'mask_decoder.transformer.layers.0.mlp.lin2.weight',\n", " 'mask_decoder.transformer.layers.0.mlp.lin2.bias',\n", " 'mask_decoder.transformer.layers.0.mlp.lin2.trainable_scale',\n", " 'mask_decoder.transformer.layers.0.mlp.lin2.trainable_shift',\n", " 'mask_decoder.transformer.layers.0.norm3.weight',\n", " 'mask_decoder.transformer.layers.0.norm3.bias',\n", " 'mask_decoder.transformer.layers.0.norm4.weight',\n", " 'mask_decoder.transformer.layers.0.norm4.bias',\n", " 'mask_decoder.transformer.layers.0.cross_attn_image_to_token.q_proj.weight',\n", " 'mask_decoder.transformer.layers.0.cross_attn_image_to_token.q_proj.bias',\n", " 'mask_decoder.transformer.layers.0.cross_attn_image_to_token.k_proj.weight',\n", " 'mask_decoder.transformer.layers.0.cross_attn_image_to_token.k_proj.bias',\n", " 'mask_decoder.transformer.layers.0.cross_attn_image_to_token.v_proj.weight',\n", " 'mask_decoder.transformer.layers.0.cross_attn_image_to_token.v_proj.bias',\n", " 'mask_decoder.transformer.layers.0.cross_attn_image_to_token.out_proj.weight',\n", " 'mask_decoder.transformer.layers.0.cross_attn_image_to_token.out_proj.bias',\n", " 'mask_decoder.transformer.layers.1.self_attn.q_proj.weight',\n", " 'mask_decoder.transformer.layers.1.self_attn.q_proj.bias',\n", " 'mask_decoder.transformer.layers.1.self_attn.k_proj.weight',\n", " 'mask_decoder.transformer.layers.1.self_attn.k_proj.bias',\n", " 'mask_decoder.transformer.layers.1.self_attn.v_proj.weight',\n", " 'mask_decoder.transformer.layers.1.self_attn.v_proj.bias',\n", " 'mask_decoder.transformer.layers.1.self_attn.out_proj.weight',\n", " 'mask_decoder.transformer.layers.1.self_attn.out_proj.bias',\n", " 'mask_decoder.transformer.layers.1.norm1.weight',\n", " 'mask_decoder.transformer.layers.1.norm1.bias',\n", " 'mask_decoder.transformer.layers.1.cross_attn_token_to_image.q_proj.weight',\n", " 'mask_decoder.transformer.layers.1.cross_attn_token_to_image.q_proj.bias',\n", " 'mask_decoder.transformer.layers.1.cross_attn_token_to_image.k_proj.weight',\n", " 'mask_decoder.transformer.layers.1.cross_attn_token_to_image.k_proj.bias',\n", " 'mask_decoder.transformer.layers.1.cross_attn_token_to_image.v_proj.weight',\n", " 'mask_decoder.transformer.layers.1.cross_attn_token_to_image.v_proj.bias',\n", " 'mask_decoder.transformer.layers.1.cross_attn_token_to_image.out_proj.weight',\n", " 'mask_decoder.transformer.layers.1.cross_attn_token_to_image.out_proj.bias',\n", " 'mask_decoder.transformer.layers.1.norm2.weight',\n", " 'mask_decoder.transformer.layers.1.norm2.bias',\n", " 'mask_decoder.transformer.layers.1.mlp.lin1.weight',\n", " 'mask_decoder.transformer.layers.1.mlp.lin1.bias',\n", " 'mask_decoder.transformer.layers.1.mlp.lin1.trainable_scale',\n", " 'mask_decoder.transformer.layers.1.mlp.lin1.trainable_shift',\n", " 'mask_decoder.transformer.layers.1.mlp.lin2.weight',\n", " 'mask_decoder.transformer.layers.1.mlp.lin2.bias',\n", " 'mask_decoder.transformer.layers.1.mlp.lin2.trainable_scale',\n", " 'mask_decoder.transformer.layers.1.mlp.lin2.trainable_shift',\n", " 'mask_decoder.transformer.layers.1.norm3.weight',\n", " 'mask_decoder.transformer.layers.1.norm3.bias',\n", " 'mask_decoder.transformer.layers.1.norm4.weight',\n", " 'mask_decoder.transformer.layers.1.norm4.bias',\n", " 'mask_decoder.transformer.layers.1.cross_attn_image_to_token.q_proj.weight',\n", " 'mask_decoder.transformer.layers.1.cross_attn_image_to_token.q_proj.bias',\n", " 'mask_decoder.transformer.layers.1.cross_attn_image_to_token.k_proj.weight',\n", " 'mask_decoder.transformer.layers.1.cross_attn_image_to_token.k_proj.bias',\n", " 'mask_decoder.transformer.layers.1.cross_attn_image_to_token.v_proj.weight',\n", " 'mask_decoder.transformer.layers.1.cross_attn_image_to_token.v_proj.bias',\n", " 'mask_decoder.transformer.layers.1.cross_attn_image_to_token.out_proj.weight',\n", " 'mask_decoder.transformer.layers.1.cross_attn_image_to_token.out_proj.bias',\n", " 'mask_decoder.transformer.final_attn_token_to_image.q_proj.weight',\n", " 'mask_decoder.transformer.final_attn_token_to_image.q_proj.bias',\n", " 'mask_decoder.transformer.final_attn_token_to_image.k_proj.weight',\n", " 'mask_decoder.transformer.final_attn_token_to_image.k_proj.bias',\n", " 'mask_decoder.transformer.final_attn_token_to_image.v_proj.weight',\n", " 'mask_decoder.transformer.final_attn_token_to_image.v_proj.bias',\n", " 'mask_decoder.transformer.final_attn_token_to_image.out_proj.weight',\n", " 'mask_decoder.transformer.final_attn_token_to_image.out_proj.bias',\n", " 'mask_decoder.transformer.norm_final_attn.weight',\n", " 'mask_decoder.transformer.norm_final_attn.bias',\n", " 'mask_decoder.iou_token.weight',\n", " 'mask_decoder.mask_tokens.weight',\n", " 'mask_decoder.output_upscaling.0.weight',\n", " 'mask_decoder.output_upscaling.0.bias',\n", " 'mask_decoder.output_upscaling.1.weight',\n", " 'mask_decoder.output_upscaling.1.bias',\n", " 'mask_decoder.output_upscaling.3.weight',\n", " 'mask_decoder.output_upscaling.3.bias',\n", " 'mask_decoder.output_hypernetworks_mlps.0.layers.0.weight',\n", " 'mask_decoder.output_hypernetworks_mlps.0.layers.0.bias',\n", " 'mask_decoder.output_hypernetworks_mlps.0.layers.1.weight',\n", " 'mask_decoder.output_hypernetworks_mlps.0.layers.1.bias',\n", " 'mask_decoder.output_hypernetworks_mlps.0.layers.2.weight',\n", " 'mask_decoder.output_hypernetworks_mlps.0.layers.2.bias',\n", " 'mask_decoder.output_hypernetworks_mlps.1.layers.0.weight',\n", " 'mask_decoder.output_hypernetworks_mlps.1.layers.0.bias',\n", " 'mask_decoder.output_hypernetworks_mlps.1.layers.1.weight',\n", " 'mask_decoder.output_hypernetworks_mlps.1.layers.1.bias',\n", " 'mask_decoder.output_hypernetworks_mlps.1.layers.2.weight',\n", " 'mask_decoder.output_hypernetworks_mlps.1.layers.2.bias',\n", " 'mask_decoder.output_hypernetworks_mlps.2.layers.0.weight',\n", " 'mask_decoder.output_hypernetworks_mlps.2.layers.0.bias',\n", " 'mask_decoder.output_hypernetworks_mlps.2.layers.1.weight',\n", " 'mask_decoder.output_hypernetworks_mlps.2.layers.1.bias',\n", " 'mask_decoder.output_hypernetworks_mlps.2.layers.2.weight',\n", " 'mask_decoder.output_hypernetworks_mlps.2.layers.2.bias',\n", " 'mask_decoder.output_hypernetworks_mlps.3.layers.0.weight',\n", " 'mask_decoder.output_hypernetworks_mlps.3.layers.0.bias',\n", " 'mask_decoder.output_hypernetworks_mlps.3.layers.1.weight',\n", " 'mask_decoder.output_hypernetworks_mlps.3.layers.1.bias',\n", " 'mask_decoder.output_hypernetworks_mlps.3.layers.2.weight',\n", " 'mask_decoder.output_hypernetworks_mlps.3.layers.2.bias',\n", " 'mask_decoder.iou_prediction_head.layers.0.weight',\n", " 'mask_decoder.iou_prediction_head.layers.0.bias',\n", " 'mask_decoder.iou_prediction_head.layers.1.weight',\n", " 'mask_decoder.iou_prediction_head.layers.1.bias',\n", " 'mask_decoder.iou_prediction_head.layers.2.weight',\n", " 'mask_decoder.iou_prediction_head.layers.2.bias',\n", " 'Text_Embedding_Affine.0.weight',\n", " 'Text_Embedding_Affine.0.bias',\n", " 'Text_Embedding_Affine.2.weight',\n", " 'Text_Embedding_Affine.2.bias',\n", " 'Text_Embedding_Affine.2.running_mean',\n", " 'Text_Embedding_Affine.2.running_var',\n", " 'Text_Embedding_Affine.2.num_batches_tracked']" ] }, "execution_count": 5, "metadata": {}, "output_type": "execute_result" } ], "source": [ "keys" ] }, { "cell_type": "code", "execution_count": 6, "metadata": {}, "outputs": [], "source": [ "def basic_stats(key):\n", " print('min: ', model_dict[key].min())\n", " print('max: ', model_dict[key].max())\n", " print('mean: ', model_dict[key].mean())\n", " print('norm: ', np.linalg.norm(model_dict[key].numpy()))\n", "\n" ] }, { "cell_type": "code", "execution_count": 8, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "min: tensor(0.)\n", "max: tensor(0.)\n", "mean: tensor(0.)\n", "norm: 0.0\n" ] } ], "source": [ "basic_stats('mask_decoder.transformer.layers.0.mlp.lin1.trainable_shift')" ] }, { "cell_type": "code", "execution_count": 10, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "'a.b.c.d'" ] }, "execution_count": 10, "metadata": {}, "output_type": "execute_result" } ], "source": [ "a = \"a.b.c.d\"\n", "a.replace(\"c\",\"e\")\n", "a" ] }, { "cell_type": "code", "execution_count": 7, "metadata": {}, "outputs": [], "source": [ "import nibabel as nib\n", "import numpy as np\n", "import os" ] }, { "cell_type": "code", "execution_count": 18, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "['case0005_slice018.npz',\n", " 'case0027_slice043.npz',\n", " 'case0027_slice044.npz',\n", " 'case0027_slice045.npz',\n", " 'case0027_slice046.npz',\n", " 'case0027_slice047.npz',\n", " 'case0027_slice048.npz',\n", " 'case0027_slice049.npz',\n", " 'case0027_slice050.npz',\n", " 'case0027_slice051.npz',\n", " 'case0027_slice052.npz',\n", " 'case0027_slice053.npz',\n", " 'case0027_slice054.npz',\n", " 'case0027_slice055.npz',\n", " 'case0027_slice056.npz',\n", " 'case0027_slice057.npz',\n", " 'case0027_slice058.npz',\n", " 'case0027_slice059.npz',\n", " 'case0027_slice060.npz',\n", " 'case0027_slice061.npz',\n", " 'case0027_slice062.npz',\n", " 'case0027_slice063.npz',\n", " 'case0027_slice064.npz',\n", " 'case0027_slice065.npz',\n", " 'case0027_slice066.npz',\n", " 'case0027_slice067.npz',\n", " 'case0027_slice068.npz',\n", " 'case0010_slice080.npz',\n", " 'case0010_slice081.npz',\n", " 'case0010_slice082.npz',\n", " 'case0010_slice083.npz',\n", " 'case0010_slice084.npz']" ] }, "execution_count": 18, "metadata": {}, "output_type": "execute_result" } ], "source": [ "os.listdir('/media/ubuntu/New Volume/jay/BTCV/train_npz')[18:50]" ] }, { "cell_type": "code", "execution_count": 9, "metadata": {}, "outputs": [], "source": [ "# im = nib.load('/media/ubuntu/New Volume/jay/LiTS/images/1.nii')\n", "im = np.load('/media/ubuntu/New Volume/jay/BTCV/train_npz/case0005_slice000.npz')" ] }, { "cell_type": "code", "execution_count": 16, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "(512, 512)\n" ] } ], "source": [ "list(im.keys())\n", "print(im['image'].shape)" ] }, { "cell_type": "code", "execution_count": 1, "metadata": {}, "outputs": [], "source": [ "import torch" ] }, { "cell_type": "code", "execution_count": 2, "metadata": {}, "outputs": [ { "data": { "text/plain": [ "(tensor([0, 0, 1, 1]), tensor([0, 4, 3, 4]))" ] }, "execution_count": 2, "metadata": {}, "output_type": "execute_result" } ], "source": [ "a = torch.Tensor([\n", " [1,0,0,0,1],\n", " [0,0,0,1,1]\n", "])\n", "torch.where(a==1)" ] }, { "cell_type": "code", "execution_count": 6, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "tensor([[315., 581.],\n", " [339., 340.]])\n" ] } ], "source": [ "a = [[torch.tensor(315), torch.tensor(581)], [torch.tensor(339), torch.tensor(340)]]\n", "print(torch.Tensor(a))" ] } ], "metadata": { "kernelspec": { "display_name": "Python 3.8.16 ('dassl')", "language": "python", "name": "python3" }, "language_info": { "codemirror_mode": { "name": "ipython", "version": 3 }, "file_extension": ".py", "mimetype": "text/x-python", "name": "python", "nbconvert_exporter": "python", "pygments_lexer": "ipython3", "version": "3.8.16" }, "orig_nbformat": 4, "vscode": { "interpreter": { "hash": "5b0d24c0401191df5ff06ef3cb04a21077c1fd7ca08d243336ea8a8a1206ff02" } } }, "nbformat": 4, "nbformat_minor": 2 }