HuggingFaceTB/SmolVLM2-256M-Video-Instruct

1{2  "architectures": [3    "SmolVLMForConditionalGeneration"4  ],5  "image_token_id": 49190,6  "model_type": "smolvlm",7  "pad_token_id": 128002,8  "scale_factor": 4,9  "text_config": {10    "_flash_attn_2_enabled": true,11    "_name_or_path": "None",12    "architectures": [13      "VLlama3ForCausalLM"14    ],15    "head_dim": 64,16    "hidden_size": 576,17    "initializer_range": 0.041666666666666664,18    "intermediate_size": 1536,19    "is_llama_config": true,20    "max_position_embeddings": 8192,21    "model_type": "llama",22    "neftune_noise_alpha": 0.0,23    "num_attention_heads": 9,24    "num_hidden_layers": 30,25    "num_key_value_heads": 3,26    "pad_token_id": 2,27    "perceiver_config": {28      "_attn_implementation_autoset": false,29      "_name_or_path": "",30      "add_cross_attention": false,31      "architectures": null,32      "attention_dropout": 0.0,33      "bad_words_ids": null,34      "begin_suppress_tokens": null,35      "bos_token_id": null,36      "chunk_size_feed_forward": 0,37      "cross_attention_hidden_size": null,38      "decoder_start_token_id": null,39      "diversity_penalty": 0.0,40      "do_sample": false,41      "early_stopping": false,42      "encoder_no_repeat_ngram_size": 0,43      "eos_token_id": null,44      "exponential_decay_length_penalty": null,45      "finetuning_task": null,46      "forced_bos_token_id": null,47      "forced_eos_token_id": null,48      "hidden_act": "silu",49      "id2label": {50        "0": "LABEL_0",51        "1": "LABEL_1"52      },53      "is_decoder": false,54      "is_encoder_decoder": false,55      "label2id": {56        "LABEL_0": 0,57        "LABEL_1": 158      },59      "length_penalty": 1.0,60      "max_length": 20,61      "min_length": 0,62      "model_type": "vllama3",63      "no_repeat_ngram_size": 0,64      "num_beam_groups": 1,65      "num_beams": 1,66      "num_key_value_heads": 1,67      "num_return_sequences": 1,68      "output_attentions": false,69      "output_hidden_states": false,70      "output_scores": false,71      "pad_token_id": null,72      "prefix": null,73      "problem_type": null,74      "pruned_heads": {},75      "qk_layer_norms_perceiver": false,76      "remove_invalid_values": false,77      "repetition_penalty": 1.0,78      "resampler_depth": 6,79      "resampler_head_dim": 96,80      "resampler_n_heads": 16,81      "resampler_n_latents": 64,82      "return_dict": true,83      "return_dict_in_generate": false,84      "sep_token_id": null,85      "suppress_tokens": null,86      "task_specific_params": null,87      "temperature": 1.0,88      "tf_legacy_loss": false,89      "tie_encoder_decoder": false,90      "tie_word_embeddings": true,91      "tokenizer_class": null,92      "top_k": 50,93      "top_p": 1.0,94      "torch_dtype": null,95      "torchscript": false,96      "transformers_version": "4.46.0",97      "typical_p": 1.0,98      "use_bfloat16": false99    },100    "pixel_shuffle_factor": 4,101    "qk_layer_norms": false,102    "rms_norm_eps": 1e-05,103    "rope_interleaved": false,104    "rope_theta": 100000,105    "torch_dtype": "bfloat16",106    "transformers.js_config": {107      "kv_cache_dtype": {108        "fp16": "float16",109        "q4f16": "float16"110      }111    },112    "use_resampler": false,113    "vocab_size": 49280114  },115  "tie_word_embeddings": false,116  "torch_dtype": "float32",117  "transformers.js_config": {118    "kv_cache_dtype": {119      "fp16": "float16",120      "q4f16": "float16"121    }122  },123  "transformers_version": "4.47.1",124  "use_cache": false,125  "use_reentrant_checkpointing": false,126  "vision_config": {127    "hidden_size": 768,128    "image_size": 512,129    "max_image_size": {130      "longest_edge": 512131    },132    "model_type": "smolvlm_vision",133    "num_attention_heads": 12,134    "patch_size": 16,135    "size": {136      "longest_edge": 2048137    },138    "tie_word_embeddings": false,139    "use_base_siglip": true140  },141  "vocab_size": 49280142}143