trl-internal-testing/tiny-SmolVLMForConditionalGeneration
06.9k
1{2 "architectures": [3 "SmolVLMForConditionalGeneration"4 ],5 "dtype": "float32",6 "image_token_id": 49190,7 "model_type": "smolvlm",8 "pad_token_id": 128002,9 "scale_factor": 3,10 "text_config": {11 "_flash_attn_2_enabled": true,12 "_name_or_path": "None",13 "architectures": [14 "VLlama3ForCausalLM"15 ],16 "attention_bias": false,17 "attention_dropout": 0.0,18 "dtype": "bfloat16",19 "head_dim": 64,20 "hidden_act": "silu",21 "hidden_size": 16,22 "initializer_range": 0.02,23 "intermediate_size": 8192,24 "layer_types": null,25 "max_position_embeddings": 8192,26 "mlp_bias": false,27 "model_type": "llama",28 "neftune_noise_alpha": 0.0,29 "num_attention_heads": 4,30 "num_hidden_layers": 2,31 "num_key_value_heads": 2,32 "pad_token_id": 2,33 "perceiver_config": {34 "_attn_implementation_autoset": false,35 "_name_or_path": "",36 "add_cross_attention": false,37 "architectures": null,38 "attention_dropout": 0.0,39 "bad_words_ids": null,40 "begin_suppress_tokens": null,41 "bos_token_id": null,42 "chunk_size_feed_forward": 0,43 "cross_attention_hidden_size": null,44 "decoder_start_token_id": null,45 "diversity_penalty": 0.0,46 "do_sample": false,47 "early_stopping": false,48 "encoder_no_repeat_ngram_size": 0,49 "eos_token_id": null,50 "exponential_decay_length_penalty": null,51 "finetuning_task": null,52 "forced_bos_token_id": null,53 "forced_eos_token_id": null,54 "hidden_act": "silu",55 "id2label": {56 "0": "LABEL_0",57 "1": "LABEL_1"58 },59 "is_decoder": false,60 "is_encoder_decoder": false,61 "label2id": {62 "LABEL_0": 0,63 "LABEL_1": 164 },65 "length_penalty": 1.0,66 "max_length": 20,67 "min_length": 0,68 "model_type": "vllama3",69 "no_repeat_ngram_size": 0,70 "num_beam_groups": 1,71 "num_beams": 1,72 "num_key_value_heads": 1,73 "num_return_sequences": 1,74 "output_attentions": false,75 "output_hidden_states": false,76 "output_scores": false,77 "pad_token_id": null,78 "prefix": null,79 "problem_type": null,80 "pruned_heads": {},81 "qk_layer_norms_perceiver": false,82 "remove_invalid_values": false,83 "repetition_penalty": 1.0,84 "resampler_depth": 6,85 "resampler_head_dim": 96,86 "resampler_n_heads": 16,87 "resampler_n_latents": 64,88 "return_dict": true,89 "return_dict_in_generate": false,90 "sep_token_id": null,91 "suppress_tokens": null,92 "task_specific_params": null,93 "temperature": 1.0,94 "tf_legacy_loss": false,95 "tie_encoder_decoder": false,96 "tie_word_embeddings": true,97 "tokenizer_class": null,98 "top_k": 50,99 "top_p": 1.0,100 "torch_dtype": null,101 "torchscript": false,102 "transformers_version": "4.46.0",103 "typical_p": 1.0,104 "use_bfloat16": false105 },106 "pixel_shuffle_factor": 3,107 "pretraining_tp": 1,108 "qk_layer_norms": false,109 "rms_norm_eps": 1e-05,110 "rope_scaling": null,111 "rope_theta": 130000,112 "transformers.js_config": {113 "kv_cache_dtype": {114 "fp16": "float16",115 "q4f16": "float16"116 }117 },118 "use_cache": true,119 "use_resampler": false,120 "vocab_size": 49280121 },122 "tie_word_embeddings": false,123 "transformers_version": "4.57.3",124 "use_cache": false,125 "use_reentrant_checkpointing": false,126 "vision_config": {127 "attention_dropout": 0.0,128 "embed_dim": 64,129 "hidden_act": "gelu_pytorch_tanh",130 "hidden_size": 16,131 "image_size": 384,132 "initializer_range": 0.02,133 "intermediate_size": 4304,134 "layer_norm_eps": 1e-06,135 "max_image_size": {136 "longest_edge": 384137 },138 "model_type": "smolvlm_vision",139 "num_attention_heads": 4,140 "num_channels": 3,141 "num_hidden_layers": 2,142 "num_key_value_heads": 2,143 "patch_size": 14,144 "size": {145 "longest_edge": 1920146 },147 "tie_word_embeddings": false,148 "use_base_siglip": false149 },150 "vocab_size": 49280151}152 