fix: kv cache support in builtin backend

This commit is contained in:
jialin
2025-10-27 14:13:47 +08:00
parent 6969cb0b9c
commit df65d3e2f6
15 changed files with 124 additions and 40 deletions
+7 -1
View File
@@ -187,6 +187,8 @@ export default {
'models.form.chunkSize': 'Size Of Cache Chunks',
'models.form.maxCPUSize': 'Maximum CPU Cache Size (GiB)',
'models.form.remoteURL': 'Remote Storage URL',
'models.form.remoteURL.tips':
'Refer to the <a href="https://docs.lmcache.ai/api_reference/configurations.html" target="_blank">configuration documentation</a> for details.',
'models.form.runCommandPlaceholder':
'e.g., vllm serve Qwen/Qwen2.5-1.5B-Instruct',
'models.accessSettings.public': 'Public',
@@ -206,5 +208,9 @@ export default {
'You can specify the exact number of GPUs per replica.',
'models.mymodels.status.inactive': 'Stopped',
'models.mymodels.status.degrade': 'Abnormal',
'models.mymodels.status.active': 'Active'
'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.',
'models.form.kvCache.tips2':
'KV cache is only supported when using built-in inference backends (vLLM or SGLang).'
};
+10 -2
View File
@@ -184,6 +184,8 @@ export default {
'models.form.chunkSize': 'Size Of Cache Chunks',
'models.form.maxCPUSize': 'Maximum CPU Cache Size (GiB)',
'models.form.remoteURL': 'Remote Storage URL',
'models.form.remoteURL.tips':
'Refer to the <a href="https://docs.lmcache.ai/api_reference/configurations.html" target="_blank">configuration documentation</a> for details.',
'models.form.runCommandPlaceholder':
'e.g., vllm serve Qwen/Qwen2.5-1.5B-Instruct',
'models.accessSettings.public': 'Public',
@@ -203,7 +205,11 @@ export default {
'You can specify the exact number of GPUs per replica.',
'models.mymodels.status.inactive': 'Stopped',
'models.mymodels.status.degrade': 'Abnormal',
'models.mymodels.status.active': 'Active'
'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.',
'models.form.kvCache.tips2':
'KV cache is only supported when using built-in inference backends (vLLM or SGLang).'
};
// ========== To-Do: Translate Keys (Remove After Translation) ==========
@@ -247,5 +253,7 @@ export default {
// 39. 'models.mymodels.status.inactive': 'Stopped',
// 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': 'KV cache is only supported when using built-in inference backends (vLLM or SGLang).'
// ========== End of To-Do List ==========
+10 -1
View File
@@ -187,6 +187,8 @@ export default {
'models.form.chunkSize': 'Размер чанков кэша',
'models.form.maxCPUSize': 'Максимальный размер CPU кэша (ГиБ)',
'models.form.remoteURL': 'URL удаленного хранилища',
'models.form.remoteURL.tips':
'Refer to the <a href="https://docs.lmcache.ai/api_reference/configurations.html" target="_blank">configuration documentation</a> for details.',
'models.form.runCommandPlaceholder':
'напр., vllm serve Qwen/Qwen2.5-1.5B-Instruct',
'models.accessSettings.public': 'Публичный',
@@ -206,11 +208,18 @@ export default {
'Вы можете указать точное количество GPU на реплику.',
'models.mymodels.status.inactive': 'Stopped',
'models.mymodels.status.degrade': 'Abnormal',
'models.mymodels.status.active': 'Active'
'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.',
'models.form.kvCache.tips2':
'KV cache is only supported when using built-in inference backends (vLLM or SGLang).'
};
// ========== To-Do: Translate Keys (Remove After Translation) ==========
// 1. 'models.mymodels.status.inactive': 'Stopped',
// 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': 'KV cache is only supported when using built-in inference backends (vLLM or SGLang).';
// ========== End of To-Do List ==========
+7 -1
View File
@@ -176,6 +176,8 @@ export default {
'models.form.chunkSize': '缓存分块大小',
'models.form.maxCPUSize': 'CPU 缓存最大占用 (GiB)',
'models.form.remoteURL': '远程存储地址',
'models.form.remoteURL.tips':
'参考 <a href="https://docs.lmcache.ai/api_reference/configurations.html" target="_blank">配置文档</a> 配置',
'models.form.runCommandPlaceholder':
'例如,vllm serve Qwen/Qwen2.5-1.5B-Instruct',
'models.accessSettings.public': '公开',
@@ -194,5 +196,9 @@ export default {
'models.form.gpusAllocationType.custom.tips': '您可以指定每个副本的 GPU 数量',
'models.mymodels.status.inactive': '已停止',
'models.mymodels.status.degrade': '异常',
'models.mymodels.status.active': '可用'
'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。'
};