Hezar: Upload model_config.yaml
Browse files- model_config.yaml +52 -0
model_config.yaml
ADDED
|
@@ -0,0 +1,52 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
name: vit_gpt2_image2text
|
| 2 |
+
config_type: model
|
| 3 |
+
encoder:
|
| 4 |
+
name: vit_gpt2_encoder
|
| 5 |
+
config_type: model
|
| 6 |
+
hidden_size: 768
|
| 7 |
+
num_hidden_layers: 12
|
| 8 |
+
num_attention_heads: 12
|
| 9 |
+
intermediate_size: 3072
|
| 10 |
+
hidden_act: gelu
|
| 11 |
+
hidden_dropout_prob: 0.0
|
| 12 |
+
attention_probs_dropout_prob: 0.0
|
| 13 |
+
initializer_range: 0.02
|
| 14 |
+
layer_norm_eps: 1.0e-12
|
| 15 |
+
image_size: 224
|
| 16 |
+
patch_size: 16
|
| 17 |
+
num_channels: 3
|
| 18 |
+
qkv_bias: true
|
| 19 |
+
encoder_stride: 16
|
| 20 |
+
decoder:
|
| 21 |
+
name: vit_gpt2_decoder
|
| 22 |
+
config_type: model
|
| 23 |
+
add_cross_attention: true
|
| 24 |
+
vocab_size: 42001
|
| 25 |
+
attn_pdrop: 0.1
|
| 26 |
+
bos_token_id: 5
|
| 27 |
+
embd_pdrop: 0.1
|
| 28 |
+
eos_token_id: 5
|
| 29 |
+
gradient_checkpointing: false
|
| 30 |
+
initializer_range: 0.02
|
| 31 |
+
layer_norm_epsilon: 1.0e-05
|
| 32 |
+
model_type: gpt2
|
| 33 |
+
n_ctx: 1024
|
| 34 |
+
n_embd: 768
|
| 35 |
+
n_head: 12
|
| 36 |
+
n_inner: null
|
| 37 |
+
n_layer: 12
|
| 38 |
+
n_positions: 1024
|
| 39 |
+
resid_pdrop: 0.1
|
| 40 |
+
summary_activation: false
|
| 41 |
+
summary_first_dropout: 0.1
|
| 42 |
+
use_cache: true
|
| 43 |
+
generation:
|
| 44 |
+
bos_token_id: 0
|
| 45 |
+
decoder_start_token_id: 0
|
| 46 |
+
early_stopping: true
|
| 47 |
+
eos_token_id: 2
|
| 48 |
+
length_penalty: 2.0
|
| 49 |
+
max_new_tokens: 24
|
| 50 |
+
no_repeat_ngram_size: 3
|
| 51 |
+
num_beams: 4
|
| 52 |
+
pad_token_id: 1
|