Upload 2 files
Browse files- config.json +84 -0
- model-00001-of-00009.safetensors +3 -0
config.json
ADDED
|
@@ -0,0 +1,84 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"GraniteSpeechForConditionalGeneration"
|
| 4 |
+
],
|
| 5 |
+
"audio_token_index": 49155,
|
| 6 |
+
"auto_map": {
|
| 7 |
+
"AutoConfig": "configuration_granite_speech.GraniteSpeechConfig",
|
| 8 |
+
"AutoFeatureExtractor": "feature_extraction_granite_speech.GraniteSpeechFeatureExtractor",
|
| 9 |
+
"AutoModelForSpeechSeq2Seq": "modeling_granite_speech.GraniteSpeechForConditionalGeneration",
|
| 10 |
+
"AutoProcessor": "processing_granite_speech.GraniteSpeechProcessor"
|
| 11 |
+
},
|
| 12 |
+
"encoder_config": {
|
| 13 |
+
"context_size": 200,
|
| 14 |
+
"conv_expansion_factor": 2,
|
| 15 |
+
"conv_kernel_size": 15,
|
| 16 |
+
"dim_head": 128,
|
| 17 |
+
"dropout": 0.1,
|
| 18 |
+
"feedforward_mult": 4,
|
| 19 |
+
"hidden_dim": 1024,
|
| 20 |
+
"input_dim": 160,
|
| 21 |
+
"model_type": "granite_speech_encoder",
|
| 22 |
+
"num_heads": 8,
|
| 23 |
+
"num_layers": 10,
|
| 24 |
+
"output_dim": 42
|
| 25 |
+
},
|
| 26 |
+
"has_lora_adapter": true,
|
| 27 |
+
"initializer_range": 0.02,
|
| 28 |
+
"model_type": "granite_speech",
|
| 29 |
+
"projector_config": {
|
| 30 |
+
"_attn_implementation_autoset": true,
|
| 31 |
+
"attention_probs_dropout_prob": 0.1,
|
| 32 |
+
"cross_attention_frequency": 1,
|
| 33 |
+
"downsample_rate": 5,
|
| 34 |
+
"encoder_hidden_size": 1024,
|
| 35 |
+
"hidden_act": "gelu",
|
| 36 |
+
"hidden_dropout_prob": 0.1,
|
| 37 |
+
"hidden_size": 1024,
|
| 38 |
+
"initializer_range": 0.02,
|
| 39 |
+
"intermediate_size": 4096,
|
| 40 |
+
"layer_norm_eps": 1e-12,
|
| 41 |
+
"llm_dim": 4096,
|
| 42 |
+
"max_position_embeddings": 2048,
|
| 43 |
+
"model_type": "granite_speech_qformer",
|
| 44 |
+
"num_attention_heads": 16,
|
| 45 |
+
"num_hidden_layers": 2,
|
| 46 |
+
"position_embedding_type": "absolute",
|
| 47 |
+
"use_qformer_text_input": false,
|
| 48 |
+
"window_size": 15
|
| 49 |
+
},
|
| 50 |
+
"text_config": {
|
| 51 |
+
"_name_or_path": "ibm-granite/granite-3.2-8b-instruct",
|
| 52 |
+
"architectures": [
|
| 53 |
+
"GraniteForCausalLM"
|
| 54 |
+
],
|
| 55 |
+
"attention_bias": false,
|
| 56 |
+
"attention_dropout": 0.0,
|
| 57 |
+
"attention_multiplier": 0.0078125,
|
| 58 |
+
"bos_token_id": 0,
|
| 59 |
+
"embedding_multiplier": 12.0,
|
| 60 |
+
"eos_token_id": 0,
|
| 61 |
+
"hidden_act": "silu",
|
| 62 |
+
"hidden_size": 4096,
|
| 63 |
+
"initializer_range": 0.02,
|
| 64 |
+
"intermediate_size": 12800,
|
| 65 |
+
"logits_scaling": 16.0,
|
| 66 |
+
"max_position_embeddings": 131072,
|
| 67 |
+
"mlp_bias": false,
|
| 68 |
+
"model_type": "granite",
|
| 69 |
+
"num_attention_heads": 32,
|
| 70 |
+
"num_hidden_layers": 40,
|
| 71 |
+
"num_key_value_heads": 8,
|
| 72 |
+
"pad_token_id": 0,
|
| 73 |
+
"residual_multiplier": 0.22,
|
| 74 |
+
"rms_norm_eps": 1e-05,
|
| 75 |
+
"rope_scaling": null,
|
| 76 |
+
"rope_theta": 10000000.0,
|
| 77 |
+
"tie_word_embeddings": true,
|
| 78 |
+
"torch_dtype": "bfloat16",
|
| 79 |
+
"use_cache": true,
|
| 80 |
+
"vocab_size": 49156
|
| 81 |
+
},
|
| 82 |
+
"torch_dtype": "bfloat16",
|
| 83 |
+
"transformers_version": "4.50.0.dev0"
|
| 84 |
+
}
|
model-00001-of-00009.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3995536316905c2e9216bea3060c531be819405bf74816a7813f9eb7750bf5f0
|
| 3 |
+
size 1996591848
|