Add app

Dockerfile

@@ -0,0 +1,18 @@
+FROM python:3.12
+COPY --from=ghcr.io/astral-sh/uv:0.4.20 /uv /bin/uv
+
+# Set up a new user named "user" with user ID 1000
+RUN useradd -m -u 1000 user
+ENV PATH="/home/user/.local/bin:$PATH"
+ENV UV_SYSTEM_PYTHON=1
+
+WORKDIR /app
+
+COPY --chown=user ./requirements.txt requirements.txt
+RUN uv pip install -r requirements.txt
+
+COPY --chown=user . /app
+# Switch to the "user" user
+USER user
+
+CMD ["solara", "run", "app.py", "--host", "0.0.0.0", "--port", "7860"]

README.md

@@ -9,4 +9,3 @@ license: apache-2.0
 short_description: Forcing a language model not to use a vowel
 ---
 
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference

app.py

@@ -0,0 +1,144 @@
+import re
+from threading import Thread
+from typing import List
+import torch
+import solara
+from unicodedata import normalize
+from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer
+from transformers.generation import LogitsProcessor
+from typing_extensions import TypedDict
+
+device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
+
+model_id = "Qwen/Qwen3-0.6B"
+model = AutoModelForCausalLM.from_pretrained(
+    model_id, cache_dir="/big_storage/llms/hf_models/"
+).to(device)
+tokenizer = AutoTokenizer.from_pretrained(model_id)
+streamer = TextIteratorStreamer(tokenizer, skip_prompt=True)
+
+def response_generator(user_input, logits_processor=[], enable_thinking=False):
+    prompt = tokenizer.apply_chat_template(
+        [{"role": "user", "content": user_input}],
+        tokenize=False,
+        add_generation_prompt=True,
+        enable_thinking=enable_thinking
+    )
+    model_inputs = tokenizer(prompt, return_tensors="pt").to(device)
+    generation_kwargs = dict(
+        model_inputs,
+        streamer=streamer,
+        logits_processor=logits_processor,
+        max_new_tokens=4 * 1024,
+        do_sample=True,
+        temperature=0.7,
+        top_p=1.0,
+        top_k=50,
+    )
+    thread = Thread(target=model.generate, kwargs=generation_kwargs)
+    thread.start()
+    for chunk in streamer:
+        if tokenizer.eos_token in chunk or tokenizer.pad_token in chunk:
+            chunk = chunk.split(tokenizer.eos_token)[0]
+            chunk = chunk.split(tokenizer.pad_token)[0]
+        yield chunk
+    thread.join()
+
+list_of_vowels = ["a", "e", "i", "o", "u"]
+tokens_per_vowel = dict()
+for vowel in list_of_vowels:
+    tokens_containing_a_given_vowel = []
+    for token_id in range(tokenizer.vocab_size):
+        if (
+            vowel in tokenizer.decode(token_id)
+            or vowel.upper() in tokenizer.decode(token_id)
+            or normalize('NFC', f"{vowel}\u0300") in tokenizer.decode(token_id)
+            or normalize('NFC', f"{vowel}\u0301") in tokenizer.decode(token_id)
+            or normalize('NFC', f"{vowel}\u0302") in tokenizer.decode(token_id)
+            or normalize('NFC', f"{vowel}\u0303") in tokenizer.decode(token_id)
+            or normalize('NFC', f"{vowel}\u0308") in tokenizer.decode(token_id)
+        ):
+            tokens_containing_a_given_vowel.append(token_id)
+    tokens_per_vowel[vowel] = tokens_containing_a_given_vowel
+
+class GeorgePerecLogitsProcessor(LogitsProcessor):
+    def __init__(self, forbidden_tokens: List[int]):
+        self.forbidden_tokens = forbidden_tokens
+
+    def __call__(
+        self, input_ids: torch.LongTensor, scores: torch.FloatTensor
+    ) -> torch.FloatTensor:
+        scores_processed = scores.clone()
+        vocab_tensor = torch.arange(scores.shape[-1], device=scores.device)
+        forbidden_tokens = torch.tensor(self.forbidden_tokens, device=scores.device)
+        forbidden_tokens_mask = torch.isin(vocab_tensor, forbidden_tokens)
+        scores_processed = torch.where(forbidden_tokens_mask, -torch.inf, scores)
+
+        return scores_processed
+
+
+def add_chunk_to_ai_message(chunk: str):
+    messages.value = [
+        *messages.value[:-1],
+        {
+            "role": "assistant",
+            "content": messages.value[-1]["content"] + chunk,
+        },
+    ]
+
+class MessageDict(TypedDict):
+    role: str
+    content: str
+
+messages: solara.Reactive[List[MessageDict]] = solara.reactive([])
+enable_thinking_options = [True, False]
+enable_thinking = solara.reactive(False)
+vowels = ["a", "e", "i", "o", "u", "None"]
+vowel = solara.reactive("e")
+@solara.component
+def Page():
+    solara.lab.theme.themes.light.primary = "#0000ff"
+    solara.lab.theme.themes.light.secondary = "#0000ff"
+    solara.lab.theme.themes.dark.primary = "#0000ff"
+    solara.lab.theme.themes.dark.secondary = "#0000ff"
+    title = "Georges Perec"
+    with solara.Head():
+        solara.Title(f"{title}")
+    with solara.Column(align="center"):
+        with solara.Sidebar():
+            solara.Markdown("# G⎵org⎵s P⎵r⎵c")
+            solara.Markdown("## Forcing a language model not to use a vowel")
+            solara.Markdown("Select a forbidden vowel:")
+            solara.ToggleButtonsSingle(value=vowel, values=vowels)
+            solara.Markdown("Enable thinking:")
+            solara.ToggleButtonsSingle(value=enable_thinking, values=enable_thinking_options)
+            if vowel.value == "None":
+                logits_processor = []
+            else:
+                logits_processor = [
+                    GeorgePerecLogitsProcessor(
+                        forbidden_tokens=tokens_per_vowel[vowel.value],
+                    )
+                ]
+        user_message_count = len([m for m in messages.value if m["role"] == "user"])
+        def send(message):
+            messages.value = [*messages.value, {"role": "user", "content": message}]
+        def response(message):
+            messages.value = [*messages.value, {"role": "assistant", "content": ""}]
+            for chunk in response_generator(message, logits_processor=logits_processor, enable_thinking=enable_thinking.value):
+                add_chunk_to_ai_message(chunk)
+        def result():
+            if messages.value != []:
+                response(messages.value[-1]["content"])
+        result = solara.lab.use_task(result, dependencies=[user_message_count])
+        with solara.lab.ChatBox(style={"position": "fixed", "overflow-y": "scroll","scrollbar-width": "none", "-ms-overflow-style": "none", "top": "0", "bottom": "10rem", "width": "60%"}):
+            for item in messages.value:
+                with solara.lab.ChatMessage(
+                    user=item["role"] == "user",
+                    name="User" if item["role"] == "user" else "Assistant",
+                    avatar_background_color="#33cccc" if item["role"] == "assistant" else "#ff991f",
+                    border_radius="20px",
+                    style="background-color:darkgrey!important;" if solara.lab.theme.dark_effective else "background-color:lightgrey!important;"
+                ):
+                    solara.Markdown(item["content"])
+        solara.lab.ChatInput(send_callback=send, style={"position": "fixed", "bottom": "3rem", "width": "60%"})

requirements.in

@@ -0,0 +1,12 @@
+jupyterlab
+jupyterlab_execute_time
+black
+isort
+ruff
+jupyterlab_code_formatter
+jupytext
+ipywidgets
+torch
+transformers
+openai
+solara

requirements.txt

@@ -0,0 +1,505 @@
+# This file was autogenerated by uv via the following command:
+#    uv pip compile requirements.in
+annotated-types==0.7.0
+    # via pydantic
+anyio==4.9.0
+    # via
+    #   httpx
+    #   jupyter-server
+    #   openai
+    #   starlette
+    #   watchfiles
+argon2-cffi==25.1.0
+    # via jupyter-server
+argon2-cffi-bindings==21.2.0
+    # via argon2-cffi
+arrow==1.3.0
+    # via isoduration
+asttokens==3.0.0
+    # via stack-data
+async-lru==2.0.5
+    # via jupyterlab
+attrs==25.3.0
+    # via
+    #   jsonschema
+    #   referencing
+babel==2.17.0
+    # via jupyterlab-server
+beautifulsoup4==4.13.4
+    # via nbconvert
+black==25.1.0
+    # via -r requirements.in
+bleach==6.2.0
+    # via nbconvert
+cachetools==6.1.0
+    # via solara-ui
+certifi==2025.7.14
+    # via
+    #   httpcore
+    #   httpx
+    #   requests
+cffi==1.17.1
+    # via argon2-cffi-bindings
+charset-normalizer==3.4.2
+    # via requests
+click==8.2.1
+    # via
+    #   black
+    #   rich-click
+    #   solara-server
+    #   uvicorn
+comm==0.2.2
+    # via
+    #   ipykernel
+    #   ipywidgets
+debugpy==1.8.15
+    # via ipykernel
+decorator==5.2.1
+    # via ipython
+defusedxml==0.7.1
+    # via nbconvert
+distro==1.9.0
+    # via openai
+executing==2.2.0
+    # via stack-data
+fastjsonschema==2.21.1
+    # via nbformat
+filelock==3.18.0
+    # via
+    #   huggingface-hub
+    #   solara-server
+    #   torch
+    #   transformers
+fqdn==1.5.1
+    # via jsonschema
+fsspec==2025.7.0
+    # via
+    #   huggingface-hub
+    #   torch
+h11==0.16.0
+    # via
+    #   httpcore
+    #   uvicorn
+hf-xet==1.1.5
+    # via huggingface-hub
+httpcore==1.0.9
+    # via httpx
+httpx==0.28.1
+    # via
+    #   jupyterlab
+    #   openai
+huggingface-hub==0.33.4
+    # via
+    #   tokenizers
+    #   transformers
+humanize==4.12.3
+    # via solara-ui
+idna==3.10
+    # via
+    #   anyio
+    #   httpx
+    #   jsonschema
+    #   requests
+ipykernel==6.29.5
+    # via
+    #   jupyterlab
+    #   solara-server
+ipython==9.4.0
+    # via
+    #   ipykernel
+    #   ipywidgets
+ipython-pygments-lexers==1.1.1
+    # via ipython
+ipyvue==1.11.2
+    # via
+    #   ipyvuetify
+    #   solara-ui
+ipyvuetify==1.11.3
+    # via solara-ui
+ipywidgets==8.1.7
+    # via
+    #   -r requirements.in
+    #   ipyvue
+    #   reacton
+    #   solara-ui
+isoduration==20.11.0
+    # via jsonschema
+isort==6.0.1
+    # via -r requirements.in
+jedi==0.19.2
+    # via ipython
+jinja2==3.1.6
+    # via
+    #   jupyter-server
+    #   jupyterlab
+    #   jupyterlab-server
+    #   nbconvert
+    #   solara-server
+    #   torch
+jiter==0.10.0
+    # via openai
+json5==0.12.0
+    # via jupyterlab-server
+jsonpointer==3.0.0
+    # via jsonschema
+jsonschema==4.24.1
+    # via
+    #   jupyter-events
+    #   jupyterlab-server
+    #   nbformat
+jsonschema-specifications==2025.4.1
+    # via jsonschema
+jupyter-client==8.6.3
+    # via
+    #   ipykernel
+    #   jupyter-server
+    #   nbclient
+    #   solara-server
+jupyter-core==5.8.1
+    # via
+    #   ipykernel
+    #   jupyter-client
+    #   jupyter-server
+    #   jupyterlab
+    #   nbclient
+    #   nbconvert
+    #   nbformat
+jupyter-events==0.12.0
+    # via jupyter-server
+jupyter-lsp==2.2.5
+    # via jupyterlab
+jupyter-server==2.16.0
+    # via
+    #   jupyter-lsp
+    #   jupyterlab
+    #   jupyterlab-code-formatter
+    #   jupyterlab-server
+    #   notebook-shim
+jupyter-server-terminals==0.5.3
+    # via jupyter-server
+jupyterlab==4.4.4
+    # via
+    #   -r requirements.in
+    #   jupyterlab-execute-time
+jupyterlab-code-formatter==3.0.2
+    # via -r requirements.in
+jupyterlab-execute-time==3.2.0
+    # via -r requirements.in
+jupyterlab-pygments==0.3.0
+    # via nbconvert
+jupyterlab-server==2.27.3
+    # via jupyterlab
+jupyterlab-widgets==3.0.15
+    # via ipywidgets
+jupytext==1.17.2
+    # via -r requirements.in
+markdown==3.8.2
+    # via
+    #   pymdown-extensions
+    #   solara-ui
+markdown-it-py==3.0.0
+    # via
+    #   jupytext
+    #   mdit-py-plugins
+    #   rich
+markupsafe==3.0.2
+    # via
+    #   jinja2
+    #   nbconvert
+matplotlib-inline==0.1.7
+    # via
+    #   ipykernel
+    #   ipython
+mdit-py-plugins==0.4.2
+    # via jupytext
+mdurl==0.1.2
+    # via markdown-it-py
+mistune==3.1.3
+    # via nbconvert
+mpmath==1.3.0
+    # via sympy
+mypy-extensions==1.1.0
+    # via black
+nbclient==0.10.2
+    # via nbconvert
+nbconvert==7.16.6
+    # via jupyter-server
+nbformat==5.10.4
+    # via
+    #   jupyter-server
+    #   jupytext
+    #   nbclient
+    #   nbconvert
+    #   solara-server
+nest-asyncio==1.6.0
+    # via ipykernel
+networkx==3.5
+    # via torch
+notebook-shim==0.2.4
+    # via jupyterlab
+numpy==2.3.1
+    # via
+    #   solara-ui
+    #   transformers
+nvidia-cublas-cu12==12.6.4.1
+    # via
+    #   nvidia-cudnn-cu12
+    #   nvidia-cusolver-cu12
+    #   torch
+nvidia-cuda-cupti-cu12==12.6.80
+    # via torch
+nvidia-cuda-nvrtc-cu12==12.6.77
+    # via torch
+nvidia-cuda-runtime-cu12==12.6.77
+    # via torch
+nvidia-cudnn-cu12==9.5.1.17
+    # via torch
+nvidia-cufft-cu12==11.3.0.4
+    # via torch
+nvidia-cufile-cu12==1.11.1.6
+    # via torch
+nvidia-curand-cu12==10.3.7.77
+    # via torch
+nvidia-cusolver-cu12==11.7.1.2
+    # via torch
+nvidia-cusparse-cu12==12.5.4.2
+    # via
+    #   nvidia-cusolver-cu12
+    #   torch
+nvidia-cusparselt-cu12==0.6.3
+    # via torch
+nvidia-nccl-cu12==2.26.2
+    # via torch
+nvidia-nvjitlink-cu12==12.6.85
+    # via
+    #   nvidia-cufft-cu12
+    #   nvidia-cusolver-cu12
+    #   nvidia-cusparse-cu12
+    #   torch
+nvidia-nvtx-cu12==12.6.77
+    # via torch
+openai==1.97.0
+    # via -r requirements.in
+overrides==7.7.0
+    # via jupyter-server
+packaging==25.0
+    # via
+    #   black
+    #   huggingface-hub
+    #   ipykernel
+    #   jupyter-events
+    #   jupyter-server
+    #   jupyterlab
+    #   jupyterlab-code-formatter
+    #   jupyterlab-server
+    #   jupytext
+    #   nbconvert
+    #   transformers
+pandocfilters==1.5.1
+    # via nbconvert
+parso==0.8.4
+    # via jedi
+pathspec==0.12.1
+    # via black
+pexpect==4.9.0
+    # via ipython
+pillow==11.3.0
+    # via solara-ui
+platformdirs==4.3.8
+    # via
+    #   black
+    #   jupyter-core
+prometheus-client==0.22.1
+    # via jupyter-server
+prompt-toolkit==3.0.51
+    # via ipython
+psutil==7.0.0
+    # via ipykernel
+ptyprocess==0.7.0
+    # via
+    #   pexpect
+    #   terminado
+pure-eval==0.2.3
+    # via stack-data
+pycparser==2.22
+    # via cffi
+pydantic==2.11.7
+    # via openai
+pydantic-core==2.33.2
+    # via pydantic
+pygments==2.19.2
+    # via
+    #   ipython
+    #   ipython-pygments-lexers
+    #   nbconvert
+    #   rich
+    #   solara-ui
+pymdown-extensions==10.16
+    # via solara-ui
+python-dateutil==2.9.0.post0
+    # via
+    #   arrow
+    #   jupyter-client
+python-json-logger==3.3.0
+    # via jupyter-events
+pyyaml==6.0.2
+    # via
+    #   huggingface-hub
+    #   jupyter-events
+    #   jupytext
+    #   pymdown-extensions
+    #   transformers
+pyzmq==27.0.0
+    # via
+    #   ipykernel
+    #   jupyter-client
+    #   jupyter-server
+reacton==1.9.1
+    # via solara-ui
+referencing==0.36.2
+    # via
+    #   jsonschema
+    #   jsonschema-specifications
+    #   jupyter-events
+regex==2024.11.6
+    # via transformers
+requests==2.32.4
+    # via
+    #   huggingface-hub
+    #   jupyterlab-server
+    #   solara-ui
+    #   transformers
+rfc3339-validator==0.1.4
+    # via
+    #   jsonschema
+    #   jupyter-events
+rfc3986-validator==0.1.1
+    # via
+    #   jsonschema
+    #   jupyter-events
+rich==14.0.0
+    # via rich-click
+rich-click==1.8.9
+    # via solara-server
+rpds-py==0.26.0
+    # via
+    #   jsonschema
+    #   referencing
+ruff==0.12.4
+    # via -r requirements.in
+safetensors==0.5.3
+    # via transformers
+send2trash==1.8.3
+    # via jupyter-server
+setuptools==80.9.0
+    # via
+    #   jupyterlab
+    #   torch
+    #   triton
+six==1.17.0
+    # via
+    #   python-dateutil
+    #   rfc3339-validator
+sniffio==1.3.1
+    # via
+    #   anyio
+    #   openai
+solara==1.50.1
+    # via -r requirements.in
+solara-server==1.50.1
+    # via solara
+solara-ui==1.50.1
+    # via
+    #   solara
+    #   solara-server
+soupsieve==2.7
+    # via beautifulsoup4
+stack-data==0.6.3
+    # via ipython
+starlette==0.47.1
+    # via solara-server
+sympy==1.14.0
+    # via torch
+terminado==0.18.1
+    # via
+    #   jupyter-server
+    #   jupyter-server-terminals
+tinycss2==1.4.0
+    # via bleach
+tokenizers==0.21.2
+    # via transformers
+torch==2.7.1
+    # via -r requirements.in
+tornado==6.5.1
+    # via
+    #   ipykernel
+    #   jupyter-client
+    #   jupyter-server
+    #   jupyterlab
+    #   terminado
+tqdm==4.67.1
+    # via
+    #   huggingface-hub
+    #   openai
+    #   transformers
+traitlets==5.14.3
+    # via
+    #   comm
+    #   ipykernel
+    #   ipython
+    #   ipywidgets
+    #   jupyter-client
+    #   jupyter-core
+    #   jupyter-events
+    #   jupyter-server
+    #   jupyterlab
+    #   matplotlib-inline
+    #   nbclient
+    #   nbconvert
+    #   nbformat
+transformers==4.53.2
+    # via -r requirements.in
+triton==3.3.1
+    # via torch
+types-python-dateutil==2.9.0.20250708
+    # via arrow
+typing-extensions==4.14.1
+    # via
+    #   anyio
+    #   beautifulsoup4
+    #   huggingface-hub
+    #   openai
+    #   pydantic
+    #   pydantic-core
+    #   reacton
+    #   referencing
+    #   rich-click
+    #   starlette
+    #   torch
+    #   typing-inspection
+typing-inspection==0.4.1
+    # via pydantic
+uri-template==1.3.0
+    # via jsonschema
+urllib3==2.5.0
+    # via requests
+uvicorn==0.35.0
+    # via solara-server
+watchdog==6.0.0
+    # via solara-server
+watchfiles==1.1.0
+    # via solara-server
+wcwidth==0.2.13
+    # via prompt-toolkit
+webcolors==24.11.1
+    # via jsonschema
+webencodings==0.5.1
+    # via
+    #   bleach
+    #   tinycss2
+websocket-client==1.8.0
+    # via jupyter-server
+websockets==15.0.1
+    # via solara-server
+widgetsnbextension==4.0.14
+    # via ipywidgets