Datasets:

---

pszemraj

/

LocalLLaMA-posts

like 0

ERROR: type should be string, got " https://arxiv.org/abs/2411.04330\n\n \nWow… just read this article and went on to do some experiments. Qwen 2.5 Coder is trained on 5.5T and as a rule of thumb simplifying what the article says we can go by the idea that if a Trillion tokens > Billion params, we can assume that the model would be sensitive to quantization. And it really does show, this is the perplexity delta compared to f16:\n\n0.5B (T/N ≈ 11,000):\n\n Q6_K: 0.187% worse\n\n Q4_K_M: 3.42% worse\n\n Q3_K_L: 3.98% worse\n\n3B (T/N ≈ 1,833):\n\n Q6_K: 0.41% worse\n\n Q4_K_M: 1.96% worse\n\n Q3_K_M: 76.09% worse (goes into 0.5b result territory while still being larger than 0.5b f16)\n\n7B (T/N ≈ 786 which is below the rule of thumb hence not affected at Q4):\n\n Q6_K: (not tested)\n\n Q4_K_M: 0.95% worse\n\n Q3_K_L: 3.28% worse\n\n\nAll compared to the best result (it is sunday and I am lazy rn so I asked sonnet to do it for me… looks correct 🤷 but dont trust it too much)\n\n7B (T/N ≈ 786):\n\n F16: baseline (8.0125)\n\n Q4_K_M: +0.95%\n\n Q3_K_L: +3.28%\n\n3B (T/N ≈ 1,833):\n\n F16: +1.66%\n\n Q6_K: +2.08%\n\n Q4_K_M: +3.65%\n\n Q3_K_M: +79.02%\n\n0.5B (T/N ≈ 11,000):\n\n F16: +93.79%\n\n Q8_0: +93.89%\n\n Q6_K: +94.15%\n\n Q4_K_M: +100.42%\n\n Q3_K_L: +101.49%\n\n"

ERROR: type should be string, got "https://preview.redd.it/2xpfatehjdfe1.png?width=2410&format=png&auto=webp&s=f21159597fd67e87756a449ec36be14cf68e99fe\n\nhttps://preview.redd.it/442ygmbkjdfe1.png?width=2860&format=png&auto=webp&s=3bb1f177314037fd647f79c3568cb8d93cf22c1c\n\nThe Baichuan-Omni-1.5 is the latest, top-performing model in the Baichuan-omni series. This model is trained and inferred in an end-to-end manner. Compared with Baichuan-omni, this model has significant improvements in text/image/audio/video understanding and text/audio generation, and supports new features such as controllable real-time voice conversations and multi-modal real-time interactions. The main features of Baichuan-Omni-1.5 include:\n\n🔥 Possess Multimodal Understanding and Interaction Capabilities. Baichuan-Omni-1.5 not only supports images, videos, text, and audio as input, and generates high-quality text and voice output, but also supports continuous video and audio streaming, and real-time voice interaction with users. In OminiBench, a comprehensive evaluation benchmark for omnimodal understanding, Baichuan-Omni-1.5 has achieved the first-class level of the open source community and surpassed GPT-4o-mini.\n\n💪 Strong Visual Capability. Baichuan-Omni-1.5 has an average score of 73.3 on the OpenCompass list (comprehensive 10 mainstream multimodal evaluation benchmarks). With the size of 7B, it surpasses mainstream commercial closed-source multimodal large models such as GPT-4o-mini, Gemini 1.5 Pro and Claude 3.5 Sonnet in single-image understanding. In addition, its video understanding performance is also better than GPT-4V and Claude 3.5 Sonnet and open source omnimodal models.\n\n🚀 Leading Medical Image Understanding Capabilities. Baichuan-Omni-1.5 achieved the best performance on GMAI-MMBench and Openmm-Medical. Using only 7B LLM, the average score exceeded Qwen2-VL-72b by 3%, i.e. 80.7% v.s 83.8%.\n\n🎙 Excellent Voice Capabilities. Baichuan-Omni-1.5 supports high-quality, controllable voice bilingual real-time conversations in Chinese and English. It outperforms GPT-4o-realtime in speech understanding tasks (such as ASR and STT, etc.), and demonstrates the highest speech generation performance among open source models in semantic and acoustic evaluation of voice conversations.\n\n🎬 Powerful Real-world Understanding and Other Features. Baichuan-Omni-1.5 further optimizes the many visual understanding capabilities of Baichuan-omni. It can process images of any aspect ratio and up to 1.8 million pixels (such as 1344x1344). It scored 68.8 points on RealWorldQA, surpassing commercial closed-source models such as GPT-4o-mini and recently open-sourced omnimodal models. It scored 85.6/83.6 on the English/Chinese evaluation subsets of MMBench, respectively, which is also in the first echelon of models with the same size.\n\n[Model Link ](https://huggingface.co/baichuan-inc/Baichuan-Omni-1d5)"