root/gpustack-ui
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

fix: tts request content-type

Browse Source
This commit is contained in:
jialin
2025-11-05 17:09:16 +08:00
parent 333b430b97
commit 6173f3b459
9 changed files with 65 additions and 33 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/locales/en-US/models.ts
+4 -3
View File
@@ -209,9 +209,9 @@ export default {
'models.mymodels.status.degrade': 'Abnormal',
'models.mymodels.status.active': 'Active',
'models.form.kvCache.tips':
'Available only with built-in backends (vLLM / SGLang) — switch backend in <span class="bold-text">Advanced</span> to enable.',
'Extended KV cache and speculative decoding are only available with built-in backends (vLLM / SGLang), Please switch the backend in the <span class="bold-text">Advanced</span> settings to enable them.',
'models.form.kvCache.tips2':
'Extended KV cache is only supported when using built-in inference backends (vLLM or SGLang).',
'Only supported when using built-in inference backends (vLLM or SGLang).',
'models.form.scheduling': 'Scheduling',
'models.form.ramRatio': 'RAM-to-VRAM Ratio',
'models.form.ramSize': 'Maximum RAM Size (GiB)',
@@ -237,5 +237,6 @@ export default {
'models.form.draftModel.placeholder': 'Please select or enter a draft model',
'models.form.draftModel.tips':
'You can enter a local path (e.g., /path/to/model) or select a model from Hugging Face or ModelScope (e.g., Tengyunw/qwen3_8b_eagle3). The system will automatically match based on the primary model source.',
'models.form.quantization': 'Quantization'
'models.form.quantization': 'Quantization',
'models.form.backend.custom': 'User Defined'
};
src/locales/ja-JP/models.ts
+8 -6
View File
@@ -209,9 +209,9 @@ export default {
'models.mymodels.status.degrade': 'Abnormal',
'models.mymodels.status.active': 'Active',
'models.form.kvCache.tips':
'Available only with built-in backends (vLLM / SGLang) — switch backend in <span class="bold-text">Advanced</span> to enable.',
'Extended KV cache and speculative decoding are only available with built-in backends (vLLM / SGLang), Please switch the backend in the <span class="bold-text">Advanced</span> settings to enable them.',
'models.form.kvCache.tips2':
'Extended KV cache is only supported when using built-in inference backends (vLLM or SGLang).',
'Only supported when using built-in inference backends (vLLM or SGLang).',
'models.form.scheduling': 'Scheduling',
'models.form.ramRatio': 'RAM-to-VRAM Ratio',
'models.form.ramSize': 'Maximum RAM Size (GiB)',
@@ -237,7 +237,8 @@ export default {
'models.form.draftModel.placeholder': 'Please select or enter a draft model',
'models.form.draftModel.tips':
'You can enter a local path (e.g., /path/to/model) or select a model from Hugging Face or ModelScope (e.g., Tengyunw/qwen3_8b_eagle3). The system will automatically match based on the primary model source.',
'models.form.quantization': 'Quantization'
'models.form.quantization': 'Quantization',
'models.form.backend.custom': 'User Defined'
};
// ========== To-Do: Translate Keys (Remove After Translation) ==========
@@ -282,8 +283,8 @@ export default {
// 41. 'models.mymodels.status.degrade': 'Abnormal',
// 42. 'models.mymodels.status.active': 'Active'
// 43. 'models.form.remoteURL.tips': 'Refer to the <a href="https://docs.lmcache.ai/api_reference/configurations.html" target="_blank">configuration documentation</a> for details.',
// 44. 'models.form.kvCache.tips': 'Available only with built-in backends (vLLM / SGLang) — switch backend in <span class="bold-text">Advanced</span> to enable.'
// 45. 'models.form.kvCache.tips2': 'Extended KV cache is only supported when using built-in inference backends (vLLM or SGLang).',
// 44. 'models.form.kvCache.tips': 'Extended KV cache and speculative decoding are only available with built-in backends (vLLM / SGLang), Please switch the backend in the <span class="bold-text">Advanced</span> settings to enable them.',
// 45. 'models.form.kvCache.tips2': 'Only supported when using built-in inference backends (vLLM or SGLang).',
// 46. 'models.form.scheduling': 'Scheduling',
// 47. 'models.form.ramRatio': 'RAM-to-VRAM Ratio',
// 48. 'models.form.ramSize': 'Maximum RAM Size (GiB)',
@@ -304,5 +305,6 @@ export default {
// 44. 'models.form.mode.baseline.tips': 'the most compatible option with full precision,
// 45. 'models.form.draftModel.placeholder': 'Please select or enter a draft model',
// 46. 'models.form.draftModel.tips': 'You can enter a local path (e.g., /path/to/model) or select a model from Hugging Face or ModelScope (e.g., Tengyunw/qwen3_8b_eagle3). The system will automatically match based on the primary model source.'
// 47. 'models.form.quantization': 'Quantization'
// 47. 'models.form.quantization': 'Quantization',
// 48. 'models.form.backend.custom': 'User Defined'
// ========== End of To-Do List ==========
src/locales/ru-RU/models.ts
+8 -6
View File
@@ -209,9 +209,9 @@ export default {
'models.mymodels.status.degrade': 'Abnormal',
'models.mymodels.status.active': 'Active',
'models.form.kvCache.tips':
'Available only with built-in backends (vLLM / SGLang) — switch backend in <span class="bold-text">Advanced</span> to enable.',
'Extended KV cache and speculative decoding are only available with built-in backends (vLLM / SGLang), Please switch the backend in the <span class="bold-text">Advanced</span> settings to enable them.',
'models.form.kvCache.tips2':
'Extended KV cache is only supported when using built-in inference backends (vLLM or SGLang).',
'Only supported when using built-in inference backends (vLLM or SGLang).',
'models.form.scheduling': 'Scheduling',
'models.form.ramRatio': 'RAM-to-VRAM Ratio',
'models.form.ramSize': 'Maximum RAM Size (GiB)',
@@ -237,7 +237,8 @@ export default {
'models.form.draftModel.placeholder': 'Please select or enter a draft model',
'models.form.draftModel.tips':
'You can enter a local path (e.g., /path/to/model) or select a model from Hugging Face or ModelScope (e.g., Tengyunw/qwen3_8b_eagle3). The system will automatically match based on the primary model source.',
'models.form.quantization': 'Quantization'
'models.form.quantization': 'Quantization',
'models.form.backend.custom': 'User Defined'
};
// ========== To-Do: Translate Keys (Remove After Translation) ==========
@@ -245,8 +246,8 @@ export default {
// 3. 'models.mymodels.status.degrade': 'Abnormal',
// 4. 'models.mymodels.status.active': 'Active'
// 5. 'models.form.remoteURL.tips': 'Refer to the <a href="https://docs.lmcache.ai/api_reference/configurations.html" target="_blank">configuration documentation</a> for details.',
// 6. 'models.form.kvCache.tips': 'Available only with built-in backends (vLLM / SGLang) — switch backend in <span class="bold-text">Advanced</span> to enable.'
// 7. 'models.form.kvCache.tips2': 'Extended KV cache is only supported when using built-in inference backends (vLLM or SGLang).';
// 6. 'models.form.kvCache.tips': 'Extended KV cache and speculative decoding are only available with built-in backends (vLLM / SGLang), Please switch the backend in the <span class="bold-text">Advanced</span> settings to enable them.',
// 7. 'models.form.kvCache.tips2': 'Only supported when using built-in inference backends (vLLM or SGLang).';
// 8. 'models.form.scheduling': 'Scheduling',
// 9. 'models.form.ramRatio': 'RAM-to-VRAM Ratio',
// 10. 'models.form.ramSize': 'Maximum RAM Size (GiB)',
@@ -267,5 +268,6 @@ export default {
// 36. 'models.form.mode.baseline.tips': 'the most compatible option with full precision.'
// 37. 'models.form.draftModel.placeholder': 'Please select or enter a draft model',
// 38. 'models.form.draftModel.tips': 'You can enter a local path (e.g., /path/to/model) or select a model from Hugging Face or ModelScope (e.g., Tengyunw/qwen3_8b_eagle3). The system will automatically match based on the primary model source.',
// 39. 'models.form.quantization': 'Quantization'
// 39. 'models.form.quantization': 'Quantization',
// 40. 'models.form.backend.custom': 'User Defined'
// ========== End of To-Do List ==========
src/locales/zh-CN/models.ts
+4 -4
View File
@@ -198,9 +198,8 @@ export default {
'models.mymodels.status.degrade': '异常',
'models.mymodels.status.active': '可用',
'models.form.kvCache.tips':
'仅在内置后端（vLLM / SGLang）可用 —— 请在<span class="bold-text">高级</span>配置中切换后端以启用。',
'models.form.kvCache.tips2':
'仅在使用内置推理后端（vLLM 或 SGLang）时支持 KV 缓存。',
'扩展 KV 缓存和推测解码仅在内置后端（vLLM / SGLang）可用，请在<span class="bold-text">高级</span>配置中切换后端以启用。',
'models.form.kvCache.tips2': '仅在使用内置推理后端（vLLM 或 SGLang）时支持。',
'models.form.scheduling': '调度',
'models.form.ramRatio': '内存与显存比例',
'models.form.ramSize': '内存最大占用 (GiB)',
@@ -225,5 +224,6 @@ export default {
'models.form.draftModel.placeholder': '请选择或输入草稿模型',
'models.form.draftModel.tips':
'可填写本地路径（如 /path/to/model），或从 Hugging Face、ModelScope 选择模型（如 Tengyunw/qwen3_8b_eagle3）。系统将根据主模型来源自动匹配。',
'models.form.quantization': '量化'
'models.form.quantization': '量化',
'models.form.backend.custom': '用户定义'
};
src/pages/llmodels/components/instance-item.tsx
+3 -1
View File
@@ -475,7 +475,9 @@ const InstanceItem: React.FC<InstanceItemProps> = ({
<div className="flex-center">
<ThunderboltFilled className="m-r-5" />
{intl.formatMessage({ id: 'models.form.backend' })}:{' '}
{modelData?.backend || ''}
{modelData?.backend === backendOptionsMap.custom
? intl.formatMessage({ id: 'models.form.backend.custom' })
: modelData?.backend || ''}
{modelData.backend_version ? `(${modelData.backend_version})` : ''}
</div>
</div>
src/pages/llmodels/forms/index.tsx
+5 -2
View File
@@ -190,12 +190,15 @@ const DataForm: React.FC<DataFormProps> = forwardRef((props, ref) => {
const updateKVCacheConfig = (backend: string, option: BackendOption) => {
if (
!option.isBuiltIn &&
[backendOptionsMap.SGLang, backendOptionsMap.vllm].includes(backend)
!option.isBuiltIn ||
![backendOptionsMap.SGLang, backendOptionsMap.vllm].includes(backend)
) {
return {
extended_kv_cache: {
enabled: false
},
speculative_config: {
enabled: false
}
};
}
src/pages/llmodels/forms/kv-cache.tsx
-9
View File
@@ -50,15 +50,6 @@ const KVCacheForm = () => {
name={['extended_kv_cache', 'enabled']}
valuePropName="checked"
style={{ marginBottom: 8 }}
extra={
!builtInBackend && (
<span
dangerouslySetInnerHTML={{
__html: intl.formatMessage({ id: 'models.form.kvCache.tips' })
}}
></span>
)
}
>
<CheckboxField
description={intl.formatMessage({
src/pages/llmodels/forms/speculative-decode.tsx
+30 -2
View File
@@ -7,7 +7,8 @@ import useAppUtils from '@/hooks/use-app-utils';
import { useIntl } from '@umijs/max';
import { Form } from 'antd';
import _ from 'lodash';
import { useRef } from 'react';
import { useMemo, useRef } from 'react';
import { backendOptionsMap } from '../config/backend-parameters';
import { useFormContext } from '../config/form-context';
import { FormData } from '../config/types';
import useQueryDraftModels from '../hooks/use-query-draftModels';
@@ -20,9 +21,10 @@ const AlgorithmMap = {
const SpeculativeDecode = () => {
const intl = useIntl();
const { source, onValuesChange } = useFormContext();
const { source, backendOptions, onValuesChange } = useFormContext();
const { getRuleMessage } = useAppUtils();
const form = Form.useFormInstance();
const backend = Form.useWatch('backend', form);
const speculativeEnabled = Form.useWatch(
['speculative_config', 'enabled'],
form
@@ -65,18 +67,44 @@ const SpeculativeDecode = () => {
}
};
const builtInBackend = useMemo(() => {
const currentBackend = backendOptions.find(
(item) => item.value === backend
);
return (
currentBackend?.isBuiltIn &&
[backendOptionsMap.SGLang, backendOptionsMap.vllm].includes(
backend as string
)
);
}, [backend, backendOptions]);
return (
<>
<Form.Item<FormData>
name={['speculative_config', 'enabled']}
valuePropName="checked"
style={{ marginBottom: 8 }}
extra={
!builtInBackend && (
<span
dangerouslySetInnerHTML={{
__html: intl.formatMessage({ id: 'models.form.kvCache.tips' })
}}
></span>
)
}
>
<CheckboxField
description={intl.formatMessage({
id: 'models.form.kvCache.tips2'
})}
label={intl.formatMessage({
id: 'models.form.enableSpeculativeDecoding'
})}
onChange={handleSpeculativeEnabledChange}
disabled={!builtInBackend}
></CheckboxField>
</Form.Item>
{speculativeEnabled && (
src/pages/playground/apis/index.ts
+3
View File
@@ -96,6 +96,9 @@ export const textToSpeech = async (params: any, options?: any) => {
const res = await fetch(AUDIO_TEXT_TO_SPEECH_API, {
method: 'POST',
body: JSON.stringify(params.data),
headers: {
'Content-Type': 'application/json'
},
signal: params.signal
});
if (!res.ok) {