Hugging Face's logo

zeeshaan-ai
/
GetSoloTech

Text Generation
Transformers
Safetensors
qwen3
solo
fine-tuned
lora
unsloth
conversational
text-generation-inference
Model card Files
xet
 Community
zeeshaan-ai commited on
Commit
c862dcb
·
verified ·
1 Parent(s): e894dc0

Upload model

Browse files
Files changed (5) hide show
  1. config.json +85 -57
  2. model.safetensors +2 -2
  3. policy_postprocessor.json +1 -1
  4. policy_preprocessor.json +17 -2
  5. train_config.json +57 -32
config.json CHANGED
@@ -1,61 +1,89 @@
1
  {
2
- "architectures": [
3
- "Qwen3ForCausalLM"
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
4
  ],
5
- "attention_bias": false,
6
- "attention_dropout": 0.0,
7
- "torch_dtype": "bfloat16",
8
- "eos_token_id": 151645,
9
- "head_dim": 128,
10
- "hidden_act": "silu",
11
- "hidden_size": 1024,
12
- "initializer_range": 0.02,
13
- "intermediate_size": 3072,
14
- "layer_types": [
15
- "full_attention",
16
- "full_attention",
17
- "full_attention",
18
- "full_attention",
19
- "full_attention",
20
- "full_attention",
21
- "full_attention",
22
- "full_attention",
23
- "full_attention",
24
- "full_attention",
25
- "full_attention",
26
- "full_attention",
27
- "full_attention",
28
- "full_attention",
29
- "full_attention",
30
- "full_attention",
31
- "full_attention",
32
- "full_attention",
33
- "full_attention",
34
- "full_attention",
35
- "full_attention",
36
- "full_attention",
37
- "full_attention",
38
- "full_attention",
39
- "full_attention",
40
- "full_attention",
41
- "full_attention",
42
- "full_attention"
43
- ],
44
- "max_position_embeddings": 40960,
45
- "max_window_layers": 28,
46
- "model_type": "qwen3",
47
- "num_attention_heads": 16,
48
- "num_hidden_layers": 28,
49
- "num_key_value_heads": 8,
50
- "pad_token_id": 151654,
51
- "rms_norm_eps": 1e-06,
52
- "rope_scaling": null,
53
- "rope_theta": 1000000,
54
- "sliding_window": null,
55
- "tie_word_embeddings": true,
56
- "unsloth_fixed": true,
57
- "unsloth_version": "2026.2.1",
58
  "use_cache": true,
59
- "use_sliding_window": false,
60
- "vocab_size": 151936
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
61
  }
 
1
  {
2
+ "type": "smolvla",
3
+ "n_obs_steps": 1,
4
+ "input_features": {
5
+ "observation.state": {
6
+ "type": "STATE",
7
+ "shape": [
8
+ 6
9
+ ]
10
+ },
11
+ "observation.images.side": {
12
+ "type": "VISUAL",
13
+ "shape": [
14
+ 3,
15
+ 1080,
16
+ 1920
17
+ ]
18
+ },
19
+ "observation.images.front": {
20
+ "type": "VISUAL",
21
+ "shape": [
22
+ 3,
23
+ 480,
24
+ 640
25
+ ]
26
+ }
27
+ },
28
+ "output_features": {
29
+ "action": {
30
+ "type": "ACTION",
31
+ "shape": [
32
+ 6
33
+ ]
34
+ }
35
+ },
36
+ "device": "cuda",
37
+ "use_amp": false,
38
+ "push_to_hub": false,
39
+ "repo_id": null,
40
+ "private": null,
41
+ "tags": null,
42
+ "license": null,
43
+ "pretrained_path": "lerobot/smolvla_base",
44
+ "chunk_size": 50,
45
+ "n_action_steps": 50,
46
+ "normalization_mapping": {
47
+ "VISUAL": "IDENTITY",
48
+ "STATE": "MEAN_STD",
49
+ "ACTION": "MEAN_STD"
50
+ },
51
+ "max_state_dim": 32,
52
+ "max_action_dim": 32,
53
+ "resize_imgs_with_padding": [
54
+ 512,
55
+ 512
56
  ],
57
+ "empty_cameras": 0,
58
+ "adapt_to_pi_aloha": false,
59
+ "use_delta_joint_actions_aloha": false,
60
+ "tokenizer_max_length": 48,
61
+ "num_steps": 10,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
62
  "use_cache": true,
63
+ "freeze_vision_encoder": true,
64
+ "train_expert_only": true,
65
+ "train_state_proj": true,
66
+ "optimizer_lr": 0.0001,
67
+ "optimizer_betas": [
68
+ 0.9,
69
+ 0.95
70
+ ],
71
+ "optimizer_eps": 1e-08,
72
+ "optimizer_weight_decay": 1e-10,
73
+ "optimizer_grad_clip_norm": 10,
74
+ "scheduler_warmup_steps": 1000,
75
+ "scheduler_decay_steps": 30000,
76
+ "scheduler_decay_lr": 2.5e-06,
77
+ "vlm_model_name": "HuggingFaceTB/SmolVLM2-500M-Video-Instruct",
78
+ "load_vlm_weights": false,
79
+ "add_image_special_tokens": false,
80
+ "attention_mode": "cross_attn",
81
+ "prefix_length": -1,
82
+ "pad_language_to": "longest",
83
+ "num_expert_layers": -1,
84
+ "num_vlm_layers": 16,
85
+ "self_attn_every_n_layers": 2,
86
+ "expert_width_multiplier": 0.75,
87
+ "min_period": 0.004,
88
+ "max_period": 4.0
89
  }
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:85cd84f50856bdea780d5e58c640b41fc127d07171f3c8025d3a5c65c12a3bcf
3
- size 1192135096
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:46b647e65da1188b7cec91c3a7f44daeacc01b6497f46acd31a3c4241c7f144e
3
+ size 1197789224
policy_postprocessor.json CHANGED
@@ -14,7 +14,7 @@
14
  }
15
  },
16
  "norm_map": {
17
- "VISUAL": "MEAN_STD",
18
  "STATE": "MEAN_STD",
19
  "ACTION": "MEAN_STD"
20
  }
 
14
  }
15
  },
16
  "norm_map": {
17
+ "VISUAL": "IDENTITY",
18
  "STATE": "MEAN_STD",
19
  "ACTION": "MEAN_STD"
20
  }
policy_preprocessor.json CHANGED
@@ -11,6 +11,21 @@
11
  "registry_name": "to_batch_processor",
12
  "config": {}
13
  },
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
14
  {
15
  "registry_name": "device_processor",
16
  "config": {
@@ -53,12 +68,12 @@
53
  }
54
  },
55
  "norm_map": {
56
- "VISUAL": "MEAN_STD",
57
  "STATE": "MEAN_STD",
58
  "ACTION": "MEAN_STD"
59
  }
60
  },
61
- "state_file": "policy_preprocessor_step_3_normalizer_processor.safetensors"
62
  }
63
  ]
64
  }
 
11
  "registry_name": "to_batch_processor",
12
  "config": {}
13
  },
14
+ {
15
+ "registry_name": "smolvla_new_line_processor",
16
+ "config": {}
17
+ },
18
+ {
19
+ "registry_name": "tokenizer_processor",
20
+ "config": {
21
+ "max_length": 48,
22
+ "task_key": "task",
23
+ "padding_side": "right",
24
+ "padding": "max_length",
25
+ "truncation": true,
26
+ "tokenizer_name": "HuggingFaceTB/SmolVLM2-500M-Video-Instruct"
27
+ }
28
+ },
29
  {
30
  "registry_name": "device_processor",
31
  "config": {
 
68
  }
69
  },
70
  "norm_map": {
71
+ "VISUAL": "IDENTITY",
72
  "STATE": "MEAN_STD",
73
  "ACTION": "MEAN_STD"
74
  }
75
  },
76
+ "state_file": "policy_preprocessor_step_5_normalizer_processor.safetensors"
77
  }
78
  ]
79
  }
train_config.json CHANGED
@@ -81,7 +81,7 @@
81
  },
82
  "env": null,
83
  "policy": {
84
- "type": "act",
85
  "n_obs_steps": 1,
86
  "input_features": {
87
  "observation.state": {
@@ -122,58 +122,83 @@
122
  "private": null,
123
  "tags": null,
124
  "license": null,
125
- "pretrained_path": null,
126
- "chunk_size": 100,
127
- "n_action_steps": 100,
128
  "normalization_mapping": {
129
- "VISUAL": "MEAN_STD",
130
  "STATE": "MEAN_STD",
131
  "ACTION": "MEAN_STD"
132
  },
133
- "vision_backbone": "resnet18",
134
- "pretrained_backbone_weights": "ResNet18_Weights.IMAGENET1K_V1",
135
- "replace_final_stride_with_dilation": false,
136
- "pre_norm": false,
137
- "dim_model": 512,
138
- "n_heads": 8,
139
- "dim_feedforward": 3200,
140
- "feedforward_activation": "relu",
141
- "n_encoder_layers": 4,
142
- "n_decoder_layers": 1,
143
- "use_vae": true,
144
- "latent_dim": 32,
145
- "n_vae_encoder_layers": 4,
146
- "temporal_ensemble_coeff": null,
147
- "dropout": 0.1,
148
- "kl_weight": 10.0,
149
- "optimizer_lr": 1e-05,
150
- "optimizer_weight_decay": 0.0001,
151
- "optimizer_lr_backbone": 1e-05
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
152
  },
153
  "output_dir": "/workspace/outputs",
154
- "job_name": "ACT",
155
  "resume": false,
156
  "seed": 1000,
157
  "num_workers": 4,
158
  "batch_size": 8,
159
- "steps": 50000,
160
  "eval_freq": 20000,
161
  "log_freq": 200,
162
  "save_checkpoint": true,
163
- "save_freq": 10000,
164
  "use_policy_training_preset": true,
165
  "optimizer": {
166
  "type": "adamw",
167
- "lr": 1e-05,
168
- "weight_decay": 0.0001,
169
- "grad_clip_norm": 10.0,
170
  "betas": [
171
  0.9,
172
- 0.999
173
  ],
174
  "eps": 1e-08
175
  },
176
- "scheduler": null,
 
 
 
 
 
 
177
  "eval": {
178
  "n_episodes": 50,
179
  "batch_size": 50,
 
81
  },
82
  "env": null,
83
  "policy": {
84
+ "type": "smolvla",
85
  "n_obs_steps": 1,
86
  "input_features": {
87
  "observation.state": {
 
122
  "private": null,
123
  "tags": null,
124
  "license": null,
125
+ "pretrained_path": "lerobot/smolvla_base",
126
+ "chunk_size": 50,
127
+ "n_action_steps": 50,
128
  "normalization_mapping": {
129
+ "VISUAL": "IDENTITY",
130
  "STATE": "MEAN_STD",
131
  "ACTION": "MEAN_STD"
132
  },
133
+ "max_state_dim": 32,
134
+ "max_action_dim": 32,
135
+ "resize_imgs_with_padding": [
136
+ 512,
137
+ 512
138
+ ],
139
+ "empty_cameras": 0,
140
+ "adapt_to_pi_aloha": false,
141
+ "use_delta_joint_actions_aloha": false,
142
+ "tokenizer_max_length": 48,
143
+ "num_steps": 10,
144
+ "use_cache": true,
145
+ "freeze_vision_encoder": true,
146
+ "train_expert_only": true,
147
+ "train_state_proj": true,
148
+ "optimizer_lr": 0.0001,
149
+ "optimizer_betas": [
150
+ 0.9,
151
+ 0.95
152
+ ],
153
+ "optimizer_eps": 1e-08,
154
+ "optimizer_weight_decay": 1e-10,
155
+ "optimizer_grad_clip_norm": 10,
156
+ "scheduler_warmup_steps": 1000,
157
+ "scheduler_decay_steps": 30000,
158
+ "scheduler_decay_lr": 2.5e-06,
159
+ "vlm_model_name": "HuggingFaceTB/SmolVLM2-500M-Video-Instruct",
160
+ "load_vlm_weights": false,
161
+ "add_image_special_tokens": false,
162
+ "attention_mode": "cross_attn",
163
+ "prefix_length": -1,
164
+ "pad_language_to": "longest",
165
+ "num_expert_layers": -1,
166
+ "num_vlm_layers": 16,
167
+ "self_attn_every_n_layers": 2,
168
+ "expert_width_multiplier": 0.75,
169
+ "min_period": 0.004,
170
+ "max_period": 4.0
171
  },
172
  "output_dir": "/workspace/outputs",
173
+ "job_name": "my-first",
174
  "resume": false,
175
  "seed": 1000,
176
  "num_workers": 4,
177
  "batch_size": 8,
178
+ "steps": 30000,
179
  "eval_freq": 20000,
180
  "log_freq": 200,
181
  "save_checkpoint": true,
182
+ "save_freq": 3000,
183
  "use_policy_training_preset": true,
184
  "optimizer": {
185
  "type": "adamw",
186
+ "lr": 0.0001,
187
+ "weight_decay": 1e-10,
188
+ "grad_clip_norm": 10,
189
  "betas": [
190
  0.9,
191
+ 0.95
192
  ],
193
  "eps": 1e-08
194
  },
195
+ "scheduler": {
196
+ "type": "cosine_decay_with_warmup",
197
+ "num_warmup_steps": 1000,
198
+ "num_decay_steps": 30000,
199
+ "peak_lr": 0.0001,
200
+ "decay_lr": 2.5e-06
201
+ },
202
  "eval": {
203
  "n_episodes": 50,
204
  "batch_size": 50,