RichardErkhov
commited on
Commit
•
d18a64f
1
Parent(s):
bd32713
uploaded model
Browse files- config.json +62 -0
config.json
ADDED
@@ -0,0 +1,62 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"_name_or_path": "bigbird-pegasus-large-K-booksum",
|
3 |
+
"activation_dropout": 0.0,
|
4 |
+
"activation_function": "gelu_new",
|
5 |
+
"architectures": [
|
6 |
+
"BigBirdPegasusForCausalLM"
|
7 |
+
],
|
8 |
+
"attention_dropout": 0.0,
|
9 |
+
"attention_type": "block_sparse",
|
10 |
+
"block_size": 64,
|
11 |
+
"bos_token_id": 2,
|
12 |
+
"classifier_dropout": 0.0,
|
13 |
+
"d_model": 1024,
|
14 |
+
"decoder_attention_heads": 16,
|
15 |
+
"decoder_ffn_dim": 4096,
|
16 |
+
"decoder_layerdrop": 0.0,
|
17 |
+
"decoder_layers": 16,
|
18 |
+
"decoder_start_token_id": 2,
|
19 |
+
"dropout": 0.1,
|
20 |
+
"early_stopping": true,
|
21 |
+
"encoder_attention_heads": 16,
|
22 |
+
"encoder_ffn_dim": 4096,
|
23 |
+
"encoder_layerdrop": 0.0,
|
24 |
+
"encoder_layers": 16,
|
25 |
+
"eos_token_id": 1,
|
26 |
+
"gradient_checkpointing": false,
|
27 |
+
"init_std": 0.02,
|
28 |
+
"is_decoder": true,
|
29 |
+
"is_encoder_decoder": false,
|
30 |
+
"length_penalty": 3.5,
|
31 |
+
"max_length": 512,
|
32 |
+
"max_position_embeddings": 4096,
|
33 |
+
"min_length": 32,
|
34 |
+
"model_type": "bigbird_pegasus",
|
35 |
+
"no_repeat_ngram_size": 3,
|
36 |
+
"num_beams": 5,
|
37 |
+
"num_hidden_layers": 16,
|
38 |
+
"num_random_blocks": 3,
|
39 |
+
"pad_token_id": 0,
|
40 |
+
"quantization_config": {
|
41 |
+
"_load_in_4bit": false,
|
42 |
+
"_load_in_8bit": true,
|
43 |
+
"bnb_4bit_compute_dtype": "float32",
|
44 |
+
"bnb_4bit_quant_storage": "uint8",
|
45 |
+
"bnb_4bit_quant_type": "fp4",
|
46 |
+
"bnb_4bit_use_double_quant": false,
|
47 |
+
"llm_int8_enable_fp32_cpu_offload": false,
|
48 |
+
"llm_int8_has_fp16_weight": false,
|
49 |
+
"llm_int8_skip_modules": null,
|
50 |
+
"llm_int8_threshold": 6.0,
|
51 |
+
"load_in_4bit": false,
|
52 |
+
"load_in_8bit": true,
|
53 |
+
"quant_method": "bitsandbytes"
|
54 |
+
},
|
55 |
+
"scale_embedding": true,
|
56 |
+
"tokenizer_class": "PegasusTokenizer",
|
57 |
+
"torch_dtype": "float16",
|
58 |
+
"transformers_version": "4.40.2",
|
59 |
+
"use_bias": false,
|
60 |
+
"use_cache": false,
|
61 |
+
"vocab_size": 96103
|
62 |
+
}
|