Karim-Gamal
commited on
Commit
•
7f8e89b
1
Parent(s):
0638b5f
Upload SwitchTransformersForConditionalGeneration
Browse files- config.json +74 -0
- pytorch_model.bin +3 -0
config.json
ADDED
@@ -0,0 +1,74 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"_name_or_path": "google/switch-base-8",
|
3 |
+
"add_router_probs": false,
|
4 |
+
"architectures": [
|
5 |
+
"SwitchTransformersForConditionalGeneration"
|
6 |
+
],
|
7 |
+
"batch_prioritized_routing": false,
|
8 |
+
"d_ff": 3072,
|
9 |
+
"d_kv": 64,
|
10 |
+
"d_model": 768,
|
11 |
+
"decoder_sparse_step": 2,
|
12 |
+
"decoder_start_token_id": 0,
|
13 |
+
"dense_act_fn": "relu",
|
14 |
+
"dropout_rate": 0.1,
|
15 |
+
"encoder_sparse_step": 2,
|
16 |
+
"eos_token_id": 1,
|
17 |
+
"expert_capacity": 64,
|
18 |
+
"feed_forward_proj": "relu",
|
19 |
+
"initializer_factor": 1.0,
|
20 |
+
"is_encoder_decoder": true,
|
21 |
+
"is_gated_act": false,
|
22 |
+
"layer_norm_epsilon": 1e-06,
|
23 |
+
"model_type": "switch_transformers",
|
24 |
+
"num_decoder_layers": 12,
|
25 |
+
"num_experts": 8,
|
26 |
+
"num_heads": 12,
|
27 |
+
"num_layers": 12,
|
28 |
+
"num_selected_experts": 1,
|
29 |
+
"num_sparse_decoder_layers": 6,
|
30 |
+
"num_sparse_encoder_layers": 6,
|
31 |
+
"pad_token_id": 0,
|
32 |
+
"relative_attention_max_distance": 128,
|
33 |
+
"relative_attention_num_buckets": 32,
|
34 |
+
"router_aux_loss_coef": 0.001,
|
35 |
+
"router_bias": false,
|
36 |
+
"router_dtype": "float32",
|
37 |
+
"router_ignore_padding_tokens": false,
|
38 |
+
"router_jitter_noise": 0,
|
39 |
+
"router_type": "tokens_masked",
|
40 |
+
"router_z_loss_coef": 0.001,
|
41 |
+
"task_specific_params": {
|
42 |
+
"summarization": {
|
43 |
+
"early_stopping": true,
|
44 |
+
"length_penalty": 2.0,
|
45 |
+
"max_length": 200,
|
46 |
+
"min_length": 30,
|
47 |
+
"no_repeat_ngram_size": 3,
|
48 |
+
"num_beams": 4,
|
49 |
+
"prefix": "summarize: "
|
50 |
+
},
|
51 |
+
"translation_en_to_de": {
|
52 |
+
"early_stopping": true,
|
53 |
+
"max_length": 300,
|
54 |
+
"num_beams": 4,
|
55 |
+
"prefix": "translate English to German: "
|
56 |
+
},
|
57 |
+
"translation_en_to_fr": {
|
58 |
+
"early_stopping": true,
|
59 |
+
"max_length": 300,
|
60 |
+
"num_beams": 4,
|
61 |
+
"prefix": "translate English to French: "
|
62 |
+
},
|
63 |
+
"translation_en_to_ro": {
|
64 |
+
"early_stopping": true,
|
65 |
+
"max_length": 300,
|
66 |
+
"num_beams": 4,
|
67 |
+
"prefix": "translate English to Romanian: "
|
68 |
+
}
|
69 |
+
},
|
70 |
+
"torch_dtype": "bfloat16",
|
71 |
+
"transformers_version": "4.25.1",
|
72 |
+
"use_cache": true,
|
73 |
+
"vocab_size": 32128
|
74 |
+
}
|
pytorch_model.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:7a9c4f0aa8b885422f9f069ae85888dd7ca5efb17d29b2729c1d53fb4dad6571
|
3 |
+
size 1238827479
|