Merge branch 'main' into feat/support-novita-async-t2v

Author: viktor2077

packages/inference/src/providers/featherless-ai.ts

@@ -7,6 +7,7 @@ import type {
 import { InferenceOutputError } from "../lib/InferenceOutputError.js";
 import type { BodyParams } from "../types.js";
 import { BaseConversationalTask, BaseTextGenerationTask } from "./providerHelper.js";
+import { omit } from "../utils/omit.js";
 
 interface FeatherlessAITextCompletionOutput extends Omit<ChatCompletionOutput, "choices"> {
 	choices: Array<{
@@ -33,9 +34,14 @@ export class FeatherlessAITextGenerationTask extends BaseTextGenerationTask {
 
 	override preparePayload(params: BodyParams<TextGenerationInput>): Record<string, unknown> {
 		return {
-			...params.args,
-			...params.args.parameters,
 			model: params.model,
+			...omit(params.args, ["inputs", "parameters"]),
+			...(params.args.parameters
+				? {
+						max_tokens: params.args.parameters.max_new_tokens,
+						...omit(params.args.parameters, "max_new_tokens"),
+				  }
+				: undefined),
 			prompt: params.args.inputs,
 		};
 	}

packages/tasks-gen/scripts/generate-snippets-fixtures.ts

@@ -19,6 +19,7 @@ import { existsSync as pathExists } from "node:fs";
 import * as fs from "node:fs/promises";
 import * as path from "node:path/posix";
 
+import type { InferenceProviderOrPolicy } from "@huggingface/inference";
 import { snippets } from "@huggingface/inference";
 import type { InferenceSnippet, ModelDataMinimal, SnippetInferenceProvider, WidgetType } from "@huggingface/tasks";
 import { inferenceSnippetLanguages } from "@huggingface/tasks";
@@ -31,7 +32,7 @@ const TEST_CASES: {
 	testName: string;
 	task: WidgetType;
 	model: ModelDataMinimal;
-	providers: SnippetInferenceProvider[];
+	providers: InferenceProviderOrPolicy[];
 	lora?: boolean;
 	opts?: snippets.InferenceSnippetOptions;
 }[] = [
@@ -306,7 +307,7 @@ function getFixtureFolder(testName: string): string {
 function generateInferenceSnippet(
 	model: ModelDataMinimal,
 	language: Language,
-	provider: SnippetInferenceProvider,
+	provider: InferenceProviderOrPolicy,
 	task: WidgetType,
 	lora: boolean = false,
 	opts?: Record<string, unknown>

packages/tasks/src/hardware.ts

@@ -100,6 +100,10 @@ export const SKUS = {
 				tflops: 38.7,
 				memory: [48],
 			},
+			"RTX A5000": {
+				tflops: 34.1,
+				memory: [24],
+			},
 			"RTX A4000": {
 				tflops: 19.2,
 				memory: [16],
@@ -533,6 +537,9 @@ export const SKUS = {
 			"Intel Core Ultra 7 265KF": {
 				tflops: 1.53,
 			},
+			"Intel Core 14th Generation (i7)": {
+				tflops: 0.8,
+			},
 			"Intel Core 13th Generation (i9)": {
 				tflops: 0.85,
 			},