DeepSpeedExamples
DeepSpeedExamples copied to clipboard
microsoft
Example models using DeepSpeed
Hi! I trained the model just as you directed, but the model generation is very very bad.It can not even speak a complete sentence...And when I train step3, its reward...
We added a new example to fine-tune LLaMA on 2080Ti-level GPUs. In my environment, with 8 2080Ti GPUs, LLama-7b can be fine-tuned on alpaca-52k dataset at the speed of 1.5...
Hello, I'm running the example script of single node reward model training in this [link](https://github.com/microsoft/DeepSpeedExamples/tree/master/applications/DeepSpeed-Chat/training/step2_reward_model_finetuning/training_scripts/single_node) and get error log like below: ```Traceback (most recent call last): File "/home/bingxing2/gpuuser183/bak/xydu/DeepSpeed-Chat/training/step2_reward_model_finetuning/main.py", line 348,...
The customized model is not in your "Supported Models" list. Can it benefit from Deepspeed chat?
After finishing install successfully, i got this error when ran this command: python train.py --actor-model facebook/opt-1.3b --reward-model facebook/opt-350m --num-gpus 1 ---=== Running Step 1 ===--- Traceback (most recent call last):...
SFT loss
https://github.com/microsoft/DeepSpeedExamples/blob/d570b2cc8a8fd4207c9424744669437d4c68ec43/applications/DeepSpeed-Chat/training/utils/data/data_utils.py#L122 ``` if self.train_phase == 1: return { "input_ids": self.chosen_dataset[idx]["input_ids"], "attention_mask": self.chosen_dataset[idx]["attention_mask"], "labels": self.chosen_dataset[idx]["input_ids"] } ``` In the SFT stage, input_ids and labels are the same, so the loss calculation...
I run the test program use "python train.py --actor-model facebook/opt-13b --reward-model facebook/opt-350m --num-gpus 8".The program can run normally.But I modified the parameter max_ answer_ seq_ len = 1024 and max_prompt_seq_len...
I get 4 4090 gpus, and I want to train the opt-6.7B by using DeepSpeed Chat. Is that possible? I mean I have no idea if I should get a...
For the step2 scoring: `python3 training/step2_reward_model_finetuning/rw_eval.py --model_name_or_path output/reward-models/350m/ ==================Eval result============================ prompt: Human: Please tell me about Microsoft in a few sentence? Assistant: good_ans: Microsoft is a software company that develops,...
