Add files using upload-large-folder tool

README.md

@@ -9,28 +9,72 @@ tags:
 library_name: transformers
 ---
 
-# GLM-5 Abliterated (BF16)
+# GLM-5 Abliterated (BF16) - Delta Weights
 
-This is an abliterated (uncensored) version of [zai-org/GLM-5](https://huggingface.co/zai-org/GLM-5) (744B MoE, 40B active parameters).
+Abliterated (uncensored) version of [zai-org/GLM-5](https://huggingface.co/zai-org/GLM-5) (744B MoE, 40B active parameters).
 
-## What is abliteration?
+**This repo contains only the 43 modified weight shards.** To use, download the base model and replace these shards.
 
-Abliteration removes the "refusal direction" from the model weights using weight orthogonalization. This allows the model to respond to a wider range of prompts without safety refusals, while preserving general capability.
+## Quick Setup
+
+```bash
+# 1. Download base model
+huggingface-cli download zai-org/GLM-5 --local-dir ./GLM-5-abliterated
+
+# 2. Download and overwrite modified shards
+huggingface-cli download skyblanket/GLM-5-abliterated --local-dir ./GLM-5-abliterated --include "*.safetensors"
+```
+
+Or use the merge script:
+
+```python
+from huggingface_hub import snapshot_download
+import json, shutil, os
+
+# Download base model
+base = snapshot_download("zai-org/GLM-5", local_dir="./GLM-5-abliterated")
+
+# Download modified shards
+delta = snapshot_download("skyblanket/GLM-5-abliterated")
+
+# Overwrite modified shards
+with open(os.path.join(delta, "modified_shards.json")) as f:
+    modified = json.load(f)["modified_shards"]
+
+for shard in modified:
+    src = os.path.join(delta, shard)
+    dst = os.path.join(base, shard)
+    if os.path.exists(src):
+        shutil.copy2(src, dst)
+        print(f"Replaced {shard}")
+
+print("Done! Model ready at ./GLM-5-abliterated")
+```
 
 ## Method
 
-1. Computed refusal directions for all 78 layers using contrastive activation pairs (harmful vs harmless prompts)
-2. Applied weight orthogonalization to layers 15-54:
+Abliteration removes the "refusal direction" from model weights using weight orthogonalization.
+
+1. Computed refusal directions for all 78 layers using contrastive activation pairs
+2. Applied weight orthogonalization (W' = W - r_hat * r_hat^T * W) to layers 15-54:
    - `self_attn.o_proj.weight` (attention output projection)
    - `mlp.shared_experts.down_proj.weight` (shared expert down projection)
-3. Alpha = 1.0, 80 weight matrices modified total
+3. Alpha = 1.0, 80 weight matrices modified across 43 safetensor shards
 
 ## Details
 
 - **Base model**: zai-org/GLM-5 (744B MoE, BF16)
 - **Modified layers**: 15-54 (40 of 78 total layers)
 - **Weights modified**: 80 (o_proj + shared_experts.down_proj per layer)
-- **Precision**: BF16 (full precision, no quantization artifacts)
+- **Precision**: BF16 (full precision, no quantization)
+- **Delta size**: ~230GB (43 modified shards out of 282 total)
+
+## Files in this repo
+
+- 43 modified `.safetensors` shards (the delta weights)
+- `modified_shards.json` - list of which shards were modified
+- `model.safetensors.index.json` - full weight map (same as base model)
+- Config and tokenizer files
 
 ## Disclaimer
 

model-00038-of-00282.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0fbd47e6fa9892fd5ffce48854901730216742f6a1c4e28601c06bdeb1952802
+size 5364851048

model-00053-of-00282.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a507d1924041c433f6949ad66f205d53de2ed37b660bd51c081c80e091f151ee
+size 5359985336

model-00057-of-00282.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:095601aebf744035086ea92540b296f9e6ff7517f6b3df39cecf300018bd2f05
+size 5359985400

model-00072-of-00282.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4a60d797843424c4bcb6e5a0d77576587fe7373d4992dff1a90adda3e6ae89cd
+size 5359985456

model-00083-of-00282.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1b12ae1f80caaa47ba97d3540f994103dfd0f775836b4511ee1e6d3b4ed1b880
+size 5359985432

model-00094-of-00282.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6b325d0032b850c4c1439d14fe4ec9d019e6117416dfa63b06ea3900a401c979
+size 5359985432

model-00097-of-00282.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:51ec639c62fa47d7aea4e0758e926c28930c96d7cd83e3d9a22d0e2104a391c8
+size 5359985296

model-00108-of-00282.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e9e16b104b25fc951f17bd40a2191f67144a1e868f289079a1139426374a5ee8
+size 5292860584

model-00116-of-00282.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:60dfdde4615b4f22ce74a972f32e161f938fe01a021ad8e0b20715b83074c337
+size 5359985280

model-00127-of-00282.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:04035a92a076fbda2f9191805dfd103ad2f5e5d1b47d9194e007b4ded90cb739
+size 5359985392

model-00134-of-00282.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a89e506124692e027dca50a860489245b2531694ec92b6acc2cef0f05ce8ec0a
+size 5359985312

model-00142-of-00282.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:db112215df777196c557320c62a92cb2f128f195006f88a8b2b4b741c3831ae9
+size 5359985440

model-00153-of-00282.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6d865f4eb13100bed6db267cea9a98c596aef760454360abf5dc58cb9a203c3e
+size 5359985432

model-00164-of-00282.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ffd29ea0c99d0abcabb93fcc07edc88c7764a3d00c03ba212a0ec36e9fd86425
+size 5359985424

model-00167-of-00282.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1614c1a2501fac16c39faca5e6a9914c6c11bdf5ae2d30bf008569bf279a9186
+size 5192196784

model-00168-of-00282.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:308ea650827f953edc81959a1857d290ce75a877a6a63a8187abe1f817172217
+size 5351974184

model-00175-of-00282.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:16c449785adf44938ee4e7649b9e29dc937c64c0f72e55d8a1f6ac6c485fc51e
+size 5359985408

modified_shards.json

@@ -0,0 +1,48 @@
+{
+  "modified_shards": [
+    "model-00024-of-00282.safetensors",
+    "model-00027-of-00282.safetensors",
+    "model-00031-of-00282.safetensors",
+    "model-00035-of-00282.safetensors",
+    "model-00038-of-00282.safetensors",
+    "model-00042-of-00282.safetensors",
+    "model-00046-of-00282.safetensors",
+    "model-00049-of-00282.safetensors",
+    "model-00050-of-00282.safetensors",
+    "model-00053-of-00282.safetensors",
+    "model-00057-of-00282.safetensors",
+    "model-00061-of-00282.safetensors",
+    "model-00064-of-00282.safetensors",
+    "model-00068-of-00282.safetensors",
+    "model-00072-of-00282.safetensors",
+    "model-00075-of-00282.safetensors",
+    "model-00083-of-00282.safetensors",
+    "model-00086-of-00282.safetensors",
+    "model-00090-of-00282.safetensors",
+    "model-00094-of-00282.safetensors",
+    "model-00097-of-00282.safetensors",
+    "model-00101-of-00282.safetensors",
+    "model-00105-of-00282.safetensors",
+    "model-00108-of-00282.safetensors",
+    "model-00109-of-00282.safetensors",
+    "model-00112-of-00282.safetensors",
+    "model-00116-of-00282.safetensors",
+    "model-00123-of-00282.safetensors",
+    "model-00127-of-00282.safetensors",
+    "model-00131-of-00282.safetensors",
+    "model-00134-of-00282.safetensors",
+    "model-00138-of-00282.safetensors",
+    "model-00142-of-00282.safetensors",
+    "model-00145-of-00282.safetensors",
+    "model-00149-of-00282.safetensors",
+    "model-00153-of-00282.safetensors",
+    "model-00156-of-00282.safetensors",
+    "model-00164-of-00282.safetensors",
+    "model-00167-of-00282.safetensors",
+    "model-00168-of-00282.safetensors",
+    "model-00171-of-00282.safetensors",
+    "model-00175-of-00282.safetensors",
+    "model-00179-of-00282.safetensors"
+  ],
+  "base_model": "zai-org/GLM-5"
+}

tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:19e773648cb4e65de8660ea6365e10acca112d42a854923df93db4a6f333a82d
+size 20217442