Instructions to use Stanford-ILIAD/minivla-vq-bridge-prismatic with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use Stanford-ILIAD/minivla-vq-bridge-prismatic with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("image-text-to-text", model="Stanford-ILIAD/minivla-vq-bridge-prismatic")# Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("Stanford-ILIAD/minivla-vq-bridge-prismatic", dtype="auto") - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- vLLM
How to use Stanford-ILIAD/minivla-vq-bridge-prismatic with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "Stanford-ILIAD/minivla-vq-bridge-prismatic" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "Stanford-ILIAD/minivla-vq-bridge-prismatic", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }'Use Docker
docker model run hf.co/Stanford-ILIAD/minivla-vq-bridge-prismatic
- SGLang
How to use Stanford-ILIAD/minivla-vq-bridge-prismatic with SGLang:
Install from pip and serve model
# Install SGLang from pip: pip install sglang # Start the SGLang server: python3 -m sglang.launch_server \ --model-path "Stanford-ILIAD/minivla-vq-bridge-prismatic" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "Stanford-ILIAD/minivla-vq-bridge-prismatic", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }'Use Docker images
docker run --gpus all \ --shm-size 32g \ -p 30000:30000 \ -v ~/.cache/huggingface:/root/.cache/huggingface \ --env "HF_TOKEN=<secret>" \ --ipc=host \ lmsysorg/sglang:latest \ python3 -m sglang.launch_server \ --model-path "Stanford-ILIAD/minivla-vq-bridge-prismatic" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "Stanford-ILIAD/minivla-vq-bridge-prismatic", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }' - Docker Model Runner
How to use Stanford-ILIAD/minivla-vq-bridge-prismatic with Docker Model Runner:
docker model run hf.co/Stanford-ILIAD/minivla-vq-bridge-prismatic
Upload folder using huggingface_hub
Browse files
.gitattributes
CHANGED
|
@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
|
|
| 33 |
*.zip filter=lfs diff=lfs merge=lfs -text
|
| 34 |
*.zst filter=lfs diff=lfs merge=lfs -text
|
| 35 |
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
|
|
|
|
|
| 33 |
*.zip filter=lfs diff=lfs merge=lfs -text
|
| 34 |
*.zst filter=lfs diff=lfs merge=lfs -text
|
| 35 |
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
| 36 |
+
prism-qwen25-dinosiglip-224px+0_5b+mx-bridge+n1+b16+x7--bridge_vq_extra_tokenizer_evenlowerlr.jsonl filter=lfs diff=lfs merge=lfs -text
|
checkpoints/step-362500-epoch-21-loss=0.2259.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2b1828f4fb96b0b7a4f3d191fde4ee96938b293c70f8616fb44dd85f5c85cadc
|
| 3 |
+
size 5554882540
|
config.json
ADDED
|
@@ -0,0 +1,48 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"data_root_dir": "/iliad/group/datasets/OXE_OCTO",
|
| 3 |
+
"hf_token": ".hf_token",
|
| 4 |
+
"image_aug": false,
|
| 5 |
+
"is_resume": true,
|
| 6 |
+
"pretrained_checkpoint": "runs/prism-qwen25-dinosiglip-224px+0_5b+mx-bridge+n1+b16+x7--bridge_vq_extra_tokenizer_lowlr/checkpoints/step-305000-epoch-18-loss=0.5682.pt",
|
| 7 |
+
"resume_epoch": 18,
|
| 8 |
+
"resume_step": 305000,
|
| 9 |
+
"run_id": "prism-qwen25-dinosiglip-224px+0_5b+mx-bridge+n1+b16+x7--bridge_vq_extra_tokenizer_evenlowerlr",
|
| 10 |
+
"run_id_note": "bridge_vq_extra_tokenizer_evenlowerlr",
|
| 11 |
+
"run_root_dir": "runs",
|
| 12 |
+
"save_interval": 2500,
|
| 13 |
+
"seed": 7,
|
| 14 |
+
"trackers": [
|
| 15 |
+
"jsonl",
|
| 16 |
+
"wandb"
|
| 17 |
+
],
|
| 18 |
+
"vla": {
|
| 19 |
+
"action_tokenizer": "bridge_vq_extra_action_tokenizer",
|
| 20 |
+
"base_vlm": "prism-qwen25-extra-dinosiglip-224px+0_5b",
|
| 21 |
+
"data_mix": "bridge_dataset",
|
| 22 |
+
"enable_gradient_checkpointing": true,
|
| 23 |
+
"enable_mixed_precision_training": true,
|
| 24 |
+
"epochs": 1000,
|
| 25 |
+
"expected_world_size": 8,
|
| 26 |
+
"freeze_llm_backbone": false,
|
| 27 |
+
"freeze_vision_backbone": false,
|
| 28 |
+
"global_batch_size": 128,
|
| 29 |
+
"image_sequence_len": 1,
|
| 30 |
+
"learning_rate": 1e-06,
|
| 31 |
+
"lr_scheduler_type": "constant",
|
| 32 |
+
"max_grad_norm": 1.0,
|
| 33 |
+
"max_steps": null,
|
| 34 |
+
"per_device_batch_size": 16,
|
| 35 |
+
"reduce_in_full_precision": true,
|
| 36 |
+
"save_every_n_steps": 25000,
|
| 37 |
+
"shuffle_buffer_size": 256000,
|
| 38 |
+
"train_strategy": "fsdp-full-shard",
|
| 39 |
+
"type": "prism-qwen25-dinosiglip-224px+0_5b+mx-bridge",
|
| 40 |
+
"unfreeze_last_llm_layer": false,
|
| 41 |
+
"use_wrist_image": false,
|
| 42 |
+
"vla_id": "prism-qwen25-dinosiglip-224px+0_5b+mx-bridge",
|
| 43 |
+
"warmup_ratio": 0.0,
|
| 44 |
+
"weight_decay": 0.0
|
| 45 |
+
},
|
| 46 |
+
"wandb_entity": null,
|
| 47 |
+
"wandb_project": "prismatic"
|
| 48 |
+
}
|
config.yaml
ADDED
|
@@ -0,0 +1,44 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
data_root_dir: /iliad/group/datasets/OXE_OCTO
|
| 2 |
+
hf_token: .hf_token
|
| 3 |
+
image_aug: false
|
| 4 |
+
is_resume: true
|
| 5 |
+
pretrained_checkpoint: runs/prism-qwen25-dinosiglip-224px+0_5b+mx-bridge+n1+b16+x7--bridge_vq_extra_tokenizer_lowlr/checkpoints/step-305000-epoch-18-loss=0.5682.pt
|
| 6 |
+
resume_epoch: 18
|
| 7 |
+
resume_step: 305000
|
| 8 |
+
run_id: prism-qwen25-dinosiglip-224px+0_5b+mx-bridge+n1+b16+x7--bridge_vq_extra_tokenizer_evenlowerlr
|
| 9 |
+
run_id_note: bridge_vq_extra_tokenizer_evenlowerlr
|
| 10 |
+
run_root_dir: runs
|
| 11 |
+
save_interval: 2500
|
| 12 |
+
seed: 7
|
| 13 |
+
trackers:
|
| 14 |
+
- jsonl
|
| 15 |
+
- wandb
|
| 16 |
+
vla:
|
| 17 |
+
action_tokenizer: bridge_vq_extra_action_tokenizer
|
| 18 |
+
base_vlm: prism-qwen25-dinosiglip-224px+0_5b
|
| 19 |
+
data_mix: bridge_dataset
|
| 20 |
+
enable_gradient_checkpointing: true
|
| 21 |
+
enable_mixed_precision_training: true
|
| 22 |
+
epochs: 1000
|
| 23 |
+
expected_world_size: 8
|
| 24 |
+
freeze_llm_backbone: false
|
| 25 |
+
freeze_vision_backbone: false
|
| 26 |
+
global_batch_size: 128
|
| 27 |
+
image_sequence_len: 1
|
| 28 |
+
learning_rate: 1.0e-06
|
| 29 |
+
lr_scheduler_type: constant
|
| 30 |
+
max_grad_norm: 1.0
|
| 31 |
+
max_steps: null
|
| 32 |
+
per_device_batch_size: 16
|
| 33 |
+
reduce_in_full_precision: true
|
| 34 |
+
save_every_n_steps: 25000
|
| 35 |
+
shuffle_buffer_size: 256000
|
| 36 |
+
train_strategy: fsdp-full-shard
|
| 37 |
+
type: prism-qwen25-dinosiglip-224px+0_5b+mx-bridge
|
| 38 |
+
unfreeze_last_llm_layer: false
|
| 39 |
+
use_wrist_image: false
|
| 40 |
+
vla_id: prism-qwen25-dinosiglip-224px+0_5b+mx-bridge
|
| 41 |
+
warmup_ratio: 0.0
|
| 42 |
+
weight_decay: 0.0
|
| 43 |
+
wandb_entity: null
|
| 44 |
+
wandb_project: prismatic
|
dataset_statistics.json
ADDED
|
@@ -0,0 +1,127 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"bridge_dataset": {
|
| 3 |
+
"action": {
|
| 4 |
+
"mean": [
|
| 5 |
+
0.00023341951600741595,
|
| 6 |
+
0.00013004841457586735,
|
| 7 |
+
-0.00012762370170094073,
|
| 8 |
+
-0.00015565504145342857,
|
| 9 |
+
-0.00040393343078903854,
|
| 10 |
+
0.00023557768145110458,
|
| 11 |
+
0.5764579176902771
|
| 12 |
+
],
|
| 13 |
+
"std": [
|
| 14 |
+
0.009765920229256153,
|
| 15 |
+
0.013689135201275349,
|
| 16 |
+
0.01266736164689064,
|
| 17 |
+
0.02853427641093731,
|
| 18 |
+
0.030637990683317184,
|
| 19 |
+
0.0769147127866745,
|
| 20 |
+
0.49736854434013367
|
| 21 |
+
],
|
| 22 |
+
"max": [
|
| 23 |
+
0.41691166162490845,
|
| 24 |
+
0.25864794850349426,
|
| 25 |
+
0.21218234300613403,
|
| 26 |
+
3.122201919555664,
|
| 27 |
+
1.8618112802505493,
|
| 28 |
+
6.280478477478027,
|
| 29 |
+
1.0
|
| 30 |
+
],
|
| 31 |
+
"min": [
|
| 32 |
+
-0.4007510244846344,
|
| 33 |
+
-0.13874775171279907,
|
| 34 |
+
-0.22553899884223938,
|
| 35 |
+
-3.2010786533355713,
|
| 36 |
+
-1.8618112802505493,
|
| 37 |
+
-6.279075622558594,
|
| 38 |
+
0.0
|
| 39 |
+
],
|
| 40 |
+
"q01": [
|
| 41 |
+
-0.02872725307941437,
|
| 42 |
+
-0.04170349963009357,
|
| 43 |
+
-0.026093858778476715,
|
| 44 |
+
-0.08092105075716972,
|
| 45 |
+
-0.09288699507713317,
|
| 46 |
+
-0.20718276381492615,
|
| 47 |
+
0.0
|
| 48 |
+
],
|
| 49 |
+
"q99": [
|
| 50 |
+
0.028309678435325586,
|
| 51 |
+
0.040855254605412394,
|
| 52 |
+
0.040161586627364146,
|
| 53 |
+
0.08192047759890528,
|
| 54 |
+
0.07792850524187081,
|
| 55 |
+
0.20382574498653397,
|
| 56 |
+
1.0
|
| 57 |
+
],
|
| 58 |
+
"mask": [
|
| 59 |
+
true,
|
| 60 |
+
true,
|
| 61 |
+
true,
|
| 62 |
+
true,
|
| 63 |
+
true,
|
| 64 |
+
true,
|
| 65 |
+
false
|
| 66 |
+
]
|
| 67 |
+
},
|
| 68 |
+
"proprio": {
|
| 69 |
+
"mean": [
|
| 70 |
+
0.0,
|
| 71 |
+
0.0,
|
| 72 |
+
0.0,
|
| 73 |
+
0.0,
|
| 74 |
+
0.0,
|
| 75 |
+
0.0,
|
| 76 |
+
0.0
|
| 77 |
+
],
|
| 78 |
+
"std": [
|
| 79 |
+
0.0,
|
| 80 |
+
0.0,
|
| 81 |
+
0.0,
|
| 82 |
+
0.0,
|
| 83 |
+
0.0,
|
| 84 |
+
0.0,
|
| 85 |
+
0.0
|
| 86 |
+
],
|
| 87 |
+
"max": [
|
| 88 |
+
0.0,
|
| 89 |
+
0.0,
|
| 90 |
+
0.0,
|
| 91 |
+
0.0,
|
| 92 |
+
0.0,
|
| 93 |
+
0.0,
|
| 94 |
+
0.0
|
| 95 |
+
],
|
| 96 |
+
"min": [
|
| 97 |
+
0.0,
|
| 98 |
+
0.0,
|
| 99 |
+
0.0,
|
| 100 |
+
0.0,
|
| 101 |
+
0.0,
|
| 102 |
+
0.0,
|
| 103 |
+
0.0
|
| 104 |
+
],
|
| 105 |
+
"q01": [
|
| 106 |
+
0.0,
|
| 107 |
+
0.0,
|
| 108 |
+
0.0,
|
| 109 |
+
0.0,
|
| 110 |
+
0.0,
|
| 111 |
+
0.0,
|
| 112 |
+
0.0
|
| 113 |
+
],
|
| 114 |
+
"q99": [
|
| 115 |
+
0.0,
|
| 116 |
+
0.0,
|
| 117 |
+
0.0,
|
| 118 |
+
0.0,
|
| 119 |
+
0.0,
|
| 120 |
+
0.0,
|
| 121 |
+
0.0
|
| 122 |
+
]
|
| 123 |
+
},
|
| 124 |
+
"num_transitions": 2135463,
|
| 125 |
+
"num_trajectories": 60064
|
| 126 |
+
}
|
| 127 |
+
}
|
prism-qwen25-dinosiglip-224px+0_5b+mx-bridge+n1+b16+x7--bridge_vq_extra_tokenizer_evenlowerlr.jsonl
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:79e30022be1b39d861bb03554272c3073a75abd6adf389733eb5b9ccc40fdfc3
|
| 3 |
+
size 76106884
|
run-metrics.jsonl
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
{"hparams": {"data_root_dir": "/iliad/group/datasets/OXE_OCTO", "hf_token": ".hf_token", "image_aug": false, "is_resume": true, "pretrained_checkpoint": "runs/prism-qwen25-dinosiglip-224px+0_5b+mx-bridge+n1+b16+x7--bridge_vq_extra_tokenizer_lowlr/checkpoints/step-305000-epoch-18-loss=0.5682.pt", "resume_epoch": 18, "resume_step": 305000, "run_id": "prism-qwen25-dinosiglip-224px+0_5b+mx-bridge+n1+b16+x7--bridge_vq_extra_tokenizer_evenlowerlr", "run_id_note": "bridge_vq_extra_tokenizer_evenlowerlr", "run_root_dir": "runs", "save_interval": 2500, "seed": 7, "trackers": ["jsonl", "wandb"], "vla": {"action_tokenizer": "bridge_vq_extra_action_tokenizer", "base_vlm": "prism-qwen25-dinosiglip-224px+0_5b", "data_mix": "bridge_dataset", "enable_gradient_checkpointing": true, "enable_mixed_precision_training": true, "epochs": 1000, "expected_world_size": 8, "freeze_llm_backbone": false, "freeze_vision_backbone": false, "global_batch_size": 128, "image_sequence_len": 1, "learning_rate": 1e-06, "lr_scheduler_type": "constant", "max_grad_norm": 1.0, "max_steps": null, "per_device_batch_size": 16, "reduce_in_full_precision": true, "save_every_n_steps": 25000, "shuffle_buffer_size": 256000, "train_strategy": "fsdp-full-shard", "type": "prism-qwen25-dinosiglip-224px+0_5b+mx-bridge", "unfreeze_last_llm_layer": false, "use_wrist_image": false, "vla_id": "prism-qwen25-dinosiglip-224px+0_5b+mx-bridge", "warmup_ratio": 0.0, "weight_decay": 0.0}, "wandb_entity": null, "wandb_project": "prismatic"}, "run_id": "prism-qwen25-dinosiglip-224px+0_5b+mx-bridge+n1+b16+x7--bridge_vq_extra_tokenizer_evenlowerlr"}
|