Qwen1.5 72B Chat GPTQ is an open-source language model by LoneStriker. Features: 72b LLM, VRAM: 45.4GB, Context: 32K, License: other, Quantized, LLM Explorer Score: 0.24, ELO: 1233, Arc: 68.5, HellaSwag: 86.4, MMLU: 77.4, GSM8K: 20.4.
Qwen1.5 72B Chat GPTQ Benchmarks
nn.n% — How the model compares to the reference models: Anthropic Sonnet 3.5 ("so35"), GPT-4o ("gpt4o") or GPT-4 ("gpt4").
Qwen1.5 72B Chat GPTQ Parameters and Internals
| Model Type | | text generation, chat model |
|
| Additional Notes | | The beta version does not include GQA and the mixture of SWA and full attention. DPO improves human preference but lowers benchmark evaluation. |
|
| Supported Languages | | en (multilingual capabilities include English) |
|
| Training Details |
| Methodology: | | Supervised finetuning and direct preference optimization (DPO) |
|
| Context Length: | |
| Model Architecture: | | Transformer architecture with SwiGLU activation, attention QKV bias, group query attention, mixture of sliding window attention and full attention |
|
|
| Input Output |
| Accepted Modalities: | |
| Performance Tips: | | Use provided hyper-parameters in 'generation_config.json' for optimal performance. |
|
|
Best Alternatives to Qwen1.5 72B Chat GPTQ