Merge remote-tracking branch 'huggingface/main' into feat/lycoris-conv1x1-support

Author: nemo

.github/ISSUE_TEMPLATE/bug-report.yml

@@ -23,30 +23,14 @@ body:
 
         Please tag fewer than 3 people.
 
-        Library: @benjaminbossan @sayakpaul
+        Library: @benjaminbossan @githubnemo
+
+        diffusers integration: @benjaminbossan @sayakpaul
 
         Documentation: @stevhliu
 
       placeholder: "@Username ..."
 
-  - type: checkboxes
-    id: information-scripts-examples
-    attributes:
-      label: Information
-      description: 'The problem arises when using:'
-      options:
-        - label: "The official example scripts"
-        - label: "My own modified scripts"
-
-  - type: checkboxes
-    id: information-tasks
-    attributes:
-      label: Tasks
-      description: "The tasks I am working on are:"
-      options:
-        - label: "An officially supported task in the `examples` folder"
-        - label: "My own task or dataset (give details below)"
-
   - type: textarea
     id: reproduction
     validations:

.github/ISSUE_TEMPLATE/feature-request.yml

@@ -11,20 +11,11 @@ body:
       description: |
         A clear and concise description of the feature proposal. Please provide a link to the paper and code in case they exist.
 
-  - type: textarea
-    id: motivation
-    validations:
-      required: true
-    attributes:
-      label: Motivation
-      description: |
-        Please outline the motivation for the proposal. Is your feature request related to a problem? 
-
   - type: textarea
     id: contribution
     validations:
       required: true
     attributes:
       label: Your contribution
       description: |
-        Is there any way that you could help, e.g. by submitting a PR? 
+        Is there any way that you could help, e.g. by submitting a PR?

.github/workflows/deploy_method_comparison_app.yml

@@ -0,0 +1,41 @@
+name: Deploy "method_comparison" Gradio to Spaces
+
+on:
+  push:
+    branches: [ main ]
+    paths:
+      - "method_comparison/**"
+  workflow_dispatch:
+
+permissions: {}
+
+jobs:
+  deploy:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout code
+        uses: actions/checkout@v4
+        with:
+          fetch-depth: 0  # full history needed for subtree
+          persist-credentials: false
+
+      - name: Authenticate via ~/.netrc
+        env:
+          HF_TOKEN: ${{ secrets.PEFT_INTERNAL_REPO_READ_WRITE }}
+        run: |
+          # netrc needs BOTH login and password entries
+          printf "machine huggingface.co\nlogin hf\npassword ${HF_TOKEN}\n" >> ~/.netrc
+          chmod 600 ~/.netrc
+
+      - name: Deploy method_comparison app to HF Spaces
+        run: |
+          cd method_comparison
+          git init
+          # Spaces expect requirements.txt
+          mv requirements-app.txt requirements.txt
+          git config user.name "github-actions[bot]"
+          git config user.email "github-actions[bot]@users.noreply.github.com"
+          git remote add gradio-app https://huggingface.co/spaces/peft-internal-testing/PEFT-method-comparison
+          git add .
+          git commit -m "🚀 Deploy method comparison app from GH action"
+          git push -f gradio-app HEAD:main

.github/workflows/test-docker-build.yml

@@ -36,7 +36,7 @@ jobs:
     needs: get_changed_files
     name: Build Docker images on modified files
     runs-on: ubuntu-latest
-    if: ${{ needs.get_changed_files.outputs.matrix }} != ''
+    if: ${{ needs.get_changed_files.outputs.matrix != '[]' }}
     strategy:
       fail-fast: false
       matrix:

.github/workflows/tests.yml

@@ -39,7 +39,6 @@ jobs:
   tests:
     needs: check_code_quality
     strategy:
-      # TODO: remove 'fail-fast' line once timeout issue from the Hub is solved
       fail-fast: false
       matrix:
         python-version: ["3.9", "3.10", "3.11", "3.12"]
@@ -87,8 +86,27 @@ jobs:
         run: |
           pip install --force-reinstall -U "numpy<2.0.0"
       - name: Test with pytest
+        # MacOS tests are currently too flaky and will fail almost each time. Thus, continue (green checkmark) even if
+        # they fail, but add a notice so that the failure is not completely silent
+        continue-on-error: ${{ matrix.os == 'macos-13' }}
+        shell: bash
         run: |
+          set +e
           make test
+          status=$?
+          # Post a notice only if this is macOS AND tests failed
+          if [ "$status" -ne 0 ] && [ "${{ matrix.os }}" = "macos-13" ]; then
+            {
+              echo "## ⚠️ macOS tests failed"
+              echo ""
+              echo "- OS: ${{ matrix.os }}"
+              echo "- Python: ${{ matrix.python-version }}"
+              echo ""
+              echo "Check the logs from this step for details."
+            } >> "$GITHUB_STEP_SUMMARY"
+          fi
+          # Return the real status. On macOS this won't fail the job because of continue-on-error.
+          exit $status
       - name: Dump cache content and diff
         # This is just debug info so that we can monitor if the model cache diverges substantially
         # over time and what the diverging model is.

.github/zizmor.yml

@@ -13,3 +13,12 @@ rules:
       - build_docker_images.yml:103:9
       - build_docker_images.yml:136:9
       - build_docker_images.yml:169:9
+  unpinned-images:
+    ignore:
+      # We want to test these images with the latest version and we're not using them
+      # to deploy anything so we deem it safe to use those, even if they are unpinned.
+      - nightly-bnb.yml:30:7
+      - nightly-bnb.yml:155:7
+      - nightly.yml:27:7
+      - nightly.yml:77:7
+      - torch_compile_tests.yml:32:7

.pre-commit-config.yaml

@@ -1,6 +1,6 @@
 repos:
   - repo: https://github.com/astral-sh/ruff-pre-commit
-    rev: v0.9.2
+    rev: v0.12.8
     hooks:
       - id: ruff
         args:

Makefile

@@ -31,8 +31,12 @@ tests_core_multi_gpu:
 tests_core_single_gpu:
 	python -m pytest -m single_gpu_tests tests/test_common_gpu.py $(if $(IS_GITHUB_CI),--report-log "core_single_gpu.log",)
 
+# exclude gemma tests, as generation fails with torch.compile, these failures
+# trigger side effects that make other tests fail with 'RuntimeError: Offset
+# increment outside graph capture encountered unexpectedly.' 
+# TODO re-enable gemma once/if it is fixed
 tests_common_gpu:
-	python -m pytest tests/test_decoder_models.py $(if $(IS_GITHUB_CI),--report-log "common_decoder.log",)
+	python -m pytest tests/test_decoder_models.py -k "not gemma" $(if $(IS_GITHUB_CI),--report-log "common_decoder.log",)
 	python -m pytest tests/test_encoder_decoder_models.py $(if $(IS_GITHUB_CI),--report-log "common_encoder_decoder.log",)
 	python -m pytest tests/test_gptqmodel.py $(if $(IS_GITHUB_CI),--report-log "gptqmodel_gpu.log",)
 

README.md

@@ -39,38 +39,43 @@ pip install peft
 Prepare a model for training with a PEFT method such as LoRA by wrapping the base model and PEFT configuration with `get_peft_model`. For the bigscience/mt0-large model, you're only training 0.19% of the parameters!
 
 ```python
-from transformers import AutoModelForSeq2SeqLM
-from peft import get_peft_config, get_peft_model, LoraConfig, TaskType
-model_name_or_path = "bigscience/mt0-large"
-tokenizer_name_or_path = "bigscience/mt0-large"
+from transformers import AutoModelForCausalLM
+from peft import LoraConfig, TaskType, get_peft_model
 
+device = torch.accelerator.current_accelerator().type if hasattr(torch, "accelerator") else "cuda"
+model_id = "Qwen/Qwen2.5-3B-Instruct"
+model = AutoModelForCausalLM.from_pretrained(model_id, device_map=device)
 peft_config = LoraConfig(
-    task_type=TaskType.SEQ_2_SEQ_LM, inference_mode=False, r=8, lora_alpha=32, lora_dropout=0.1
+    r=16,
+    lora_alpha=32,
+    task_type=TaskType.CAUSAL_LM,
+    # target_modules=["q_proj", "v_proj", ...]  # optionally indicate target modules
 )
-
-model = AutoModelForSeq2SeqLM.from_pretrained(model_name_or_path)
 model = get_peft_model(model, peft_config)
 model.print_trainable_parameters()
-"trainable params: 2359296 || all params: 1231940608 || trainable%: 0.19151053100118282"
+# prints: trainable params: 3,686,400 || all params: 3,089,625,088 || trainable%: 0.1193
+
+# now perform training on your dataset, e.g. using transformers Trainer, then save the model
+model.save_pretrained("qwen2.5-3b-lora")
 ```
 
 To load a PEFT model for inference:
 
-```py
-from peft import AutoPeftModelForCausalLM
-from transformers import AutoTokenizer
-import torch
+```python
+from transformers import AutoModelForCausalLM, AutoTokenizer
+from peft import PeftModel
 
-model = AutoPeftModelForCausalLM.from_pretrained("ybelkada/opt-350m-lora").to("cuda")
-tokenizer = AutoTokenizer.from_pretrained("facebook/opt-350m")
+device = torch.accelerator.current_accelerator().type if hasattr(torch, "accelerator") else "cuda"
+model_id = "Qwen/Qwen2.5-3B-Instruct"
+tokenizer = AutoTokenizer.from_pretrained(model_id)
+model = AutoModelForCausalLM.from_pretrained(model_id, device_map=device)
+model = PeftModel.from_pretrained(model, "qwen2.5-3b-lora")
 
-model.eval()
 inputs = tokenizer("Preheat the oven to 350 degrees and place the cookie dough", return_tensors="pt")
+outputs = model.generate(**inputs.to(device), max_new_tokens=50)
+print(tokenizer.decode(outputs[0], skip_special_tokens=True))
 
-outputs = model.generate(input_ids=inputs["input_ids"].to("cuda"), max_new_tokens=50)
-print(tokenizer.batch_decode(outputs, skip_special_tokens=True)[0])
-
-"Preheat the oven to 350 degrees and place the cookie dough in the center of the oven. In a large bowl, combine the flour, baking powder, baking soda, salt, and cinnamon. In a separate bowl, combine the egg yolks, sugar, and vanilla."
+# prints something like: Preheat the oven to 350 degrees and place the cookie dough in a baking dish [...]
 ```
 
 ## Why you should use PEFT
@@ -124,6 +129,32 @@ The iterative diffusion process consumes a lot of memory which can make it diffi
 > [!TIP]
 > Take a look at the [examples/lora_dreambooth/train_dreambooth.py](examples/lora_dreambooth/train_dreambooth.py) training script to try training your own Stable Diffusion model with LoRA, and play around with the [smangrul/peft-lora-sd-dreambooth](https://huggingface.co/spaces/smangrul/peft-lora-sd-dreambooth) Space which is running on a T4 instance. Learn more about the PEFT integration in Diffusers in this [tutorial](https://huggingface.co/docs/peft/main/en/tutorial/peft_integrations#diffusers).
 
+### Transformers
+
+PEFT is directly integrated with [Transformers](https://huggingface.co/docs/transformers/main/en/peft). After loading a model, call `add_adapter` to add a new PEFT adapter to the model:
+
+```python
+from peft import LoraConfig
+model = ...  # transformers model
+peft_config = LoraConfig(...)
+model.add_adapter(lora_config, adapter_name="lora_1")
+```
+
+To load a trained PEFT adapter, call `load_adapter`:
+
+```python
+model = ...  # transformers model
+model.load_adapter(<path-to-adapter>, adapter_name="lora_1")
+```
+
+And to switch between different adapters, call `set_adapter`:
+
+```python
+model.set_adapter("lora_2")
+```
+
+The Transformers integration doesn't include all the functionalities offered in PEFT, such as methods for merging the adapter into the base model.
+
 ### Accelerate
 
 [Accelerate](https://huggingface.co/docs/accelerate/index) is a library for distributed training and inference on various training setups and hardware (GPUs, TPUs, Apple Silicon, etc.). PEFT models work with Accelerate out of the box, making it really convenient to train really large models or use them for inference on consumer hardware with limited resources.
@@ -150,9 +181,9 @@ To use 🤗 PEFT in your publication, please cite it by using the following BibT
 
 ```bibtex
 @Misc{peft,
-  title =        {PEFT: State-of-the-art Parameter-Efficient Fine-Tuning methods},
+  title =        {{PEFT}: State-of-the-art Parameter-Efficient Fine-Tuning methods},
   author =       {Sourab Mangrulkar and Sylvain Gugger and Lysandre Debut and Younes Belkada and Sayak Paul and Benjamin Bossan},
   howpublished = {\url{https://github.com/huggingface/peft}},
   year =         {2022}
 }
-```
+```

docs/source/_toctree.yml

@@ -126,6 +126,14 @@
       title: Trainable Tokens
     - local: package_reference/randlora
       title: RandLora
+    - local: package_reference/shira
+      title: SHiRA
+    - local: package_reference/c3a
+      title: C3A
+    - local: package_reference/miss
+      title: MiSS
+    - local: package_reference/road
+      title: RoAd
 
     title: Adapters
   - sections: