Spaces:
Sleeping
Sleeping
model: | |
arch: mini_gpt4_llama_v2 | |
model_type: pretrain_vicuna | |
freeze_vit: True | |
freeze_qformer: True | |
max_txt_len: 256 | |
low_resource: False | |
image_size: 224 | |
end_sym: "</s>" | |
llama_model: "meta-llama/Llama-2-7b-chat-hf" | |
ckpt: "checkpoints/image_llama2_checkpoint.pth" | |
use_grad_checkpoint: True | |
chat_template: True | |
lora_r: 64 | |
lora_alpha: 16 | |
length: 50 | |
use_grad_checkpoint_llm: True | |
max_context_len: 3600 | |
token_pooling: True | |
datasets: | |
cmd_video: # 15938 | |
batch_size: 4 | |
vis_processor: | |
train: | |
name: "blip2_image_train" | |
image_size: 224 | |
text_processor: | |
train: | |
name: "blip_caption" | |
sample_ratio: 100 | |
webvid: # 42387 | |
batch_size: 4 | |
vis_processor: | |
train: | |
name: "blip2_image_train" | |
image_size: 224 | |
text_processor: | |
train: | |
name: "blip_caption" | |
sample_ratio: 50 | |
run: | |
task: image_text_pretrain | |
# optimizer | |
lr_sched: "linear_warmup_cosine_lr" | |
init_lr: 1e-4 | |
min_lr: 8e-5 | |
warmup_lr: 1e-6 | |
weight_decay: 0.05 | |
max_epoch: 50 | |
num_workers: 16 | |
warmup_steps: 1000 | |
iters_per_epoch: 1000 | |
seed: 42 | |
output_dir: "training_output/cmd_webvid_pretrain" | |
amp: True | |
resume_ckpt_path: null | |
evaluate: False | |
train_splits: ["train"] | |
device: "cuda" | |
world_size: 1 | |
dist_url: "env://" | |
distributed: True | |