theluyuan/rwkv-runner
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

expose global_penalty

Browse Source
This commit is contained in:
josc146
2024-03-02 17:50:41 +08:00
parent 53a5574080
commit 4f14074a75
6 changed files with 92 additions and 31 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
frontend/src/_locales/ja/main.json
View File

@@ -345,5 +345,7 @@
"Quantized Layers": "量子化されたレイヤー",
"Number of the neural network layers quantized with current precision, the more you quantize, the lower the VRAM usage, but the quality correspondingly decreases.": "現在の精度で量子化されたニューラルネットワークのレイヤーの数、量子化するほどVRAMの使用量が低くなりますが、品質も相応に低下します。",
"Parallel Token Chunk Size": "並列トークンチャンクサイズ",
"Maximum tokens to be processed in parallel at once. For high end GPUs, this could be 64 or 128 (faster).": "一度に並列で処理される最大トークン数。高性能なGPUの場合、64または128になります（高速）。"
"Maximum tokens to be processed in parallel at once. For high end GPUs, this could be 64 or 128 (faster).": "一度に並列で処理される最大トークン数。高性能なGPUの場合、64または128になります（高速）。",
"Global Penalty": "グローバルペナルティ",
"When generating a response, whether to include the submitted prompt as a penalty factor. By turning this off, you will get the same generated results as official RWKV Gradio. If you find duplicate results in the generated results, turning this on can help avoid generating duplicates.": "レスポンスを生成する際、提出されたプロンプトをペナルティ要因として含めるかどうか。これをオフにすると、公式RWKV Gradioと同じ生成結果を得ることができます。生成された結果に重複がある場合、これをオンにすることで重複の生成を回避するのに役立ちます。"
}

4
frontend/src/_locales/zh-hans/main.json
View File

@@ -345,5 +345,7 @@
"Quantized Layers": "量化层数",
"Number of the neural network layers quantized with current precision, the more you quantize, the lower the VRAM usage, but the quality correspondingly decreases.": "神经网络以当前精度量化的层数, 量化越多, 占用显存越低, 但质量相应下降",
"Parallel Token Chunk Size": "并行Token块大小",
"Maximum tokens to be processed in parallel at once. For high end GPUs, this could be 64 or 128 (faster).": "一次最多可以并行处理的token数量. 对于高端显卡, 这可以是64或128 (更快)"
"Maximum tokens to be processed in parallel at once. For high end GPUs, this could be 64 or 128 (faster).": "一次最多可以并行处理的token数量. 对于高端显卡, 这可以是64或128 (更快)",
"Global Penalty": "全局惩罚",
"When generating a response, whether to include the submitted prompt as a penalty factor. By turning this off, you will get the same generated results as official RWKV Gradio. If you find duplicate results in the generated results, turning this on can help avoid generating duplicates.": "生成响应时, 是否将提交的prompt也纳入到惩罚项. 关闭此项将得到与RWKV官方Gradio完全一致的生成结果. 如果你发现生成结果出现重复, 那么开启此项有助于避免生成重复"
}