#FluxArt👅🌐
gloomy tundraBOT
Flux_Agent
sand spruce
Widget: Claude3.5 Sonnet
Reason: ChatCompletion error, req: &openai.ChatCompletionRequest{Model:"gpt-4o", Messages:[]openai.ChatCompletionMessage{openai.ChatCompletionMessage{Role:"system", Content:"First, translate any non-English user input to English. You should always write your prompts in English only.\nAs an AI image prompt engineering expert, generate enhanced prompts based on user input (highest priority) and reference images (if provided).\nNever interact with user but take \"script (⏱,💰)\" and \"https://image.myshell.ai/image/chat/embed_obj/13896/202409181622/Bad Kid #3770.jpeg\" as a whole to generate your prompt.\nFollow these steps: 1. Analyze user input thoroughly. 2. If a reference image is provided, conduct an in-depth analysis: a) Describe overall style, mood, and composition b) Identify and detail specific elements, textures, and color schemes c) Note any unique or standout features 3. Determine image type: photograph, painting, digital art, or 3D rendering. 4. Create an enhanced prompt (strictly limited to 300 words) using techniques optimized for tools like Midjourney: a) User's specific requests (top priority) b) Detailed description of main elements, their relationships, and key features from the reference image (if provided) c) Artistic style, techniques, and any relevant famous artworks or masters with similar styles d) Composition, framing, lighting, and color palette, matching the reference image if applicable e) Texture, material properties, mood, and atmosphere f) Technical parameters relevant to the image type g) Specific details that enhance the overall concept, ensuring consistency with the reference image 5. Ensure prompt coherence and style consistency throughout. 6. Adapt complexity to user's expertise level. 7. Emphasize high-resolution, vivid, and detail-rich image generation. 8. Balance essential concepts with nuanced details, prioritizing user requests while providing clear guidance for AI interpretation. The final prompt must not exceed 300 words. Always aim for high-quality, detailed, and vibrant image generation that closely aligns with both user input and reference image (if provided).", MultiContent:[]openai.ChatMessagePart(nil), Name:"", FunctionCall:(*openai.FunctionCall)(nil)}, openai.ChatCompletionMessage{Role:"user", Content:"", MultiContent:[]openai.ChatMessagePart{openai.ChatMessagePart{Type:"text", Text:"script (⏱,💰)", ImageURL:(*openai.ChatMessageImageURL)(nil)}, openai.ChatMessagePart{Type:"image_url", Text:"", ImageURL:(*openai.ChatMessageImageURL)(0xc013b7ea20)}}, Name:"", FunctionCall:(*openai.FunctionCall)(nil)}}, MaxTokens:5000, Temperature:(*float32)(0xc00ae9f610), TopP:(*float32)(0xc00ae9f614), N:(*int)(nil), Stream:false, Stop:[]string(nil), PresencePenalty:(*float32)(0xc00ae9f618), FrequencyPenalty:(*float32)(0xc00ae9f61c), LogitBias:map[string]int(nil), ResponseFormat:map[string]string(nil), User:"", Functions:[]openai.FunctionDefinition(nil), FunctionCall:interface {}(nil)}: error, status code: 400, message: Value of max_tokens must be less than or equal to 4096.
报错了 @silk river
silk river
@sand spruce脚哥 我也遇到几次
没理解这个错误是如何产生的 ChatCompletion error, req: &openai.ChatCompletionRequest{Model:"gpt-4o"。。怎么还搞来了4o，明明只用了claude 3.5
sand spruce
<@790804607973457940>脚哥 我也遇到几次
没理解这个错误是如何产生的 ChatCompletion error, req: &openai....
log里有个 MaxTokens:5000，但是llm只支持4096
感觉得问问官方的开发了
silk river
那我先改为4096试试吧
gloomy tundraBOT
Flux (Multilingual)
gloomy tundraBOT
FluxMultilingual👅🌐
silk river
log里有个 MaxTokens:5000，但是llm只支持4096
今天没碰到那个报错情况了哦 就按你说了改了MaxToken
多谢指导 🫡
gloomy tundraBOT
FluxArt👅🌐