Black Forest Labs Image Models

[saksham36]

No description provided.

[SBrandeis]

Let's add some tests!

You can find examples here:

huggingface.js/packages/inference/test/HfInference.spec.ts

Lines 776 to 861 in f85f5f7

	describe.concurrent(
	"Fal AI",
	() => {
	const client = new HfInference(env.HF_FAL_KEY);
	it(`textToImage - black-forest-labs/FLUX.1-schnell`, async () => {
	const res = await client.textToImage({
	model: "black-forest-labs/FLUX.1-schnell",
	provider: "fal-ai",
	inputs:
	"Extreme close-up of a single tiger eye, direct frontal view. Detailed iris and pupil. Sharp focus on eye texture and color. Natural lighting to capture authentic eye shine and depth.",
	});
	expect(res).toBeInstanceOf(Blob);
	});
	it(`automaticSpeechRecognition - openai/whisper-large-v3`, async () => {
	const res = await client.automaticSpeechRecognition({
	model: "openai/whisper-large-v3",
	provider: "fal-ai",
	data: new Blob([readTestFile("sample2.wav")], { type: "audio/x-wav" }),
	});
	expect(res).toMatchObject({
	text: " he has grave doubts whether sir frederick leighton's work is really greek after all and can discover in it but little of rocky ithaca",
	});
	});
	it("textToVideo - genmo/mochi-1-preview", async () => {
	const res = await textToVideo({
	model: "genmo/mochi-1-preview",
	inputs: "A running dog",
	parameters: {
	seed: 176,
	},
	provider: "fal-ai",
	accessToken: env.HF_FAL_KEY,
	});
	expect(res).toBeInstanceOf(Blob);
	});
	it("textToVideo - HunyuanVideo", async () => {
	const res = await textToVideo({
	model: "tencent/HunyuanVideo",
	inputs: "A running dog",
	parameters: {
	seed: 176,
	num_inference_steps: 2,
	num_frames: 85,
	resolution: "480p",
	},
	provider: "fal-ai",
	accessToken: env.HF_FAL_KEY,
	});
	expect(res).toBeInstanceOf(Blob);
	});
	it("textToVideo - CogVideoX-5b", async () => {
	const res = await textToVideo({
	model: "THUDM/CogVideoX-5b",
	inputs: "A running dog",
	parameters: {
	seed: 176,
	num_frames: 2,
	},
	provider: "fal-ai",
	accessToken: env.HF_FAL_KEY,
	});
	expect(res).toBeInstanceOf(Blob);
	});
	it("textToVideo - LTX-Video", async () => {
	const res = await textToVideo({
	model: "Lightricks/LTX-Video",
	inputs: "A running dog",
	parameters: {
	seed: 176,
	num_inference_steps: 2,
	},
	provider: "fal-ai",
	accessToken: env.HF_FAL_KEY,
	});
	expect(res).toBeInstanceOf(Blob);
	});
	},
	TIMEOUT
	);

You will need to override the hf model id -> black forest labs id mapping similarily to what is done here:

huggingface.js/packages/inference/test/HfInference.spec.ts

Lines 1072 to 1076 in f85f5f7

	HARDCODED_MODEL_ID_MAPPING.nebius = {
	"meta-llama/Llama-3.1-8B-Instruct": "meta-llama/Meta-Llama-3.1-8B-Instruct",
	"meta-llama/Llama-3.1-70B-Instruct": "meta-llama/Meta-Llama-3.1-70B-Instruct",
	"black-forest-labs/FLUX.1-schnell": "black-forest-labs/flux-schnell",
	};

^ keys of this dict must match the name of a model on the HF Hub

---

Once you've added tests, you can update the VCR tapes (= cached API responses for offline testing) by running this command:

VCR_MODE=cache pnpm run test

[julien-c]

@saksham36 let us know if any help is needed!

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

[julien-c]

👍

[julien-c]

Thanks @saksham36 for starting the PR and thanks @SBrandeis for pushing it over the finish line! let's go!!