Upload folder using huggingface_hub

models/breeds_australian shepherd dog_uce_000/README.md

@@ -0,0 +1,121 @@
+---
+hyperparameters:
+  pretrained_model_name_or_path: CompVis/stable-diffusion-v1-4
+  erase_scale: 230.0
+  preserve_scale: 1.2
+  lamb: 0.1
+  save_entire_model: false
+  edit_concepts: an australian shepherd dog
+  guide_concepts: breeds
+  preserve_concepts: breeds
+  expand_prompts: false
+  output_dir: assets/models/breeds_australian shepherd dog_uce_000
+  device: cuda
+  compute_runtimes: true
+  hub_model_id: null
+model-index:
+- name: None
+  results:
+  - task:
+      type: text-to-image
+    dataset:
+      name: Forget set
+      type: inline-prompts
+    metrics:
+    - type: clip
+      value: 37.21034049987793
+      name: ForgetSet clip score of original model mean (~↑)
+    - type: clip
+      value: 0.9309128811809773
+      name: ForgetSet clip score of original model std (~↓)
+    - type: clip
+      value: 29.654611110687256
+      name: ForgetSet clip score of unlearned model mean (↓)
+    - type: clip
+      value: 0.994837151027172
+      name: ForgetSet clip score of unlearned model std (~↓)
+    - type: clip
+      value: 7.555729389190674
+      name: ForgetSet clip score difference between original and unlearned mean (↑)
+    - type: clip
+      value: 0.6035309497264993
+      name: ForgetSet clip score difference between original and unlearned std (~↓)
+    - type: clip
+      value: 36.85027599334717
+      name: RetainSet clip score of original model mean (~↑)
+    - type: clip
+      value: 0.7795019365176555
+      name: RetainSet clip score of original model std (~↓)
+    - type: clip
+      value: 37.475037574768066
+      name: RetainSet clip score of unlearned model mean (↑)
+    - type: clip
+      value: 1.5376670393896208
+      name: RetainSet clip score of unlearned model std (~↓)
+    - type: clip
+      value: -0.6247615814208984
+      name: RetainSet clip score difference between original and unlearned mean (↓)
+    - type: clip
+      value: 1.7536265861353804
+      name: RetainSet clip score difference between original and unlearned std (~↓)
+    - type: runtime
+      value: 4.235275089740754
+      name: Inference latency seconds mean (↓)
+    - type: runtime
+      value: 0.02736567477892634
+      name: Inference latency seconds std (~↓)
+  - task:
+      type: text-to-image
+    dataset:
+      name: '[''an australian shepherd dog''] (forget) and [''breeds''] (retain) sets'
+      type: forget-and-retain-together
+    metrics:
+    - type: runtime
+      value: 0.00038552284240722656
+      name: Runtime init seconds (~↓)
+    - type: runtime
+      value: 1.0013580322265625e-05
+      name: Runtime data loading seconds (~↓)
+    - type: runtime
+      value: 2.500298023223877
+      name: Runtime training seconds (↓)
+    - type: runtime
+      value: 101.3922221660614
+      name: Runtime eval seconds (~↓)
+---
+
+<!-- This model card has been generated automatically according to the information the training script had access to. You
+should probably proofread and complete it, then remove this comment. -->
+
+
+# LoRA text2image fine-tuning - None
+These are LoRA adaption weights for CompVis/stable-diffusion-v1-4.
+The weights were fine-tuned for forgetting an australian shepherd dog dataset, while retaining breeds.
+You can find some example images in the following.
+
+![img](images/Forget - An image of an australian shepherd dog.png)
+![img](images/Forget - Photograph of an australian shepherd dog; high definition.png)
+![img](images/Forget - An picture of an australian shepherd dog in the rain.png)
+![img](images/Forget - An picture of an australian shepherd dog running.png)
+![img](images/Retain - An image of an american water spaniel dog.png)
+![img](images/Retain - Photograph of an american water spaniel dog; high definition.png)
+![img](images/Retain - An picture of an american water spaniel dog in the rain.png)
+![img](images/Retain - An picture of an american water spaniel dog running.png)
+
+
+
+## Intended uses & limitations
+
+#### How to use
+
+```python
+# TODO: add an example code snippet for running this diffusion pipeline
+```
+
+#### Limitations and bias
+
+[TODO: provide examples of latent issues and potential remediations]
+
+## Training details
+
+[TODO: describe the data used to train the model]

models/breeds_australian shepherd dog_uce_000/uce_sd_weights.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9ecaa118cf30b5da9f63215fc06292172af47b87f2d47fe99a8c09f99538a9a1
+size 76681464