Add HF compatible checkpoints (#3)

- Upload folder using huggingface_hub (545cedc82d1e8c090dfab9c189334960e3e34f33)
- Remove old file (fb256d38b0379522815acaef3afcce8c3491a913)

config.json

@@ -1,6 +1,6 @@
 {
   "architectures": [
-    "AIMv2Model"
+    "Aimv2VisionModel"
   ],
   "attention_dropout": 0.0,
   "auto_map": {
@@ -8,10 +8,14 @@
     "AutoModel": "modeling_aimv2.AIMv2Model",
     "FlaxAutoModel": "modeling_flax_aimv2.FlaxAIMv2Model"
   },
+  "hidden_act": "silu",
   "hidden_size": 3072,
   "image_size": 448,
+  "initializer_range": 0.02,
   "intermediate_size": 8192,
-  "model_type": "aimv2",
+  "is_native": false,
+  "mlp_bias": false,
+  "model_type": "aimv2_vision_model",
   "num_attention_heads": 24,
   "num_channels": 3,
   "num_hidden_layers": 24,
@@ -20,6 +24,7 @@
   "qkv_bias": false,
   "rms_norm_eps": 1e-05,
   "torch_dtype": "float32",
-  "transformers_version": "4.46.3",
-  "use_bias": false
-}
+  "transformers_version": "4.54.0.dev0",
+  "use_bias": false,
+  "use_head": false
+}

model.safetensors → model-00001-of-00003.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2663c7027af21885e56160748f449343be0dabe9e3cc8d447d18e1f7b814cdc0
-size 10892089304
+oid sha256:dd819ea84275237a07324447ef4e300e816f66219ef7e6f68c03429ece33058e
+size 4902259840

model-00002-of-00003.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5fbcf46425d1e93cd74f55744d819de39d29ee184ee6ad6ab4216fbf90f11bde
+size 4983114904

model-00003-of-00003.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d480a95d8c05261a74af6b500902af63b62e446beae3379277f6dd7f8f10daab
+size 1006721472

model.safetensors.index.json

@@ -0,0 +1,229 @@
+{
+  "metadata": {
+    "total_parameters": 2723017728,
+    "total_size": 10892070912
+  },
+  "weight_map": {
+    "embeddings.patch_embed.bias": "model-00001-of-00003.safetensors",
+    "embeddings.patch_embed.weight": "model-00001-of-00003.safetensors",
+    "embeddings.position_embedding.weight": "model-00001-of-00003.safetensors",
+    "embeddings.rms_norm.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.0.attention.k_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.0.attention.out_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.0.attention.q_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.0.attention.v_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.0.ffn.down_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.0.ffn.gate_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.0.ffn.up_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.0.rms_norm1.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.0.rms_norm2.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.1.attention.k_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.1.attention.out_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.1.attention.q_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.1.attention.v_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.1.ffn.down_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.1.ffn.gate_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.1.ffn.up_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.1.rms_norm1.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.1.rms_norm2.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.10.attention.k_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.10.attention.out_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.10.attention.q_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.10.attention.v_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.10.ffn.down_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.10.ffn.gate_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.10.ffn.up_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.10.rms_norm1.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.10.rms_norm2.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.11.attention.k_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.11.attention.out_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.11.attention.q_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.11.attention.v_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.11.ffn.down_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.11.ffn.gate_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.11.ffn.up_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.11.rms_norm1.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.11.rms_norm2.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.12.attention.k_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.12.attention.out_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.12.attention.q_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.12.attention.v_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.12.ffn.down_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.12.ffn.gate_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.12.ffn.up_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.12.rms_norm1.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.12.rms_norm2.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.13.attention.k_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.13.attention.out_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.13.attention.q_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.13.attention.v_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.13.ffn.down_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.13.ffn.gate_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.13.ffn.up_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.13.rms_norm1.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.13.rms_norm2.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.14.attention.k_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.14.attention.out_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.14.attention.q_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.14.attention.v_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.14.ffn.down_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.14.ffn.gate_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.14.ffn.up_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.14.rms_norm1.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.14.rms_norm2.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.15.attention.k_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.15.attention.out_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.15.attention.q_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.15.attention.v_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.15.ffn.down_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.15.ffn.gate_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.15.ffn.up_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.15.rms_norm1.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.15.rms_norm2.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.16.attention.k_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.16.attention.out_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.16.attention.q_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.16.attention.v_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.16.ffn.down_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.16.ffn.gate_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.16.ffn.up_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.16.rms_norm1.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.16.rms_norm2.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.17.attention.k_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.17.attention.out_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.17.attention.q_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.17.attention.v_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.17.ffn.down_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.17.ffn.gate_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.17.ffn.up_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.17.rms_norm1.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.17.rms_norm2.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.18.attention.k_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.18.attention.out_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.18.attention.q_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.18.attention.v_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.18.ffn.down_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.18.ffn.gate_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.18.ffn.up_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.18.rms_norm1.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.18.rms_norm2.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.19.attention.k_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.19.attention.out_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.19.attention.q_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.19.attention.v_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.19.ffn.down_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.19.ffn.gate_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.19.ffn.up_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.19.rms_norm1.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.19.rms_norm2.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.2.attention.k_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.2.attention.out_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.2.attention.q_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.2.attention.v_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.2.ffn.down_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.2.ffn.gate_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.2.ffn.up_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.2.rms_norm1.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.2.rms_norm2.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.20.attention.k_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.20.attention.out_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.20.attention.q_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.20.attention.v_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.20.ffn.down_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.20.ffn.gate_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.20.ffn.up_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.20.rms_norm1.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.20.rms_norm2.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.21.attention.k_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.21.attention.out_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.21.attention.q_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.21.attention.v_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.21.ffn.down_proj.weight": "model-00003-of-00003.safetensors",
+    "encoder.layers.21.ffn.gate_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.21.ffn.up_proj.weight": "model-00002-of-00003.safetensors",
+    "encoder.layers.21.rms_norm1.weight": "model-00003-of-00003.safetensors",
+    "encoder.layers.21.rms_norm2.weight": "model-00003-of-00003.safetensors",
+    "encoder.layers.22.attention.k_proj.weight": "model-00003-of-00003.safetensors",
+    "encoder.layers.22.attention.out_proj.weight": "model-00003-of-00003.safetensors",
+    "encoder.layers.22.attention.q_proj.weight": "model-00003-of-00003.safetensors",
+    "encoder.layers.22.attention.v_proj.weight": "model-00003-of-00003.safetensors",
+    "encoder.layers.22.ffn.down_proj.weight": "model-00003-of-00003.safetensors",
+    "encoder.layers.22.ffn.gate_proj.weight": "model-00003-of-00003.safetensors",
+    "encoder.layers.22.ffn.up_proj.weight": "model-00003-of-00003.safetensors",
+    "encoder.layers.22.rms_norm1.weight": "model-00003-of-00003.safetensors",
+    "encoder.layers.22.rms_norm2.weight": "model-00003-of-00003.safetensors",
+    "encoder.layers.23.attention.k_proj.weight": "model-00003-of-00003.safetensors",
+    "encoder.layers.23.attention.out_proj.weight": "model-00003-of-00003.safetensors",
+    "encoder.layers.23.attention.q_proj.weight": "model-00003-of-00003.safetensors",
+    "encoder.layers.23.attention.v_proj.weight": "model-00003-of-00003.safetensors",
+    "encoder.layers.23.ffn.down_proj.weight": "model-00003-of-00003.safetensors",
+    "encoder.layers.23.ffn.gate_proj.weight": "model-00003-of-00003.safetensors",
+    "encoder.layers.23.ffn.up_proj.weight": "model-00003-of-00003.safetensors",
+    "encoder.layers.23.rms_norm1.weight": "model-00003-of-00003.safetensors",
+    "encoder.layers.23.rms_norm2.weight": "model-00003-of-00003.safetensors",
+    "encoder.layers.3.attention.k_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.3.attention.out_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.3.attention.q_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.3.attention.v_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.3.ffn.down_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.3.ffn.gate_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.3.ffn.up_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.3.rms_norm1.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.3.rms_norm2.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.4.attention.k_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.4.attention.out_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.4.attention.q_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.4.attention.v_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.4.ffn.down_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.4.ffn.gate_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.4.ffn.up_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.4.rms_norm1.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.4.rms_norm2.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.5.attention.k_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.5.attention.out_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.5.attention.q_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.5.attention.v_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.5.ffn.down_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.5.ffn.gate_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.5.ffn.up_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.5.rms_norm1.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.5.rms_norm2.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.6.attention.k_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.6.attention.out_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.6.attention.q_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.6.attention.v_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.6.ffn.down_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.6.ffn.gate_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.6.ffn.up_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.6.rms_norm1.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.6.rms_norm2.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.7.attention.k_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.7.attention.out_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.7.attention.q_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.7.attention.v_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.7.ffn.down_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.7.ffn.gate_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.7.ffn.up_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.7.rms_norm1.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.7.rms_norm2.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.8.attention.k_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.8.attention.out_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.8.attention.q_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.8.attention.v_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.8.ffn.down_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.8.ffn.gate_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.8.ffn.up_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.8.rms_norm1.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.8.rms_norm2.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.9.attention.k_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.9.attention.out_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.9.attention.q_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.9.attention.v_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.9.ffn.down_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.9.ffn.gate_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.9.ffn.up_proj.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.9.rms_norm1.weight": "model-00001-of-00003.safetensors",
+    "encoder.layers.9.rms_norm2.weight": "model-00001-of-00003.safetensors",
+    "rms_norm.weight": "model-00003-of-00003.safetensors"
+  }
+}

preprocessor_config.json

@@ -3,6 +3,10 @@
     "height": 448,
     "width": 448
   },
+  "data_format": "channels_first",
+  "default_to_square": false,
+  "device": null,
+  "disable_grouping": null,
   "do_center_crop": true,
   "do_convert_rgb": true,
   "do_normalize": true,
@@ -13,14 +17,16 @@
     0.4578275,
     0.40821073
   ],
-  "image_processor_type": "CLIPImageProcessor",
+  "image_processor_type": "CLIPImageProcessorFast",
   "image_std": [
     0.26862954,
     0.26130258,
     0.27577711
   ],
+  "input_data_format": null,
   "resample": 3,
   "rescale_factor": 0.00392156862745098,
+  "return_tensors": null,
   "size": {
     "shortest_edge": 448
   }