Rebuild tiny HyperCLOVAX Vision V2 from native upstream config

config.json

@@ -1,69 +1,101 @@
 {
+  "anyres": false,
   "architectures": [
     "HyperCLOVAXVisionV2ForConditionalGeneration"
   ],
+  "audio_projector_type": "mlp",
   "dtype": "float32",
-  "image_token_id": 3,
+  "eos_token_id": 128001,
+  "hidden_size": 32,
+  "ignore_index": -100,
+  "image_end_token_id": 128057,
+  "image_start_token_id": 128056,
+  "image_token_id": 128060,
+  "max_num_grids": -1,
   "model_type": "hyperclovax_vision_v2",
+  "num_queries_vis_abstractor": -1,
   "pad_token_id": 0,
+  "possible_resolutions": [],
+  "proj_pos_emb": true,
+  "proj_prenorm": false,
   "text_config": {
+    "architectures": [
+      "HyperCLOVAXForCausalLM"
+    ],
     "attention_bias": false,
     "attention_dropout": 0.0,
     "attention_multiplier": 0.25,
-    "bos_token_id": 0,
+    "bos_token_id": 128000,
     "embedding_multiplier": 1.0,
     "eos_token_id": 128001,
     "head_dim": 16,
-    "hidden_act": "gelu",
+    "hidden_act": "silu",
     "hidden_size": 32,
-    "initializer_range": 0.02,
-    "intermediate_size": 32,
+    "initializer_range": 0.006,
+    "intermediate_size": 64,
     "logits_scaling": 1.0,
-    "max_position_embeddings": 512,
+    "max_position_embeddings": 131072,
     "mlp_bias": false,
     "model_type": "hyperclovax",
     "num_attention_heads": 2,
     "num_hidden_layers": 2,
-    "num_key_value_heads": 2,
+    "num_key_value_heads": 1,
     "pad_token_id": 0,
+    "pretraining_tp": 1,
+    "resid_pdrop": 0.2,
     "residual_multiplier": 1.0,
-    "rms_norm_eps": 1e-06,
+    "rms_norm_eps": 1e-05,
     "rope_parameters": {
-      "rope_theta": 10000.0,
+      "rope_theta": 50000000,
       "rope_type": "default"
     },
     "tie_word_embeddings": false,
     "use_cache": true,
-    "use_post_norm": true,
+    "use_post_norm": false,
     "vocab_size": 128256
   },
-  "tie_word_embeddings": false,
+  "tie_word_embeddings": true,
   "transformers_version": "5.18.0.dev0",
-  "video_token_id": 4,
+  "unpad": false,
+  "use_1x1_grid": false,
+  "video_end_token_id": 128059,
+  "video_first_last_frames_slows": null,
+  "video_max_num_frames": null,
+  "video_num_queries_fast": null,
+  "video_num_queries_slow": null,
+  "video_start_token_id": 128058,
+  "video_token_id": 128061,
   "vision_config": {
+    "anyres": false,
+    "architectures": [
+      "Qwen2_5_VisionTransformerPretrainedModel"
+    ],
     "depth": 2,
     "fullatt_block_indexes": [
-      7,
-      15,
-      23,
-      31
+      1
     ],
-    "hidden_act": "gelu",
+    "hidden_act": "silu",
     "hidden_size": 32,
     "in_channels": 3,
+    "in_chans": 3,
     "initializer_range": 0.02,
-    "intermediate_size": 32,
+    "intermediate_size": 64,
+    "max_num_grids": -1,
     "model_type": "qwen2_5_vl_vision",
-    "num_heads": 4,
+    "num_heads": 2,
     "out_hidden_size": 32,
     "patch_size": 14,
     "rope_parameters": {
       "rope_theta": 10000.0,
       "rope_type": "axial"
     },
-    "spatial_merge_size": 1,
+    "spatial_merge_size": 2,
+    "spatial_patch_size": 14,
     "temporal_patch_size": 2,
-    "tokens_per_second": 1,
+    "tokens_per_second": 2,
     "window_size": 112
-  }
+  },
+  "vision_feature_layer": -2,
+  "vision_input_chunk_size": null,
+  "vision_projector_type": "linear"
 }

generation_config.json

@@ -1,10 +1,12 @@
 {
   "_from_model_config": true,
   "bos_token_id": 0,
+  "do_sample": true,
   "eos_token_id": 128001,
-  "output_attentions": false,
-  "output_hidden_states": false,
-  "pad_token_id": 0,
-  "transformers_version": "5.18.0.dev0",
-  "use_cache": true
+  "num_beams": 1,
+  "repetition_penalty": 1.05,
+  "temperature": 0.5,
+  "top_k": 0,
+  "top_p": 0.6,
+  "transformers_version": "5.18.0.dev0"
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c1adfaf1e9e589f2bce5b1ccc27e7dd6352de4ef07b898735ad1e0545ff1a335
-size 33121552
+oid sha256:b46f5d6a27f50d8c78edd618b4d707f7bf8b224a26e46cf4ae3a24b21147e814
+size 16819288

processor_config.json

@@ -0,0 +1,118 @@
+{
+  "image_processor": {
+    "discrete_image_end_token": "<|discrete_image_end|>",
+    "discrete_image_ratios": [
+      [
+        1,
+        1
+      ],
+      [
+        1,
+        2
+      ],
+      [
+        2,
+        1
+      ],
+      [
+        3,
+        4
+      ],
+      [
+        4,
+        3
+      ],
+      [
+        9,
+        16
+      ],
+      [
+        16,
+        9
+      ],
+      [
+        1,
+        3
+      ],
+      [
+        3,
+        1
+      ]
+    ],
+    "discrete_image_start_token": "<|discrete_image_start|>",
+    "discrete_image_token": "<|DISCRETE_IMAGE_PAD|>",
+    "discrete_token_size": 27,
+    "do_center_crop": false,
+    "do_convert_rgb": true,
+    "do_normalize": true,
+    "do_rescale": true,
+    "do_resize": true,
+    "image_end_token": "<|image_end|>",
+    "image_mean": [
+      0.48145466,
+      0.4578275,
+      0.40821073
+    ],
+    "image_processor_type": "Qwen2VLImageProcessor",
+    "image_start_token": "<|image_start|>",
+    "image_std": [
+      0.26862954,
+      0.26130258,
+      0.27577711
+    ],
+    "image_token": "<|IMAGE_PAD|>",
+    "merge_size": 2,
+    "patch_size": 14,
+    "resample": 3,
+    "rescale_factor": 0.00392156862745098,
+    "resize_mode": "smart_resize",
+    "size": {
+      "longest_edge": 2073600,
+      "shortest_edge": 3136
+    },
+    "temporal_patch_size": 2,
+    "use_discrete_token": false,
+    "vision_eof_token": "<|vision_eof|>",
+    "vision_eol_token": "<|vision_eol|>"
+  },
+  "processor_class": "HyperCLOVAXVisionV2Processor",
+  "video_processor": {
+    "do_convert_rgb": true,
+    "do_normalize": true,
+    "do_rescale": true,
+    "do_resize": true,
+    "do_sample_frames": false,
+    "image_mean": [
+      0.48145466,
+      0.4578275,
+      0.40821073
+    ],
+    "image_std": [
+      0.26862954,
+      0.26130258,
+      0.27577711
+    ],
+    "max_frames": 120,
+    "max_video_tokens": 128000,
+    "merge_size": 2,
+    "min_frames": 4,
+    "patch_size": 14,
+    "resample": 3,
+    "rescale_factor": 0.00392156862745098,
+    "return_metadata": false,
+    "size": {
+      "longest_edge": 285768,
+      "shortest_edge": 3136
+    },
+    "temporal_patch_size": 2,
+    "use_audio_in_video": false,
+    "use_discrete_token": false,
+    "video_audio_token": "<|VIDEO_AUDIO_PAD|>",
+    "video_end_token": "<|video_end|>",
+    "video_processor_type": "Qwen2VLVideoProcessor",
+    "video_start_token": "<|video_start|>",
+    "video_token": "<|VIDEO_PAD|>",
+    "vision_eof_token": "<|vision_eof|>",
+    "vision_eol_token": "<|vision_eol|>"
+  }
+}

tokenizer_config.json

@@ -7,7 +7,7 @@
   "errors": "replace",
   "fix_mistral_regex": true,
   "image_token": "<|IMAGE_PAD|>",
-  "is_local": true,
+  "is_local": false,
   "local_files_only": false,
   "model_max_length": 1000000000000000019884624838656,
   "model_specific_special_tokens": {
@@ -15,7 +15,7 @@
     "video_token": "<|VIDEO_PAD|>"
   },
   "pad_token": "<|endoftext|>",
-  "processor_class": "Exaone4_5_Processor",
+  "processor_class": "HyperCLOVAXVisionV2Processor",
   "sep_token": "<|endoftext|>",
   "tokenizer_class": "GPT2Tokenizer",
   "unk_token": "<|endoftext|>",