Add UniMVU release checkpoint: unimvu_0.5B_avsd
Browse files
unimvu_0.5B_avsd/adapter_config.json
ADDED
|
@@ -0,0 +1,195 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"alpha_pattern": {},
|
| 3 |
+
"auto_mapping": null,
|
| 4 |
+
"base_model_name_or_path": "lmms-lab/llava-onevision-qwen2-0.5b-ov",
|
| 5 |
+
"bias": "none",
|
| 6 |
+
"fan_in_fan_out": false,
|
| 7 |
+
"inference_mode": true,
|
| 8 |
+
"init_lora_weights": true,
|
| 9 |
+
"layer_replication": null,
|
| 10 |
+
"layers_pattern": null,
|
| 11 |
+
"layers_to_transform": null,
|
| 12 |
+
"loftq_config": {},
|
| 13 |
+
"lora_alpha": 128,
|
| 14 |
+
"lora_dropout": 0.05,
|
| 15 |
+
"megatron_config": null,
|
| 16 |
+
"megatron_core": "megatron.core",
|
| 17 |
+
"modules_to_save": null,
|
| 18 |
+
"peft_type": "LORA",
|
| 19 |
+
"r": 64,
|
| 20 |
+
"rank_pattern": {},
|
| 21 |
+
"revision": null,
|
| 22 |
+
"target_modules": [
|
| 23 |
+
"model.layers.14.mlp.gate_proj",
|
| 24 |
+
"model.layers.10.mlp.gate_proj",
|
| 25 |
+
"model.layers.4.mlp.up_proj",
|
| 26 |
+
"model.layers.17.mlp.up_proj",
|
| 27 |
+
"model.layers.4.self_attn.o_proj",
|
| 28 |
+
"model.layers.23.mlp.down_proj",
|
| 29 |
+
"model.layers.11.mlp.up_proj",
|
| 30 |
+
"model.layers.0.self_attn.v_proj",
|
| 31 |
+
"model.layers.7.self_attn.q_proj",
|
| 32 |
+
"model.layers.17.self_attn.o_proj",
|
| 33 |
+
"model.layers.15.mlp.up_proj",
|
| 34 |
+
"model.layers.2.mlp.gate_proj",
|
| 35 |
+
"model.layers.0.self_attn.k_proj",
|
| 36 |
+
"model.layers.10.self_attn.o_proj",
|
| 37 |
+
"model.layers.6.mlp.down_proj",
|
| 38 |
+
"model.layers.3.mlp.up_proj",
|
| 39 |
+
"model.layers.4.self_attn.v_proj",
|
| 40 |
+
"model.layers.19.self_attn.q_proj",
|
| 41 |
+
"model.layers.5.self_attn.v_proj",
|
| 42 |
+
"model.layers.6.mlp.gate_proj",
|
| 43 |
+
"model.layers.14.self_attn.q_proj",
|
| 44 |
+
"model.layers.21.self_attn.v_proj",
|
| 45 |
+
"model.layers.2.mlp.down_proj",
|
| 46 |
+
"model.layers.22.mlp.gate_proj",
|
| 47 |
+
"model.layers.22.self_attn.o_proj",
|
| 48 |
+
"model.layers.18.self_attn.k_proj",
|
| 49 |
+
"model.layers.19.mlp.down_proj",
|
| 50 |
+
"model.layers.5.self_attn.o_proj",
|
| 51 |
+
"model.layers.8.mlp.down_proj",
|
| 52 |
+
"model.layers.10.mlp.up_proj",
|
| 53 |
+
"model.layers.7.self_attn.k_proj",
|
| 54 |
+
"model.layers.5.self_attn.q_proj",
|
| 55 |
+
"model.layers.16.mlp.up_proj",
|
| 56 |
+
"model.layers.14.self_attn.o_proj",
|
| 57 |
+
"model.layers.13.self_attn.v_proj",
|
| 58 |
+
"model.layers.23.self_attn.v_proj",
|
| 59 |
+
"model.layers.16.self_attn.k_proj",
|
| 60 |
+
"model.layers.15.self_attn.v_proj",
|
| 61 |
+
"model.layers.21.mlp.up_proj",
|
| 62 |
+
"model.layers.17.self_attn.q_proj",
|
| 63 |
+
"model.layers.20.mlp.up_proj",
|
| 64 |
+
"model.layers.13.self_attn.o_proj",
|
| 65 |
+
"model.layers.4.mlp.gate_proj",
|
| 66 |
+
"model.layers.13.mlp.down_proj",
|
| 67 |
+
"model.layers.1.self_attn.k_proj",
|
| 68 |
+
"model.layers.7.self_attn.v_proj",
|
| 69 |
+
"model.layers.14.self_attn.k_proj",
|
| 70 |
+
"model.layers.17.self_attn.v_proj",
|
| 71 |
+
"model.layers.22.self_attn.v_proj",
|
| 72 |
+
"model.layers.7.mlp.up_proj",
|
| 73 |
+
"model.layers.0.self_attn.o_proj",
|
| 74 |
+
"model.layers.12.self_attn.v_proj",
|
| 75 |
+
"model.layers.12.self_attn.q_proj",
|
| 76 |
+
"model.layers.19.self_attn.o_proj",
|
| 77 |
+
"model.layers.18.self_attn.q_proj",
|
| 78 |
+
"model.layers.17.mlp.down_proj",
|
| 79 |
+
"model.layers.18.mlp.gate_proj",
|
| 80 |
+
"model.layers.0.mlp.gate_proj",
|
| 81 |
+
"model.layers.9.self_attn.k_proj",
|
| 82 |
+
"model.layers.22.mlp.up_proj",
|
| 83 |
+
"model.layers.6.self_attn.q_proj",
|
| 84 |
+
"model.layers.18.self_attn.o_proj",
|
| 85 |
+
"model.layers.8.mlp.up_proj",
|
| 86 |
+
"model.layers.9.self_attn.q_proj",
|
| 87 |
+
"model.layers.19.mlp.gate_proj",
|
| 88 |
+
"model.layers.5.mlp.gate_proj",
|
| 89 |
+
"model.layers.20.self_attn.v_proj",
|
| 90 |
+
"model.layers.20.mlp.gate_proj",
|
| 91 |
+
"model.layers.23.self_attn.q_proj",
|
| 92 |
+
"model.layers.2.mlp.up_proj",
|
| 93 |
+
"model.layers.14.self_attn.v_proj",
|
| 94 |
+
"model.layers.18.mlp.down_proj",
|
| 95 |
+
"model.layers.0.mlp.up_proj",
|
| 96 |
+
"model.layers.0.self_attn.q_proj",
|
| 97 |
+
"model.layers.8.mlp.gate_proj",
|
| 98 |
+
"model.layers.5.self_attn.k_proj",
|
| 99 |
+
"model.layers.15.mlp.gate_proj",
|
| 100 |
+
"model.layers.10.self_attn.k_proj",
|
| 101 |
+
"model.layers.1.mlp.gate_proj",
|
| 102 |
+
"model.layers.20.self_attn.q_proj",
|
| 103 |
+
"model.layers.3.self_attn.o_proj",
|
| 104 |
+
"model.layers.7.self_attn.o_proj",
|
| 105 |
+
"model.layers.21.mlp.gate_proj",
|
| 106 |
+
"model.layers.6.self_attn.v_proj",
|
| 107 |
+
"model.layers.18.self_attn.v_proj",
|
| 108 |
+
"model.layers.6.mlp.up_proj",
|
| 109 |
+
"model.layers.11.mlp.down_proj",
|
| 110 |
+
"model.layers.1.self_attn.q_proj",
|
| 111 |
+
"model.layers.11.mlp.gate_proj",
|
| 112 |
+
"model.layers.13.mlp.up_proj",
|
| 113 |
+
"model.layers.3.self_attn.v_proj",
|
| 114 |
+
"model.layers.17.mlp.gate_proj",
|
| 115 |
+
"model.layers.14.mlp.up_proj",
|
| 116 |
+
"model.layers.16.mlp.gate_proj",
|
| 117 |
+
"model.layers.16.self_attn.q_proj",
|
| 118 |
+
"model.layers.9.mlp.down_proj",
|
| 119 |
+
"model.layers.9.mlp.up_proj",
|
| 120 |
+
"model.layers.2.self_attn.o_proj",
|
| 121 |
+
"model.layers.16.mlp.down_proj",
|
| 122 |
+
"model.layers.21.mlp.down_proj",
|
| 123 |
+
"model.layers.1.self_attn.v_proj",
|
| 124 |
+
"model.layers.8.self_attn.k_proj",
|
| 125 |
+
"model.layers.16.self_attn.v_proj",
|
| 126 |
+
"model.layers.12.self_attn.k_proj",
|
| 127 |
+
"model.layers.22.self_attn.k_proj",
|
| 128 |
+
"model.layers.23.self_attn.k_proj",
|
| 129 |
+
"model.layers.13.self_attn.k_proj",
|
| 130 |
+
"model.layers.7.mlp.gate_proj",
|
| 131 |
+
"model.layers.7.mlp.down_proj",
|
| 132 |
+
"model.layers.2.self_attn.q_proj",
|
| 133 |
+
"model.layers.4.self_attn.q_proj",
|
| 134 |
+
"model.layers.11.self_attn.q_proj",
|
| 135 |
+
"model.layers.23.mlp.up_proj",
|
| 136 |
+
"model.layers.19.self_attn.k_proj",
|
| 137 |
+
"model.layers.12.mlp.down_proj",
|
| 138 |
+
"model.layers.6.self_attn.o_proj",
|
| 139 |
+
"model.layers.9.mlp.gate_proj",
|
| 140 |
+
"model.layers.14.mlp.down_proj",
|
| 141 |
+
"model.layers.1.mlp.up_proj",
|
| 142 |
+
"model.layers.15.mlp.down_proj",
|
| 143 |
+
"model.layers.5.mlp.up_proj",
|
| 144 |
+
"model.layers.15.self_attn.k_proj",
|
| 145 |
+
"model.layers.16.self_attn.o_proj",
|
| 146 |
+
"model.layers.19.mlp.up_proj",
|
| 147 |
+
"model.layers.10.self_attn.q_proj",
|
| 148 |
+
"model.layers.2.self_attn.v_proj",
|
| 149 |
+
"model.layers.1.mlp.down_proj",
|
| 150 |
+
"model.layers.23.self_attn.o_proj",
|
| 151 |
+
"model.layers.10.mlp.down_proj",
|
| 152 |
+
"model.layers.15.self_attn.o_proj",
|
| 153 |
+
"model.layers.11.self_attn.o_proj",
|
| 154 |
+
"model.layers.1.self_attn.o_proj",
|
| 155 |
+
"model.layers.8.self_attn.q_proj",
|
| 156 |
+
"model.layers.12.mlp.up_proj",
|
| 157 |
+
"model.layers.3.mlp.down_proj",
|
| 158 |
+
"model.layers.17.self_attn.k_proj",
|
| 159 |
+
"model.layers.21.self_attn.q_proj",
|
| 160 |
+
"model.layers.3.self_attn.q_proj",
|
| 161 |
+
"model.layers.5.mlp.down_proj",
|
| 162 |
+
"model.layers.0.mlp.down_proj",
|
| 163 |
+
"model.layers.4.self_attn.k_proj",
|
| 164 |
+
"model.layers.11.self_attn.v_proj",
|
| 165 |
+
"model.layers.11.self_attn.k_proj",
|
| 166 |
+
"model.layers.21.self_attn.o_proj",
|
| 167 |
+
"model.layers.20.self_attn.o_proj",
|
| 168 |
+
"model.layers.8.self_attn.o_proj",
|
| 169 |
+
"model.layers.23.mlp.gate_proj",
|
| 170 |
+
"model.layers.6.self_attn.k_proj",
|
| 171 |
+
"model.layers.15.self_attn.q_proj",
|
| 172 |
+
"model.layers.19.self_attn.v_proj",
|
| 173 |
+
"model.layers.22.mlp.down_proj",
|
| 174 |
+
"model.layers.2.self_attn.k_proj",
|
| 175 |
+
"model.layers.13.self_attn.q_proj",
|
| 176 |
+
"model.layers.8.self_attn.v_proj",
|
| 177 |
+
"model.layers.12.self_attn.o_proj",
|
| 178 |
+
"model.layers.9.self_attn.v_proj",
|
| 179 |
+
"model.layers.9.self_attn.o_proj",
|
| 180 |
+
"model.layers.20.self_attn.k_proj",
|
| 181 |
+
"model.layers.3.self_attn.k_proj",
|
| 182 |
+
"model.layers.12.mlp.gate_proj",
|
| 183 |
+
"model.layers.21.self_attn.k_proj",
|
| 184 |
+
"model.layers.4.mlp.down_proj",
|
| 185 |
+
"model.layers.3.mlp.gate_proj",
|
| 186 |
+
"model.layers.20.mlp.down_proj",
|
| 187 |
+
"model.layers.22.self_attn.q_proj",
|
| 188 |
+
"model.layers.10.self_attn.v_proj",
|
| 189 |
+
"model.layers.13.mlp.gate_proj",
|
| 190 |
+
"model.layers.18.mlp.up_proj"
|
| 191 |
+
],
|
| 192 |
+
"task_type": "CAUSAL_LM",
|
| 193 |
+
"use_dora": false,
|
| 194 |
+
"use_rslora": false
|
| 195 |
+
}
|
unimvu_0.5B_avsd/adapter_model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:31151f711f92f70adf43fc79b76e21025c121fe17b8a8edeb41e00b730d77301
|
| 3 |
+
size 70430368
|
unimvu_0.5B_avsd/config.json
ADDED
|
@@ -0,0 +1,392 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_name_or_path": "lmms-lab/llava-onevision-qwen2-0.5b-ov",
|
| 3 |
+
"architectures": [
|
| 4 |
+
"LlavaQwenForCausalLM"
|
| 5 |
+
],
|
| 6 |
+
"attention_dropout": 0.0,
|
| 7 |
+
"bos_token_id": 151643,
|
| 8 |
+
"eos_token_id": 151645,
|
| 9 |
+
"feat_combine_method": "concat",
|
| 10 |
+
"hidden_act": "silu",
|
| 11 |
+
"hidden_size": 896,
|
| 12 |
+
"ignore_index": -100,
|
| 13 |
+
"image_aspect_ratio": "anyres_max_9",
|
| 14 |
+
"image_crop_resolution": null,
|
| 15 |
+
"image_grid_pinpoints": [
|
| 16 |
+
[
|
| 17 |
+
384,
|
| 18 |
+
384
|
| 19 |
+
],
|
| 20 |
+
[
|
| 21 |
+
384,
|
| 22 |
+
768
|
| 23 |
+
],
|
| 24 |
+
[
|
| 25 |
+
384,
|
| 26 |
+
1152
|
| 27 |
+
],
|
| 28 |
+
[
|
| 29 |
+
384,
|
| 30 |
+
1536
|
| 31 |
+
],
|
| 32 |
+
[
|
| 33 |
+
384,
|
| 34 |
+
1920
|
| 35 |
+
],
|
| 36 |
+
[
|
| 37 |
+
384,
|
| 38 |
+
2304
|
| 39 |
+
],
|
| 40 |
+
[
|
| 41 |
+
768,
|
| 42 |
+
384
|
| 43 |
+
],
|
| 44 |
+
[
|
| 45 |
+
768,
|
| 46 |
+
768
|
| 47 |
+
],
|
| 48 |
+
[
|
| 49 |
+
768,
|
| 50 |
+
1152
|
| 51 |
+
],
|
| 52 |
+
[
|
| 53 |
+
768,
|
| 54 |
+
1536
|
| 55 |
+
],
|
| 56 |
+
[
|
| 57 |
+
768,
|
| 58 |
+
1920
|
| 59 |
+
],
|
| 60 |
+
[
|
| 61 |
+
768,
|
| 62 |
+
2304
|
| 63 |
+
],
|
| 64 |
+
[
|
| 65 |
+
1152,
|
| 66 |
+
384
|
| 67 |
+
],
|
| 68 |
+
[
|
| 69 |
+
1152,
|
| 70 |
+
768
|
| 71 |
+
],
|
| 72 |
+
[
|
| 73 |
+
1152,
|
| 74 |
+
1152
|
| 75 |
+
],
|
| 76 |
+
[
|
| 77 |
+
1152,
|
| 78 |
+
1536
|
| 79 |
+
],
|
| 80 |
+
[
|
| 81 |
+
1152,
|
| 82 |
+
1920
|
| 83 |
+
],
|
| 84 |
+
[
|
| 85 |
+
1152,
|
| 86 |
+
2304
|
| 87 |
+
],
|
| 88 |
+
[
|
| 89 |
+
1536,
|
| 90 |
+
384
|
| 91 |
+
],
|
| 92 |
+
[
|
| 93 |
+
1536,
|
| 94 |
+
768
|
| 95 |
+
],
|
| 96 |
+
[
|
| 97 |
+
1536,
|
| 98 |
+
1152
|
| 99 |
+
],
|
| 100 |
+
[
|
| 101 |
+
1536,
|
| 102 |
+
1536
|
| 103 |
+
],
|
| 104 |
+
[
|
| 105 |
+
1536,
|
| 106 |
+
1920
|
| 107 |
+
],
|
| 108 |
+
[
|
| 109 |
+
1536,
|
| 110 |
+
2304
|
| 111 |
+
],
|
| 112 |
+
[
|
| 113 |
+
1920,
|
| 114 |
+
384
|
| 115 |
+
],
|
| 116 |
+
[
|
| 117 |
+
1920,
|
| 118 |
+
768
|
| 119 |
+
],
|
| 120 |
+
[
|
| 121 |
+
1920,
|
| 122 |
+
1152
|
| 123 |
+
],
|
| 124 |
+
[
|
| 125 |
+
1920,
|
| 126 |
+
1536
|
| 127 |
+
],
|
| 128 |
+
[
|
| 129 |
+
1920,
|
| 130 |
+
1920
|
| 131 |
+
],
|
| 132 |
+
[
|
| 133 |
+
1920,
|
| 134 |
+
2304
|
| 135 |
+
],
|
| 136 |
+
[
|
| 137 |
+
2304,
|
| 138 |
+
384
|
| 139 |
+
],
|
| 140 |
+
[
|
| 141 |
+
2304,
|
| 142 |
+
768
|
| 143 |
+
],
|
| 144 |
+
[
|
| 145 |
+
2304,
|
| 146 |
+
1152
|
| 147 |
+
],
|
| 148 |
+
[
|
| 149 |
+
2304,
|
| 150 |
+
1536
|
| 151 |
+
],
|
| 152 |
+
[
|
| 153 |
+
2304,
|
| 154 |
+
1920
|
| 155 |
+
],
|
| 156 |
+
[
|
| 157 |
+
2304,
|
| 158 |
+
2304
|
| 159 |
+
]
|
| 160 |
+
],
|
| 161 |
+
"image_split_resolution": null,
|
| 162 |
+
"image_token_index": 151646,
|
| 163 |
+
"initializer_range": 0.02,
|
| 164 |
+
"input_dim": 1024,
|
| 165 |
+
"intermediate_size": 4864,
|
| 166 |
+
"max_position_embeddings": 32768,
|
| 167 |
+
"max_window_layers": 24,
|
| 168 |
+
"mm_hidden_size": 1152,
|
| 169 |
+
"mm_newline_position": "grid",
|
| 170 |
+
"mm_patch_merge_type": "spatial_unpad",
|
| 171 |
+
"mm_projector_lr": null,
|
| 172 |
+
"mm_projector_type": "mlp2x_gelu",
|
| 173 |
+
"mm_resampler_type": null,
|
| 174 |
+
"mm_spatial_pool_mode": "bilinear",
|
| 175 |
+
"mm_spatial_pool_out_channels": null,
|
| 176 |
+
"mm_spatial_pool_stride": 2,
|
| 177 |
+
"mm_tunable_parts": "mm_vision_tower,mm_mlp_adapter,mm_language_model",
|
| 178 |
+
"mm_use_im_patch_token": false,
|
| 179 |
+
"mm_use_im_start_end": false,
|
| 180 |
+
"mm_video_tower": null,
|
| 181 |
+
"mm_vision_select_feature": "patch",
|
| 182 |
+
"mm_vision_select_layer": -2,
|
| 183 |
+
"mm_vision_tower": "google/siglip-so400m-patch14-384",
|
| 184 |
+
"mm_vision_tower_lr": 2e-06,
|
| 185 |
+
"modality_aggregator_attention_dropout": 0.0,
|
| 186 |
+
"modality_aggregator_config": {
|
| 187 |
+
"attention_dropout": 0.0,
|
| 188 |
+
"hidden_size": 896,
|
| 189 |
+
"modality_token_num": 1,
|
| 190 |
+
"num_heads": 14,
|
| 191 |
+
"num_key_value_heads": 14,
|
| 192 |
+
"rope_theta": 250000
|
| 193 |
+
},
|
| 194 |
+
"modality_aggregator_hidden_size": 896,
|
| 195 |
+
"modality_aggregator_modality_token_num": 1,
|
| 196 |
+
"modality_aggregator_num_heads": 14,
|
| 197 |
+
"modality_aggregator_num_key_value_heads": 14,
|
| 198 |
+
"modality_aggregator_rope_theta": 250000,
|
| 199 |
+
"modality_input_dims": null,
|
| 200 |
+
"model_name_or_path": "lmms-lab/llava-onevision-qwen2-0.5b-ov",
|
| 201 |
+
"model_type": "unimvu_v3_qwen2",
|
| 202 |
+
"num_attention_heads": 14,
|
| 203 |
+
"num_cross_modality_hidden_layers": 1,
|
| 204 |
+
"num_hidden_layers": 24,
|
| 205 |
+
"num_key_value_heads": 2,
|
| 206 |
+
"pos_skipping_range": 4096,
|
| 207 |
+
"pretrain_mm_mlp_adapter": null,
|
| 208 |
+
"projector_hidden_act": "gelu",
|
| 209 |
+
"rms_norm_eps": 1e-06,
|
| 210 |
+
"rope_scaling": null,
|
| 211 |
+
"rope_theta": 1000000.0,
|
| 212 |
+
"sliding_window": 32768,
|
| 213 |
+
"support_modalities": [
|
| 214 |
+
"video",
|
| 215 |
+
"audio"
|
| 216 |
+
],
|
| 217 |
+
"text_config": {
|
| 218 |
+
"_name_or_path": "",
|
| 219 |
+
"add_cross_attention": false,
|
| 220 |
+
"architectures": null,
|
| 221 |
+
"attention_bias": false,
|
| 222 |
+
"attention_dropout": 0.0,
|
| 223 |
+
"bad_words_ids": null,
|
| 224 |
+
"begin_suppress_tokens": null,
|
| 225 |
+
"bos_token_id": 1,
|
| 226 |
+
"chunk_size_feed_forward": 0,
|
| 227 |
+
"cross_attention_hidden_size": null,
|
| 228 |
+
"decoder_start_token_id": null,
|
| 229 |
+
"diversity_penalty": 0.0,
|
| 230 |
+
"do_sample": false,
|
| 231 |
+
"early_stopping": false,
|
| 232 |
+
"encoder_no_repeat_ngram_size": 0,
|
| 233 |
+
"eos_token_id": 2,
|
| 234 |
+
"exponential_decay_length_penalty": null,
|
| 235 |
+
"finetuning_task": null,
|
| 236 |
+
"forced_bos_token_id": null,
|
| 237 |
+
"forced_eos_token_id": null,
|
| 238 |
+
"hidden_act": "silu",
|
| 239 |
+
"hidden_size": 4096,
|
| 240 |
+
"id2label": {
|
| 241 |
+
"0": "LABEL_0",
|
| 242 |
+
"1": "LABEL_1"
|
| 243 |
+
},
|
| 244 |
+
"initializer_range": 0.02,
|
| 245 |
+
"intermediate_size": 11008,
|
| 246 |
+
"is_decoder": false,
|
| 247 |
+
"is_encoder_decoder": false,
|
| 248 |
+
"label2id": {
|
| 249 |
+
"LABEL_0": 0,
|
| 250 |
+
"LABEL_1": 1
|
| 251 |
+
},
|
| 252 |
+
"length_penalty": 1.0,
|
| 253 |
+
"max_length": 20,
|
| 254 |
+
"max_position_embeddings": 2048,
|
| 255 |
+
"min_length": 0,
|
| 256 |
+
"model_type": "llama",
|
| 257 |
+
"no_repeat_ngram_size": 0,
|
| 258 |
+
"num_attention_heads": 32,
|
| 259 |
+
"num_beam_groups": 1,
|
| 260 |
+
"num_beams": 1,
|
| 261 |
+
"num_hidden_layers": 32,
|
| 262 |
+
"num_key_value_heads": 32,
|
| 263 |
+
"num_return_sequences": 1,
|
| 264 |
+
"output_attentions": false,
|
| 265 |
+
"output_hidden_states": false,
|
| 266 |
+
"output_scores": false,
|
| 267 |
+
"pad_token_id": null,
|
| 268 |
+
"prefix": null,
|
| 269 |
+
"pretraining_tp": 1,
|
| 270 |
+
"problem_type": null,
|
| 271 |
+
"pruned_heads": {},
|
| 272 |
+
"remove_invalid_values": false,
|
| 273 |
+
"repetition_penalty": 1.0,
|
| 274 |
+
"return_dict": true,
|
| 275 |
+
"return_dict_in_generate": false,
|
| 276 |
+
"rms_norm_eps": 1e-06,
|
| 277 |
+
"rope_scaling": null,
|
| 278 |
+
"rope_theta": 10000.0,
|
| 279 |
+
"sep_token_id": null,
|
| 280 |
+
"suppress_tokens": null,
|
| 281 |
+
"task_specific_params": null,
|
| 282 |
+
"temperature": 1.0,
|
| 283 |
+
"tf_legacy_loss": false,
|
| 284 |
+
"tie_encoder_decoder": false,
|
| 285 |
+
"tie_word_embeddings": false,
|
| 286 |
+
"tokenizer_class": null,
|
| 287 |
+
"top_k": 50,
|
| 288 |
+
"top_p": 1.0,
|
| 289 |
+
"torch_dtype": null,
|
| 290 |
+
"torchscript": false,
|
| 291 |
+
"typical_p": 1.0,
|
| 292 |
+
"use_bfloat16": false,
|
| 293 |
+
"use_cache": true,
|
| 294 |
+
"vocab_size": 32000
|
| 295 |
+
},
|
| 296 |
+
"tie_word_embeddings": true,
|
| 297 |
+
"tokenizer_model_max_length": 32768,
|
| 298 |
+
"tokenizer_padding_side": "right",
|
| 299 |
+
"torch_dtype": "bfloat16",
|
| 300 |
+
"transformers_version": "4.37.2",
|
| 301 |
+
"tune_addition_token_embeddings": false,
|
| 302 |
+
"tune_mm_mlp_adapter": false,
|
| 303 |
+
"unfreeze_mm_vision_tower": false,
|
| 304 |
+
"use_cache": true,
|
| 305 |
+
"use_mm_proj": true,
|
| 306 |
+
"use_pos_skipping": false,
|
| 307 |
+
"use_sliding_window": false,
|
| 308 |
+
"version": "conv_llava_ov_qwen",
|
| 309 |
+
"video_tower": null,
|
| 310 |
+
"vision_config": {
|
| 311 |
+
"_name_or_path": "",
|
| 312 |
+
"add_cross_attention": false,
|
| 313 |
+
"architectures": null,
|
| 314 |
+
"attention_dropout": 0.0,
|
| 315 |
+
"bad_words_ids": null,
|
| 316 |
+
"begin_suppress_tokens": null,
|
| 317 |
+
"bos_token_id": null,
|
| 318 |
+
"chunk_size_feed_forward": 0,
|
| 319 |
+
"cross_attention_hidden_size": null,
|
| 320 |
+
"decoder_start_token_id": null,
|
| 321 |
+
"diversity_penalty": 0.0,
|
| 322 |
+
"do_sample": false,
|
| 323 |
+
"early_stopping": false,
|
| 324 |
+
"encoder_no_repeat_ngram_size": 0,
|
| 325 |
+
"eos_token_id": null,
|
| 326 |
+
"exponential_decay_length_penalty": null,
|
| 327 |
+
"finetuning_task": null,
|
| 328 |
+
"forced_bos_token_id": null,
|
| 329 |
+
"forced_eos_token_id": null,
|
| 330 |
+
"hidden_act": "quick_gelu",
|
| 331 |
+
"hidden_size": 1024,
|
| 332 |
+
"id2label": {
|
| 333 |
+
"0": "LABEL_0",
|
| 334 |
+
"1": "LABEL_1"
|
| 335 |
+
},
|
| 336 |
+
"image_size": 336,
|
| 337 |
+
"initializer_factor": 1.0,
|
| 338 |
+
"initializer_range": 0.02,
|
| 339 |
+
"intermediate_size": 4096,
|
| 340 |
+
"is_decoder": false,
|
| 341 |
+
"is_encoder_decoder": false,
|
| 342 |
+
"label2id": {
|
| 343 |
+
"LABEL_0": 0,
|
| 344 |
+
"LABEL_1": 1
|
| 345 |
+
},
|
| 346 |
+
"layer_norm_eps": 1e-05,
|
| 347 |
+
"length_penalty": 1.0,
|
| 348 |
+
"max_length": 20,
|
| 349 |
+
"min_length": 0,
|
| 350 |
+
"model_type": "clip_vision_model",
|
| 351 |
+
"no_repeat_ngram_size": 0,
|
| 352 |
+
"num_attention_heads": 16,
|
| 353 |
+
"num_beam_groups": 1,
|
| 354 |
+
"num_beams": 1,
|
| 355 |
+
"num_channels": 3,
|
| 356 |
+
"num_hidden_layers": 24,
|
| 357 |
+
"num_return_sequences": 1,
|
| 358 |
+
"output_attentions": false,
|
| 359 |
+
"output_hidden_states": false,
|
| 360 |
+
"output_scores": false,
|
| 361 |
+
"pad_token_id": null,
|
| 362 |
+
"patch_size": 14,
|
| 363 |
+
"prefix": null,
|
| 364 |
+
"problem_type": null,
|
| 365 |
+
"projection_dim": 768,
|
| 366 |
+
"pruned_heads": {},
|
| 367 |
+
"remove_invalid_values": false,
|
| 368 |
+
"repetition_penalty": 1.0,
|
| 369 |
+
"return_dict": true,
|
| 370 |
+
"return_dict_in_generate": false,
|
| 371 |
+
"sep_token_id": null,
|
| 372 |
+
"suppress_tokens": null,
|
| 373 |
+
"task_specific_params": null,
|
| 374 |
+
"temperature": 1.0,
|
| 375 |
+
"tf_legacy_loss": false,
|
| 376 |
+
"tie_encoder_decoder": false,
|
| 377 |
+
"tie_word_embeddings": true,
|
| 378 |
+
"tokenizer_class": null,
|
| 379 |
+
"top_k": 50,
|
| 380 |
+
"top_p": 1.0,
|
| 381 |
+
"torch_dtype": null,
|
| 382 |
+
"torchscript": false,
|
| 383 |
+
"typical_p": 1.0,
|
| 384 |
+
"use_bfloat16": false,
|
| 385 |
+
"vocab_size": 32000
|
| 386 |
+
},
|
| 387 |
+
"vision_feature_layer": -2,
|
| 388 |
+
"vision_feature_select_strategy": "default",
|
| 389 |
+
"vision_tower": "google/siglip-so400m-patch14-384",
|
| 390 |
+
"vision_tower_pretrained": null,
|
| 391 |
+
"vocab_size": 151647
|
| 392 |
+
}
|
unimvu_0.5B_avsd/non_lora_trainables.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:424a0ccaa2ffdfea4f21220b29da1039b8af6dcb6c88c23c80403243c9a0d257
|
| 3 |
+
size 25955038
|