update max tokens based on 30sec (max trained sample) * 6.5 toks/sec

config.json

@@ -5,6 +5,7 @@
   "attention_bias": false,
   "attention_dropout": 0.0,
   "bos_token_id": 1,
+  "pad_token_id": 2,
   "decoder_hidden_act": "silu",
   "decoder_num_attention_heads": 8,
   "decoder_num_hidden_layers": 6,
@@ -19,7 +20,7 @@
   "initializer_range": 0.02,
   "intermediate_size": 1152,
   "is_encoder_decoder": true,
-  "max_position_embeddings": 512,
+  "max_position_embeddings": 194,
   "model_type": "moonshine",
   "partial_rotary_factor": 0.9,
   "rope_scaling": null,

generation_config.json

@@ -2,6 +2,8 @@
   "_from_model_config": true,
   "bos_token_id": 1,
   "decoder_start_token_id": 1,
+  "pad_token_id": 2,
   "eos_token_id": 2,
-  "transformers_version": "4.48.0.dev0"
+  "transformers_version": "4.48.0.dev0",
+  "max_length": 194
 }