diff --git a/docs/inference-providers/_toctree.yml b/docs/inference-providers/_toctree.yml index e5f6caf55..437c8b916 100644 --- a/docs/inference-providers/_toctree.yml +++ b/docs/inference-providers/_toctree.yml @@ -29,6 +29,8 @@ title: VS Code with GitHub Copilot - local: guides/github-actions-code-review title: Automating Code Review with GitHub Actions + - local: guides/evaluation-inspect-ai + title: Evaluating Models with Inspect - local: tasks/index title: Inference Tasks diff --git a/docs/inference-providers/guides/evaluation-inspect-ai.md b/docs/inference-providers/guides/evaluation-inspect-ai.md new file mode 100644 index 000000000..3a2fe2c2f --- /dev/null +++ b/docs/inference-providers/guides/evaluation-inspect-ai.md @@ -0,0 +1,253 @@ +# Evaluating models with Inspect +In this guide, we'll learn how to evaluate models using [Inspect](https://inspect.aisi.org.uk/), an open-source framework for language model evaluations created by the UK AI Security Institute. + +## Installation +To get started: + +1. Install Inspect: + +```bash +pip install inspect-ai +``` + +2. If you're using VS Code or a compatible IDE, consider installing the [Inspect VS Code Extension](https://inspect.aisi.org.uk/vscode.html). + + +3. Set your `HF_TOKEN` as an environment variable and install the `openai` package to call models using Inference Providers. + +```bash +export HF_TOKEN="your_token_here" +pip install openai +``` + +> [!TIP] +> The Hugging Face token will be used to authenticate your requests. If this is the first time you use Inference Providers, [create a token in your settings](https://huggingface.co/settings/tokens/new?ownUserPermissions=inference.serverless.write&tokenType=fineGrained) or check [this section](https://huggingface.co/docs/inference-providers/index#authentication) to learn more details. + +## Run your first evaluation +Let's start by running the ["Hello, Inspect"](https://inspect.aisi.org.uk/#sec-hello-inspect) example from Inspect's documentation. This example evaluates models on the Sally-Anne test, which assesses the ability to infer false beliefs in others. The test format is as follows: + +| input | target | +|-------|--------| +| Jackson entered the hall. Chloe entered the hall. The boots is in the bathtub. Jackson exited the hall. Jackson entered the dining_room. Chloe moved the boots to the pantry. Where was the boots at the beginning? | bathtub | +| Hannah entered the patio. Noah entered the patio. The sweater is in the bucket. Noah exited the patio. Ethan entered the study. Ethan exited the study. Hannah moved the sweater to the pantry. Where will Hannah look for the sweater? |pantry | + +Evaluations in Inspect are written in Python. The following code implements our evaluation. It will run the model over the inputs of the `theory_of_mind` dataset using our model of choice and grade the responses using the same model as `scorer` and the targets in the dataset. + +```python +from inspect_ai import Task, task +from inspect_ai.dataset import example_dataset +from inspect_ai.scorer import model_graded_fact +from inspect_ai.solver import generate + +@task +def theory_of_mind(): + return Task( + dataset=example_dataset("theory_of_mind"), + solver=generate(), + scorer=model_graded_fact() + ) +``` +If we save the above to a file `theory-of-mind.py`, we can use the `inspect eval` command from the terminal. Let's evaluate the [`gpt-oss-20b` model by OpenAI](https://huggingface.co/openai/gpt-oss-20b): + +```bash +inspect eval theory-of-mind.py --model hf-inference-providers/openai/gpt-oss-20b +``` +If everything went well we will see the following beautiful TUI: + +![Screenshot of theory of mind's eval with gpt-oss-20b](https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/inference-providers-guides/evals-guide-first-eval.png) + +To see the evaluation samples and inference in real time you can check the `Running samples` tab: + +![Screenshot of running samples with gpt-oss-20b](https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/inference-providers-guides/evals-guide-running-samples.png) + +Once it finishes, we'll see the evaluation results: + +![Screenshot of theory of mind's eval results with gpt-oss-20b](https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/inference-providers-guides/evals-guide-first-eval-results.png) + +Besides the command line report, Inspect comes with a nice viewer UI. We can launch it with the following command: + +```bash +inspect viewer +``` +![Screenshot of inspect viewer results with gpt-oss-20b](https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/inference-providers-guides/evals-guide-first-eval-viewer.png) + +Nice! We have just evaluated our first model with Inspect and Inference Providers. Now let's look at more advanced examples. + +## Example: Benchmarking several models for a task +In this section, we will evaluate several models for a specific task. This is useful for selecting the most suitable model for your project and establishing a baseline if you plan to fine-tune a model for your use case. Fortunately, Inspect and Inference Providers make this task very easy. We need to: + +1. **Select a list of target models**. The best place to select a model is the "Models" page on the Hub, where you can sort and filter models by size, task, languages, and many other features. You can use [this link](https://huggingface.co/models?pipeline_tag=text-generation&inference_provider=all&sort=trending) to browse all `text-generation` models with Inference Providers' support. For this guide, let's use the following models: `MiniMaxAI/MiniMax-M2`, `openai/gpt-oss-20b`, `openai/gpt-oss-120b`, and `moonshotai/Kimi-K2-Instruct-0905`. + +2. **Write and run the evaluation**. To run the evaluation across different models, we can use the `eval-set` command and provide the list of models separated by commas, as follows: + +```bash +inspect eval-set theory-of-mind.py --model \ + "hf-inference-providers/MiniMaxAI/MiniMax-M2,\ + hf-inference-providers/openai/gpt-oss-20b,\ + hf-inference-providers/openai/gpt-oss-120b,\ + hf-inference-providers/moonshotai/Kimi-K2-Instruct-0905" +``` + +If everything went well we will see the evaluations running in parallel for each model on the list. To analyze the results, we can use the viewer: + + +```bash +inspect viewer +``` +![Screenshot of inspect viewer results with gpt-oss-20b](https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/inference-providers-guides/evals-guide-model-bench-viewer.png) + + +## Example: Comparing several inference providers for a task +In this section, we will evaluate the same model across different providers. Inference Providers gives us access to many providers for the same model. Performance might vary across providers, so this is a useful factor, in addition to speed and cost, to choose the most appropriate inference provider for your task. + +If we don't specify a provider, like we did in our previous examples, the system automatically routes your request to the first available provider for the specified model, following your preference order in [Inference Provider settings](https://hf.co/settings/inference-providers). But we can also select the provider by appending the provider name to the model id (e.g. `openai/gpt-oss-120b:sambanova`). + +Let's run the evaluations for `gpt-oss-120b` across several providers. Please note that this time we are using the `eval_set` function directly in Python for extra flexibility (e.g., changing the list of providers): + +```python +from inspect_ai import eval_set + +target_providers = [ + "together", + "sambanova", + "groq", + "novita", + "nebius", + "cerebras", + "nscale", + "hyperbolic", + "fireworks-ai", + "scaleway" +] + +models = [f"hf-inference-providers/openai/gpt-oss-120b:{provider}" for provider in target_providers] + +success, logs = eval_set( + tasks=["theory-of-mind.py"], + model=models +) +``` + +If we save the above to a file `theory-of-mind-providers.py`, we can run the evaluation set with Python as follows: + +```bash +python theory-of-mind-providers.py +``` + +![Screenshot of theory of mind's eval results across providers](https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/inference-providers-guides/evals-guide-providers-benchmarking.png) + +Launching the viewer and sorting by score, we can compare the performance across providers, similar to the table below: + +| Model | Provider | Score | +|-------|----------|-------| +| openai/gpt-oss-120b | hyperbolic | 0.84 | +| openai/gpt-oss-120b | fireworks-ai | 0.82 | +| openai/gpt-oss-120b | nscale | 0.82 | +| openai/gpt-oss-120b | nebius | 0.82 | +| openai/gpt-oss-120b | scaleway | 0.81 | +| openai/gpt-oss-120b | together | 0.81 | +| openai/gpt-oss-120b | cerebras | 0.8 | +| openai/gpt-oss-120b | novita | 0.8 | +| openai/gpt-oss-120b | groq | 0.8 | +| openai/gpt-oss-120b | sambanova | 0.8 | + +> [!TIP] +> **Why performance varies across providers**: As seen above, the same model can produce different results when served by different inference providers due to several factors: variations in inference implementations, differences in hardware (GPU generations, optimizations), and non-determinism introduced by load balancing and batching strategies. Performance can vary across the matrix of provider-model combinations and may change with updates to inference stacks, GPU generations, and model versions. Evaluating across multiple providers helps identify the best-performing combinations for your specific use case. + +As mentioned earlier, two additional factors for choosing a model are speed and cost. Luckily, Inference Providers give you another selection policy by appending `:fastest` (selects the provider with highest throughput) or `:cheapest` (selects the provider with lowest price per output token) to the model id (e.g., `openai/gpt-oss-120b:fastest`). Using Inspect you can compare the performance between these two options as follows: + +```bash +inspect eval-set theory-of-mind.py --model \ + "hf-inference-providers/openai/gpt-oss-120b:fastest,\ + hf-inference-providers/openai/gpt-oss-120b:cheapest" +``` +After the evaluation completes, we see the following report: +![Screenshot of theory of mind's eval results with gpt-oss-120b for the fastest and cheapest providers](https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/inference-providers-guides/evals-guide-providers-fastest-vs-cheapest.png) + +In this case, for this tiny benchmark the fastest provider gets better accuracy while being significantly faster. + + +## Example: Writing a custom evaluation for Vision Language Models +In this section, we will write custom evaluation code and learn how to evaluate Vision Language Models (VLMs). Inference Providers give us access to dozens of VLMs. You can use [this link](https://huggingface.co/models?pipeline_tag=image-text-to-text&inference_provider=all&sort=trending) to browse all `image-text-to-text` models (VLMs) with Inference Providers' support. + +In order to write a custom evaluation, we need to find or create a dataset. For this evaluation, we have created a new dataset called `animals_or_else`. This dataset is inspired by the popular "chihuahua or muffin" challenge that demonstrates how even state-of-the-art computer vision models can struggle with visually similar objects. The task tests whether VLMs can correctly identify and count animals in images, distinguishing them from similar-looking objects like food items. This fun but challenging benchmark helps assess a model's ability to handle ambiguous visual inputs. We can browse the dataset below: + + + +As this dataset is available on the Hub, we can leverage Inspect's utility to read and get the samples from the dataset. The rest of the code is very similar to our previous examples, highlighting how easy it is to run custom evaluations with Inference Providers and Inspect: + +```python +from typing import Any +import tempfile +from io import BytesIO +from PIL import Image + +from inspect_ai import Task, task +from inspect_ai.dataset import Sample, hf_dataset +from inspect_ai.model import ChatMessageUser, ContentText, ContentImage +from inspect_ai.scorer import model_graded_fact +from inspect_ai.solver import generate + +@task +def animal_or_else(): + return Task( + dataset=hf_dataset( + path="dvilasuero/animal_or_else", + split="train", + sample_fields=record_to_sample, + shuffle=True, + ), + solver=generate(), + scorer=model_graded_fact() + ) + +def record_to_sample(record: dict[str, Any]) -> Sample: + + # Resize and save the image + img = Image.open(BytesIO(record["image"]["bytes"])) + img.thumbnail((1024, 1024)) + with tempfile.NamedTemporaryFile(delete=False, suffix='.jpg') as tmp_file: + img.save(tmp_file.name, format='JPEG') + image_path = tmp_file.name + + # We ask the VLM to count the number of images containing animals + message = [ + ChatMessageUser( + content=[ + ContentText(text="Count the number of images containing animals"), + ContentImage(image=image_path), + ] + ) + ] + + # The scorer will compare the VLM response with the target values in the dataset + return Sample( + input=message, + target=record["target"] + ) +``` +We can now run the `inspect eval` command as before: + +```bash +inspect eval animal_or_else.py --model hf-inference-providers/Qwen/Qwen3-VL-8B-Instruct +``` + +And compare the above model with a larger reasoning model: +```bash +inspect eval animal_or_else.py --model hf-inference-providers/Qwen/Qwen3-VL-30B-A3B-Thinking +``` + +| Model | Accuracy | +|-------|----------| +| Qwen/Qwen3-VL-8B-Instruct | 0.7 | +| Qwen/Qwen3-VL-30B-A3B-Thinking | 0.9 | + +# Next Steps +* Explore [Inspect's documentation](https://inspect.aisi.org.uk/) to learn more about model evaluation. +* Check out the [lighteval](https://github.com/huggingface/lighteval) library. It comes with over [1,000 tasks](https://huggingface.co/spaces/OpenEvals/open_benchmark_index), so you don't have to write any code, and it gives you several quality-of-life features for quickly running evaluations. +* Browse models available through Inference Providers to find the best model for your needs and run your own evaluations. \ No newline at end of file