Robotics
Safetensors
vision-language-action-model
Jia-Zeng commited on
Commit
ecfef92
·
verified ·
1 Parent(s): 3c7aedd

Update checkpoint: achieve the success rates of 89.4 and 89.6 in the Easy and Hard settings.

Browse files
Files changed (4) hide show
  1. README.md +3 -3
  2. config.json +1 -1
  3. model.safetensors +1 -1
  4. train_config.json +4 -4
README.md CHANGED
@@ -25,7 +25,7 @@ datasets:
25
 
26
  Building upon InternVL3 and Qwen3-VL, we instantiate InternVLA-A1 at 2B and 3B parameter scales. Covering different model scales and pre-training data configurations, we release the InternVLA-A1 series:
27
 
28
- - [x] [InternVLA-A1-3B](https://huggingface.co/InternRobotics/InternVLA-A1-3B): pretrained on the large-scale, high-fidelity simulation data [InternData-A1](https://huggingface.co/datasets/InternRobotics/InternData-A1), together with open-source robot data (e.g. Agibot-World)
29
  - [x] [InternVLA-A1-3B-RoboTwin](https://huggingface.co/InternRobotics/InternVLA-A1-3B-RoboTwin): finetuned on RoboTwin 2.0 benchmark
30
  - [ ] [InternVLA-A1-3B-Pretrain-InternData-A1](https://huggingface.co/InternRobotics/InternVLA-A1-3B-Pretrain-InternData-A1): pretrained on InternData-A1 only
31
  - [ ] [InternVLA-A1-2B-Pretrain-InternData-A1](https://huggingface.co/InternRobotics/InternVLA-A1-2B-Pretrain-InternData-A1): pretrained on InternData-A1 only
@@ -38,8 +38,8 @@ Building upon InternVL3 and Qwen3-VL, we instantiate InternVLA-A1 at 2B and 3B p
38
 
39
  | Metric | pi0 | pi0.5 | **InternVLA-A1-3B** |
40
  | :--- | :---: | :---: | :---: |
41
- | Avg. Success (Easy) | 79.98% | 86.76% | **88.30%** |
42
- | Avg. Success (Hard) | 79.50% | 86.96% | **88.48%** |
43
 
44
  ## 🔑 Key Features
45
 
 
25
 
26
  Building upon InternVL3 and Qwen3-VL, we instantiate InternVLA-A1 at 2B and 3B parameter scales. Covering different model scales and pre-training data configurations, we release the InternVLA-A1 series:
27
 
28
+ - [x] [InternVLA-A1-3B](https://huggingface.co/InternRobotics/InternVLA-A1-3B): pretrained on heterogeneous data sources over real-robot data, synthetic simulation data, and human videos
29
  - [x] [InternVLA-A1-3B-RoboTwin](https://huggingface.co/InternRobotics/InternVLA-A1-3B-RoboTwin): finetuned on RoboTwin 2.0 benchmark
30
  - [ ] [InternVLA-A1-3B-Pretrain-InternData-A1](https://huggingface.co/InternRobotics/InternVLA-A1-3B-Pretrain-InternData-A1): pretrained on InternData-A1 only
31
  - [ ] [InternVLA-A1-2B-Pretrain-InternData-A1](https://huggingface.co/InternRobotics/InternVLA-A1-2B-Pretrain-InternData-A1): pretrained on InternData-A1 only
 
38
 
39
  | Metric | pi0 | pi0.5 | **InternVLA-A1-3B** |
40
  | :--- | :---: | :---: | :---: |
41
+ | Avg. Success (Easy) | 79.98% | 86.76% | **89.40%** |
42
+ | Avg. Success (Hard) | 79.50% | 86.96% | **89.64%** |
43
 
44
  ## 🔑 Key Features
45
 
config.json CHANGED
@@ -24,7 +24,7 @@
24
  "private": null,
25
  "tags": null,
26
  "license": null,
27
- "pretrained_path": "/mnt/shared-storage-user/internvla/Users/caijunhao/lerobot/outputs/qwena1/2026_01_16_20_15_44-qwena1-a1_robotwin-delta-pretrain/checkpoints/01050000/pretrained_model",
28
  "qwen3_vl_variant": "qwen3_vl_28l",
29
  "action_expert_variant": "qwen3_28l",
30
  "dtype": "bfloat16",
 
24
  "private": null,
25
  "tags": null,
26
  "license": null,
27
+ "pretrained_path": "/mnt/shared-storage-user/internvla/Users/caijunhao/lerobot/outputs/qwena1/2026_01_27_19_50_57-qwena1-a1_robotwin_egodex_agibotworld_robomind-delta-pretrain-agibotworld_robomind_wo_act_loss/checkpoints/01420000/pretrained_model",
28
  "qwen3_vl_variant": "qwen3_vl_28l",
29
  "action_expert_variant": "qwen3_28l",
30
  "dtype": "bfloat16",
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:427abfdf76cef779fbe0d3cc04b9b0079c0507e7fbbebdcd75491f251283f2e0
3
  size 6721209208
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a5548b6aebef6787ee74d8b0c957da2ad05cadf19cb6c6e4165d17db6b5bd079
3
  size 6721209208
train_config.json CHANGED
@@ -178,7 +178,7 @@
178
  "private": null,
179
  "tags": null,
180
  "license": null,
181
- "pretrained_path": "/mnt/shared-storage-user/internvla/Users/caijunhao/lerobot/outputs/qwena1/2026_01_16_20_15_44-qwena1-a1_robotwin-delta-pretrain/checkpoints/01050000/pretrained_model",
182
  "qwen3_vl_variant": "qwen3_vl_28l",
183
  "action_expert_variant": "qwen3_28l",
184
  "dtype": "bfloat16",
@@ -224,8 +224,8 @@
224
  "scale_factor": 8,
225
  "lambda_gen": 0.01
226
  },
227
- "output_dir": "/mnt/shared-storage-user/internvla/Users/mahaoxiang/lerobot_exp/qwena1_sim_full_1050k/2026_01_20_16_01_40",
228
- "job_name": "2026_01_20_16_30_25-qwena1-robotwin-full-remix-delta",
229
  "resume": false,
230
  "seed": 42,
231
  "num_workers": 8,
@@ -265,7 +265,7 @@
265
  "project": "lerobot_qwena1_robotwin_full",
266
  "entity": null,
267
  "notes": null,
268
- "run_id": "fj1c8atu",
269
  "mode": "offline"
270
  },
271
  "checkpoint_path": null,
 
178
  "private": null,
179
  "tags": null,
180
  "license": null,
181
+ "pretrained_path": "/mnt/shared-storage-user/internvla/Users/caijunhao/lerobot/outputs/qwena1/2026_01_27_19_50_57-qwena1-a1_robotwin_egodex_agibotworld_robomind-delta-pretrain-agibotworld_robomind_wo_act_loss/checkpoints/01420000/pretrained_model",
182
  "qwen3_vl_variant": "qwen3_vl_28l",
183
  "action_expert_variant": "qwen3_28l",
184
  "dtype": "bfloat16",
 
224
  "scale_factor": 8,
225
  "lambda_gen": 0.01
226
  },
227
+ "output_dir": "/mnt/shared-storage-user/internvla/Users/mahaoxiang/lerobot_exp/qwena1_robotwin_a1_agibotworld_robomind_wo_act_loss/2026_01_29_11_13_13",
228
+ "job_name": "2026_01_29_11_14_18-qwena1-robotwin-full-remix-delta",
229
  "resume": false,
230
  "seed": 42,
231
  "num_workers": 8,
 
265
  "project": "lerobot_qwena1_robotwin_full",
266
  "entity": null,
267
  "notes": null,
268
+ "run_id": "uu88o32c",
269
  "mode": "offline"
270
  },
271
  "checkpoint_path": null,