Text Generation
Transformers
Safetensors
PyTorch
llama4
image-text-to-text
facebook
meta
llama
fine-tuned
conversational-ai
large-language-model
huggingface
open-source-llm
generative-ai
meta-llama
instruct-tuning
chat-model
llm
artificial-intelligence
deep-learning
tensorrt-llm
gpu-optimized
multilingual
instruction-following
conversational
text-generation-inference
Uploading optimized model files
Browse files- config.json +80 -0
config.json
ADDED
|
@@ -0,0 +1,80 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"Llama4ForConditionalGeneration"
|
| 4 |
+
],
|
| 5 |
+
"boi_token_index": 200080,
|
| 6 |
+
"eoi_token_index": 200081,
|
| 7 |
+
"image_token_index": 200092,
|
| 8 |
+
"model_type": "llama4",
|
| 9 |
+
"text_config": {
|
| 10 |
+
"_attn_implementation_autoset": true,
|
| 11 |
+
"attention_bias": false,
|
| 12 |
+
"attention_chunk_size": 8192,
|
| 13 |
+
"attention_dropout": 0.0,
|
| 14 |
+
"bos_token_id": 200000,
|
| 15 |
+
"eos_token_id": [
|
| 16 |
+
200001,
|
| 17 |
+
200007,
|
| 18 |
+
200008
|
| 19 |
+
],
|
| 20 |
+
"for_llm_compressor": false,
|
| 21 |
+
"head_dim": 128,
|
| 22 |
+
"hidden_act": "silu",
|
| 23 |
+
"hidden_size": 5120,
|
| 24 |
+
"initializer_range": 0.02,
|
| 25 |
+
"interleave_moe_layer_step": 1,
|
| 26 |
+
"intermediate_size": 8192,
|
| 27 |
+
"intermediate_size_mlp": 16384,
|
| 28 |
+
"max_position_embeddings": 10485760,
|
| 29 |
+
"model_type": "llama4_text",
|
| 30 |
+
"no_rope_layers": [],
|
| 31 |
+
"num_attention_heads": 40,
|
| 32 |
+
"num_experts_per_tok": 1,
|
| 33 |
+
"num_hidden_layers": 48,
|
| 34 |
+
"num_key_value_heads": 8,
|
| 35 |
+
"num_local_experts": 16,
|
| 36 |
+
"output_router_logits": false,
|
| 37 |
+
"pad_token_id": 200018,
|
| 38 |
+
"rms_norm_eps": 1e-05,
|
| 39 |
+
"rope_scaling": {
|
| 40 |
+
"factor": 16.0,
|
| 41 |
+
"high_freq_factor": 1.0,
|
| 42 |
+
"low_freq_factor": 1.0,
|
| 43 |
+
"original_max_position_embeddings": 8192,
|
| 44 |
+
"rope_type": "llama3"
|
| 45 |
+
},
|
| 46 |
+
"rope_theta": 500000.0,
|
| 47 |
+
"router_aux_loss_coef": 0.001,
|
| 48 |
+
"router_jitter_noise": 0.0,
|
| 49 |
+
"torch_dtype": "bfloat16",
|
| 50 |
+
"use_cache": true,
|
| 51 |
+
"use_qk_norm": true,
|
| 52 |
+
"vocab_size": 202048
|
| 53 |
+
},
|
| 54 |
+
"torch_dtype": "bfloat16",
|
| 55 |
+
"transformers_version": "4.51.0.dev0",
|
| 56 |
+
"vision_config": {
|
| 57 |
+
"_attn_implementation_autoset": true,
|
| 58 |
+
"attention_dropout": 0.0,
|
| 59 |
+
"hidden_act": "gelu",
|
| 60 |
+
"hidden_size": 1408,
|
| 61 |
+
"image_size": 336,
|
| 62 |
+
"initializer_range": 0.02,
|
| 63 |
+
"intermediate_size": 5632,
|
| 64 |
+
"model_type": "llama4_vision_model",
|
| 65 |
+
"multi_modal_projector_bias": false,
|
| 66 |
+
"norm_eps": 1e-05,
|
| 67 |
+
"num_attention_heads": 16,
|
| 68 |
+
"num_channels": 3,
|
| 69 |
+
"num_hidden_layers": 34,
|
| 70 |
+
"patch_size": 14,
|
| 71 |
+
"pixel_shuffle_ratio": 0.5,
|
| 72 |
+
"projector_dropout": 0.0,
|
| 73 |
+
"projector_input_dim": 4096,
|
| 74 |
+
"projector_output_dim": 4096,
|
| 75 |
+
"rope_theta": 10000,
|
| 76 |
+
"vision_feature_layer": -1,
|
| 77 |
+
"vision_feature_select_strategy": "default",
|
| 78 |
+
"vision_output_dim": 4096
|
| 79 |
+
}
|
| 80 |
+
}
|