INT4 quantization of LiquidAI/LFM2-VL-3B (LLM backbone quantized, vision encoder fp16)

chat_template.jinja

@@ -0,0 +1,4 @@
+{{bos_token}}{% for message in messages %}{{'<|im_start|>' + message['role'] + '
+'}}{% if message['content'] is string %}{{ message['content'] }}{% else %}{% for content in message['content'] %}{% if content['type'] == 'image' %}{{ '<image>' }}{% elif content['type'] == 'text' %}{{ content['text'] }}{% endif %}{% endfor %}{% endif %}{{'<|im_end|>
+'}}{% endfor %}{% if add_generation_prompt %}{{'<|im_start|>assistant
+' }}{% endif %}

config.json

@@ -0,0 +1,117 @@
+{
+  "architectures": [
+    "Lfm2VlForConditionalGeneration"
+  ],
+  "do_image_splitting": true,
+  "downsample_factor": 2,
+  "dtype": "float16",
+  "encoder_patch_size": 16,
+  "image_token_id": 396,
+  "max_image_tokens": 256,
+  "max_pixels_tolerance": 2.0,
+  "max_tiles": 10,
+  "min_image_tokens": 64,
+  "min_tiles": 2,
+  "model_type": "lfm2_vl",
+  "projector_bias": true,
+  "projector_hidden_act": "gelu",
+  "projector_hidden_size": 2560,
+  "projector_use_layernorm": true,
+  "text_config": {
+    "_name_or_path": "LiquidAI/LFM2-2.6B",
+    "architectures": [
+      "Lfm2ForCausalLM"
+    ],
+    "block_auto_adjust_ff_dim": false,
+    "block_dim": 2048,
+    "block_ff_dim": 10752,
+    "block_ffn_dim_multiplier": 1.0,
+    "block_mlp_init_scale": 1.0,
+    "block_multiple_of": 256,
+    "block_norm_eps": 1e-05,
+    "block_out_init_scale": 1.0,
+    "block_use_swiglu": true,
+    "block_use_xavier_init": true,
+    "bos_token_id": 1,
+    "conv_L_cache": 3,
+    "conv_bias": false,
+    "conv_dim": 2048,
+    "conv_dim_out": 2048,
+    "conv_use_xavier_init": true,
+    "dtype": "float16",
+    "eos_token_id": 7,
+    "hidden_size": 2048,
+    "initializer_range": 0.02,
+    "intermediate_size": 10752,
+    "layer_types": [
+      "conv",
+      "conv",
+      "full_attention",
+      "conv",
+      "conv",
+      "full_attention",
+      "conv",
+      "conv",
+      "conv",
+      "full_attention",
+      "conv",
+      "conv",
+      "conv",
+      "full_attention",
+      "conv",
+      "conv",
+      "conv",
+      "full_attention",
+      "conv",
+      "conv",
+      "conv",
+      "full_attention",
+      "conv",
+      "conv",
+      "full_attention",
+      "conv",
+      "conv",
+      "full_attention",
+      "conv",
+      "conv"
+    ],
+    "max_position_embeddings": 128000,
+    "model_type": "lfm2",
+    "norm_eps": 1e-05,
+    "num_attention_heads": 32,
+    "num_heads": 32,
+    "num_hidden_layers": 30,
+    "num_key_value_heads": 8,
+    "pad_token_id": 0,
+    "rope_parameters": {
+      "rope_theta": 1000000.0,
+      "rope_type": "default"
+    },
+    "theta": 1000000.0,
+    "tie_embedding": true,
+    "tie_word_embeddings": true,
+    "use_cache": true,
+    "use_pos_enc": true,
+    "vocab_size": 65536
+  },
+  "tie_word_embeddings": true,
+  "tile_size": 512,
+  "transformers_version": "5.3.0",
+  "use_image_special_tokens": true,
+  "use_thumbnail": true,
+  "vision_config": {
+    "attention_dropout": 0.0,
+    "dtype": "float16",
+    "hidden_act": "gelu_pytorch_tanh",
+    "hidden_size": 1152,
+    "intermediate_size": 4304,
+    "layer_norm_eps": 1e-06,
+    "model_type": "siglip2_vision_model",
+    "num_attention_heads": 16,
+    "num_channels": 3,
+    "num_hidden_layers": 27,
+    "num_patches": 256,
+    "patch_size": 16,
+    "vision_use_head": false
+  }
+}

generation_config.json

@@ -0,0 +1,7 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 1,
+  "eos_token_id": 7,
+  "pad_token_id": 0,
+  "transformers_version": "5.3.0"
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:08cc31871a9217753251d7d1344d95ae47f0e0f9f57ad8535eab17831452e7d8
+size 5998047656

processor_config.json

@@ -0,0 +1,41 @@
+{
+  "image_processor": {
+    "data_format": "channels_first",
+    "default_to_square": true,
+    "do_image_splitting": true,
+    "do_normalize": true,
+    "do_pad": true,
+    "do_rescale": true,
+    "do_resize": false,
+    "downsample_factor": 2,
+    "encoder_patch_size": 16,
+    "image_mean": [
+      0.5,
+      0.5,
+      0.5
+    ],
+    "image_processor_type": "Lfm2VlImageProcessorFast",
+    "image_std": [
+      0.5,
+      0.5,
+      0.5
+    ],
+    "max_image_tokens": 256,
+    "max_num_patches": 1024,
+    "max_pixels_tolerance": 2.0,
+    "max_tiles": 10,
+    "min_image_tokens": 64,
+    "min_tiles": 2,
+    "resample": 3,
+    "rescale_factor": 0.00392156862745098,
+    "return_row_col_info": true,
+    "size": {
+      "height": 512,
+      "width": 512
+    },
+    "tile_size": 512,
+    "use_numpy_backend": true,
+    "use_thumbnail": true
+  },
+  "processor_class": "Lfm2VlProcessor"
+}

quantization_config.json

@@ -0,0 +1,10 @@
+{
+  "base_model": "LiquidAI/LFM2-VL-3B",
+  "quantization": "int4_per_group_symmetric",
+  "group_size": 128,
+  "bits": 4,
+  "method": "static_int4_dequantized",
+  "description": "INT4 per-group symmetric quantization of LLM backbone weights. Vision encoder kept in fp16. Weights stored as dequantized fp16 for maximum compatibility.",
+  "quantized_layers": 167,
+  "skipped_vision_layers": 165
+}

tokenizer_config.json

@@ -0,0 +1,27 @@
+{
+  "backend": "tokenizers",
+  "bos_token": "<|startoftext|>",
+  "clean_up_tokenization_spaces": true,
+  "eos_token": "<|im_end|>",
+  "image_end_token": "<|image_end|>",
+  "image_start_token": "<|image_start|>",
+  "image_thumbnail": "<|img_thumbnail|>",
+  "image_token": "<image>",
+  "is_local": false,
+  "legacy": false,
+  "model_max_length": 1000000000000000019884624838656,
+  "model_specific_special_tokens": {
+    "image_end_token": "<|image_end|>",
+    "image_start_token": "<|image_start|>",
+    "image_thumbnail": "<|img_thumbnail|>",
+    "image_token": "<image>"
+  },
+  "pad_token": "<|pad|>",
+  "processor_class": "Lfm2VlProcessor",
+  "return_token_type_ids": false,
+  "sp_model_kwargs": {},
+  "spaces_between_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "use_default_system_prompt": false,
+  "use_fast": true
+}