feat: update for mosaic

This commit is contained in:
zanussbaum 2023-04-07 16:54:29 -04:00
parent 1b5e660476
commit aa8dd7a636

View File

@ -2,14 +2,14 @@
model_name: "EleutherAI/gpt-j-6B" model_name: "EleutherAI/gpt-j-6B"
tokenizer_name: "EleutherAI/gpt-j-6B" tokenizer_name: "EleutherAI/gpt-j-6B"
gradient_checkpointing: true gradient_checkpointing: true
save_name: "nomic-ai/gpt4all-gptj-multinode-deepspeed" save_name: "nomic-ai/gpt4all-mosaic"
# dataset # dataset
streaming: false streaming: false
num_proc: 64 num_proc: 64
dataset_path: "data_multiplus" dataset_path: "nomic-ai/turbo-500k-multi"
max_length: 1024 max_length: 1024
batch_size: 32 batch_size: 16
# train dynamics # train dynamics
lr: 2.0e-5 lr: 2.0e-5
@ -23,7 +23,7 @@ output_dir: "ckpts/gpt4all-gptj-multinode"
checkpoint: null checkpoint: null
lora: false lora: false
warmup_steps: 500 warmup_steps: 500
num_epochs: 4 num_epochs: 2
# logging # logging
wandb: true wandb: true