Upload folder using ModelScope SDK (batch 1/1)

.gitattributes

@@ -45,3 +45,7 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.wasm filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+
+split_files/text_encoders/qwen_3_4b.safetensors filter=lfs diff=lfs merge=lfs -text
+split_files/vae/ae.safetensors filter=lfs diff=lfs merge=lfs -text
+split_files/diffusion_models/z_image_turbo_bf16.safetensors filter=lfs diff=lfs merge=lfs -text

README.md

@@ -1,48 +1,7 @@
 ---
-license: Apache License 2.0
+tags:
-tags: []
+- diffusion-single-file
-
+- comfyui
-#model-type:
-##如 gpt、phi、llama、chatglm、baichuan 等
-#- gpt
-
-#domain:
-##如 nlp、cv、audio、multi-modal
-#- nlp
-
-#language:
-##语言代码列表 https://help.aliyun.com/document_detail/215387.html?spm=a2c4g.11186623.0.0.9f8d7467kni6Aa
-#- cn 
-
-#metrics:
-##如 CIDEr、Blue、ROUGE 等
-#- CIDEr
-
-#tags:
-##各种自定义，包括 pretrained、fine-tuned、instruction-tuned、RL-tuned 等训练方法和其他
-#- pretrained
-
-#tools:
-##如 vllm、fastchat、llamacpp、AdaSeq 等
-#- vllm
 ---
-### 当前模型的贡献者未提供更加详细的模型介绍。模型文件和权重，可浏览“模型文件”页面获取。
-#### 您可以通过如下git clone命令，或者ModelScope SDK来下载模型
 
-SDK下载
+Workflows: https://comfyanonymous.github.io/ComfyUI_examples/z_image/
-```bash
-#安装ModelScope
-pip install modelscope
-```
-```python
-#SDK模型下载
-from modelscope import snapshot_download
-model_dir = snapshot_download('Comfy-Org/z_image_turbo')
-```
-Git下载
-```
-#Git模型下载
-git clone https://www.modelscope.cn/Comfy-Org/z_image_turbo.git
-```
-
-<p style="color: lightgrey;">如果您是本模型的贡献者，我们邀请您根据<a href="https://modelscope.cn/docs/ModelScope%E6%A8%A1%E5%9E%8B%E6%8E%A5%E5%85%A5%E6%B5%81%E7%A8%8B%E6%A6%82%E8%A7%88" style="color: lightgrey; text-decoration: underline;">模型贡献文档</a>，及时完善模型卡片内容。</p>

configuration.json

@@ -0,0 +1 @@
+{"framework": "pytorch", "task": "text-to-image-synthesis", "allow_remote": true}

split_files/diffusion_models/z_image_turbo_bf16.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2407613050b809ffdff18a4ac99af83ea6b95443ecebdf80e064a79c825574a6
+size 12309866400

split_files/text_encoders/qwen_3_4b.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6c671498573ac2f7a5501502ccce8d2b08ea6ca2f661c458e708f36b36edfc5a
+size 8044982048

z_image_convert_original_to_comfy.py

@@ -0,0 +1,52 @@
+import safetensors.torch
+import torch
+import sys
+
+# Usage: python z_image_convert_original_to_comfy.py output.safetensors diffusion_model*.safetensors
+
+cast_to = None
+if "fp8_e4m3fn" in sys.argv[1]:
+    cast_to = torch.float8_e4m3fn
+elif "fp16" in sys.argv[1]:
+    cast_to = torch.float16
+elif "bf16" in sys.argv[1]:
+    cast_to = torch.bfloat16
+
+replace_keys = {"all_final_layer.2-1.": "final_layer.",
+                "all_x_embedder.2-1.": "x_embedder.",
+                ".attention.to_out.0.bias": ".attention.out.bias",
+                ".attention.norm_k.weight": ".attention.k_norm.weight",
+                ".attention.norm_q.weight": ".attention.q_norm.weight",
+                ".attention.to_out.0.weight": ".attention.out.weight"
+                }
+
+out_sd = {}
+for f in sys.argv[2:]:
+    sd = safetensors.torch.load_file(f)
+    cc = None
+    for k in sd:
+        w = sd[k]
+
+        if cast_to is not None:
+            w = w.to(cast_to)
+        k_out = k
+        if k_out.endswith(".attention.to_out.0.bias"):
+            continue
+        if k_out.endswith(".attention.to_k.weight"):
+            cc = [w]
+            continue
+        if k_out.endswith(".attention.to_q.weight"):
+            cc = [w] + cc
+            continue
+        if k_out.endswith(".attention.to_v.weight"):
+            cc = cc + [w]
+            w = torch.cat(cc, dim=0)
+            k_out = k_out.replace(".attention.to_v.weight", ".attention.qkv.weight")
+
+        for r, rr in replace_keys.items():
+            k_out = k_out.replace(r, rr)
+        out_sd[k_out] = w
+
+
+
+safetensors.torch.save_file(out_sd, sys.argv[1])