Upload folder using huggingface_hub

train_vae.py

@@ -28,12 +28,12 @@ from collections import deque
 
 # --------------------------- Параметры ---------------------------
 ds_path            = "/workspace/d23"
-project            = "vae"
+project            = "vae2"
 batch_size         = 1
-base_learning_rate = 2e-6
+base_learning_rate = 6e-6
 min_learning_rate  = 8e-8
-num_epochs         = 25
-sample_interval_share = 2
+num_epochs         = 20
+sample_interval_share = 10
 use_wandb          = True
 save_model         = True
 use_decay          = True
@@ -52,7 +52,7 @@ clip_grad_norm     = 1.0
 mixed_precision    = "no"
 gradient_accumulation_steps = 16
 generated_folder   = "samples"
-save_as            = "vae"
+save_as            = "vae2"
 num_workers        = 0
 device = None
 
@@ -171,10 +171,9 @@ else:
     if hasattr(core, "decoder"):
         if hasattr(core.decoder, "up_blocks") and len(core.decoder.up_blocks) > 0:
             # --- только 0-й up_block ---
-            # for name, p in core.decoder.up_blocks[0].named_parameters():
-            for name, p in core.decoder.up_blocks.named_parameters():
+            for name, p in core.decoder.up_blocks[0].named_parameters():
                 p.requires_grad = True
-                unfrozen_param_names.append(f"decoder.up_blocks.0.{name}")
+                unfrozen_param_names.append(f"{name}")
         else:
             print("[WARN] Decoder has no up_blocks — fallback to full decoder")
             for name, p in core.decoder.named_parameters():
@@ -232,12 +231,12 @@ class PngFolderDataset(Dataset):
             scale = resize_long_side / float(long)
             new_w = int(round(w * scale))
             new_h = int(round(h * scale))
-            return img.resize((new_w, new_h), Image.LANCZOS)
+            return img.resize((new_w, new_h), Image.BICUBIC)
 
 def random_crop(img, sz):
     w, h = img.size
     if w < sz or h < sz:
-        img = img.resize((max(sz, w), max(sz, h)), Image.LANCZOS)
+        img = img.resize((max(sz, w), max(sz, h)), Image.BICUBIC)
     x = random.randint(0, max(1, img.width - sz))
     y = random.randint(0, max(1, img.height - sz))
     return img.crop((x, y, x + sz, y + sz))

transfer_simplevae.ipynb

@@ -82,7 +82,7 @@
     "        \"DownEncoderBlock2D\",\n",
     "    ],\n",
     "    \"latent_channels\": 16,\n",
-    "    \"up_block_out_channels\": [128, 256, 512, 512, 512],  # +1 блок\n",
+    "    \"up_block_out_channels\": [64, 128, 256, 512, 512],  # +1 блок\n",
     "    \"up_block_types\": [\n",
     "        \"UpDecoderBlock2D\",\n",
     "        \"UpDecoderBlock2D\",\n",

transfer_simplevae2.ipynb

@@ -0,0 +1,216 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": 1,
+   "id": "c15deb04-94a0-4073-a174-adcd22af10b8",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "✅ Создана новая модель: <class 'diffusers.models.autoencoders.autoencoder_asym_kl.AsymmetricAutoencoderKL'>\n"
+     ]
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "The config attributes {'block_out_channels': [64, 128, 256, 512, 512], 'force_upcast': False} were passed to AsymmetricAutoencoderKL, but are not expected and will be ignored. Please verify your config.json configuration file.\n"
+     ]
+    },
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "\n",
+      "--- Перенос весов ---\n"
+     ]
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "100%|██████████| 248/248 [00:00<00:00, 114887.05it/s]"
+     ]
+    },
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "\n",
+      "✅ Перенос завершён.\n",
+      "Статистика:\n",
+      "  перенесено: 216\n",
+      "  дублировано: 26\n",
+      "  пропущено: 0\n"
+     ]
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "\n"
+     ]
+    }
+   ],
+   "source": [
+    "from diffusers.models import AsymmetricAutoencoderKL, AutoencoderKL\n",
+    "import torch\n",
+    "from tqdm import tqdm\n",
+    "\n",
+    "# ---- Конфиг новой модели ----\n",
+    "config = {\n",
+    "    \"_class_name\": \"AsymmetricAutoencoderKL\",\n",
+    "    \"act_fn\": \"silu\",\n",
+    "    \"in_channels\": 3,\n",
+    "    \"out_channels\": 3,\n",
+    "    \"scaling_factor\": 1.0,\n",
+    "    \"norm_num_groups\": 32,\n",
+    "    \"down_block_out_channels\": [128, 256, 512, 512],\n",
+    "    \"down_block_types\": [\n",
+    "        \"DownEncoderBlock2D\",\n",
+    "        \"DownEncoderBlock2D\",\n",
+    "        \"DownEncoderBlock2D\",\n",
+    "        \"DownEncoderBlock2D\",\n",
+    "    ],\n",
+    "    \"latent_channels\": 16,\n",
+    "    # Новый UpDecoderBlock добавлен в начало\n",
+    "    \"up_block_out_channels\": [64, 128, 256, 512, 512],\n",
+    "    \"up_block_types\": [\n",
+    "        \"UpDecoderBlock2D\",\n",
+    "        \"UpDecoderBlock2D\",\n",
+    "        \"UpDecoderBlock2D\",\n",
+    "        \"UpDecoderBlock2D\",\n",
+    "        \"UpDecoderBlock2D\",\n",
+    "    ],\n",
+    "}\n",
+    "\n",
+    "# ---- Создание пустой асимметричной модели ----\n",
+    "vae = AsymmetricAutoencoderKL(\n",
+    "    act_fn=config[\"act_fn\"],\n",
+    "    down_block_out_channels=config[\"down_block_out_channels\"],\n",
+    "    down_block_types=config[\"down_block_types\"],\n",
+    "    latent_channels=config[\"latent_channels\"],\n",
+    "    up_block_out_channels=config[\"up_block_out_channels\"],\n",
+    "    up_block_types=config[\"up_block_types\"],\n",
+    "    in_channels=config[\"in_channels\"],\n",
+    "    out_channels=config[\"out_channels\"],\n",
+    "    scaling_factor=config[\"scaling_factor\"],\n",
+    "    norm_num_groups=config[\"norm_num_groups\"],\n",
+    "    layers_per_down_block=2,\n",
+    "    layers_per_up_block=2,\n",
+    "    sample_size=1024\n",
+    ")\n",
+    "\n",
+    "vae.save_pretrained(\"asymmetric_vae_empty\")\n",
+    "print(\"✅ Создана новая модель:\", type(vae))\n",
+    "\n",
+    "# ---- Функция переноса весов старого VAE ----\n",
+    "def transfer_weights(old_path, new_path, save_path=\"asymmetric_vae\", device=\"cuda\", dtype=torch.float16):\n",
+    "    old_vae = AutoencoderKL.from_pretrained(old_path, subfolder=\"vae\").to(device, dtype=dtype)\n",
+    "    new_vae = AsymmetricAutoencoderKL.from_pretrained(new_path).to(device, dtype=dtype)\n",
+    "\n",
+    "    old_sd = old_vae.state_dict()\n",
+    "    new_sd = new_vae.state_dict()\n",
+    "\n",
+    "    transferred_keys = set()\n",
+    "    transfer_stats = {\"перенесено\": 0, \"дублировано\": 0, \"пропущено\": 0}\n",
+    "\n",
+    "    print(\"\\n--- Перенос весов ---\")\n",
+    "    for k, v in tqdm(old_sd.items()):\n",
+    "        # Копирование энкодера и прочих совпадающих ключей\n",
+    "        if (\"encoder\" in k) or (\"quant_conv\" in k) or (\"post_quant_conv\" in k):\n",
+    "            if k in new_sd and new_sd[k].shape == v.shape:\n",
+    "                new_sd[k] = v.clone()\n",
+    "                transferred_keys.add(k)\n",
+    "                transfer_stats[\"перенесено\"] += 1\n",
+    "            continue\n",
+    "\n",
+    "        # Копирование декодера (без сдвига)\n",
+    "        if \"decoder.up_blocks\" in k:\n",
+    "            if k in new_sd and new_sd[k].shape == v.shape:\n",
+    "                new_sd[k] = v.clone()\n",
+    "                transferred_keys.add(k)\n",
+    "                transfer_stats[\"перенесено\"] += 1\n",
+    "            continue\n",
+    "\n",
+    "    # Дублирование весов старого первого 512→512 блока в новый блок 64→128 для апскейла\n",
+    "    ref_prefix = \"decoder.up_blocks.1\"\n",
+    "    new_prefix = \"decoder.up_blocks.0\"\n",
+    "    for k, v in old_sd.items():\n",
+    "        if k.startswith(ref_prefix) and new_prefix + k[len(ref_prefix):] in new_sd:\n",
+    "            new_k = k.replace(ref_prefix, new_prefix)\n",
+    "            if new_sd[new_k].shape == v.shape:\n",
+    "                new_sd[new_k] = v.clone()\n",
+    "                transferred_keys.add(new_k)\n",
+    "                transfer_stats[\"дублировано\"] += 1\n",
+    "\n",
+    "    # Загрузка и сохранение\n",
+    "    new_vae.load_state_dict(new_sd, strict=False)\n",
+    "    new_vae.save_pretrained(save_path)\n",
+    "\n",
+    "    print(\"\\n✅ Перенос завершён.\")\n",
+    "    print(\"Статистика:\")\n",
+    "    for k, v in transfer_stats.items():\n",
+    "        print(f\"  {k}: {v}\")\n",
+    "\n",
+    "# ---- Запуск переноса ----\n",
+    "transfer_weights(\"AiArtLab/simplevae\", \"asymmetric_vae_empty\", save_path=\"vae2\")\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 8,
+   "id": "65653a65-e7c2-4b67-bc17-62c21cfd1db8",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "Collecting hf_transfer\n",
+      "  Downloading hf_transfer-0.1.9-cp38-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (1.7 kB)\n",
+      "Downloading hf_transfer-0.1.9-cp38-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (3.6 MB)\n",
+      "\u001b[2K   \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m3.6/3.6 MB\u001b[0m \u001b[31m34.5 MB/s\u001b[0m  \u001b[33m0:00:00\u001b[0m\n",
+      "\u001b[?25hInstalling collected packages: hf_transfer\n",
+      "Successfully installed hf_transfer-0.1.9\n"
+     ]
+    }
+   ],
+   "source": [
+    "!pip install hf_transfer"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "59fcafb9-6d89-49b4-8362-b4891f591687",
+   "metadata": {},
+   "outputs": [],
+   "source": []
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Python 3 (ipykernel)",
+   "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": "3.12.3"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}

vae/config.json

@@ -3,13 +3,6 @@
   "_diffusers_version": "0.35.1",
   "_name_or_path": "vae",
   "act_fn": "silu",
-  "block_out_channels": [
-    128,
-    256,
-    512,
-    512,
-    512
-  ],
   "down_block_out_channels": [
     128,
     256,
@@ -22,7 +15,6 @@
     "DownEncoderBlock2D",
     "DownEncoderBlock2D"
   ],
-  "force_upcast": false,
   "in_channels": 3,
   "latent_channels": 16,
   "layers_per_down_block": 2,

vae2/config.json

@@ -0,0 +1,48 @@
+{
+  "_class_name": "AsymmetricAutoencoderKL",
+  "_diffusers_version": "0.35.1",
+  "_name_or_path": "vae2",
+  "act_fn": "silu",
+  "block_out_channels": [
+    64,
+    128,
+    256,
+    512,
+    512
+  ],
+  "down_block_out_channels": [
+    128,
+    256,
+    512,
+    512
+  ],
+  "down_block_types": [
+    "DownEncoderBlock2D",
+    "DownEncoderBlock2D",
+    "DownEncoderBlock2D",
+    "DownEncoderBlock2D"
+  ],
+  "force_upcast": false,
+  "in_channels": 3,
+  "latent_channels": 16,
+  "layers_per_down_block": 2,
+  "layers_per_up_block": 2,
+  "norm_num_groups": 32,
+  "out_channels": 3,
+  "sample_size": 1024,
+  "scaling_factor": 1.0,
+  "up_block_out_channels": [
+    64,
+    128,
+    256,
+    512,
+    512
+  ],
+  "up_block_types": [
+    "UpDecoderBlock2D",
+    "UpDecoderBlock2D",
+    "UpDecoderBlock2D",
+    "UpDecoderBlock2D",
+    "UpDecoderBlock2D"
+  ]
+}

vae2/diffusion_pytorch_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:22865ffa9f97886f766a9ee38537a8e4cab2b9ad1672a2a3c600152235ce392b
+size 364540148