-
### Describe the bug
I'm trying to use OPEN AI LIKE API on Bolt but don't seem to work at at all.
I set the environment variables and modified constants.ts file to include Open Ai Like, but still…
-
## Problem
Most of our current marketing focuses on specific products, this leads to several problems.
1. We don't communicate the numerous use cases of our products.
2. We don't communicate ho…
-
Hi, thanks for your great work! I have read your MovieChat+ paper and noticed that the Zero-shot QA Evaluation result of MovieChat on EgoSchema is 53.5, while the evaluation result in this CVPR paper(…
-
### Problem Statement
I'm frustrated when the app tells me to reuse threads instead of creating new ones. This is completely arbitrary since its all hosted on my machine, and I want to be able to cre…
-
For the example in this page: https://github.com/mit-han-lab/llm-awq/tree/main/tinychat#usage
You can easily inference on images:
python vlm_demo_new.py \
--model-path VILA1.5-13b-AWQ \
…
-
### Context:
Using video_chat
### Error:
`**openai.error.InvalidRequestError: This is a chat model and not supported in the v1/completions endpoint. Did you mean to use v1/chat/completions?**`
…
-
### Problem Description
runTrace.sh the vLLM benchmark failed
### Operating System
Ubuntu22.04 in the docker image rocm/vllm-dev:20241025-tuned
### CPU
AMD EPYC 9654 96-Core Processor
### GPU
A…
-
Great work! I notice the LLaVA-NeXT-Qwen2 (image model) can achieve a surprising 49.5 Video-MME results. In contrast, the LLaVA-NeXT-Video (Llama3) can only achieve a 30+ Video-MME score (according to…
-
### 插件版本号 | Version
1.7.5
### 平台 | Platform
macOS
### 浏览器 | Browsers
Arc 1.49.1
### 插件类型 | Extension Type
浏览器插件 | Browser Extension
### 请描述这个Bug | Describe the bug
The vide…
-
Is it not supported by streaming output?