gpt-tokenizer 2.0.0 → 2.1.1
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/README.md +54 -8
- package/cjs/GptEncoding.d.ts +33 -6
- package/cjs/GptEncoding.js +79 -34
- package/cjs/GptEncoding.js.map +1 -1
- package/cjs/GptEncoding.test.js +54 -0
- package/cjs/GptEncoding.test.js.map +1 -1
- package/cjs/encoding/cl100k_base.d.ts +2 -2
- package/cjs/encoding/cl100k_base.js +4 -2
- package/cjs/encoding/cl100k_base.js.map +1 -1
- package/cjs/encoding/p50k_base.d.ts +1 -1
- package/cjs/encoding/p50k_edit.d.ts +1 -1
- package/cjs/encoding/r50k_base.d.ts +1 -1
- package/cjs/main.d.ts +1 -0
- package/cjs/main.js +7 -0
- package/cjs/main.js.map +1 -1
- package/cjs/mapping.d.ts +37 -0
- package/cjs/mapping.js +34 -2
- package/cjs/mapping.js.map +1 -1
- package/cjs/model/ada.d.ts +1 -0
- package/cjs/model/ada.js +7 -0
- package/cjs/model/ada.js.map +1 -1
- package/cjs/model/babbage.d.ts +1 -0
- package/cjs/model/babbage.js +7 -0
- package/cjs/model/babbage.js.map +1 -1
- package/cjs/model/code-cushman-001.d.ts +1 -0
- package/cjs/model/code-cushman-001.js +7 -0
- package/cjs/model/code-cushman-001.js.map +1 -1
- package/cjs/model/code-cushman-002.d.ts +1 -0
- package/cjs/model/code-cushman-002.js +7 -0
- package/cjs/model/code-cushman-002.js.map +1 -1
- package/cjs/model/code-davinci-001.d.ts +1 -0
- package/cjs/model/code-davinci-001.js +7 -0
- package/cjs/model/code-davinci-001.js.map +1 -1
- package/cjs/model/code-davinci-002.d.ts +1 -0
- package/cjs/model/code-davinci-002.js +7 -0
- package/cjs/model/code-davinci-002.js.map +1 -1
- package/cjs/model/code-davinci-edit-001.d.ts +1 -0
- package/cjs/model/code-davinci-edit-001.js +7 -0
- package/cjs/model/code-davinci-edit-001.js.map +1 -1
- package/cjs/model/code-search-ada-code-001.d.ts +1 -0
- package/cjs/model/code-search-ada-code-001.js +7 -0
- package/cjs/model/code-search-ada-code-001.js.map +1 -1
- package/cjs/model/code-search-babbage-code-001.d.ts +1 -0
- package/cjs/model/code-search-babbage-code-001.js +7 -0
- package/cjs/model/code-search-babbage-code-001.js.map +1 -1
- package/cjs/model/curie.d.ts +1 -0
- package/cjs/model/curie.js +7 -0
- package/cjs/model/curie.js.map +1 -1
- package/cjs/model/cushman-codex.d.ts +1 -0
- package/cjs/model/cushman-codex.js +7 -0
- package/cjs/model/cushman-codex.js.map +1 -1
- package/cjs/model/davinci-codex.d.ts +1 -0
- package/cjs/model/davinci-codex.js +7 -0
- package/cjs/model/davinci-codex.js.map +1 -1
- package/cjs/model/davinci.d.ts +1 -0
- package/cjs/model/davinci.js +7 -0
- package/cjs/model/davinci.js.map +1 -1
- package/cjs/model/gpt-3.5-turbo-0301.d.ts +6 -0
- package/cjs/model/gpt-3.5-turbo-0301.js +38 -0
- package/cjs/model/gpt-3.5-turbo-0301.js.map +1 -0
- package/cjs/model/gpt-3.5-turbo.d.ts +6 -1
- package/cjs/model/gpt-3.5-turbo.js +21 -1
- package/cjs/model/gpt-3.5-turbo.js.map +1 -1
- package/cjs/model/gpt-4-0314.d.ts +6 -0
- package/cjs/model/gpt-4-0314.js +38 -0
- package/cjs/model/gpt-4-0314.js.map +1 -0
- package/cjs/model/gpt-4-32k-0314.d.ts +6 -0
- package/cjs/model/gpt-4-32k-0314.js +38 -0
- package/cjs/model/gpt-4-32k-0314.js.map +1 -0
- package/cjs/model/gpt-4-32k.d.ts +6 -0
- package/cjs/model/gpt-4-32k.js +38 -0
- package/cjs/model/gpt-4-32k.js.map +1 -0
- package/cjs/model/gpt-4.d.ts +6 -1
- package/cjs/model/gpt-4.js +21 -1
- package/cjs/model/gpt-4.js.map +1 -1
- package/cjs/model/text-ada-001.d.ts +1 -0
- package/cjs/model/text-ada-001.js +7 -0
- package/cjs/model/text-ada-001.js.map +1 -1
- package/cjs/model/text-babbage-001.d.ts +1 -0
- package/cjs/model/text-babbage-001.js +7 -0
- package/cjs/model/text-babbage-001.js.map +1 -1
- package/cjs/model/text-curie-001.d.ts +1 -0
- package/cjs/model/text-curie-001.js +7 -0
- package/cjs/model/text-curie-001.js.map +1 -1
- package/cjs/model/text-davinci-001.d.ts +1 -0
- package/cjs/model/text-davinci-001.js +7 -0
- package/cjs/model/text-davinci-001.js.map +1 -1
- package/cjs/model/text-davinci-002.d.ts +1 -0
- package/cjs/model/text-davinci-002.js +7 -0
- package/cjs/model/text-davinci-002.js.map +1 -1
- package/cjs/model/text-davinci-003.d.ts +1 -0
- package/cjs/model/text-davinci-003.js +7 -0
- package/cjs/model/text-davinci-003.js.map +1 -1
- package/cjs/model/text-davinci-edit-001.d.ts +1 -0
- package/cjs/model/text-davinci-edit-001.js +7 -0
- package/cjs/model/text-davinci-edit-001.js.map +1 -1
- package/cjs/model/text-embedding-ada-002.d.ts +1 -0
- package/cjs/model/text-embedding-ada-002.js +7 -0
- package/cjs/model/text-embedding-ada-002.js.map +1 -1
- package/cjs/model/text-search-ada-doc-001.d.ts +1 -0
- package/cjs/model/text-search-ada-doc-001.js +7 -0
- package/cjs/model/text-search-ada-doc-001.js.map +1 -1
- package/cjs/model/text-search-babbage-doc-001.d.ts +1 -0
- package/cjs/model/text-search-babbage-doc-001.js +7 -0
- package/cjs/model/text-search-babbage-doc-001.js.map +1 -1
- package/cjs/model/text-search-curie-doc-001.d.ts +1 -0
- package/cjs/model/text-search-curie-doc-001.js +7 -0
- package/cjs/model/text-search-curie-doc-001.js.map +1 -1
- package/cjs/model/text-search-davinci-doc-001.d.ts +1 -0
- package/cjs/model/text-search-davinci-doc-001.js +7 -0
- package/cjs/model/text-search-davinci-doc-001.js.map +1 -1
- package/cjs/model/text-similarity-ada-001.d.ts +1 -0
- package/cjs/model/text-similarity-ada-001.js +7 -0
- package/cjs/model/text-similarity-ada-001.js.map +1 -1
- package/cjs/model/text-similarity-babbage-001.d.ts +1 -0
- package/cjs/model/text-similarity-babbage-001.js +7 -0
- package/cjs/model/text-similarity-babbage-001.js.map +1 -1
- package/cjs/model/text-similarity-curie-001.d.ts +1 -0
- package/cjs/model/text-similarity-curie-001.js +7 -0
- package/cjs/model/text-similarity-curie-001.js.map +1 -1
- package/cjs/model/text-similarity-davinci-001.d.ts +1 -0
- package/cjs/model/text-similarity-davinci-001.js +7 -0
- package/cjs/model/text-similarity-davinci-001.js.map +1 -1
- package/cjs/modelParams.d.ts +5 -4
- package/cjs/modelParams.js +7 -4
- package/cjs/modelParams.js.map +1 -1
- package/cjs/specialTokens.d.ts +3 -0
- package/cjs/specialTokens.js +4 -1
- package/cjs/specialTokens.js.map +1 -1
- package/cjs/util.d.ts +2 -0
- package/cjs/util.js +19 -0
- package/cjs/util.js.map +1 -0
- package/dist/cl100k_base.js +1 -1
- package/dist/cl100k_base.js.map +1 -1
- package/dist/p50k_base.js +1 -1
- package/dist/p50k_base.js.map +1 -1
- package/dist/p50k_edit.js +1 -1
- package/dist/p50k_edit.js.map +1 -1
- package/dist/r50k_base.js +1 -1
- package/dist/r50k_base.js.map +1 -1
- package/esm/GptEncoding.d.ts +33 -6
- package/esm/GptEncoding.js +81 -36
- package/esm/GptEncoding.js.map +1 -1
- package/esm/GptEncoding.test.js +55 -1
- package/esm/GptEncoding.test.js.map +1 -1
- package/esm/codegen/generateByModel.js +19 -0
- package/esm/codegen/generateByModel.js.map +1 -0
- package/esm/{generateJsModels.js → codegen/generateJsEncodings.js} +4 -4
- package/esm/codegen/generateJsEncodings.js.map +1 -0
- package/esm/encoding/cl100k_base.d.ts +2 -2
- package/esm/encoding/cl100k_base.js +2 -2
- package/esm/encoding/cl100k_base.js.map +1 -1
- package/esm/encoding/p50k_base.d.ts +1 -1
- package/esm/encoding/p50k_edit.d.ts +1 -1
- package/esm/encoding/r50k_base.d.ts +1 -1
- package/esm/main.d.ts +1 -0
- package/esm/main.js +2 -0
- package/esm/main.js.map +1 -1
- package/esm/mapping.d.ts +37 -0
- package/esm/mapping.js +32 -0
- package/esm/mapping.js.map +1 -1
- package/esm/model/ada.d.ts +1 -0
- package/esm/model/ada.js +2 -0
- package/esm/model/ada.js.map +1 -1
- package/esm/model/babbage.d.ts +1 -0
- package/esm/model/babbage.js +2 -0
- package/esm/model/babbage.js.map +1 -1
- package/esm/model/code-cushman-001.d.ts +1 -0
- package/esm/model/code-cushman-001.js +2 -0
- package/esm/model/code-cushman-001.js.map +1 -1
- package/esm/model/code-cushman-002.d.ts +1 -0
- package/esm/model/code-cushman-002.js +2 -0
- package/esm/model/code-cushman-002.js.map +1 -1
- package/esm/model/code-davinci-001.d.ts +1 -0
- package/esm/model/code-davinci-001.js +2 -0
- package/esm/model/code-davinci-001.js.map +1 -1
- package/esm/model/code-davinci-002.d.ts +1 -0
- package/esm/model/code-davinci-002.js +2 -0
- package/esm/model/code-davinci-002.js.map +1 -1
- package/esm/model/code-davinci-edit-001.d.ts +1 -0
- package/esm/model/code-davinci-edit-001.js +2 -0
- package/esm/model/code-davinci-edit-001.js.map +1 -1
- package/esm/model/code-search-ada-code-001.d.ts +1 -0
- package/esm/model/code-search-ada-code-001.js +2 -0
- package/esm/model/code-search-ada-code-001.js.map +1 -1
- package/esm/model/code-search-babbage-code-001.d.ts +1 -0
- package/esm/model/code-search-babbage-code-001.js +2 -0
- package/esm/model/code-search-babbage-code-001.js.map +1 -1
- package/esm/model/curie.d.ts +1 -0
- package/esm/model/curie.js +2 -0
- package/esm/model/curie.js.map +1 -1
- package/esm/model/cushman-codex.d.ts +1 -0
- package/esm/model/cushman-codex.js +2 -0
- package/esm/model/cushman-codex.js.map +1 -1
- package/esm/model/davinci-codex.d.ts +1 -0
- package/esm/model/davinci-codex.js +2 -0
- package/esm/model/davinci-codex.js.map +1 -1
- package/esm/model/davinci.d.ts +1 -0
- package/esm/model/davinci.js +2 -0
- package/esm/model/davinci.js.map +1 -1
- package/esm/model/gpt-3.5-turbo-0301.d.ts +6 -0
- package/esm/model/gpt-3.5-turbo-0301.js +11 -0
- package/esm/model/gpt-3.5-turbo-0301.js.map +1 -0
- package/esm/model/gpt-3.5-turbo.d.ts +6 -1
- package/esm/model/gpt-3.5-turbo.js +10 -1
- package/esm/model/gpt-3.5-turbo.js.map +1 -1
- package/esm/model/gpt-4-0314.d.ts +6 -0
- package/esm/model/gpt-4-0314.js +11 -0
- package/esm/model/gpt-4-0314.js.map +1 -0
- package/esm/model/gpt-4-32k-0314.d.ts +6 -0
- package/esm/model/gpt-4-32k-0314.js +11 -0
- package/esm/model/gpt-4-32k-0314.js.map +1 -0
- package/esm/model/gpt-4-32k.d.ts +6 -0
- package/esm/model/gpt-4-32k.js +11 -0
- package/esm/model/gpt-4-32k.js.map +1 -0
- package/esm/model/gpt-4.d.ts +6 -1
- package/esm/model/gpt-4.js +10 -1
- package/esm/model/gpt-4.js.map +1 -1
- package/esm/model/text-ada-001.d.ts +1 -0
- package/esm/model/text-ada-001.js +2 -0
- package/esm/model/text-ada-001.js.map +1 -1
- package/esm/model/text-babbage-001.d.ts +1 -0
- package/esm/model/text-babbage-001.js +2 -0
- package/esm/model/text-babbage-001.js.map +1 -1
- package/esm/model/text-curie-001.d.ts +1 -0
- package/esm/model/text-curie-001.js +2 -0
- package/esm/model/text-curie-001.js.map +1 -1
- package/esm/model/text-davinci-001.d.ts +1 -0
- package/esm/model/text-davinci-001.js +2 -0
- package/esm/model/text-davinci-001.js.map +1 -1
- package/esm/model/text-davinci-002.d.ts +1 -0
- package/esm/model/text-davinci-002.js +2 -0
- package/esm/model/text-davinci-002.js.map +1 -1
- package/esm/model/text-davinci-003.d.ts +1 -0
- package/esm/model/text-davinci-003.js +2 -0
- package/esm/model/text-davinci-003.js.map +1 -1
- package/esm/model/text-davinci-edit-001.d.ts +1 -0
- package/esm/model/text-davinci-edit-001.js +2 -0
- package/esm/model/text-davinci-edit-001.js.map +1 -1
- package/esm/model/text-embedding-ada-002.d.ts +1 -0
- package/esm/model/text-embedding-ada-002.js +2 -0
- package/esm/model/text-embedding-ada-002.js.map +1 -1
- package/esm/model/text-search-ada-doc-001.d.ts +1 -0
- package/esm/model/text-search-ada-doc-001.js +2 -0
- package/esm/model/text-search-ada-doc-001.js.map +1 -1
- package/esm/model/text-search-babbage-doc-001.d.ts +1 -0
- package/esm/model/text-search-babbage-doc-001.js +2 -0
- package/esm/model/text-search-babbage-doc-001.js.map +1 -1
- package/esm/model/text-search-curie-doc-001.d.ts +1 -0
- package/esm/model/text-search-curie-doc-001.js +2 -0
- package/esm/model/text-search-curie-doc-001.js.map +1 -1
- package/esm/model/text-search-davinci-doc-001.d.ts +1 -0
- package/esm/model/text-search-davinci-doc-001.js +2 -0
- package/esm/model/text-search-davinci-doc-001.js.map +1 -1
- package/esm/model/text-similarity-ada-001.d.ts +1 -0
- package/esm/model/text-similarity-ada-001.js +2 -0
- package/esm/model/text-similarity-ada-001.js.map +1 -1
- package/esm/model/text-similarity-babbage-001.d.ts +1 -0
- package/esm/model/text-similarity-babbage-001.js +2 -0
- package/esm/model/text-similarity-babbage-001.js.map +1 -1
- package/esm/model/text-similarity-curie-001.d.ts +1 -0
- package/esm/model/text-similarity-curie-001.js +2 -0
- package/esm/model/text-similarity-curie-001.js.map +1 -1
- package/esm/model/text-similarity-davinci-001.d.ts +1 -0
- package/esm/model/text-similarity-davinci-001.js +2 -0
- package/esm/model/text-similarity-davinci-001.js.map +1 -1
- package/esm/modelParams.d.ts +5 -4
- package/esm/modelParams.js +6 -3
- package/esm/modelParams.js.map +1 -1
- package/esm/specialTokens.d.ts +3 -0
- package/esm/specialTokens.js +3 -0
- package/esm/specialTokens.js.map +1 -1
- package/esm/util.d.ts +2 -0
- package/esm/util.js +14 -0
- package/esm/util.js.map +1 -0
- package/package.json +8 -6
- package/src/GptEncoding.test.ts +78 -1
- package/src/GptEncoding.ts +125 -44
- package/src/__snapshots__/GptEncoding.test.ts.snap +811 -0
- package/src/codegen/generateByModel.ts +35 -0
- package/src/{generateJsModels.ts → codegen/generateJsEncodings.ts} +3 -3
- package/src/encoding/cl100k_base.ts +4 -0
- package/src/main.ts +2 -0
- package/src/mapping.ts +42 -0
- package/src/model/ada.ts +2 -0
- package/src/model/babbage.ts +2 -0
- package/src/model/code-cushman-001.ts +2 -0
- package/src/model/code-cushman-002.ts +2 -0
- package/src/model/code-davinci-001.ts +2 -0
- package/src/model/code-davinci-002.ts +2 -0
- package/src/model/code-davinci-edit-001.ts +2 -0
- package/src/model/code-search-ada-code-001.ts +2 -0
- package/src/model/code-search-babbage-code-001.ts +2 -0
- package/src/model/curie.ts +2 -0
- package/src/model/cushman-codex.ts +2 -0
- package/src/model/davinci-codex.ts +2 -0
- package/src/model/davinci.ts +2 -0
- package/src/model/gpt-3.5-turbo-0301.ts +32 -0
- package/src/model/gpt-3.5-turbo.ts +32 -1
- package/src/model/gpt-4-0314.ts +32 -0
- package/src/model/gpt-4-32k-0314.ts +32 -0
- package/src/model/gpt-4-32k.ts +32 -0
- package/src/model/gpt-4.ts +32 -1
- package/src/model/text-ada-001.ts +2 -0
- package/src/model/text-babbage-001.ts +2 -0
- package/src/model/text-curie-001.ts +2 -0
- package/src/model/text-davinci-001.ts +2 -0
- package/src/model/text-davinci-002.ts +2 -0
- package/src/model/text-davinci-003.ts +2 -0
- package/src/model/text-davinci-edit-001.ts +2 -0
- package/src/model/text-embedding-ada-002.ts +2 -0
- package/src/model/text-search-ada-doc-001.ts +2 -0
- package/src/model/text-search-babbage-doc-001.ts +2 -0
- package/src/model/text-search-curie-doc-001.ts +2 -0
- package/src/model/text-search-davinci-doc-001.ts +2 -0
- package/src/model/text-similarity-ada-001.ts +2 -0
- package/src/model/text-similarity-babbage-001.ts +2 -0
- package/src/model/text-similarity-curie-001.ts +2 -0
- package/src/model/text-similarity-davinci-001.ts +2 -0
- package/src/modelParams.ts +17 -10
- package/src/specialTokens.ts +3 -0
- package/src/util.ts +15 -0
- package/esm/generateByModel.js +0 -13
- package/esm/generateByModel.js.map +0 -1
- package/esm/generateJsModels.js.map +0 -1
- package/src/generateByModel.ts +0 -20
- /package/esm/{generateByModel.d.ts → codegen/generateByModel.d.ts} +0 -0
- /package/esm/{generateJsModels.d.ts → codegen/generateJsEncodings.d.ts} +0 -0
package/esm/GptEncoding.d.ts
CHANGED
|
@@ -1,6 +1,18 @@
|
|
|
1
1
|
import { type EncodingName, type ModelName } from './mapping.js';
|
|
2
2
|
import { type GetMergeableRanksAsyncFn, type GetMergeableRanksFn } from './modelParams.js';
|
|
3
|
-
export declare const
|
|
3
|
+
export declare const ALL_SPECIAL_TOKENS = "all";
|
|
4
|
+
export interface EncodeOptions {
|
|
5
|
+
allowedSpecial?: Set<string>;
|
|
6
|
+
disallowedSpecial?: Set<string>;
|
|
7
|
+
}
|
|
8
|
+
export interface ChatMessage {
|
|
9
|
+
role?: 'system' | 'user' | 'assistant';
|
|
10
|
+
name?: string;
|
|
11
|
+
content: string;
|
|
12
|
+
}
|
|
13
|
+
export interface EncodeChatOptions {
|
|
14
|
+
primeWithAssistantResponse?: string;
|
|
15
|
+
}
|
|
4
16
|
export declare class GptEncoding {
|
|
5
17
|
static EndOfPrompt: string;
|
|
6
18
|
static EndOfText: string;
|
|
@@ -8,6 +20,7 @@ export declare class GptEncoding {
|
|
|
8
20
|
static FimPrefix: string;
|
|
9
21
|
static FimSuffix: string;
|
|
10
22
|
decoder: TextDecoder;
|
|
23
|
+
modelName?: ModelName;
|
|
11
24
|
private bytePairEncodingCoreProcessor;
|
|
12
25
|
private specialTokenMapping;
|
|
13
26
|
private constructor();
|
|
@@ -15,14 +28,28 @@ export declare class GptEncoding {
|
|
|
15
28
|
static getEncodingApiForModel(modelName: ModelName, getMergeableRanks: GetMergeableRanksFn): GptEncoding;
|
|
16
29
|
static getEncodingApiAsync(encodingName: EncodingName, getMergeableRanks: GetMergeableRanksAsyncFn): Promise<GptEncoding>;
|
|
17
30
|
static getEncodingApiForModelAsync(modelName: ModelName, getMergeableRanks: GetMergeableRanksAsyncFn): Promise<GptEncoding>;
|
|
18
|
-
|
|
19
|
-
|
|
20
|
-
|
|
21
|
-
|
|
31
|
+
encodeGenerator(lineToEncode: string, { allowedSpecial, disallowedSpecial, }?: EncodeOptions): Generator<number[], number, undefined>;
|
|
32
|
+
encode(lineToEncode: string, encodeOptions?: EncodeOptions): number[];
|
|
33
|
+
/**
|
|
34
|
+
* Progressively tokenizes an OpenAI chat.
|
|
35
|
+
* Warning: gpt-3.5-turbo and gpt-4 chat format may change over time.
|
|
36
|
+
* Returns tokens assuming the 'gpt-3.5-turbo-0301' / 'gpt-4-0314' format.
|
|
37
|
+
* Based on OpenAI's guidelines: https://github.com/openai/openai-python/blob/main/chatml.md
|
|
38
|
+
* Also mentioned in section 6 of this document: https://github.com/openai/openai-cookbook/blob/main/examples/How_to_count_tokens_with_tiktoken.ipynb
|
|
39
|
+
*/
|
|
40
|
+
encodeChatGenerator(chat: Iterable<ChatMessage>, model?: "gpt-4" | "gpt-4-32k" | "gpt-4-0314" | "gpt-4-32k-0314" | "gpt-3.5-turbo" | "gpt-3.5-turbo-0301" | "text-davinci-003" | "text-davinci-002" | "text-davinci-001" | "text-curie-001" | "text-babbage-001" | "text-ada-001" | "davinci" | "curie" | "babbage" | "ada" | "code-davinci-002" | "code-davinci-001" | "code-cushman-002" | "code-cushman-001" | "davinci-codex" | "cushman-codex" | "text-davinci-edit-001" | "code-davinci-edit-001" | "text-embedding-ada-002" | "text-similarity-davinci-001" | "text-similarity-curie-001" | "text-similarity-babbage-001" | "text-similarity-ada-001" | "text-search-davinci-doc-001" | "text-search-curie-doc-001" | "text-search-babbage-doc-001" | "text-search-ada-doc-001" | "code-search-babbage-code-001" | "code-search-ada-code-001" | undefined): Generator<number[], void, undefined>;
|
|
41
|
+
/**
|
|
42
|
+
* Encodes a chat into a single array of tokens.
|
|
43
|
+
* Warning: gpt-3.5-turbo and gpt-4 chat format may change over time.
|
|
44
|
+
* Returns tokens assuming the 'gpt-3.5-turbo-0301' / 'gpt-4-0314' format.
|
|
45
|
+
* Based on OpenAI's guidelines: https://github.com/openai/openai-python/blob/main/chatml.md
|
|
46
|
+
* Also mentioned in section 6 of this document: https://github.com/openai/openai-cookbook/blob/main/examples/How_to_count_tokens_with_tiktoken.ipynb
|
|
47
|
+
*/
|
|
48
|
+
encodeChat(chat: readonly ChatMessage[], model?: "gpt-4" | "gpt-4-32k" | "gpt-4-0314" | "gpt-4-32k-0314" | "gpt-3.5-turbo" | "gpt-3.5-turbo-0301" | "text-davinci-003" | "text-davinci-002" | "text-davinci-001" | "text-curie-001" | "text-babbage-001" | "text-ada-001" | "davinci" | "curie" | "babbage" | "ada" | "code-davinci-002" | "code-davinci-001" | "code-cushman-002" | "code-cushman-001" | "davinci-codex" | "cushman-codex" | "text-davinci-edit-001" | "code-davinci-edit-001" | "text-embedding-ada-002" | "text-similarity-davinci-001" | "text-similarity-curie-001" | "text-similarity-babbage-001" | "text-similarity-ada-001" | "text-search-davinci-doc-001" | "text-search-curie-doc-001" | "text-search-babbage-doc-001" | "text-search-ada-doc-001" | "code-search-babbage-code-001" | "code-search-ada-code-001" | undefined): number[];
|
|
22
49
|
/**
|
|
23
50
|
* @returns {false | number} false if token limit is exceeded, otherwise the number of tokens
|
|
24
51
|
*/
|
|
25
|
-
isWithinTokenLimit(
|
|
52
|
+
isWithinTokenLimit(input: string | Iterable<ChatMessage>, tokenLimit: number): false | number;
|
|
26
53
|
decodeGenerator(inputTokensToDecode: Iterable<number>): Generator<string, void>;
|
|
27
54
|
decodeAsyncGenerator(inputTokensToDecode: AsyncIterable<number>): AsyncGenerator<string, void>;
|
|
28
55
|
decode(inputTokensToDecode: Iterable<number>): string;
|
package/esm/GptEncoding.js
CHANGED
|
@@ -1,11 +1,11 @@
|
|
|
1
1
|
/* eslint-disable no-param-reassign */
|
|
2
2
|
import { BytePairEncodingCore } from './BytePairEncodingCore.js';
|
|
3
|
-
import {
|
|
4
|
-
import {
|
|
5
|
-
import {
|
|
6
|
-
import { EndOfPrompt, EndOfText, FimMiddle, FimPrefix, FimSuffix, } from './specialTokens.js';
|
|
3
|
+
import { chatModelParams, modelToEncodingMap, } from './mapping.js';
|
|
4
|
+
import { getEncodingParams, getModelParamsAsync, } from './modelParams.js';
|
|
5
|
+
import { EndOfPrompt, EndOfText, FimMiddle, FimPrefix, FimSuffix, ImEnd, ImSep, ImStart, } from './specialTokens.js';
|
|
7
6
|
import { endsWithIncompleteUtfPairSurrogate } from './utfUtil.js';
|
|
8
|
-
|
|
7
|
+
import { getMaxValueFromMap, getSpecialTokenRegex } from './util.js';
|
|
8
|
+
export const ALL_SPECIAL_TOKENS = 'all';
|
|
9
9
|
class GptEncoding {
|
|
10
10
|
static EndOfPrompt = EndOfPrompt;
|
|
11
11
|
static EndOfText = EndOfText;
|
|
@@ -13,10 +13,11 @@ class GptEncoding {
|
|
|
13
13
|
static FimPrefix = FimPrefix;
|
|
14
14
|
static FimSuffix = FimSuffix;
|
|
15
15
|
decoder = new TextDecoder('utf8');
|
|
16
|
+
modelName;
|
|
16
17
|
bytePairEncodingCoreProcessor;
|
|
17
18
|
specialTokenMapping;
|
|
18
|
-
constructor({ tokenSplitRegex, mergeableBytePairRanks, specialTokenMapping, expectedVocabularySize, }) {
|
|
19
|
-
const maxTokenValue = Math.max(
|
|
19
|
+
constructor({ tokenSplitRegex, mergeableBytePairRanks, specialTokenMapping, expectedVocabularySize, modelName, }) {
|
|
20
|
+
const maxTokenValue = Math.max(getMaxValueFromMap(mergeableBytePairRanks), getMaxValueFromMap(specialTokenMapping));
|
|
20
21
|
this.specialTokenMapping = specialTokenMapping;
|
|
21
22
|
if (expectedVocabularySize !== undefined) {
|
|
22
23
|
if (mergeableBytePairRanks.size + specialTokenMapping.size !==
|
|
@@ -38,14 +39,16 @@ class GptEncoding {
|
|
|
38
39
|
this.decodeGenerator = this.decodeGenerator.bind(this);
|
|
39
40
|
this.decodeAsyncGenerator = this.decodeAsyncGenerator.bind(this);
|
|
40
41
|
this.isWithinTokenLimit = this.isWithinTokenLimit.bind(this);
|
|
42
|
+
this.modelName = modelName;
|
|
41
43
|
}
|
|
42
44
|
static getEncodingApi(encodingName, getMergeableRanks) {
|
|
43
|
-
const modelParams =
|
|
45
|
+
const modelParams = getEncodingParams(encodingName, getMergeableRanks);
|
|
44
46
|
return new GptEncoding(modelParams);
|
|
45
47
|
}
|
|
46
48
|
static getEncodingApiForModel(modelName, getMergeableRanks) {
|
|
47
49
|
const encodingName = modelToEncodingMap[modelName];
|
|
48
|
-
|
|
50
|
+
const modelParams = getEncodingParams(encodingName, getMergeableRanks);
|
|
51
|
+
return new GptEncoding({ ...modelParams, modelName });
|
|
49
52
|
}
|
|
50
53
|
static async getEncodingApiAsync(encodingName, getMergeableRanks) {
|
|
51
54
|
const modelParams = await getModelParamsAsync(encodingName, getMergeableRanks);
|
|
@@ -53,33 +56,21 @@ class GptEncoding {
|
|
|
53
56
|
}
|
|
54
57
|
static async getEncodingApiForModelAsync(modelName, getMergeableRanks) {
|
|
55
58
|
const encodingName = modelToEncodingMap[modelName];
|
|
56
|
-
|
|
57
|
-
|
|
58
|
-
static getMaxValueFromMap(map) {
|
|
59
|
-
let max = 0;
|
|
60
|
-
map.forEach((val) => {
|
|
61
|
-
max = Math.max(max, val);
|
|
62
|
-
});
|
|
63
|
-
return max;
|
|
64
|
-
}
|
|
65
|
-
static specialTokenRegex(tokens) {
|
|
66
|
-
const escapedTokens = [...tokens].map(escapeRegExp);
|
|
67
|
-
const inner = escapedTokens.join('|');
|
|
68
|
-
return new RegExp(`(${inner})`);
|
|
59
|
+
const modelParams = await getModelParamsAsync(encodingName, getMergeableRanks);
|
|
60
|
+
return new GptEncoding({ ...modelParams, modelName });
|
|
69
61
|
}
|
|
70
|
-
encodeGenerator(lineToEncode, allowedSpecial = new Set(), disallowedSpecial = new Set([
|
|
71
|
-
DISALLOW_ALL_SPECIAL_TOKENS,
|
|
72
|
-
])) {
|
|
62
|
+
encodeGenerator(lineToEncode, { allowedSpecial = new Set(), disallowedSpecial = new Set([ALL_SPECIAL_TOKENS]), } = {}) {
|
|
73
63
|
const specialTokensSet = new Set(this.specialTokenMapping.keys());
|
|
74
|
-
if (disallowedSpecial.has(
|
|
64
|
+
if (disallowedSpecial.has(ALL_SPECIAL_TOKENS)) {
|
|
75
65
|
disallowedSpecial = new Set(specialTokensSet);
|
|
66
|
+
allowedSpecial.forEach((val) => disallowedSpecial.delete(val));
|
|
76
67
|
disallowedSpecial.forEach((val) => allowedSpecial.delete(val));
|
|
77
68
|
}
|
|
78
|
-
if (allowedSpecial.has(
|
|
69
|
+
if (allowedSpecial.has(ALL_SPECIAL_TOKENS)) {
|
|
79
70
|
allowedSpecial = specialTokensSet;
|
|
80
71
|
}
|
|
81
72
|
if (disallowedSpecial.size > 0) {
|
|
82
|
-
const regexPattern =
|
|
73
|
+
const regexPattern = getSpecialTokenRegex(disallowedSpecial);
|
|
83
74
|
const match = lineToEncode.match(regexPattern);
|
|
84
75
|
if (match !== null) {
|
|
85
76
|
throw new Error(`Disallowed special token found: ${match[0]}`);
|
|
@@ -87,18 +78,72 @@ class GptEncoding {
|
|
|
87
78
|
}
|
|
88
79
|
return this.bytePairEncodingCoreProcessor.encodeNative(lineToEncode, allowedSpecial);
|
|
89
80
|
}
|
|
90
|
-
encode(lineToEncode,
|
|
91
|
-
|
|
92
|
-
|
|
93
|
-
|
|
94
|
-
|
|
95
|
-
|
|
81
|
+
encode(lineToEncode, encodeOptions = {}) {
|
|
82
|
+
return [...this.encodeGenerator(lineToEncode, encodeOptions)].flat();
|
|
83
|
+
}
|
|
84
|
+
/**
|
|
85
|
+
* Progressively tokenizes an OpenAI chat.
|
|
86
|
+
* Warning: gpt-3.5-turbo and gpt-4 chat format may change over time.
|
|
87
|
+
* Returns tokens assuming the 'gpt-3.5-turbo-0301' / 'gpt-4-0314' format.
|
|
88
|
+
* Based on OpenAI's guidelines: https://github.com/openai/openai-python/blob/main/chatml.md
|
|
89
|
+
* Also mentioned in section 6 of this document: https://github.com/openai/openai-cookbook/blob/main/examples/How_to_count_tokens_with_tiktoken.ipynb
|
|
90
|
+
*/
|
|
91
|
+
*encodeChatGenerator(chat, model = this.modelName) {
|
|
92
|
+
if (!model) {
|
|
93
|
+
throw new Error('Model name must be provided either during initialization or passed in to the method.');
|
|
94
|
+
}
|
|
95
|
+
const params = chatModelParams[model];
|
|
96
|
+
const chatStartToken = this.specialTokenMapping.get(ImStart);
|
|
97
|
+
const chatEndToken = this.specialTokenMapping.get(ImEnd);
|
|
98
|
+
if (!params || chatStartToken === undefined || chatEndToken === undefined) {
|
|
99
|
+
throw new Error(`Model '${model}' does not support chat.`);
|
|
100
|
+
}
|
|
101
|
+
const allowedSpecial = new Set([ImSep]);
|
|
102
|
+
const { messageSeparator, roleSeparator } = params;
|
|
103
|
+
const encodedMessageSeparator = messageSeparator.length > 0 ? this.encode(messageSeparator) : [];
|
|
104
|
+
const encodedRoleSeparator = roleSeparator.length > 0
|
|
105
|
+
? this.encode(roleSeparator, { allowedSpecial })
|
|
106
|
+
: [];
|
|
107
|
+
const nameCache = new Map();
|
|
108
|
+
for (const { role = 'system', name = role, content } of chat) {
|
|
109
|
+
if (content === undefined) {
|
|
110
|
+
throw new Error('Content must be defined for all messages.');
|
|
111
|
+
}
|
|
112
|
+
yield [chatStartToken];
|
|
113
|
+
const encodedName = nameCache.get(name) ?? this.encode(name);
|
|
114
|
+
nameCache.set(name, encodedName);
|
|
115
|
+
yield encodedName;
|
|
116
|
+
if (encodedRoleSeparator.length > 0) {
|
|
117
|
+
yield encodedRoleSeparator;
|
|
118
|
+
}
|
|
119
|
+
yield* this.encodeGenerator(content);
|
|
120
|
+
yield [chatEndToken];
|
|
121
|
+
yield encodedMessageSeparator;
|
|
122
|
+
}
|
|
123
|
+
// every reply is primed with <|start|>assistant<|message|>
|
|
124
|
+
yield [chatStartToken];
|
|
125
|
+
yield* this.encodeGenerator('assistant');
|
|
126
|
+
if (encodedRoleSeparator.length > 0) {
|
|
127
|
+
yield encodedRoleSeparator;
|
|
128
|
+
}
|
|
129
|
+
}
|
|
130
|
+
/**
|
|
131
|
+
* Encodes a chat into a single array of tokens.
|
|
132
|
+
* Warning: gpt-3.5-turbo and gpt-4 chat format may change over time.
|
|
133
|
+
* Returns tokens assuming the 'gpt-3.5-turbo-0301' / 'gpt-4-0314' format.
|
|
134
|
+
* Based on OpenAI's guidelines: https://github.com/openai/openai-python/blob/main/chatml.md
|
|
135
|
+
* Also mentioned in section 6 of this document: https://github.com/openai/openai-cookbook/blob/main/examples/How_to_count_tokens_with_tiktoken.ipynb
|
|
136
|
+
*/
|
|
137
|
+
encodeChat(chat, model = this.modelName) {
|
|
138
|
+
return [...this.encodeChatGenerator(chat, model)].flat();
|
|
96
139
|
}
|
|
97
140
|
/**
|
|
98
141
|
* @returns {false | number} false if token limit is exceeded, otherwise the number of tokens
|
|
99
142
|
*/
|
|
100
|
-
isWithinTokenLimit(
|
|
101
|
-
const tokenGenerator =
|
|
143
|
+
isWithinTokenLimit(input, tokenLimit) {
|
|
144
|
+
const tokenGenerator = typeof input === 'string'
|
|
145
|
+
? this.encodeGenerator(input)
|
|
146
|
+
: this.encodeChatGenerator(input);
|
|
102
147
|
let count = 0;
|
|
103
148
|
for (const tokens of tokenGenerator) {
|
|
104
149
|
count += tokens.length;
|
package/esm/GptEncoding.js.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"GptEncoding.js","sourceRoot":"","sources":["../src/GptEncoding.ts"],"names":[],"mappings":"AAAA,sCAAsC;AACtC,OAAO,EAAE,oBAAoB,EAAE,MAAM,2BAA2B,CAAA;AAChE,OAAO,
|
|
1
|
+
{"version":3,"file":"GptEncoding.js","sourceRoot":"","sources":["../src/GptEncoding.ts"],"names":[],"mappings":"AAAA,sCAAsC;AACtC,OAAO,EAAE,oBAAoB,EAAE,MAAM,2BAA2B,CAAA;AAChE,OAAO,EAGL,eAAe,EACf,kBAAkB,GACnB,MAAM,cAAc,CAAA;AACrB,OAAO,EAIL,iBAAiB,EACjB,mBAAmB,GACpB,MAAM,kBAAkB,CAAA;AACzB,OAAO,EACL,WAAW,EACX,SAAS,EACT,SAAS,EACT,SAAS,EACT,SAAS,EACT,KAAK,EACL,KAAK,EACL,OAAO,GACR,MAAM,oBAAoB,CAAA;AAC3B,OAAO,EAAE,kCAAkC,EAAE,MAAM,cAAc,CAAA;AACjE,OAAO,EAAE,kBAAkB,EAAE,oBAAoB,EAAE,MAAM,WAAW,CAAA;AAEpE,MAAM,CAAC,MAAM,kBAAkB,GAAG,KAAK,CAAA;AAiBvC,MAAa,WAAW;IACtB,MAAM,CAAC,WAAW,GAAG,WAAW,CAAA;IAChC,MAAM,CAAC,SAAS,GAAG,SAAS,CAAA;IAC5B,MAAM,CAAC,SAAS,GAAG,SAAS,CAAA;IAC5B,MAAM,CAAC,SAAS,GAAG,SAAS,CAAA;IAC5B,MAAM,CAAC,SAAS,GAAG,SAAS,CAAA;IAE5B,OAAO,GAAG,IAAI,WAAW,CAAC,MAAM,CAAC,CAAA;IACjC,SAAS,CAAY;IACb,6BAA6B,CAAsB;IACnD,mBAAmB,CAAqB;IAEhD,YAAoB,EAClB,eAAe,EACf,sBAAsB,EACtB,mBAAmB,EACnB,sBAAsB,EACtB,SAAS,GACM;QACf,MAAM,aAAa,GAAG,IAAI,CAAC,GAAG,CAC5B,kBAAkB,CAAC,sBAAsB,CAAC,EAC1C,kBAAkB,CAAC,mBAAmB,CAAC,CACxC,CAAA;QACD,IAAI,CAAC,mBAAmB,GAAG,mBAAmB,CAAA;QAE9C,IAAI,sBAAsB,KAAK,SAAS,EAAE;YACxC,IACE,sBAAsB,CAAC,IAAI,GAAG,mBAAmB,CAAC,IAAI;gBACtD,sBAAsB,EACtB;gBACA,MAAM,IAAI,KAAK,CACb,sFAAsF,CACvF,CAAA;aACF;YAED,IAAI,aAAa,KAAK,sBAAsB,GAAG,CAAC,EAAE;gBAChD,MAAM,IAAI,KAAK,CACb,gEAAgE,CACjE,CAAA;aACF;SACF;QAED,IAAI,CAAC,6BAA6B,GAAG,IAAI,oBAAoB,CAAC;YAC5D,eAAe,EAAE,sBAAsB;YACvC,mBAAmB,EAAE,mBAAmB;YACxC,eAAe;SAChB,CAAC,CAAA;QAEF,IAAI,CAAC,MAAM,GAAG,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,IAAI,CAAC,CAAA;QACpC,IAAI,CAAC,MAAM,GAAG,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,IAAI,CAAC,CAAA;QACpC,IAAI,CAAC,eAAe,GAAG,IAAI,CAAC,eAAe,CAAC,IAAI,CAAC,IAAI,CAAC,CAAA;QACtD,IAAI,CAAC,eAAe,GAAG,IAAI,CAAC,eAAe,CAAC,IAAI,CAAC,IAAI,CAAC,CAAA;QACtD,IAAI,CAAC,oBAAoB,GAAG,IAAI,CAAC,oBAAoB,CAAC,IAAI,CAAC,IAAI,CAAC,CAAA;QAChE,IAAI,CAAC,kBAAkB,GAAG,IAAI,CAAC,kBAAkB,CAAC,IAAI,CAAC,IAAI,CAAC,CAAA;QAC5D,IAAI,CAAC,SAAS,GAAG,SAAS,CAAA;IAC5B,CAAC;IAED,MAAM,CAAC,cAAc,CACnB,YAA0B,EAC1B,iBAAsC;QAEtC,MAAM,WAAW,GAAG,iBAAiB,CAAC,YAAY,EAAE,iBAAiB,CAAC,CAAA;QACtE,OAAO,IAAI,WAAW,CAAC,WAAW,CAAC,CAAA;IACrC,CAAC;IAED,MAAM,CAAC,sBAAsB,CAC3B,SAAoB,EACpB,iBAAsC;QAEtC,MAAM,YAAY,GAAG,kBAAkB,CAAC,SAAS,CAAC,CAAA;QAClD,MAAM,WAAW,GAAG,iBAAiB,CAAC,YAAY,EAAE,iBAAiB,CAAC,CAAA;QACtE,OAAO,IAAI,WAAW,CAAC,EAAE,GAAG,WAAW,EAAE,SAAS,EAAE,CAAC,CAAA;IACvD,CAAC;IAED,MAAM,CAAC,KAAK,CAAC,mBAAmB,CAC9B,YAA0B,EAC1B,iBAA2C;QAE3C,MAAM,WAAW,GAAG,MAAM,mBAAmB,CAC3C,YAAY,EACZ,iBAAiB,CAClB,CAAA;QACD,OAAO,IAAI,WAAW,CAAC,WAAW,CAAC,CAAA;IACrC,CAAC;IAED,MAAM,CAAC,KAAK,CAAC,2BAA2B,CACtC,SAAoB,EACpB,iBAA2C;QAE3C,MAAM,YAAY,GAAG,kBAAkB,CAAC,SAAS,CAAC,CAAA;QAClD,MAAM,WAAW,GAAG,MAAM,mBAAmB,CAC3C,YAAY,EACZ,iBAAiB,CAClB,CAAA;QACD,OAAO,IAAI,WAAW,CAAC,EAAE,GAAG,WAAW,EAAE,SAAS,EAAE,CAAC,CAAA;IACvD,CAAC;IAED,eAAe,CACb,YAAoB,EACpB,EACE,cAAc,GAAG,IAAI,GAAG,EAAU,EAClC,iBAAiB,GAAG,IAAI,GAAG,CAAS,CAAC,kBAAkB,CAAC,CAAC,MACxC,EAAE;QAErB,MAAM,gBAAgB,GAAG,IAAI,GAAG,CAAS,IAAI,CAAC,mBAAmB,CAAC,IAAI,EAAE,CAAC,CAAA;QAEzE,IAAI,iBAAiB,CAAC,GAAG,CAAC,kBAAkB,CAAC,EAAE;YAC7C,iBAAiB,GAAG,IAAI,GAAG,CAAS,gBAAgB,CAAC,CAAA;YACrD,cAAc,CAAC,OAAO,CAAC,CAAC,GAAG,EAAE,EAAE,CAAC,iBAAiB,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC,CAAA;YAC9D,iBAAiB,CAAC,OAAO,CAAC,CAAC,GAAG,EAAE,EAAE,CAAC,cAAc,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC,CAAA;SAC/D;QAED,IAAI,cAAc,CAAC,GAAG,CAAC,kBAAkB,CAAC,EAAE;YAC1C,cAAc,GAAG,gBAAgB,CAAA;SAClC;QAED,IAAI,iBAAiB,CAAC,IAAI,GAAG,CAAC,EAAE;YAC9B,MAAM,YAAY,GAAG,oBAAoB,CAAC,iBAAiB,CAAC,CAAA;YAC5D,MAAM,KAAK,GAAG,YAAY,CAAC,KAAK,CAAC,YAAY,CAAC,CAAA;YAC9C,IAAI,KAAK,KAAK,IAAI,EAAE;gBAClB,MAAM,IAAI,KAAK,CAAC,mCAAmC,KAAK,CAAC,CAAC,CAAC,EAAE,CAAC,CAAA;aAC/D;SACF;QAED,OAAO,IAAI,CAAC,6BAA6B,CAAC,YAAY,CACpD,YAAY,EACZ,cAAc,CACf,CAAA;IACH,CAAC;IAED,MAAM,CAAC,YAAoB,EAAE,gBAA+B,EAAE;QAC5D,OAAO,CAAC,GAAG,IAAI,CAAC,eAAe,CAAC,YAAY,EAAE,aAAa,CAAC,CAAC,CAAC,IAAI,EAAE,CAAA;IACtE,CAAC;IAED;;;;;;OAMG;IACH,CAAC,mBAAmB,CAClB,IAA2B,EAC3B,KAAK,GAAG,IAAI,CAAC,SAAS;QAEtB,IAAI,CAAC,KAAK,EAAE;YACV,MAAM,IAAI,KAAK,CACb,sFAAsF,CACvF,CAAA;SACF;QACD,MAAM,MAAM,GAAG,eAAe,CAAC,KAAK,CAAC,CAAA;QACrC,MAAM,cAAc,GAAG,IAAI,CAAC,mBAAmB,CAAC,GAAG,CAAC,OAAO,CAAC,CAAA;QAC5D,MAAM,YAAY,GAAG,IAAI,CAAC,mBAAmB,CAAC,GAAG,CAAC,KAAK,CAAC,CAAA;QAExD,IAAI,CAAC,MAAM,IAAI,cAAc,KAAK,SAAS,IAAI,YAAY,KAAK,SAAS,EAAE;YACzE,MAAM,IAAI,KAAK,CAAC,UAAU,KAAK,0BAA0B,CAAC,CAAA;SAC3D;QACD,MAAM,cAAc,GAAG,IAAI,GAAG,CAAC,CAAC,KAAK,CAAC,CAAC,CAAA;QACvC,MAAM,EAAE,gBAAgB,EAAE,aAAa,EAAE,GAAG,MAAM,CAAA;QAClD,MAAM,uBAAuB,GAC3B,gBAAgB,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC,MAAM,CAAC,gBAAgB,CAAC,CAAC,CAAC,CAAC,EAAE,CAAA;QAClE,MAAM,oBAAoB,GACxB,aAAa,CAAC,MAAM,GAAG,CAAC;YACtB,CAAC,CAAC,IAAI,CAAC,MAAM,CAAC,aAAa,EAAE,EAAE,cAAc,EAAE,CAAC;YAChD,CAAC,CAAC,EAAE,CAAA;QACR,MAAM,SAAS,GAAG,IAAI,GAAG,EAAoB,CAAA;QAE7C,KAAK,MAAM,EAAE,IAAI,GAAG,QAAQ,EAAE,IAAI,GAAG,IAAI,EAAE,OAAO,EAAE,IAAI,IAAI,EAAE;YAC5D,IAAI,OAAO,KAAK,SAAS,EAAE;gBACzB,MAAM,IAAI,KAAK,CAAC,2CAA2C,CAAC,CAAA;aAC7D;YAED,MAAM,CAAC,cAAc,CAAC,CAAA;YACtB,MAAM,WAAW,GAAG,SAAS,CAAC,GAAG,CAAC,IAAI,CAAC,IAAI,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,CAAA;YAC5D,SAAS,CAAC,GAAG,CAAC,IAAI,EAAE,WAAW,CAAC,CAAA;YAChC,MAAM,WAAW,CAAA;YACjB,IAAI,oBAAoB,CAAC,MAAM,GAAG,CAAC,EAAE;gBACnC,MAAM,oBAAoB,CAAA;aAC3B;YACD,KAAK,CAAC,CAAC,IAAI,CAAC,eAAe,CAAC,OAAO,CAAC,CAAA;YACpC,MAAM,CAAC,YAAY,CAAC,CAAA;YACpB,MAAM,uBAAuB,CAAA;SAC9B;QAED,2DAA2D;QAC3D,MAAM,CAAC,cAAc,CAAC,CAAA;QACtB,KAAK,CAAC,CAAC,IAAI,CAAC,eAAe,CAAC,WAAW,CAAC,CAAA;QACxC,IAAI,oBAAoB,CAAC,MAAM,GAAG,CAAC,EAAE;YACnC,MAAM,oBAAoB,CAAA;SAC3B;IACH,CAAC;IAED;;;;;;OAMG;IACH,UAAU,CAAC,IAA4B,EAAE,KAAK,GAAG,IAAI,CAAC,SAAS;QAC7D,OAAO,CAAC,GAAG,IAAI,CAAC,mBAAmB,CAAC,IAAI,EAAE,KAAK,CAAC,CAAC,CAAC,IAAI,EAAE,CAAA;IAC1D,CAAC;IAED;;OAEG;IACH,kBAAkB,CAChB,KAAqC,EACrC,UAAkB;QAElB,MAAM,cAAc,GAClB,OAAO,KAAK,KAAK,QAAQ;YACvB,CAAC,CAAC,IAAI,CAAC,eAAe,CAAC,KAAK,CAAC;YAC7B,CAAC,CAAC,IAAI,CAAC,mBAAmB,CAAC,KAAK,CAAC,CAAA;QACrC,IAAI,KAAK,GAAG,CAAC,CAAA;QACb,KAAK,MAAM,MAAM,IAAI,cAAc,EAAE;YACnC,KAAK,IAAI,MAAM,CAAC,MAAM,CAAA;YACtB,IAAI,KAAK,GAAG,UAAU,EAAE;gBACtB,OAAO,KAAK,CAAA;aACb;SACF;QACD,OAAO,KAAK,CAAA;IACd,CAAC;IAED,CAAC,eAAe,CACd,mBAAqC;QAErC,MAAM,oBAAoB,GACxB,IAAI,CAAC,6BAA6B,CAAC,YAAY,CAAC,mBAAmB,CAAC,CAAA;QAEtE,IAAI,MAAM,GAAG,EAAE,CAAA;QAEf,KAAK,MAAM,WAAW,IAAI,oBAAoB,EAAE;YAC9C,MAAM,IAAI,IAAI,CAAC,OAAO,CAAC,MAAM,CAAC,WAAW,EAAE,EAAE,MAAM,EAAE,IAAI,EAAE,CAAC,CAAA;YAE5D,IAAI,MAAM,CAAC,MAAM,KAAK,CAAC,IAAI,kCAAkC,CAAC,MAAM,CAAC,EAAE;gBACrE,yEAAyE;gBACzE,uCAAuC;gBACvC,SAAQ;aACT;iBAAM;gBACL,MAAM,MAAM,CAAA;gBACZ,eAAe;gBACf,MAAM,GAAG,EAAE,CAAA;aACZ;SACF;QAED,+CAA+C;QAC/C,IAAI,MAAM,CAAC,MAAM,GAAG,CAAC,EAAE;YACrB,MAAM,MAAM,CAAA;SACb;IACH,CAAC;IAED,KAAK,CAAC,CAAC,oBAAoB,CACzB,mBAA0C;QAE1C,MAAM,oBAAoB,GACxB,IAAI,CAAC,6BAA6B,CAAC,iBAAiB,CAAC,mBAAmB,CAAC,CAAA;QAE3E,IAAI,MAAM,GAAG,EAAE,CAAA;QAEf,IAAI,KAAK,EAAE,MAAM,WAAW,IAAI,oBAAoB,EAAE;YACpD,MAAM,IAAI,IAAI,CAAC,OAAO,CAAC,MAAM,CAAC,WAAW,EAAE,EAAE,MAAM,EAAE,IAAI,EAAE,CAAC,CAAA;YAE5D,IAAI,MAAM,CAAC,MAAM,KAAK,CAAC,IAAI,kCAAkC,CAAC,MAAM,CAAC,EAAE;gBACrE,yEAAyE;gBACzE,uCAAuC;gBACvC,SAAQ;aACT;iBAAM;gBACL,MAAM,MAAM,CAAA;gBACZ,eAAe;gBACf,MAAM,GAAG,EAAE,CAAA;aACZ;SACF;QAED,+CAA+C;QAC/C,IAAI,MAAM,CAAC,MAAM,GAAG,CAAC,EAAE;YACrB,MAAM,MAAM,CAAA;SACb;IACH,CAAC;IAED,MAAM,CAAC,mBAAqC;QAC1C,OAAO,CAAC,GAAG,IAAI,CAAC,eAAe,CAAC,mBAAmB,CAAC,CAAC,CAAC,IAAI,CAAC,EAAE,CAAC,CAAA;IAChE,CAAC;;SA1RU,WAAW"}
|
package/esm/GptEncoding.test.js
CHANGED
|
@@ -1,7 +1,7 @@
|
|
|
1
1
|
import fs from 'fs';
|
|
2
2
|
import path from 'path';
|
|
3
3
|
import { GptEncoding } from './GptEncoding.js';
|
|
4
|
-
import { encodingNames } from './mapping.js';
|
|
4
|
+
import { chatModelParams, encodingNames, } from './mapping.js';
|
|
5
5
|
import { resolveEncoding } from './resolveEncoding.js';
|
|
6
6
|
const sharedResults = {
|
|
7
7
|
space: [220],
|
|
@@ -129,6 +129,60 @@ describe.each(encodingNames)('%s', (encodingName) => {
|
|
|
129
129
|
});
|
|
130
130
|
});
|
|
131
131
|
});
|
|
132
|
+
const chatModelNames = Object.keys(chatModelParams);
|
|
133
|
+
const exampleMessages = [
|
|
134
|
+
{
|
|
135
|
+
role: 'system',
|
|
136
|
+
content: 'You are a helpful, pattern-following assistant that translates corporate jargon into plain English.',
|
|
137
|
+
},
|
|
138
|
+
{
|
|
139
|
+
role: 'system',
|
|
140
|
+
name: 'example_user',
|
|
141
|
+
content: 'New synergies will help drive top-line growth.',
|
|
142
|
+
},
|
|
143
|
+
{
|
|
144
|
+
role: 'system',
|
|
145
|
+
name: 'example_assistant',
|
|
146
|
+
content: 'Things working well together will increase revenue.',
|
|
147
|
+
},
|
|
148
|
+
{
|
|
149
|
+
role: 'system',
|
|
150
|
+
name: 'example_user',
|
|
151
|
+
content: "Let's circle back when we have more bandwidth to touch base on opportunities for increased leverage.",
|
|
152
|
+
},
|
|
153
|
+
{
|
|
154
|
+
role: 'system',
|
|
155
|
+
name: 'example_assistant',
|
|
156
|
+
content: "Let's talk later when we're less busy about how to do better.",
|
|
157
|
+
},
|
|
158
|
+
{
|
|
159
|
+
role: 'user',
|
|
160
|
+
content: "This late pivot means we don't have time to boil the ocean for the client deliverable.",
|
|
161
|
+
},
|
|
162
|
+
];
|
|
163
|
+
describe.each(chatModelNames)('%s', (modelName) => {
|
|
164
|
+
const encoding = GptEncoding.getEncodingApiForModel(modelName, resolveEncoding);
|
|
165
|
+
const expectedEncodedLength = modelName.startsWith('gpt-3.5-turbo')
|
|
166
|
+
? 127
|
|
167
|
+
: 121;
|
|
168
|
+
describe('chat functionality', () => {
|
|
169
|
+
it('encodes a chat correctly', () => {
|
|
170
|
+
const encoded = encoding.encodeChat(exampleMessages);
|
|
171
|
+
expect(encoded).toMatchSnapshot();
|
|
172
|
+
expect(encoded).toHaveLength(expectedEncodedLength);
|
|
173
|
+
const decoded = encoding.decode(encoded);
|
|
174
|
+
expect(decoded).toMatchSnapshot();
|
|
175
|
+
});
|
|
176
|
+
it('isWithinTokenLimit: false', () => {
|
|
177
|
+
const isWithinTokenLimit = encoding.isWithinTokenLimit(exampleMessages, 50);
|
|
178
|
+
expect(isWithinTokenLimit).toBe(false);
|
|
179
|
+
});
|
|
180
|
+
it('isWithinTokenLimit: true (number)', () => {
|
|
181
|
+
const isWithinTokenLimit = encoding.isWithinTokenLimit(exampleMessages, 150);
|
|
182
|
+
expect(isWithinTokenLimit).toBe(expectedEncodedLength);
|
|
183
|
+
});
|
|
184
|
+
});
|
|
185
|
+
});
|
|
132
186
|
function loadTestPlans() {
|
|
133
187
|
const testPlanPath = path.join(__dirname, '../data/TestPlans.txt');
|
|
134
188
|
const testPlanData = fs.readFileSync(testPlanPath, 'utf8');
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"GptEncoding.test.js","sourceRoot":"","sources":["../src/GptEncoding.test.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,MAAM,IAAI,CAAA;AACnB,OAAO,IAAI,MAAM,MAAM,CAAA;AACvB,OAAO,EAAE,WAAW,EAAE,MAAM,kBAAkB,CAAA;AAC9C,OAAO,
|
|
1
|
+
{"version":3,"file":"GptEncoding.test.js","sourceRoot":"","sources":["../src/GptEncoding.test.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,MAAM,IAAI,CAAA;AACnB,OAAO,IAAI,MAAM,MAAM,CAAA;AACvB,OAAO,EAAE,WAAW,EAAE,MAAM,kBAAkB,CAAA;AAC9C,OAAO,EAGL,eAAe,EACf,aAAa,GACd,MAAM,cAAc,CAAA;AACrB,OAAO,EAAE,eAAe,EAAE,MAAM,sBAAsB,CAAA;AAEtD,MAAM,aAAa,GAAG;IACpB,KAAK,EAAE,CAAC,GAAG,CAAC;IACZ,GAAG,EAAE,CAAC,GAAG,CAAC;IACV,mBAAmB,EAAE,CAAC,IAAK,EAAE,GAAG,EAAE,GAAG,EAAE,IAAK,CAAC;IAC7C,WAAW,EAAE,CAAC,GAAG,EAAE,GAAG,EAAE,KAAM,CAAC;IAC/B,mBAAmB,EAAE,CAAC,KAAM,EAAE,KAAM,EAAE,GAAG,EAAE,GAAG,EAAE,KAAM,EAAE,GAAG,EAAE,GAAG,CAAC;IACjE,uBAAuB,EAAE,CAAC,OAAO,EAAE,GAAG,EAAE,IAAI,EAAE,QAAQ,EAAE,GAAG,EAAE,IAAI,CAAC;IAClE,6CAA6C,EAAE;QAC7C,IAAK,EAAE,KAAM,EAAE,KAAM,EAAE,GAAG,EAAE,KAAM,EAAE,KAAM,EAAE,IAAK,EAAE,IAAK;KACzD;IACD,8CAA8C,EAAE;QAC9C,KAAM,EAAE,EAAE,EAAE,GAAG,EAAE,GAAG,EAAE,GAAG,EAAE,IAAK,EAAE,EAAE,EAAE,GAAG,EAAE,GAAG,EAAE,GAAG,EAAE,GAAG,EAAE,GAAG,EAAE,EAAE,EAAE,GAAG,EAAE,EAAE;QAC1E,EAAE,EAAE,EAAE,EAAE,EAAE;KACX;CACF,CAAA;AAED,MAAM,OAAO,GAAG;IACd,WAAW,EAAE;QACX,KAAK,EAAE,CAAC,GAAG,CAAC;QACZ,GAAG,EAAE,CAAC,GAAG,CAAC;QACV,mBAAmB,EAAE,CAAC,IAAK,EAAE,GAAG,EAAE,IAAK,EAAE,IAAK,CAAC;QAC/C,WAAW,EAAE,CAAC,GAAG,EAAE,GAAG,EAAE,KAAM,CAAC;QAC/B,mBAAmB,EAAE,CAAC,KAAM,EAAE,KAAM,EAAE,GAAG,EAAE,IAAK,EAAE,KAAM,EAAE,GAAG,EAAE,GAAG,CAAC;QACnE,uBAAuB,EAAE,CAAC,OAAO,EAAE,GAAG,EAAE,IAAI,EAAE,QAAQ,EAAE,GAAG,EAAE,IAAI,CAAC;QAClE,6CAA6C,EAAE;YAC7C,IAAK,EAAE,IAAK,EAAE,GAAG,EAAE,KAAM,EAAE,GAAG,EAAE,IAAK;SACtC;QACD,8CAA8C,EAAE;YAC9C,KAAM,EAAE,EAAE,EAAE,GAAG,EAAE,IAAK,EAAE,GAAG,EAAE,IAAK,EAAE,EAAE,EAAE,GAAG,EAAE,GAAG,EAAE,GAAG,EAAE,KAAM,EAAE,EAAE,EAAE,GAAG;YACtE,IAAK,EAAE,KAAM;SACd;KACF;IACD,SAAS,EAAE,aAAa;IACxB,SAAS,EAAE,aAAa;IACxB,SAAS,EAAE,aAAa;CACe,CAAA;AAEzC,mEAAmE;AACnE,MAAM,SAAS,GAAG,aAAa,EAAE,CAAA;AAEjC,QAAQ,CAAC,IAAI,CAAC,aAAa,CAAC,CAAC,IAAI,EAAE,CAAC,YAA0B,EAAE,EAAE;IAChE,MAAM,QAAQ,GAAG,WAAW,CAAC,cAAc,CAAC,YAAY,EAAE,eAAe,CAAC,CAAA;IAE1E,MAAM,EACJ,MAAM,EACN,eAAe,EACf,oBAAoB,EACpB,MAAM,EACN,kBAAkB,GACnB,GAAG,QAAQ,CAAA;IAEZ,MAAM,MAAM,GAAG,OAAO,CAAC,YAAY,CAAC,CAAA;IAEpC,QAAQ,CAAC,qBAAqB,EAAE,GAAG,EAAE;QACnC,EAAE,CAAC,cAAc,EAAE,GAAG,EAAE;YACtB,MAAM,GAAG,GAAG,EAAE,CAAA;YACd,MAAM,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC,CAAC,OAAO,CAAC,EAAE,CAAC,CAAA;YAC/B,MAAM,CAAC,MAAM,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,OAAO,CAAC,GAAG,CAAC,CAAA;YACxC,MAAM,CAAC,kBAAkB,CAAC,GAAG,EAAE,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,CAAC,CAAA;YAC1C,MAAM,CAAC,kBAAkB,CAAC,GAAG,EAAE,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,CAAC,CAAA;QAC5C,CAAC,CAAC,CAAA;QAEF,EAAE,CAAC,OAAO,EAAE,GAAG,EAAE;YACf,MAAM,GAAG,GAAG,GAAG,CAAA;YACf,MAAM,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC,CAAC,OAAO,CAAC,MAAM,CAAC,KAAK,CAAC,CAAA;YACzC,MAAM,CAAC,MAAM,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,OAAO,CAAC,GAAG,CAAC,CAAA;YACxC,MAAM,CAAC,kBAAkB,CAAC,GAAG,EAAE,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,CAAC,CAAA;YAC1C,MAAM,CAAC,kBAAkB,CAAC,GAAG,EAAE,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC,KAAK,CAAC,CAAA;QAChD,CAAC,CAAC,CAAA;QAEF,EAAE,CAAC,KAAK,EAAE,GAAG,EAAE;YACb,MAAM,GAAG,GAAG,IAAI,CAAA;YAChB,MAAM,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC,CAAC,OAAO,CAAC,MAAM,CAAC,GAAG,CAAC,CAAA;YACvC,MAAM,CAAC,MAAM,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,OAAO,CAAC,GAAG,CAAC,CAAA;QAC1C,CAAC,CAAC,CAAA;QAEF,EAAE,CAAC,aAAa,EAAE,GAAG,EAAE;YACrB,MAAM,GAAG,GAAG,mBAAmB,CAAA;YAC/B,MAAM,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC,CAAC,OAAO,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC,CAAA;YACxC,MAAM,CAAC,MAAM,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,OAAO,CAAC,GAAG,CAAC,CAAA;YACxC,MAAM,CAAC,kBAAkB,CAAC,GAAG,EAAE,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC,KAAK,CAAC,CAAA;YAC9C,MAAM,CAAC,kBAAkB,CAAC,GAAG,EAAE,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC,MAAM,CAAC,CAAA;QAC7D,CAAC,CAAC,CAAA;QAEF,EAAE,CAAC,kBAAkB,EAAE,GAAG,EAAE;YAC1B,MAAM,GAAG,GAAG,aAAa,CAAA;YACzB,MAAM,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC,CAAC,OAAO,CAAC,MAAM,CAAC,WAAW,CAAC,CAAA;YAC/C,MAAM,CAAC,MAAM,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,OAAO,CAAC,GAAG,CAAC,CAAA;YACxC,MAAM,CAAC,kBAAkB,CAAC,GAAG,EAAE,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC,MAAM,CAAC,WAAW,CAAC,MAAM,CAAC,CAAA;QACpE,CAAC,CAAC,CAAA;QAEF,EAAE,CAAC,QAAQ,EAAE,GAAG,EAAE;YAChB,MAAM,GAAG,GAAG,mBAAmB,CAAA;YAC/B,MAAM,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC,CAAC,OAAO,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC,CAAA;YACxC,MAAM,CAAC,MAAM,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,OAAO,CAAC,GAAG,CAAC,CAAA;YACxC,MAAM,CAAC,kBAAkB,CAAC,GAAG,EAAE,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC,KAAK,CAAC,CAAA;YAC9C,MAAM,CAAC,kBAAkB,CAAC,GAAG,EAAE,GAAG,CAAC,CAAC,CAAC,IAAI,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC,MAAM,CAAC,CAAA;QAC/D,CAAC,CAAC,CAAA;QAEF,EAAE,CAAC,qCAAqC,EAAE,GAAG,EAAE;YAC7C,MAAM,GAAG,GAAG,mBAAmB,CAAA;YAC/B,MAAM,SAAS,GAAG,eAAe,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC,CAAA;YAC9C,MAAM,CAAC,uBAAuB,CAAC,OAAO,CAAC,CAAC,KAAK,EAAE,EAAE;gBAC/C,MAAM,CAAC,SAAS,CAAC,IAAI,EAAE,CAAC,KAAK,CAAC,CAAC,IAAI,CAAC,KAAK,CAAC,CAAA;YAC5C,CAAC,CAAC,CAAA;QACJ,CAAC,CAAC,CAAA;QAEF,KAAK,SAAS,CAAC,CAAC,wBAAwB;YACtC,MAAM,GAAG,GAAG,mBAAmB,CAAA;YAC/B,KAAK,MAAM,KAAK,IAAI,MAAM,CAAC,GAAG,CAAC,EAAE;gBAC/B,4CAA4C;gBAC5C,MAAM,MAAM,OAAO,CAAC,OAAO,CAAC,KAAK,CAAC,CAAA;aACnC;QACH,CAAC;QAED,EAAE,CAAC,2CAA2C,EAAE,KAAK,IAAI,EAAE;YACzD,MAAM,SAAS,GAAG,oBAAoB,CAAC,wBAAwB,EAAE,CAAC,CAAA;YAClE,MAAM,OAAO,GAAG,CAAC,GAAG,MAAM,CAAC,uBAAuB,CAAC,CAAA;YACnD,IAAI,KAAK,EAAE,MAAM,KAAK,IAAI,SAAS,EAAE;gBACnC,MAAM,CAAC,KAAK,CAAC,CAAC,OAAO,CAAC,OAAO,CAAC,KAAK,EAAE,CAAC,CAAA;aACvC;QACH,CAAC,CAAC,CAAA;QAEF,EAAE,CAAC,sBAAsB,EAAE,GAAG,EAAE;YAC9B,MAAM,GAAG,GAAG,6CAA6C,CAAA;YAEzD,MAAM,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC,CAAC,OAAO,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC,CAAA;YACxC,MAAM,CAAC,MAAM,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,OAAO,CAAC,GAAG,CAAC,CAAA;QAC1C,CAAC,CAAC,CAAA;QAEF,EAAE,CAAC,kBAAkB,EAAE,GAAG,EAAE;YAC1B,MAAM,GAAG,GAAG,8CAA8C,CAAA;YAC1D,MAAM,CAAC,MAAM,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,OAAO,CAAC,GAAG,CAAC,CAAA;YACxC,MAAM,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC,CAAC,aAAa,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC,CAAA;YAC9C,MAAM,CAAC,kBAAkB,CAAC,GAAG,EAAE,MAAM,CAAC,GAAG,CAAC,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC,KAAK,CAAC,CAAA;YACnE,MAAM,CAAC,kBAAkB,CAAC,GAAG,EAAE,GAAG,CAAC,CAAC,CAAC,IAAI,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC,MAAM,CAAC,CAAA;QAC/D,CAAC,CAAC,CAAA;IACJ,CAAC,CAAC,CAAA;IAEF,QAAQ,CAAC,WAAW,EAAE,GAAG,EAAE;QACzB,SAAS,CAAC,YAAY,CAAC,CAAC,OAAO,CAAC,CAAC,EAAE,MAAM,EAAE,OAAO,EAAE,EAAE,EAAE;YACtD,EAAE,CAAC,WAAW,MAAM,EAAE,EAAE,GAAG,EAAE;gBAC3B,MAAM,CAAC,MAAM,CAAC,MAAM,CAAC,CAAC,CAAC,OAAO,CAAC,OAAO,CAAC,CAAA;YACzC,CAAC,CAAC,CAAA;YACF,EAAE,CAAC,WAAW,MAAM,EAAE,EAAE,GAAG,EAAE;gBAC3B,MAAM,CAAC,MAAM,CAAC,OAAO,CAAC,CAAC,CAAC,OAAO,CAAC,MAAM,CAAC,CAAA;YACzC,CAAC,CAAC,CAAA;QACJ,CAAC,CAAC,CAAA;IACJ,CAAC,CAAC,CAAA;AACJ,CAAC,CAAC,CAAA;AAEF,MAAM,cAAc,GAAG,MAAM,CAAC,IAAI,CAAC,eAAe,CAA6B,CAAA;AAE/E,MAAM,eAAe,GAAG;IACtB;QACE,IAAI,EAAE,QAAQ;QACd,OAAO,EACL,qGAAqG;KACxG;IACD;QACE,IAAI,EAAE,QAAQ;QACd,IAAI,EAAE,cAAc;QACpB,OAAO,EAAE,gDAAgD;KAC1D;IACD;QACE,IAAI,EAAE,QAAQ;QACd,IAAI,EAAE,mBAAmB;QACzB,OAAO,EAAE,qDAAqD;KAC/D;IACD;QACE,IAAI,EAAE,QAAQ;QACd,IAAI,EAAE,cAAc;QACpB,OAAO,EACL,sGAAsG;KACzG;IACD;QACE,IAAI,EAAE,QAAQ;QACd,IAAI,EAAE,mBAAmB;QACzB,OAAO,EAAE,+DAA+D;KACzE;IACD;QACE,IAAI,EAAE,MAAM;QACZ,OAAO,EACL,wFAAwF;KAC3F;CACO,CAAA;AAEV,QAAQ,CAAC,IAAI,CAAC,cAAc,CAAC,CAAC,IAAI,EAAE,CAAC,SAAS,EAAE,EAAE;IAChD,MAAM,QAAQ,GAAG,WAAW,CAAC,sBAAsB,CACjD,SAAS,EACT,eAAe,CAChB,CAAA;IACD,MAAM,qBAAqB,GAAG,SAAS,CAAC,UAAU,CAAC,eAAe,CAAC;QACjE,CAAC,CAAC,GAAG;QACL,CAAC,CAAC,GAAG,CAAA;IAEP,QAAQ,CAAC,oBAAoB,EAAE,GAAG,EAAE;QAClC,EAAE,CAAC,0BAA0B,EAAE,GAAG,EAAE;YAClC,MAAM,OAAO,GAAG,QAAQ,CAAC,UAAU,CAAC,eAAe,CAAC,CAAA;YACpD,MAAM,CAAC,OAAO,CAAC,CAAC,eAAe,EAAE,CAAA;YACjC,MAAM,CAAC,OAAO,CAAC,CAAC,YAAY,CAAC,qBAAqB,CAAC,CAAA;YAEnD,MAAM,OAAO,GAAG,QAAQ,CAAC,MAAM,CAAC,OAAO,CAAC,CAAA;YACxC,MAAM,CAAC,OAAO,CAAC,CAAC,eAAe,EAAE,CAAA;QACnC,CAAC,CAAC,CAAA;QAEF,EAAE,CAAC,2BAA2B,EAAE,GAAG,EAAE;YACnC,MAAM,kBAAkB,GAAG,QAAQ,CAAC,kBAAkB,CACpD,eAAe,EACf,EAAE,CACH,CAAA;YACD,MAAM,CAAC,kBAAkB,CAAC,CAAC,IAAI,CAAC,KAAK,CAAC,CAAA;QACxC,CAAC,CAAC,CAAA;QACF,EAAE,CAAC,mCAAmC,EAAE,GAAG,EAAE;YAC3C,MAAM,kBAAkB,GAAG,QAAQ,CAAC,kBAAkB,CACpD,eAAe,EACf,GAAG,CACJ,CAAA;YACD,MAAM,CAAC,kBAAkB,CAAC,CAAC,IAAI,CAAC,qBAAqB,CAAC,CAAA;QACxD,CAAC,CAAC,CAAA;IACJ,CAAC,CAAC,CAAA;AACJ,CAAC,CAAC,CAAA;AAEF,SAAS,aAAa;IACpB,MAAM,YAAY,GAAG,IAAI,CAAC,IAAI,CAAC,SAAS,EAAE,uBAAuB,CAAC,CAAA;IAClE,MAAM,YAAY,GAAG,EAAE,CAAC,YAAY,CAAC,YAAY,EAAE,MAAM,CAAC,CAAA;IAC1D,MAAM,KAAK,GAGP;QACF,WAAW,EAAE,EAAE;QACf,SAAS,EAAE,EAAE;QACb,SAAS,EAAE,EAAE;QACb,SAAS,EAAE,EAAE;KACd,CAAA;IACD,YAAY,CAAC,KAAK,CAAC,MAAM,CAAC,CAAC,OAAO,CAAC,CAAC,QAAQ,EAAE,EAAE;QAC9C,MAAM,CAAC,gBAAgB,EAAE,UAAU,EAAE,WAAW,CAAC,GAAG,QAAQ,CAAC,KAAK,CAAC,IAAI,CAAC,CAAA;QACxE,IAAI,CAAC,gBAAgB,IAAI,CAAC,UAAU,IAAI,CAAC,WAAW;YAAE,OAAM;QAC5D,MAAM,YAAY,GAAG,gBAAgB,CAAC,KAAK,CAAC,IAAI,CAAC,CAAC,CAAC,CAAiB,CAAA;QACpE,KAAK,CAAC,YAAY,CAAC,CAAC,IAAI,CAAC;YACvB,MAAM,EAAE,UAAU,CAAC,KAAK,CAAC,IAAI,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC,IAAI,CAAC,IAAI,EAAE;YACxD,OAAO,EAAE,IAAI,CAAC,KAAK,CAAC,WAAW,CAAC,KAAK,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC,IAAI,IAAI,CAAC;SACxD,CAAC,CAAA;IACJ,CAAC,CAAC,CAAA;IACF,OAAO,KAAK,CAAA;AACd,CAAC"}
|
|
@@ -0,0 +1,19 @@
|
|
|
1
|
+
import * as fs from 'fs/promises';
|
|
2
|
+
import * as path from 'path';
|
|
3
|
+
import { fileURLToPath } from 'url';
|
|
4
|
+
import { chatModelParams, modelToEncodingMap } from '../mapping.js';
|
|
5
|
+
// eslint-disable-next-line no-underscore-dangle
|
|
6
|
+
const __dirname = path.dirname(fileURLToPath(import.meta.url));
|
|
7
|
+
const chatModels = Object.keys(chatModelParams);
|
|
8
|
+
const template = await fs.readFile(path.join(__dirname, '../encoding/cl100k_base.ts'), 'utf8');
|
|
9
|
+
await Promise.all(Object.entries(modelToEncodingMap).map(async ([modelName, encoding]) => {
|
|
10
|
+
await fs.mkdir(path.join(__dirname, '../model'), { recursive: true });
|
|
11
|
+
const isChatModel = chatModels.includes(modelName);
|
|
12
|
+
const content = isChatModel
|
|
13
|
+
? template.replace(`getEncodingApi('cl100k_base'`, `getEncodingApiForModel('${modelName}'`)
|
|
14
|
+
: `// eslint-disable-next-line no-restricted-exports, import/no-default-export\nexport { default } from '../encoding/${encoding}.js'\nexport * from '../encoding/${encoding}.js'\n`;
|
|
15
|
+
await fs.writeFile(path.join(__dirname, `../model/${modelName}.ts`), content);
|
|
16
|
+
// eslint-disable-next-line no-console
|
|
17
|
+
console.log(`encoding/${modelName}.ts`);
|
|
18
|
+
}));
|
|
19
|
+
//# sourceMappingURL=generateByModel.js.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"generateByModel.js","sourceRoot":"","sources":["../../src/codegen/generateByModel.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,MAAM,aAAa,CAAA;AACjC,OAAO,KAAK,IAAI,MAAM,MAAM,CAAA;AAC5B,OAAO,EAAE,aAAa,EAAE,MAAM,KAAK,CAAA;AACnC,OAAO,EAAE,eAAe,EAAE,kBAAkB,EAAE,MAAM,eAAe,CAAA;AAEnE,gDAAgD;AAChD,MAAM,SAAS,GAAG,IAAI,CAAC,OAAO,CAAC,aAAa,CAAC,MAAM,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC,CAAA;AAE9D,MAAM,UAAU,GAAG,MAAM,CAAC,IAAI,CAAC,eAAe,CAAC,CAAA;AAE/C,MAAM,QAAQ,GAAG,MAAM,EAAE,CAAC,QAAQ,CAChC,IAAI,CAAC,IAAI,CAAC,SAAS,EAAE,4BAA4B,CAAC,EAClD,MAAM,CACP,CAAA;AAED,MAAM,OAAO,CAAC,GAAG,CACf,MAAM,CAAC,OAAO,CAAC,kBAAkB,CAAC,CAAC,GAAG,CAAC,KAAK,EAAE,CAAC,SAAS,EAAE,QAAQ,CAAC,EAAE,EAAE;IACrE,MAAM,EAAE,CAAC,KAAK,CAAC,IAAI,CAAC,IAAI,CAAC,SAAS,EAAE,UAAU,CAAC,EAAE,EAAE,SAAS,EAAE,IAAI,EAAE,CAAC,CAAA;IACrE,MAAM,WAAW,GAAG,UAAU,CAAC,QAAQ,CAAC,SAAS,CAAC,CAAA;IAElD,MAAM,OAAO,GAAG,WAAW;QACzB,CAAC,CAAC,QAAQ,CAAC,OAAO,CACd,8BAA8B,EAC9B,2BAA2B,SAAS,GAAG,CACxC;QACH,CAAC,CAAC,qHAAqH,QAAQ,oCAAoC,QAAQ,QAAQ,CAAA;IACrL,MAAM,EAAE,CAAC,SAAS,CAChB,IAAI,CAAC,IAAI,CAAC,SAAS,EAAE,YAAY,SAAS,KAAK,CAAC,EAChD,OAAO,CACR,CAAA;IAED,sCAAsC;IACtC,OAAO,CAAC,GAAG,CAAC,YAAY,SAAS,KAAK,CAAC,CAAA;AACzC,CAAC,CAAC,CACH,CAAA"}
|
|
@@ -20,7 +20,7 @@ const processFilesInDirectory = async (directoryPath, fn) => {
|
|
|
20
20
|
};
|
|
21
21
|
// eslint-disable-next-line no-underscore-dangle
|
|
22
22
|
const __dirname = path.dirname(fileURLToPath(import.meta.url));
|
|
23
|
-
await processFilesInDirectory(path.join(__dirname, '
|
|
23
|
+
await processFilesInDirectory(path.join(__dirname, '../../data'), async (filePath) => {
|
|
24
24
|
if (!filePath.endsWith('.tiktoken'))
|
|
25
25
|
return;
|
|
26
26
|
const modelName = path.basename(filePath, '.tiktoken');
|
|
@@ -30,9 +30,9 @@ await processFilesInDirectory(path.join(__dirname, '../data'), async (filePath)
|
|
|
30
30
|
const [token, rank] = x.split(' ');
|
|
31
31
|
return [token, Number.parseInt(rank, 10)];
|
|
32
32
|
});
|
|
33
|
-
await fs.mkdir(path.join(__dirname, '
|
|
34
|
-
await fs.writeFile(path.join(__dirname,
|
|
33
|
+
await fs.mkdir(path.join(__dirname, '../encodings'), { recursive: true });
|
|
34
|
+
await fs.writeFile(path.join(__dirname, `../encodings/${modelName}.js`), `/* eslint-disable */\n// @ts-nocheck\n// prettier-ignore\n/** @type {[string, number][]} */\nconst encoder = ${JSON.stringify(encoder)};\nexport default encoder;`);
|
|
35
35
|
// eslint-disable-next-line no-console
|
|
36
36
|
console.log(`Wrote ${modelName}.js`);
|
|
37
37
|
});
|
|
38
|
-
//# sourceMappingURL=
|
|
38
|
+
//# sourceMappingURL=generateJsEncodings.js.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"generateJsEncodings.js","sourceRoot":"","sources":["../../src/codegen/generateJsEncodings.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,MAAM,aAAa,CAAA;AACjC,OAAO,KAAK,IAAI,MAAM,MAAM,CAAA;AAC5B,OAAO,EAAE,aAAa,EAAE,MAAM,KAAK,CAAA;AAInC,MAAM,uBAAuB,GAAG,KAAK,EACnC,aAAqB,EACrB,EAAoB,EACL,EAAE;IACjB,IAAI;QACF,MAAM,KAAK,GAAG,MAAM,EAAE,CAAC,OAAO,CAAC,aAAa,EAAE,EAAE,aAAa,EAAE,IAAI,EAAE,CAAC,CAAA;QAEtE,KAAK,MAAM,IAAI,IAAI,KAAK,EAAE;YACxB,uCAAuC;YACvC,IAAI,CAAC,IAAI,CAAC,MAAM,EAAE;gBAAE,SAAQ;YAC5B,MAAM,QAAQ,GAAG,IAAI,CAAC,IAAI,CAAC,aAAa,EAAE,IAAI,CAAC,IAAI,CAAC,CAAA;YACpD,4CAA4C;YAC5C,MAAM,EAAE,CAAC,QAAQ,CAAC,CAAA;SACnB;KACF;IAAC,OAAO,KAAK,EAAE;QACd,sCAAsC;QACtC,OAAO,CAAC,KAAK,CAAC,oBAAoB,EAAE,KAAK,CAAC,CAAA;KAC3C;AACH,CAAC,CAAA;AAED,gDAAgD;AAChD,MAAM,SAAS,GAAG,IAAI,CAAC,OAAO,CAAC,aAAa,CAAC,MAAM,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC,CAAA;AAE9D,MAAM,uBAAuB,CAC3B,IAAI,CAAC,IAAI,CAAC,SAAS,EAAE,YAAY,CAAC,EAClC,KAAK,EAAE,QAAQ,EAAE,EAAE;IACjB,IAAI,CAAC,QAAQ,CAAC,QAAQ,CAAC,WAAW,CAAC;QAAE,OAAM;IAE3C,MAAM,SAAS,GAAG,IAAI,CAAC,QAAQ,CAAC,QAAQ,EAAE,WAAW,CAAC,CAAA;IACtD,MAAM,OAAO,GAAG,MAAM,EAAE,CAAC,QAAQ,CAAC,QAAQ,EAAE,MAAM,CAAC,CAAA;IACnD,MAAM,KAAK,GAAG,OAAO,CAAC,KAAK,CAAC,IAAI,CAAC,CAAA;IACjC,MAAM,OAAO,GAAG,KAAK,CAAC,KAAK,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE;QAC3C,MAAM,CAAC,KAAK,EAAE,IAAI,CAAC,GAAG,CAAC,CAAC,KAAK,CAAC,GAAG,CAAC,CAAA;QAClC,OAAO,CAAC,KAAK,EAAE,MAAM,CAAC,QAAQ,CAAC,IAAK,EAAE,EAAE,CAAC,CAAC,CAAA;IAC5C,CAAC,CAAC,CAAA;IAEF,MAAM,EAAE,CAAC,KAAK,CAAC,IAAI,CAAC,IAAI,CAAC,SAAS,EAAE,cAAc,CAAC,EAAE,EAAE,SAAS,EAAE,IAAI,EAAE,CAAC,CAAA;IACzE,MAAM,EAAE,CAAC,SAAS,CAChB,IAAI,CAAC,IAAI,CAAC,SAAS,EAAE,gBAAgB,SAAS,KAAK,CAAC,EACpD,gHAAgH,IAAI,CAAC,SAAS,CAC5H,OAAO,CACR,4BAA4B,CAC9B,CAAA;IAED,sCAAsC;IACtC,OAAO,CAAC,GAAG,CAAC,SAAS,SAAS,KAAK,CAAC,CAAA;AACtC,CAAC,CACF,CAAA"}
|
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
import { GptEncoding } from '../GptEncoding.js';
|
|
2
2
|
export * from '../specialTokens.js';
|
|
3
3
|
declare const api: GptEncoding;
|
|
4
|
-
declare const decode: (inputTokensToDecode: Iterable<number>) => string, decodeAsyncGenerator: (inputTokensToDecode: AsyncIterable<number>) => AsyncGenerator<string, void, unknown>, decodeGenerator: (inputTokensToDecode: Iterable<number>) => Generator<string, void, unknown>, encode: (lineToEncode: string,
|
|
5
|
-
export { decode, decodeAsyncGenerator, decodeGenerator, encode, encodeGenerator, isWithinTokenLimit, };
|
|
4
|
+
declare const decode: (inputTokensToDecode: Iterable<number>) => string, decodeAsyncGenerator: (inputTokensToDecode: AsyncIterable<number>) => AsyncGenerator<string, void, unknown>, decodeGenerator: (inputTokensToDecode: Iterable<number>) => Generator<string, void, unknown>, encode: (lineToEncode: string, encodeOptions?: import("../GptEncoding.js").EncodeOptions) => number[], encodeGenerator: (lineToEncode: string, { allowedSpecial, disallowedSpecial, }?: import("../GptEncoding.js").EncodeOptions) => Generator<number[], number, undefined>, isWithinTokenLimit: (input: string | Iterable<import("../GptEncoding.js").ChatMessage>, tokenLimit: number) => number | false, encodeChat: (chat: readonly import("../GptEncoding.js").ChatMessage[], model?: "gpt-4" | "gpt-4-32k" | "gpt-4-0314" | "gpt-4-32k-0314" | "gpt-3.5-turbo" | "gpt-3.5-turbo-0301" | "text-davinci-003" | "text-davinci-002" | "text-davinci-001" | "text-curie-001" | "text-babbage-001" | "text-ada-001" | "davinci" | "curie" | "babbage" | "ada" | "code-davinci-002" | "code-davinci-001" | "code-cushman-002" | "code-cushman-001" | "davinci-codex" | "cushman-codex" | "text-davinci-edit-001" | "code-davinci-edit-001" | "text-embedding-ada-002" | "text-similarity-davinci-001" | "text-similarity-curie-001" | "text-similarity-babbage-001" | "text-similarity-ada-001" | "text-search-davinci-doc-001" | "text-search-curie-doc-001" | "text-search-babbage-doc-001" | "text-search-ada-doc-001" | "code-search-babbage-code-001" | "code-search-ada-code-001" | undefined) => number[], encodeChatGenerator: (chat: Iterable<import("../GptEncoding.js").ChatMessage>, model?: "gpt-4" | "gpt-4-32k" | "gpt-4-0314" | "gpt-4-32k-0314" | "gpt-3.5-turbo" | "gpt-3.5-turbo-0301" | "text-davinci-003" | "text-davinci-002" | "text-davinci-001" | "text-curie-001" | "text-babbage-001" | "text-ada-001" | "davinci" | "curie" | "babbage" | "ada" | "code-davinci-002" | "code-davinci-001" | "code-cushman-002" | "code-cushman-001" | "davinci-codex" | "cushman-codex" | "text-davinci-edit-001" | "code-davinci-edit-001" | "text-embedding-ada-002" | "text-similarity-davinci-001" | "text-similarity-curie-001" | "text-similarity-babbage-001" | "text-similarity-ada-001" | "text-search-davinci-doc-001" | "text-search-curie-doc-001" | "text-search-babbage-doc-001" | "text-search-ada-doc-001" | "code-search-babbage-code-001" | "code-search-ada-code-001" | undefined) => Generator<number[], void, undefined>;
|
|
5
|
+
export { decode, decodeAsyncGenerator, decodeGenerator, encode, encodeChat, encodeChatGenerator, encodeGenerator, isWithinTokenLimit, };
|
|
6
6
|
export default api;
|
|
@@ -4,8 +4,8 @@ import encoder from '../encodings/cl100k_base.js';
|
|
|
4
4
|
import { GptEncoding } from '../GptEncoding.js';
|
|
5
5
|
export * from '../specialTokens.js';
|
|
6
6
|
const api = GptEncoding.getEncodingApi('cl100k_base', () => convertTokenBytePairEncodingFromTuples(encoder));
|
|
7
|
-
const { decode, decodeAsyncGenerator, decodeGenerator, encode, encodeGenerator, isWithinTokenLimit, } = api;
|
|
8
|
-
export { decode, decodeAsyncGenerator, decodeGenerator, encode, encodeGenerator, isWithinTokenLimit, };
|
|
7
|
+
const { decode, decodeAsyncGenerator, decodeGenerator, encode, encodeGenerator, isWithinTokenLimit, encodeChat, encodeChatGenerator, } = api;
|
|
8
|
+
export { decode, decodeAsyncGenerator, decodeGenerator, encode, encodeChat, encodeChatGenerator, encodeGenerator, isWithinTokenLimit, };
|
|
9
9
|
// eslint-disable-next-line import/no-default-export
|
|
10
10
|
export default api;
|
|
11
11
|
//# sourceMappingURL=cl100k_base.js.map
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"cl100k_base.js","sourceRoot":"","sources":["../../src/encoding/cl100k_base.ts"],"names":[],"mappings":"AAAA,sCAAsC;AACtC,OAAO,EAAE,sCAAsC,EAAE,MAAM,8CAA8C,CAAA;AACrG,OAAO,OAAO,MAAM,6BAA6B,CAAA;AACjD,OAAO,EAAE,WAAW,EAAE,MAAM,mBAAmB,CAAA;AAE/C,cAAc,qBAAqB,CAAA;AAEnC,MAAM,GAAG,GAAG,WAAW,CAAC,cAAc,CAAC,aAAa,EAAE,GAAG,EAAE,CACzD,sCAAsC,CAAC,OAAO,CAAC,CAChD,CAAA;AACD,MAAM,EACJ,MAAM,EACN,oBAAoB,EACpB,eAAe,EACf,MAAM,EACN,eAAe,EACf,kBAAkB,
|
|
1
|
+
{"version":3,"file":"cl100k_base.js","sourceRoot":"","sources":["../../src/encoding/cl100k_base.ts"],"names":[],"mappings":"AAAA,sCAAsC;AACtC,OAAO,EAAE,sCAAsC,EAAE,MAAM,8CAA8C,CAAA;AACrG,OAAO,OAAO,MAAM,6BAA6B,CAAA;AACjD,OAAO,EAAE,WAAW,EAAE,MAAM,mBAAmB,CAAA;AAE/C,cAAc,qBAAqB,CAAA;AAEnC,MAAM,GAAG,GAAG,WAAW,CAAC,cAAc,CAAC,aAAa,EAAE,GAAG,EAAE,CACzD,sCAAsC,CAAC,OAAO,CAAC,CAChD,CAAA;AACD,MAAM,EACJ,MAAM,EACN,oBAAoB,EACpB,eAAe,EACf,MAAM,EACN,eAAe,EACf,kBAAkB,EAClB,UAAU,EACV,mBAAmB,GACpB,GAAG,GAAG,CAAA;AACP,OAAO,EACL,MAAM,EACN,oBAAoB,EACpB,eAAe,EACf,MAAM,EACN,UAAU,EACV,mBAAmB,EACnB,eAAe,EACf,kBAAkB,GACnB,CAAA;AACD,oDAAoD;AACpD,eAAe,GAAG,CAAA"}
|
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
import { GptEncoding } from '../GptEncoding.js';
|
|
2
2
|
export * from '../specialTokens.js';
|
|
3
3
|
declare const api: GptEncoding;
|
|
4
|
-
declare const decode: (inputTokensToDecode: Iterable<number>) => string, decodeAsyncGenerator: (inputTokensToDecode: AsyncIterable<number>) => AsyncGenerator<string, void, unknown>, decodeGenerator: (inputTokensToDecode: Iterable<number>) => Generator<string, void, unknown>, encode: (lineToEncode: string,
|
|
4
|
+
declare const decode: (inputTokensToDecode: Iterable<number>) => string, decodeAsyncGenerator: (inputTokensToDecode: AsyncIterable<number>) => AsyncGenerator<string, void, unknown>, decodeGenerator: (inputTokensToDecode: Iterable<number>) => Generator<string, void, unknown>, encode: (lineToEncode: string, encodeOptions?: import("../GptEncoding.js").EncodeOptions) => number[], encodeGenerator: (lineToEncode: string, { allowedSpecial, disallowedSpecial, }?: import("../GptEncoding.js").EncodeOptions) => Generator<number[], number, undefined>, isWithinTokenLimit: (input: string | Iterable<import("../GptEncoding.js").ChatMessage>, tokenLimit: number) => number | false;
|
|
5
5
|
export { decode, decodeAsyncGenerator, decodeGenerator, encode, encodeGenerator, isWithinTokenLimit, };
|
|
6
6
|
export default api;
|
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
import { GptEncoding } from '../GptEncoding.js';
|
|
2
2
|
export * from '../specialTokens.js';
|
|
3
3
|
declare const api: GptEncoding;
|
|
4
|
-
declare const decode: (inputTokensToDecode: Iterable<number>) => string, decodeAsyncGenerator: (inputTokensToDecode: AsyncIterable<number>) => AsyncGenerator<string, void, unknown>, decodeGenerator: (inputTokensToDecode: Iterable<number>) => Generator<string, void, unknown>, encode: (lineToEncode: string,
|
|
4
|
+
declare const decode: (inputTokensToDecode: Iterable<number>) => string, decodeAsyncGenerator: (inputTokensToDecode: AsyncIterable<number>) => AsyncGenerator<string, void, unknown>, decodeGenerator: (inputTokensToDecode: Iterable<number>) => Generator<string, void, unknown>, encode: (lineToEncode: string, encodeOptions?: import("../GptEncoding.js").EncodeOptions) => number[], encodeGenerator: (lineToEncode: string, { allowedSpecial, disallowedSpecial, }?: import("../GptEncoding.js").EncodeOptions) => Generator<number[], number, undefined>, isWithinTokenLimit: (input: string | Iterable<import("../GptEncoding.js").ChatMessage>, tokenLimit: number) => number | false;
|
|
5
5
|
export { decode, decodeAsyncGenerator, decodeGenerator, encode, encodeGenerator, isWithinTokenLimit, };
|
|
6
6
|
export default api;
|
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
import { GptEncoding } from '../GptEncoding.js';
|
|
2
2
|
export * from '../specialTokens.js';
|
|
3
3
|
declare const api: GptEncoding;
|
|
4
|
-
declare const decode: (inputTokensToDecode: Iterable<number>) => string, decodeAsyncGenerator: (inputTokensToDecode: AsyncIterable<number>) => AsyncGenerator<string, void, unknown>, decodeGenerator: (inputTokensToDecode: Iterable<number>) => Generator<string, void, unknown>, encode: (lineToEncode: string,
|
|
4
|
+
declare const decode: (inputTokensToDecode: Iterable<number>) => string, decodeAsyncGenerator: (inputTokensToDecode: AsyncIterable<number>) => AsyncGenerator<string, void, unknown>, decodeGenerator: (inputTokensToDecode: Iterable<number>) => Generator<string, void, unknown>, encode: (lineToEncode: string, encodeOptions?: import("../GptEncoding.js").EncodeOptions) => number[], encodeGenerator: (lineToEncode: string, { allowedSpecial, disallowedSpecial, }?: import("../GptEncoding.js").EncodeOptions) => Generator<number[], number, undefined>, isWithinTokenLimit: (input: string | Iterable<import("../GptEncoding.js").ChatMessage>, tokenLimit: number) => number | false;
|
|
5
5
|
export { decode, decodeAsyncGenerator, decodeGenerator, encode, encodeGenerator, isWithinTokenLimit, };
|
|
6
6
|
export default api;
|
package/esm/main.d.ts
CHANGED
package/esm/main.js
CHANGED
package/esm/main.js.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"main.js","sourceRoot":"","sources":["../src/main.ts"],"names":[],"mappings":"AAAA,cAAc,2BAA2B,CAAA"}
|
|
1
|
+
{"version":3,"file":"main.js","sourceRoot":"","sources":["../src/main.ts"],"names":[],"mappings":"AAAA,2EAA2E;AAC3E,OAAO,EAAE,OAAO,EAAE,MAAM,2BAA2B,CAAA;AACnD,cAAc,2BAA2B,CAAA"}
|
package/esm/mapping.d.ts
CHANGED
|
@@ -5,7 +5,11 @@ export declare const r50k_base = "r50k_base";
|
|
|
5
5
|
export declare const encodingNames: readonly ["cl100k_base", "p50k_base", "r50k_base", "p50k_edit"];
|
|
6
6
|
export declare const modelToEncodingMap: {
|
|
7
7
|
readonly 'gpt-4': "cl100k_base";
|
|
8
|
+
readonly 'gpt-4-32k': "cl100k_base";
|
|
9
|
+
readonly 'gpt-4-0314': "cl100k_base";
|
|
10
|
+
readonly 'gpt-4-32k-0314': "cl100k_base";
|
|
8
11
|
readonly 'gpt-3.5-turbo': "cl100k_base";
|
|
12
|
+
readonly 'gpt-3.5-turbo-0301': "cl100k_base";
|
|
9
13
|
readonly 'text-davinci-003': "p50k_base";
|
|
10
14
|
readonly 'text-davinci-002': "p50k_base";
|
|
11
15
|
readonly 'text-davinci-001': "r50k_base";
|
|
@@ -36,5 +40,38 @@ export declare const modelToEncodingMap: {
|
|
|
36
40
|
readonly 'code-search-babbage-code-001': "r50k_base";
|
|
37
41
|
readonly 'code-search-ada-code-001': "r50k_base";
|
|
38
42
|
};
|
|
43
|
+
export interface ChatParameters {
|
|
44
|
+
messageSeparator: string;
|
|
45
|
+
roleSeparator: string;
|
|
46
|
+
}
|
|
47
|
+
declare const internalChatModelParams: {
|
|
48
|
+
'gpt-3.5-turbo': {
|
|
49
|
+
messageSeparator: string;
|
|
50
|
+
roleSeparator: string;
|
|
51
|
+
};
|
|
52
|
+
'gpt-3.5-turbo-0301': {
|
|
53
|
+
messageSeparator: string;
|
|
54
|
+
roleSeparator: string;
|
|
55
|
+
};
|
|
56
|
+
'gpt-4': {
|
|
57
|
+
messageSeparator: string;
|
|
58
|
+
roleSeparator: string;
|
|
59
|
+
};
|
|
60
|
+
'gpt-4-0314': {
|
|
61
|
+
messageSeparator: string;
|
|
62
|
+
roleSeparator: string;
|
|
63
|
+
};
|
|
64
|
+
'gpt-4-32k': {
|
|
65
|
+
messageSeparator: string;
|
|
66
|
+
roleSeparator: string;
|
|
67
|
+
};
|
|
68
|
+
'gpt-4-32k-0314': {
|
|
69
|
+
messageSeparator: string;
|
|
70
|
+
roleSeparator: string;
|
|
71
|
+
};
|
|
72
|
+
};
|
|
73
|
+
export declare const chatModelParams: Partial<Record<ModelName, ChatParameters>>;
|
|
39
74
|
export type ModelName = keyof typeof modelToEncodingMap;
|
|
75
|
+
export type ChatModelName = keyof typeof internalChatModelParams;
|
|
40
76
|
export type EncodingName = (typeof modelToEncodingMap)[ModelName];
|
|
77
|
+
export {};
|