{ | |
"folder": "phi_1_5_672_final_every_2_v3", | |
"cuda": 0, | |
"encoder": "apple/DFN5B-CLIP-ViT-H-14", | |
"tokenizer": "microsoft/phi-1_5", | |
"decoder": "miguelcarv/phi-1_5-slimorca", | |
"train_decoder": false, | |
"use_dropout": true, | |
"first_level": true, | |
"level": 3, | |
"cross_interval": 2, | |
"from_layer": 0, | |
"reduce": 1, | |
"beams": 3, | |
"max_tokens": 512, | |
"lr": 5e-05, | |
"weight_decay": 0, | |
"encoder_dtype": "torch.bfloat16", | |
"decoder_dtype": "torch.bfloat16", | |
"batch": 2, | |
"acc": 64, | |
"debug": false, | |
"llavar": "/tmp/llavar", | |
"textcaps": "/tmp/textcaps", | |
"pt": "/tmp/stage_3", | |
"from_model": "phi_1_5_672_stage_2_every_2", | |
"notes": null | |
} |