Release Symbol v1.1 repair refresh

Update the public LoRA to Symbol v1.1 from the v3.2 step880 repair-refresh checkpoint. Replace the v1.0 weight, update the model card and release metadata, and retain the existing examples.

CIVITAI-DESCRIPTION.md

@@ -1,4 +1,4 @@
-# Symbol v1.0 — Civitai 发布文案
+# Symbol v1.1 — Civitai 发布文案
 
 ## 简介
 
@@ -6,7 +6,7 @@ Symbol（符号）是一位蓝色到粉／珊瑚色渐变短发、蓝眼睛的
 她最初源于网友随手 AI 生成并分享的一张头像；在获得二次创作授权后，角色经过了
 重新设计、补全、精修、错误修复、多角度扩展和 LoRA 训练。
 
-本次对外发布版本为 **Symbol v1.0**，支持：
+本次对外发布版本为 **Symbol v1.1**，支持：
 
 - 原设服装；
 - 多种换装；

DATASET.md

@@ -11,7 +11,7 @@
 - 没有文件级或解码像素级精确重复；
 - Trigger `Symbol` 在每份 caption 中恰好出现一次；
 - 聚合 pair SHA-256：
-  `83552F2C58E534E532B6399A40FBDC1369391428C4CD3BBC242D5D67D78A1089`；
+  `E77583B6EFE24B25E6B637BFC51565B9A538F8D1369AECE8DA0BC848C203B93A`；
 - 安全标签分布包含 safe、sensitive 和 explicit 内容，角色按成年虚构角色处理。
 
 训练集覆盖原设服装、多种换装、正面/侧面/背面/三分之四角度、常见动作与表情、

NOTICE.md

@@ -1,6 +1,6 @@
 # Notices
 
-Symbol v1.0 is a modified derivative of the CircleStone Labs Anima model.
+Symbol v1.1 is a modified derivative of the CircleStone Labs Anima model.
 It is an independently produced community LoRA and is not an official CircleStone Labs,
 Comfy Org, or NVIDIA product, and has not been endorsed or validated by those organizations.
 

README.md

@@ -15,56 +15,35 @@ license_link: https://huggingface.co/circlestone-labs/Anima/blob/main/LICENSE.md
 
 <div align="center">
   <picture>
-    <img src="assets/symbol-logo.png" width="46%" alt="Symbol v1.0">
+    <img src="assets/symbol-logo.png" width="42%" alt="Symbol v1.1">
   </picture>
-  <h1>Symbol v1.0</h1>
-  <p><strong>Anima Character LoRA</strong></p>
-  <p>
-    <code>Trigger: Symbol</code>
-    ·
-    <code>LoRA strength: 1.0</code>
-    ·
-    <code>Safetensors</code>
-  </p>
-  <p>
-    <a href="https://huggingface.co/circlestone-labs/Anima">Anima Base v1.0</a>
-    ·
-    <a href="examples/PROMPTS.md">完整样图参数</a>
-    ·
-    <a href="PROVENANCE.md">角色来源</a>
-  </p>
 </div>
+<hr>
 
----
-
-`Symbol` 是一位发色独特、蓝眼睛的原创二次元女孩，她经常穿着那件蓝色夹克，看上去对一切都保持乐观✌。
+# Symbol v1.1 — Anima Character LoRA
 
-本仓库发布的是 **Symbol 单角色 LoRA**，适用于
+`Symbol` 是一位蓝色到粉／珊瑚色渐变短发、蓝眼睛的原创二次元女孩角色。
+本仓库发布的是 **Symbol v1.1 单角色 LoRA**，适用于
 [Anima Base v1.0](https://huggingface.co/circlestone-labs/Anima)。
 
 该角色源于一张由网友 AI 生成的头像。发布者在获得二次创作授权后，对角色进行了重新设计、补全、精修、错误修复与多角度扩展，形成了现在的角色设定。
 训练内容覆盖原设服装、多种换装，以及一套可通过提示词调用的单丸子头替代发型。
 更完整的来源说明见 [PROVENANCE.md](PROVENANCE.md)。
 
-> 理论上本模型可生成成人向内容；Symbol 被设定为成年虚构角色。
+v1.1 是一次小修复刷新：7 张训练图经过修复，101 份 caption 与训练配方保持不变。
+仓库示例图沿用 v1.0，本次未重新进行跨版本 A/B。
 
-## 快速开始
-
-| 项目 | 推荐值 |
-| --- | --- |
-| 模型文件 | [`symbol-v1.0.safetensors`](https://huggingface.co/sssmartmita/symbol-lora/resolve/main/symbol-v1.0.safetensors) |
-| Trigger | `Symbol`（请保持这个大小写形式） |
-| LoRA strength | `1.0` |
-| 基模 | [`circlestone-labs/Anima`](https://huggingface.co/circlestone-labs/Anima) / `anima-base-v1.0.safetensors` |
+> 理论上本模型可生成成人向内容；Symbol 被设定为成年虚构角色。
 
-<details>
-<summary><strong>文件校验信息</strong></summary>
+## 模型文件
 
-- LoRA SHA-256：`32B6CFDD1A7390BE1259798FA77E5B647F7E136DD7CA3F50ED798C9A0C97EE14`
+- 文件：`symbol-v1.1.safetensors`
+- SHA-256：`BF1AE9658AD66D3B01B5F640D024C454DA162A2844CCBA88EF318B437303986E`
+- Trigger：`Symbol`（请保持这个大小写形式）
+- 推荐 LoRA strength：`1.0`
+- 基模：`circlestone-labs/Anima` / `anima-base-v1.0.safetensors`
 - 基模 SHA-256：`BD43B7CFFE1ED1153D9C41E7BEB2F18CB1273EAFBAA3AF3EDD6A173DC90A006E`
 
-</details>
-
 ## 安装
 
 把 `.safetensors` 文件放入：
@@ -98,7 +77,7 @@ black shorts, fingerless gloves, loose socks, high tops
 hair bun, single hair bun, hair up, sidelocks
 ```
 
-经过本版本验证的基础设置：
+仓库示例所用的基础设置：
 
 - CFG：`4.5`；
 - Sampler：`dpmpp_2m_sde_gpu`；
@@ -112,8 +91,7 @@ hair bun, single hair bun, hair up, sidelocks
 `beta57` 和本仓库样图使用的 FLS Sampler 属于可选自定义节点，并不是 LoRA 的
 硬依赖。使用其他采样器时，请参照 Anima Base 的官方建议自行验证。
 
-<details>
-<summary><strong>起步推荐负向提示</strong></summary>
+起步推荐以下负向提示：
 
 ```text
 oldest, low quality, score_1, score_2, score_3, watermark, logo, text,
@@ -121,30 +99,22 @@ bad anatomy, bad hands, bad feet, extra fingers, missing fingers,
 distorted face, blurry, extra limbs, missing limbs, broken accessories
 ```
 
-</details>
-
 ## 示例
 
-### 原设与场景
-
-| 原设正面 | 三分之四视角 | 公交站场景 | 海边场景 |
-| --- | --- | --- | --- |
+| 原设正面                         | 三分之四视角                             | 公交站场景                                 | 海边场景                                  |
+| -------------------------------- | ---------------------------------------- | ------------------------------------------ | ----------------------------------------- |
 | ![](examples/original-front.png) | ![](examples/original-three-quarter.png) | ![](examples/original-outfit-bus-stop.png) | ![](examples/original-outfit-seaside.png) |
 
-### 发型变化
-
-| 替代丸子头 | 丸子头晨间场景 |
-| --- | --- |
-| ![](examples/alternative-bun.png) | ![](examples/alternative-bun-morning.png) |
-
-### 换装测试
+| 背面回望                        | 替代丸子头                        | 丸子头晨间场景                            |
+| ------------------------------- | --------------------------------- | ----------------------------------------- |
+| ![](examples/original-rear.png) | ![](examples/alternative-bun.png) | ![](examples/alternative-bun-morning.png) |
 
-| 婚纱换装 | 职业装换装 |
-| --- | --- |
-| ![](examples/wedding-dress-meadow.png) | ![](examples/office-outfit.png) |
+| 婚纱换装                               | 职业装换装                      | 休闲换装                          |
+| -------------------------------------- | ------------------------------- | --------------------------------- |
+| ![](examples/wedding-dress-meadow.png) | ![](examples/office-outfit.png) | ![](examples/casual-cardigan.png) |
 
-完整 prompt、seed 与生成参数见 [examples/PROMPTS.md](examples/PROMPTS.md)。当前展示的
-八张样图中，四张经过 RTX Video Super Resolution 2× 放大，四张为原生尺寸输出；
+完整 prompt、seed 与生成参数见 [examples/PROMPTS.md](examples/PROMPTS.md)。其中六张
+样图经过 RTX Video Super Resolution 2× 放大，四张新样图为原生尺寸输出；
 放大器不包含在 LoRA 中，也不是使用 LoRA 的必要条件。
 
 ## 局限性
@@ -179,8 +149,8 @@ distorted face, blurry, extra limbs, missing limbs, broken accessories
 - 880 optimizer updates；每 80 steps 保存；
 - Trainer：`sd-scripts v0.10.5`，commit
   `a1b48df430a3690aeb5c9b6e7b19025afe8fb518`；
-- 版本：Symbol v1.0；对应内部训练 checkpoint step880；
-- step880：840 个 finite tensor；记录的最终 `avr_loss=0.0535`。
+- 版本：Symbol v1.1；对应内部训练 checkpoint step880；
+- step880：840 个 finite tensor；记录的最终 `avr_loss=0.0544`。
 
 完整参数见 [training/TRAINING.md](training/TRAINING.md)。
 说明见 [DATASET.md](DATASET.md)。
@@ -196,7 +166,7 @@ License 及适用的 NVIDIA Open Model License 条款约束。请阅读本仓库
 ## English summary
 
 Symbol is an original adult fictional anime character with short blue-to-pink gradient hair
-and blue eyes. This repository contains the Symbol v1.0 character LoRA for Anima Base v1.0;
+and blue eyes. This repository contains the Symbol v1.1 character LoRA for Anima Base v1.0;
 the public weight corresponds to internal training checkpoint step880. Use the exact trigger
 `Symbol` at strength `1.0`. The model covers
 the default outfit, outfit changes, and an optional single-bun hairstyle. It may generate

RELEASE-MANIFEST.json

@@ -1,11 +1,11 @@
 {
   "schema_version": 1,
   "status": "PASS",
-  "release": "Symbol v1.0 open-weights repository",
+  "release": "Symbol v1.1 open-weights repository",
   "weight": {
-    "path": "symbol-v1.0.safetensors",
-    "sha256": "32b6cfdd1a7390be1259798fa77e5b647f7e136dd7ca3f50ed798c9a0c97ee14",
-    "bytes": 183617152,
+    "path": "symbol-v1.1.safetensors",
+    "sha256": "bf1ae9658ad66d3b01b5f640d024c454da162a2844ccba88ef318b437303986e",
+    "bytes": 183617272,
     "tensor_count": 840,
     "metadata_private_or_stale_hits": []
   },
@@ -16,12 +16,12 @@
     {
       "path": "CIVITAI-DESCRIPTION.md",
       "bytes": 1835,
-      "sha256": "56fa20186dde4153cd1d112fb5ce95084b559a1d5ad22ea72ecd8250d9bc9356"
+      "sha256": "838e1b02ab58970f22b235201dfd78d1be957c8d57233aba311df78ab0d0d802"
     },
     {
       "path": "DATASET.md",
       "bytes": 1332,
-      "sha256": "550e2212306e37f200df0b787c974c35b8d5ecb7a961b58e443fc2e21b76e506"
+      "sha256": "6932c1d04e294a2552571a8c8f75f9cb8f80af5e027a5c0f0b1e39054c52ac69"
     },
     {
       "path": "LICENSE.md",
@@ -31,27 +31,27 @@
     {
       "path": "NOTICE.md",
       "bytes": 805,
-      "sha256": "53c4e18645a4d8f2c270e3ca6ab364ba3c5f4c358bded02d245689b3d0672301"
+      "sha256": "b2df039c09c2db1cd44d3a78eb8aa2b2b254bc7a2c82f022d4db656d24e64ac4"
     },
     {
       "path": "PROVENANCE.md",
       "bytes": 1019,
       "sha256": "61231ff3d83385d4f349c908c3bc6af2365e1eddf8e78601b547e81e3525dc6f"
     },
-    {
-      "path": "README.md",
-      "bytes": 7946,
-      "sha256": "f085a7230d118de4dc044b53c4810c457e7ff71e514cda9cd00afccf511f9c4a"
+    {
+      "path": "README.md",
+      "bytes": 8042,
+      "sha256": "5773eda2841f6cfb094af4d83528a7913db347258761cd6c98ac1b165d9f87d8"
     },
     {
       "path": "assets/symbol-logo.png",
       "bytes": 416381,
       "sha256": "3393320c19d03be2a843f33d9863d68ff011646d6bd1e9d3f565910c5eb47946"
     },
-    {
-      "path": "examples/PROMPTS.md",
-      "bytes": 4397,
-      "sha256": "61b07a192451d050e580e9854af4b8ef0f83953b26b5d59a446a0b0b101e4851"
+    {
+      "path": "examples/PROMPTS.md",
+      "bytes": 5015,
+      "sha256": "67d4a7005b91c2616858525d17c86cac62229aff0bd7ccdb61c8e80fbf4f04aa"
     },
     {
       "path": "examples/alternative-bun-morning.png",
@@ -114,14 +114,14 @@
       "sha256": "ebde29ab1dee01fd9010038fa972bfb8f24441f53ff54e4efc18c009789fdeab"
     },
     {
-      "path": "symbol-v1.0.safetensors",
-      "bytes": 183617152,
-      "sha256": "32b6cfdd1a7390be1259798fa77e5b647f7e136dd7ca3f50ed798c9a0c97ee14"
+      "path": "symbol-v1.1.safetensors",
+      "bytes": 183617272,
+      "sha256": "bf1ae9658ad66d3b01b5f640d024c454da162a2844ccba88ef318b437303986e"
     },
     {
       "path": "training/TRAINING.md",
       "bytes": 1975,
-      "sha256": "c24ecb29e20023d387ecae993249a159064ca8fa90f5ba9891f310d61e710e6a"
+      "sha256": "1ec462d457442bb9b00848c9a41d7d74a1da1bfb7a17752e6b7910aa43ab2c47"
     },
     {
       "path": "training/dataset-config.example.toml",
@@ -131,12 +131,12 @@
     {
       "path": "training/training-config.public.json",
       "bytes": 1612,
-      "sha256": "359fb7415aab23ef3119b1da1d3b27cceda4eff36f9e6225bd1be9f568c8b6db"
+      "sha256": "1a9915ba57754b9affc56991e56345aa930430bc69f513f08548af915bbd4b01"
     },
-    {
-      "path": "SHA256SUMS.txt",
-      "bytes": 2239,
-      "sha256": "b31430a72adfc5bce7d5741d1c2514605fbc6b6d3f3cce392a5e29315b890aee"
+    {
+      "path": "SHA256SUMS.txt",
+      "bytes": 2241,
+      "sha256": "47a8a506390bc102504d5db660189b07ddb957851e1d6be8a1accca5a54aa596"
     }
   ]
 }

SHA256SUMS.txt

@@ -1,11 +1,11 @@
-56fa20186dde4153cd1d112fb5ce95084b559a1d5ad22ea72ecd8250d9bc9356  CIVITAI-DESCRIPTION.md
-550e2212306e37f200df0b787c974c35b8d5ecb7a961b58e443fc2e21b76e506  DATASET.md
+838e1b02ab58970f22b235201dfd78d1be957c8d57233aba311df78ab0d0d802  CIVITAI-DESCRIPTION.md
+6932c1d04e294a2552571a8c8f75f9cb8f80af5e027a5c0f0b1e39054c52ac69  DATASET.md
 ee956174133d7c2cdcf220440c7726187eaf4b50e8e48ee32194353a22164d15  LICENSE.md
-53c4e18645a4d8f2c270e3ca6ab364ba3c5f4c358bded02d245689b3d0672301  NOTICE.md
+b2df039c09c2db1cd44d3a78eb8aa2b2b254bc7a2c82f022d4db656d24e64ac4  NOTICE.md
 61231ff3d83385d4f349c908c3bc6af2365e1eddf8e78601b547e81e3525dc6f  PROVENANCE.md
-f085a7230d118de4dc044b53c4810c457e7ff71e514cda9cd00afccf511f9c4a  README.md
+5773eda2841f6cfb094af4d83528a7913db347258761cd6c98ac1b165d9f87d8  README.md
 3393320c19d03be2a843f33d9863d68ff011646d6bd1e9d3f565910c5eb47946  assets/symbol-logo.png
-61b07a192451d050e580e9854af4b8ef0f83953b26b5d59a446a0b0b101e4851  examples/PROMPTS.md
+67d4a7005b91c2616858525d17c86cac62229aff0bd7ccdb61c8e80fbf4f04aa  examples/PROMPTS.md
 966752ab5aab79e45ed2c53998e7835d104828455c377c0431b9434c333a9adb  examples/alternative-bun-morning.png
 29b99a0a98c47d3980b403edaea6fb38fbd1256fa2c21dadc34710bf1988d473  examples/alternative-bun.png
 f1afa7911deef16fd787a72b68e9972374860a782ef61a8026101a01e8dd3fcc  examples/casual-cardigan.png
@@ -18,7 +18,7 @@ b7ca4aa71b60564a3822d82f573eb7a0906d827c2dac397251eaf70be629c817  examples/origi
 bd03462b7bb9736c3c4315f84039e36b007b24f8cf282c74714d7da085b1b899  examples/wedding-dress-meadow.png
 4d2fb590aa9b30c47f2058bff17291df7fa2aa0c1bd775a20703da9bb267cfab  licenses/NVIDIA-Open-Model-License-2025-10-24.pdf
 ebde29ab1dee01fd9010038fa972bfb8f24441f53ff54e4efc18c009789fdeab  licenses/SOURCES.md
-32b6cfdd1a7390be1259798fa77e5b647f7e136dd7ca3f50ed798c9a0c97ee14  symbol-v1.0.safetensors
-c24ecb29e20023d387ecae993249a159064ca8fa90f5ba9891f310d61e710e6a  training/TRAINING.md
+bf1ae9658ad66d3b01b5f640d024c454da162a2844ccba88ef318b437303986e  symbol-v1.1.safetensors
+1ec462d457442bb9b00848c9a41d7d74a1da1bfb7a17752e6b7910aa43ab2c47  training/TRAINING.md
 eb39cb24ef0a3fd8eff3f9fa381df85274063a79fef5167aeb1848717b476070  training/dataset-config.example.toml
-359fb7415aab23ef3119b1da1d3b27cceda4eff36f9e6225bd1be9f568c8b6db  training/training-config.public.json
+1a9915ba57754b9affc56991e56345aa930430bc69f513f08548af915bbd4b01  training/training-config.public.json

examples/PROMPTS.md

@@ -13,9 +13,9 @@ bad anatomy, bad hands, bad feet, extra fingers, missing fingers,
 distorted face, blurry, extra limbs, missing limbs, broken accessories
 ```
 
-本文收录的八张样图中，四张较早选出的样图使用 `768×1024` latent、30 steps，
-随后通过 RTX Video Super Resolution 2× ULTRA 放大到 `1536×2048`；另外四张
-样图使用原生输出尺寸、60 steps，没有额外放大。每张图的具体参数如下。
+较早选出的六张样图使用 `768×1024` latent、30 steps，随后通过 RTX Video Super
+Resolution 2× ULTRA 放大到 `1536×2048`。四张新样图使用原生输出尺寸、60 steps，
+没有额外放大。每张图的具体参数如下。
 
 ## original-front.png
 
@@ -40,6 +40,17 @@ three quarter view, standing, hand on own hip, looking at viewer, smile,
 closed mouth, gradient background
 ```
 
+## original-rear.png
+
+- Seed：`2026080514`
+- 生成：`768×1024`，30 steps；发布图为 2× 放大后的 `1536×2048`
+
+```text
+newest, safe, 1girl, Symbol, solo, blue jacket, white shirt, black shirt,
+black shorts, fingerless gloves, loose socks, high tops, full body, from behind,
+looking back, standing, arms at sides, gradient background
+```
+
 ## original-outfit-bus-stop.png
 
 - Seed：`0`
@@ -115,3 +126,14 @@ collared shirt, black skirt, pencil skirt, black pantyhose, cowboy shot,
 three quarter view, standing, arms behind back, looking at viewer, serious,
 simple background
 ```
+
+## casual-cardigan.png
+
+- Seed：`2026080523`
+- 生成：`768×1024`，30 steps；发布图为 2× 放大后的 `1536×2048`
+
+```text
+newest, safe, 1girl, Symbol, solo, cardigan, white tank top, white pants,
+full body, from side, walking, looking to the side, smile, closed mouth,
+simple background
+```

symbol-v1.0.safetensors → symbol-v1.1.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:32b6cfdd1a7390be1259798fa77e5b647f7e136dd7ca3f50ed798c9a0c97ee14
-size 183617152
+oid sha256:bf1ae9658ad66d3b01b5f640d024c454da162a2844ccba88ef318b437303986e
+size 183617272

training/TRAINING.md

@@ -44,11 +44,11 @@
 - Seed：`20260725`
 - Max train steps：880 optimizer updates
 - Checkpoint interval：80 updates
-- Recorded final average loss：`0.0535`
+- Recorded final average loss：`0.0544`
 
 ## Mechanical audit
 
-- Public release version：Symbol v1.0
+- Public release version：Symbol v1.1
 - Internal source checkpoint：step880
 - 11 step checkpoints plus one final weight audited
 - Each weight contains 840 tensors

training/training-config.public.json

@@ -1,6 +1,6 @@
 {
   "schema_version": 1,
-  "model": "Symbol v1.0",
+  "model": "Symbol v1.1",
   "base_model": {
     "repository": "circlestone-labs/Anima",
     "file": "anima-base-v1.0.safetensors",
@@ -29,7 +29,7 @@
     "flip_aug": false,
     "color_aug": false,
     "random_crop": false,
-    "aggregate_pair_sha256": "83552f2c58e534e532b6399a40fbdc1369391428c4cd3bbc242d5d67d78a1089"
+    "aggregate_pair_sha256": "e77583b6efe24b25e6b637bfc51565b9a538f8d1369aece8da0bc848c203b93a"
   },
   "lora": {
     "network_module": "networks.lora_anima",
@@ -54,6 +54,6 @@
     "checkpoint": "step880",
     "tensor_count": 840,
     "all_finite": true,
-    "recorded_final_average_loss": 0.0535
+    "recorded_final_average_loss": 0.0544
   }
 }