Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

chore(adapters): add llama-3-1-8b to vllm, remove qwen2 templates #172

Merged
merged 1 commit into from
Nov 15, 2024
Merged

chore(adapters): add llama-3-1-8b to vllm, remove qwen2 templates #172

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
16 changes: 0 additions & 16 deletions src/adapters/bam/chatPreset.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -62,22 +62,6 @@ export const BAMChatLLMPreset = {
},
};
},
"qwen/qwen2-72b-instruct": (): BAMChatLLMPreset => {
const { template, parameters, messagesToPrompt } = LLMChatTemplates.get("qwen2");

return {
base: {
parameters: {
decoding_method: "greedy",
include_stop_sequence: false,
stop_sequences: [...parameters.stop_sequence],
},
},
chat: {
messagesToPrompt: messagesToPrompt(template),
},
};
},
} as const;

export type BAMChatLLMPresetModel = keyof typeof BAMChatLLMPreset;
5 changes: 1 addition & 4 deletions src/adapters/bam/llm.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -203,10 +203,7 @@ export class BAMLLM extends LLM<BAMLLMOutput, BAMLLMGenerateOptions> {
};
} catch {
// TODO: remove once retrieval gets fixed on the API
if (
this.modelId === "qwen/qwen2-72b-instruct" ||
this.modelId === "meta-llama/llama-3-1-70b-instruct"
) {
if (this.modelId === "meta-llama/llama-3-1-70b-instruct") {
return {
tokenLimit: 131_072,
};
Expand Down
10 changes: 5 additions & 5 deletions src/adapters/ibm-vllm/chatPreset.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -26,7 +26,7 @@ interface IBMVllmChatLLMPreset {
export const IBMVllmModel = {
LLAMA_3_1_405B_INSTRUCT_FP8: "meta-llama/llama-3-1-405b-instruct-fp8",
LLAMA_3_1_70B_INSTRUCT: "meta-llama/llama-3-1-70b-instruct",
QWEN2_72B_INSTRUCT: "qwen/qwen2-72b-instruct",
LLAMA_3_1_8B_INSTRUCT: "meta-llama/llama-3-1-8b-instruct",
GRANITE_INSTRUCT: "ibm/granite-instruct", // Generic model ID is used for ease of development, ground it once stable
} as const;
export type IBMVllmModel = (typeof IBMVllmModel)[keyof typeof IBMVllmModel];
Expand Down Expand Up @@ -76,17 +76,17 @@ export const IBMVllmChatLLMPreset = {
},
};
},
[IBMVllmModel.QWEN2_72B_INSTRUCT]: (): IBMVllmChatLLMPreset => {
const { template, parameters, messagesToPrompt } = LLMChatTemplates.get("qwen2");
[IBMVllmModel.LLAMA_3_1_8B_INSTRUCT]: (): IBMVllmChatLLMPreset => {
const { template, parameters, messagesToPrompt } = LLMChatTemplates.get("llama3");
return {
base: {
modelId: IBMVllmModel.QWEN2_72B_INSTRUCT,
modelId: IBMVllmModel.LLAMA_3_1_8B_INSTRUCT,
parameters: {
method: "GREEDY",
stopping: {
stop_sequences: [...parameters.stop_sequence],
include_stop_sequence: false,
max_new_tokens: 1024,
max_new_tokens: 2048,
},
},
},
Expand Down
19 changes: 0 additions & 19 deletions src/adapters/shared/llmChatTemplates.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -114,24 +114,6 @@ const llama3: LLMChatTemplate = {
},
};

const qwen2: LLMChatTemplate = {
template: new PromptTemplate({
schema: templateSchemaFactory(["system", "user", "assistant"] as const),
template: `{{#messages}}{{#system}}<|im_start|>system
{{system}}<|im_end|>
{{ end }}{{/system}}{{#user}}<|im_start|>user
{{user}}<|im_end|>
{{ end }}{{/user}}{{#assistant}}<|im_start|>assistant
{{assistant}}<|im_end|>
{{ end }}{{/assistant}}{{/messages}}<|im_start|>assistant
`,
}),
messagesToPrompt: messagesToPromptFactory(),
parameters: {
stop_sequence: ["<|im_end|>"],
},
};

const granite3Instruct: LLMChatTemplate = {
template: new PromptTemplate({
schema: templateSchemaFactory([
Expand Down Expand Up @@ -171,7 +153,6 @@ export class LLMChatTemplates {
protected static readonly registry = {
"llama3.1": llama31,
"llama3": llama3,
"qwen2": qwen2,
"granite3Instruct": granite3Instruct,
};

Expand Down