瑞
瑞
I tried to modify this function, ```c static int mp4_h265_write_nal(mp4_h26x_writer_t *h, const unsigned char *nal, int sizeof_nal, unsigned timeStamp90kHz_next) { int payload_type = (nal[0] >> 1) & 0x3f; int slice_type;...
你好,deepspeed_config是没有做改动的。 下面是完整的日志 ``` (Belle) ➜ train git:(main) deepspeed --num_gpus=1 finetune.py --model_config_file run_config/Llama_config.json --deepspeed run_config/deepspeed_config.json [2023-04-08 12:57:32,667] [WARNING] [runner.py:186:fetch_hostfile] Unable to find hostfile, will proceed with training with local resources only....
基模型是,基于llama-7b-hf合并了chinese_alpaca_plus_lora_7b和chinese_llama_plus_lora_7b的模型llama-7b-hf-chinese。 要合并的模型是,使用alpaca-lora基于上面的llama-7b-hf-chinese为基础模型,搭配一个自己整理的中文数据集alpaca_data_chinese.json用Lora模式训练的权重。 ```bash python finetune.py \ --base_model '/srv/hdd/datasets/llama-7b-hf-chinese' \ --data_path './alpaca_data_chinese.json' \ --num_epochs 3 \ --val_set_size 2000 \ --output_dir './lora-alpaca' ``` 最后的合并命令是 ```bash python /srv/ssd/github/Chinese-LLaMA-Alpaca/scripts/merge_llama_with_chinese_lora.py \ --base_model /srv/hdd/datasets/llama-7b-hf-chinese \ --lora_model...
好的,我明天再试试,另外有一个疑问请教一下,新训练出来的Lora权重是没有tokenizer的,我是该复制llama-7b-hf中的tokenizer.model,还是复制复制llama-7b-hf-chinese中的呢?
降peft版本到0.2.0试试,也会报错: