Delete config.json
Browse files- config.json +0 -72
config.json
DELETED
|
@@ -1,72 +0,0 @@
|
|
| 1 |
-
{
|
| 2 |
-
"activation_type": "swiglu",
|
| 3 |
-
"alibi": false,
|
| 4 |
-
"alibi_bias_max": 8.0,
|
| 5 |
-
"architectures": [
|
| 6 |
-
"LlavaOLMoBitnet1BForCausalLM"
|
| 7 |
-
],
|
| 8 |
-
"attention_dropout": 0.0,
|
| 9 |
-
"attention_layer_norm": false,
|
| 10 |
-
"attention_layer_norm_with_affine": false,
|
| 11 |
-
"auto_map": {
|
| 12 |
-
"AutoConfig": "configuration_olmo.OLMoConfig",
|
| 13 |
-
"AutoModelForCausalLM": "modeling_olmo.OLMoForCausalLM"
|
| 14 |
-
},
|
| 15 |
-
"bias_for_layer_norm": false,
|
| 16 |
-
"block_group_size": 1,
|
| 17 |
-
"block_type": "sequential",
|
| 18 |
-
"bos_token_id": 50279,
|
| 19 |
-
"clip_qkv": null,
|
| 20 |
-
"d_model": 2048,
|
| 21 |
-
"embedding_dropout": 0.0,
|
| 22 |
-
"embedding_size": 50304,
|
| 23 |
-
"eos_token_id": 50279,
|
| 24 |
-
"flash_attention": true,
|
| 25 |
-
"freeze_mm_mlp_adapter": false,
|
| 26 |
-
"freeze_mm_vision_resampler": false,
|
| 27 |
-
"image_aspect_ratio": "pad",
|
| 28 |
-
"include_bias": false,
|
| 29 |
-
"inference_mode": false,
|
| 30 |
-
"init_cutoff_factor": null,
|
| 31 |
-
"init_device": "cpu",
|
| 32 |
-
"init_fn": "mitchell",
|
| 33 |
-
"init_std": 0.02,
|
| 34 |
-
"layer_norm_type": "rms",
|
| 35 |
-
"layer_norm_with_affine": true,
|
| 36 |
-
"max_sequence_length": 2048,
|
| 37 |
-
"mlp_hidden_size": null,
|
| 38 |
-
"mlp_ratio": 8,
|
| 39 |
-
"mm_hidden_size": 1024,
|
| 40 |
-
"mm_patch_merge_type": "flat",
|
| 41 |
-
"mm_projector_lr": null,
|
| 42 |
-
"mm_projector_type": "mlp2x_gelu",
|
| 43 |
-
"mm_resampler_type": null,
|
| 44 |
-
"mm_use_im_patch_token": false,
|
| 45 |
-
"mm_use_im_start_end": false,
|
| 46 |
-
"mm_vision_select_feature": "patch",
|
| 47 |
-
"mm_vision_select_layer": -2,
|
| 48 |
-
"mm_vision_tower": "openai/clip-vit-large-patch14-336",
|
| 49 |
-
"model_type": "IntelLabs/LlavaOLMoBitnet1B",
|
| 50 |
-
"multi_query_attention": false,
|
| 51 |
-
"n_heads": 16,
|
| 52 |
-
"n_kv_heads": null,
|
| 53 |
-
"n_layers": 16,
|
| 54 |
-
"pad_token_id": 1,
|
| 55 |
-
"precision": "amp_bf16",
|
| 56 |
-
"residual_dropout": 0.0,
|
| 57 |
-
"rope": true,
|
| 58 |
-
"rope_full_precision": true,
|
| 59 |
-
"scale_logits": false,
|
| 60 |
-
"ternary": true,
|
| 61 |
-
"tokenizer_model_max_length": 2048,
|
| 62 |
-
"tokenizer_padding_side": "right",
|
| 63 |
-
"torch_dtype": "float32",
|
| 64 |
-
"transformers_version": "4.37.2",
|
| 65 |
-
"tune_mm_mlp_adapter": false,
|
| 66 |
-
"tune_mm_vision_resampler": false,
|
| 67 |
-
"unfreeze_mm_vision_tower": false,
|
| 68 |
-
"use_cache": true,
|
| 69 |
-
"use_mm_proj": true,
|
| 70 |
-
"vocab_size": 50280,
|
| 71 |
-
"weight_tying": true
|
| 72 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|