@genai-fi/nanogpt 0.20.4 → 0.20.6

package/dist/Generator.js

@@ -1,2 +1,2 @@
-import { c as e, s as t } from "./main-BqRjvJ8B.js";
+import { c as e, s as t } from "./main-C0YUnw98.js";
 export { t as default, e as isConversation };

package/dist/TeachableLLM.js

@@ -1,2 +1,2 @@
-import { i as e } from "./main-BqRjvJ8B.js";
+import { i as e } from "./main-C0YUnw98.js";
 export { e as default };

package/dist/Trainer.d.ts

@@ -27,6 +27,7 @@ export default class Trainer extends EE<'start' | 'stop' | 'log'> {
     constructor(trainer: Trainer, options?: TrainingOptions);
     get model(): Model<ModelForwardAttributes>;
     get optimizer(): AdamWOptimizer;
+    get isTraining(): boolean;
     stop(): void;
     reset(): void;
     dispose(): void;

package/dist/Trainer.js

@@ -1,2 +1,2 @@
-import { a as e } from "./main-BqRjvJ8B.js";
+import { a as e } from "./main-C0YUnw98.js";
 export { e as default };

package/dist/loader/load.js

@@ -1,2 +1,2 @@
-import { d as e, u as t } from "../main-BqRjvJ8B.js";
+import { d as e, u as t } from "../main-C0YUnw98.js";
 export { t as VERSION, e as loadModel };

package/dist/loader/loadHF.js

@@ -1,2 +1,2 @@
-import { f as e } from "../main-BqRjvJ8B.js";
+import { f as e } from "../main-C0YUnw98.js";
 export { e as default };

package/dist/loader/loadTransformers.js

@@ -1,2 +1,2 @@
-import { h as e, m as t } from "../main-BqRjvJ8B.js";
+import { h as e, m as t } from "../main-C0YUnw98.js";
 export { t as default, e as mapTransformersConfigToGPTConfig };

package/dist/loader/newZipLoad.js

@@ -1,2 +1,2 @@
-import { p as e } from "../main-BqRjvJ8B.js";
+import { p as e } from "../main-C0YUnw98.js";
 export { e as default };

package/dist/loader/oldZipLoad.js

@@ -1,2 +1,2 @@
-import { g as e } from "../main-BqRjvJ8B.js";
+import { g as e } from "../main-C0YUnw98.js";
 export { e as default };

package/dist/loader/save.js

@@ -1,2 +1,2 @@
-import { l as e } from "../main-BqRjvJ8B.js";
+import { l as e } from "../main-C0YUnw98.js";
 export { e as saveModel };

package/dist/{main-BqRjvJ8B.js → main-C0YUnw98.js}

@@ -12829,7 +12829,7 @@ var bS = class extends hc {
 	}
 	async tokenisePrompt(e, t, n) {
 		if (t) {
-			let r = t.length > 0 && t[t.length - 1].role === "assistant", i = [];
+			let r = t.length > 0 && t[t.length - 1].role === "text", i = [];
 			return i = n?.nonConversational ? r && n?.continuation ? [e.bosToken, ...e.encode(t[t.length - 1].content)] : e.encodeAsSequence(t, !0) : e.encodeConversation(t, !0), i.length > this.model.config.blockSize && (i = i.slice(-this.model.config.blockSize)), ce([i], [1, i.length], "int32");
 		} else {
 			let t = n?.nonConversational ? void 0 : e.getSpecialTokenIndex("<|assistant_start|>"), r = t === void 0 ? [e.bosToken] : [e.bosToken, t];
@@ -12947,32 +12947,32 @@ var bS = class extends hc {
 			multinomialRand: p
 		};
 	}
-	async _generate(e) {
-		let t = !1;
-		this.outputConversation.length === 0 || this.outputConversation[this.outputConversation.length - 1]._completed || this.outputConversation[this.outputConversation.length - 1].role !== "assistant" ? (this.outputConversation.push({
-			role: "assistant",
+	async _generate(e, t) {
+		let n = !1;
+		this.outputConversation.length === 0 || this.outputConversation[this.outputConversation.length - 1]._completed || this.outputConversation[this.outputConversation.length - 1].role !== "assistant" && e?.nonConversational !== !0 || this.outputConversation[this.outputConversation.length - 1].role !== "text" && e?.nonConversational === !0 ? (this.outputConversation.push({
+			role: e?.nonConversational === !0 ? "text" : "assistant",
 			content: "",
 			_timestamp: Date.now()
-		}), t = !0, this.resetCache(!e?.noCache)) : this.lastToken < 0 && this.resetCache(!e?.noCache);
-		let n = this.lastToken >= 0 && this.cache ? ce([this.lastToken], [1, 1], "int32") : await this.tokenisePrompt(this.actualTokeniser, t ? this.outputConversation.slice(0, -1) : this.outputConversation, e), r = e?.maxLength ?? 1e3;
-		for (let t = 0; t < r && this.active; t++) {
-			let { output: i, probabilities: a, attention: o, loss: s, multinomialRand: c } = await this._generateToken(n, this.cache ? this.cache : void 0, {
+		}), n = !0, this.resetCache(!e?.noCache)) : (this.lastToken < 0 || t) && this.resetCache(!e?.noCache);
+		let r = this.lastToken >= 0 && this.cache ? ce([this.lastToken], [1, 1], "int32") : await this.tokenisePrompt(this.actualTokeniser, t ? n ? this.outputConversation.slice(0, -1) : this.outputConversation : void 0, e), i = e?.maxLength ?? 1e3;
+		for (let t = 0; t < i && this.active; t++) {
+			let { output: n, probabilities: a, attention: o, loss: s, multinomialRand: c } = await this._generateToken(r, this.cache ? this.cache : void 0, {
 				...e,
 				usePadding: !this.cache
 			});
-			if (this.lastMultinomialRand = c, s !== void 0 && (this.lastLoss = s), this.cache) n.dispose(), n = i;
+			if (this.lastMultinomialRand = c, s !== void 0 && (this.lastLoss = s), this.cache) r.dispose(), r = n;
 			else {
-				let e = n;
-				n = kt([n, i], 1), e.dispose();
+				let e = r;
+				r = kt([r, n], 1), e.dispose();
 			}
-			let l = await this.processResponse(this.actualTokeniser, i, e?.allowSpecial ?? !1, o, a);
-			if (this.cache || i.dispose(), l === null) {
+			let l = await this.processResponse(this.actualTokeniser, n, e?.allowSpecial ?? !1, o, a);
+			if (this.cache || n.dispose(), l === null) {
 				this.outputConversation[this.outputConversation.length - 1]._completed = !0;
 				break;
 			}
-			t === r - 1 && r > 1 && (this.outputConversation[this.outputConversation.length - 1]._completed = !0), this.outputConversation[this.outputConversation.length - 1].content += l;
+			t === i - 1 && i > 1 && (this.outputConversation[this.outputConversation.length - 1]._completed = !0), this.outputConversation[this.outputConversation.length - 1].content += l;
 		}
-		return n.dispose(), this.outputConversation;
+		return r.dispose(), this.outputConversation;
 	}
 	resetCache(e) {
 		this.cache && (this.cache.forEach((e) => {
@@ -13036,7 +13036,7 @@ var bS = class extends hc {
 	}
 	async startJob(e, t) {
 		this.initialise(e, t), this.active = !0, this.model.metaData.generationSettings = t, t?.maxLength !== 1 && this.emit("start");
-		let n = await this._generate(t);
+		let n = await this._generate(t, !!e);
 		if (this.active = !1, this.startTime !== null) {
 			let e = Date.now(), n = e - this.startTime;
 			this.startTime = null, this.model.metaData.actionLog = this.model.metaData.actionLog || [], this.model.metaData.actionLog.push({
@@ -13178,6 +13178,9 @@ var DS = class e extends hc {
 	get optimizer() {
 		return this.trainer.optimizer;
 	}
+	get isTraining() {
+		return this.trainer.isRunning;
+	}
 	stop() {
 		this.trainer.stop();
 	}

package/dist/main.js

@@ -1,4 +1,4 @@
-import { a as e, b as t, i as n, n as r, r as i, s as a, t as o, y as s } from "./main-BqRjvJ8B.js";
+import { a as e, b as t, i as n, n as r, r as i, s as a, t as o, y as s } from "./main-C0YUnw98.js";
 import c from "./tokeniser/CharTokeniser.js";
 import l from "./tokeniser/bpe.js";
 import { AdamWOptimizer as u } from "./training/AdamW.js";

package/dist/models/NanoGPTV1.js

@@ -1,2 +1,2 @@
-import { y as e } from "../main-BqRjvJ8B.js";
+import { y as e } from "../main-C0YUnw98.js";
 export { e as default };

package/dist/models/NanoGPTV2.js

@@ -1,2 +1,2 @@
-import { v as e } from "../main-BqRjvJ8B.js";
+import { v as e } from "../main-C0YUnw98.js";
 export { e as default };

package/dist/models/factory.js

@@ -1,2 +1,2 @@
-import { _ as e } from "../main-BqRjvJ8B.js";
+import { _ as e } from "../main-C0YUnw98.js";
 export { e as default };

package/dist/models/model.js

@@ -1,2 +1,2 @@
-import { b as e } from "../main-BqRjvJ8B.js";
+import { b as e } from "../main-C0YUnw98.js";
 export { e as default };

package/dist/training/validation.js

@@ -1,2 +1,2 @@
-import { o as e } from "../main-BqRjvJ8B.js";
+import { o as e } from "../main-C0YUnw98.js";
 export { e as createTrainValidationSplit };

package/package.json

@@ -1,6 +1,6 @@
 {
     "name": "@genai-fi/nanogpt",
-    "version": "0.20.4",
+    "version": "0.20.6",
     "type": "module",
     "main": "dist/main.js",
     "types": "dist/main.d.ts",