Spaces:
Runtime error
Runtime error
temp: hardcode custom maxTokens
Browse files
src/lib/components/InferencePlayground/InferencePlaygroundGenerationConfig.svelte
CHANGED
|
@@ -6,7 +6,54 @@
|
|
| 6 |
export let conversation: Conversation;
|
| 7 |
export let classNames = "";
|
| 8 |
|
| 9 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 10 |
$: maxTokens = Math.min(modelMaxLength ?? GENERATION_CONFIG_SETTINGS["max_tokens"].max, 64_000);
|
| 11 |
</script>
|
| 12 |
|
|
|
|
| 6 |
export let conversation: Conversation;
|
| 7 |
export let classNames = "";
|
| 8 |
|
| 9 |
+
const customMaxTokens: { [key: string]: number } = {
|
| 10 |
+
"01-ai/Yi-1.5-34B-Chat": 2048,
|
| 11 |
+
"HuggingFaceM4/idefics-9b-instruct": 2048,
|
| 12 |
+
"deepseek-ai/DeepSeek-Coder-V2-Instruct": 16384,
|
| 13 |
+
"bigcode/starcoder": 8192,
|
| 14 |
+
"bigcode/starcoderplus": 8192,
|
| 15 |
+
"HuggingFaceH4/starcoderbase-finetuned-oasst1": 8192,
|
| 16 |
+
"google/gemma-7b": 8192,
|
| 17 |
+
"google/gemma-1.1-7b-it": 8192,
|
| 18 |
+
"google/gemma-2b": 8192,
|
| 19 |
+
"google/gemma-1.1-2b-it": 8192,
|
| 20 |
+
"google/gemma-2-27b-it": 8192,
|
| 21 |
+
"google/gemma-2-9b-it": 4096,
|
| 22 |
+
"google/gemma-2-2b-it": 8192,
|
| 23 |
+
"tiiuae/falcon-7b": 8192,
|
| 24 |
+
"tiiuae/falcon-7b-instruct": 8192,
|
| 25 |
+
"timdettmers/guanaco-33b-merged": 2048,
|
| 26 |
+
"mistralai/Mixtral-8x7B-Instruct-v0.1": 32768,
|
| 27 |
+
"Qwen/Qwen2.5-72B-Instruct": 32768,
|
| 28 |
+
"meta-llama/Meta-Llama-3-70B-Instruct": 8192,
|
| 29 |
+
"CohereForAI/c4ai-command-r-plus-08-2024": 32768,
|
| 30 |
+
"NousResearch/Nous-Hermes-2-Mixtral-8x7B-DPO": 32768,
|
| 31 |
+
"meta-llama/Llama-2-70b-chat-hf": 8192,
|
| 32 |
+
"HuggingFaceH4/zephyr-7b-alpha": 17432,
|
| 33 |
+
"HuggingFaceH4/zephyr-7b-beta": 32768,
|
| 34 |
+
"mistralai/Mistral-7B-Instruct-v0.1": 32768,
|
| 35 |
+
"mistralai/Mistral-7B-Instruct-v0.2": 32768,
|
| 36 |
+
"mistralai/Mistral-7B-Instruct-v0.3": 32768,
|
| 37 |
+
"mistralai/Mistral-Nemo-Instruct-2407": 32768,
|
| 38 |
+
"meta-llama/Meta-Llama-3-8B-Instruct": 8192,
|
| 39 |
+
"mistralai/Mistral-7B-v0.1": 32768,
|
| 40 |
+
"bigcode/starcoder2-3b": 16384,
|
| 41 |
+
"bigcode/starcoder2-15b": 16384,
|
| 42 |
+
"HuggingFaceH4/starchat2-15b-v0.1": 16384,
|
| 43 |
+
"codellama/CodeLlama-7b-hf": 8192,
|
| 44 |
+
"codellama/CodeLlama-13b-hf": 8192,
|
| 45 |
+
"codellama/CodeLlama-34b-Instruct-hf": 8192,
|
| 46 |
+
"meta-llama/Llama-2-7b-chat-hf": 8192,
|
| 47 |
+
"meta-llama/Llama-2-13b-chat-hf": 8192,
|
| 48 |
+
"OpenAssistant/oasst-sft-6-llama-30b": 2048,
|
| 49 |
+
"TheBloke/vicuna-7B-v1.5-GPTQ": 2048,
|
| 50 |
+
"HuggingFaceH4/starchat-beta": 8192,
|
| 51 |
+
"bigcode/octocoder": 8192,
|
| 52 |
+
"vwxyzjn/starcoderbase-triviaqa": 8192,
|
| 53 |
+
"lvwerra/starcoderbase-gsm8k": 8192,
|
| 54 |
+
} as const;
|
| 55 |
+
|
| 56 |
+
$: modelMaxLength = customMaxTokens[conversation.model.id] ?? conversation.model.tokenizerConfig.model_max_length;
|
| 57 |
$: maxTokens = Math.min(modelMaxLength ?? GENERATION_CONFIG_SETTINGS["max_tokens"].max, 64_000);
|
| 58 |
</script>
|
| 59 |
|