Peng
commited on
Upload folder using huggingface_hub
Browse files- checkpoints/latest-checkpoint.pt +3 -0
- config.json +59 -0
checkpoints/latest-checkpoint.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:617036333bd331dcd883691d78f08e723cbd6583deaea993a672441144bc9c26
|
| 3 |
+
size 13520988270
|
config.json
ADDED
|
@@ -0,0 +1,59 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"dataset": {
|
| 3 |
+
"align_stage_components": [
|
| 4 |
+
"download/llava-laion-cc-sbu-558k/chat.json",
|
| 5 |
+
"download/llava-laion-cc-sbu-558k"
|
| 6 |
+
],
|
| 7 |
+
"dataset_id": "llava-v15",
|
| 8 |
+
"dataset_root_dir": "data",
|
| 9 |
+
"finetune_stage_components": [
|
| 10 |
+
"download/llava-v1.5-instruct/llava_v1_5_mix665k.json",
|
| 11 |
+
"download/llava-v1.5-instruct"
|
| 12 |
+
],
|
| 13 |
+
"type": "llava-v15"
|
| 14 |
+
},
|
| 15 |
+
"hf_token": ".hf_token",
|
| 16 |
+
"model": {
|
| 17 |
+
"align_epochs": 1,
|
| 18 |
+
"align_global_batch_size": 256,
|
| 19 |
+
"align_learning_rate": 0.001,
|
| 20 |
+
"align_lr_scheduler_type": "linear-warmup+cosine-decay",
|
| 21 |
+
"align_max_grad_norm": 1.0,
|
| 22 |
+
"align_max_steps": null,
|
| 23 |
+
"align_per_device_batch_size": 16,
|
| 24 |
+
"align_train_strategy": "fsdp-shard-grad-op",
|
| 25 |
+
"align_warmup_ratio": 0.03,
|
| 26 |
+
"align_weight_decay": 0.0,
|
| 27 |
+
"arch_specifier": "no-align+gelu-mlp",
|
| 28 |
+
"enable_gradient_checkpointing": true,
|
| 29 |
+
"enable_mixed_precision_training": true,
|
| 30 |
+
"finetune_epochs": 1,
|
| 31 |
+
"finetune_global_batch_size": 128,
|
| 32 |
+
"finetune_learning_rate": 2e-05,
|
| 33 |
+
"finetune_lr_scheduler_type": "linear-warmup+cosine-decay",
|
| 34 |
+
"finetune_max_grad_norm": 1.0,
|
| 35 |
+
"finetune_max_steps": null,
|
| 36 |
+
"finetune_per_device_batch_size": 16,
|
| 37 |
+
"finetune_train_strategy": "fsdp-full-shard",
|
| 38 |
+
"finetune_warmup_ratio": 0.03,
|
| 39 |
+
"finetune_weight_decay": 0.1,
|
| 40 |
+
"image_resize_strategy": "letterbox",
|
| 41 |
+
"llm_backbone_id": "vicuna-v15-7b",
|
| 42 |
+
"llm_max_length": 2048,
|
| 43 |
+
"model_id": "custom-convnext+7b",
|
| 44 |
+
"reduce_in_full_precision": false,
|
| 45 |
+
"type": "custom-convnext+7b",
|
| 46 |
+
"vision_backbone_id": "convnext-xxlarge-clip-laion2b"
|
| 47 |
+
},
|
| 48 |
+
"pretrained_checkpoint": null,
|
| 49 |
+
"run_id": "custom-convnext+7b",
|
| 50 |
+
"run_root_dir": "runs",
|
| 51 |
+
"seed": 7,
|
| 52 |
+
"stage": "finetune",
|
| 53 |
+
"trackers": [
|
| 54 |
+
"jsonl",
|
| 55 |
+
"wandb"
|
| 56 |
+
],
|
| 57 |
+
"wandb_entity": null,
|
| 58 |
+
"wandb_project": "prismatic"
|
| 59 |
+
}
|