AI is happening. The future is updating.Llama-3.1-1-million-cxt-Dark-Planet-8B-GGUF
It is a LLama 3.1 model, max context of 1 million tokens, with additional steps to improve generation and re-mastered source and ggufs in float 32 (32 bit precision). There are TWO quant sets here, "reg" and "V1.01" (noted in filename). The latter has modifications to address some issues related to non-stop/overly long gen and/or repeat "end paragraph" issues. I am keeping the org quants too, because of the difference in creative generation between the two versions is very strong. I am not saying "reg" is better than "v1.01", they are just different, and you should have the choice between both in my opinion. This model was converted to Nvidia's new "UltraLong8B" long context Llama 3.1 model structure (https://huggingface.co/nvidia/Llama-3.1-8B-UltraLong-1M-Instruct) which allowed full transfer of "Dark Planet 8B" in all it's "glory" so to speak. Due to Nvidia's structure, the new Dark Planet has attained far greater long generation not only in terms of context, but also coherence too. There is a also a bump in overall performance as well.
It is a LLama 3.1 model, max context of 1 million tokens, with additional steps to improve generation and re-mastered source and ggufs in float 32 (32 bit precision). There are TWO quant sets here, "reg" and "V1.01" (noted in filename). The latter has modifications to address some issues related to non-stop/overly long gen and/or repeat "end paragraph" issues. I am keeping the org quants too, because of the difference in creative generation between the two versions is very strong. I am not saying "reg" is better than "v1.01", they are just different, and you should have the choice between both in my opinion. This model was converted to Nvidia's new "UltraLong8B" long context Llama 3.1 model structure (https://huggingface.co/nvidia/Llama-3.1-8B-UltraLong-1M-Instruct) which allowed full transfer of "Dark Planet 8B" in all it's "glory" so to speak. Due to Nvidia's structure, the new Dark Planet has attained far greater long generation not only in terms of context, but also coherence too. There is a also a bump in overall performance as well.