trfms integration

config.json

@@ -1,50 +1,37 @@
 {
   "architectures": [
-    "ParakeetCTC"
+    "ParakeetForCTC"
   ],
-  "blank_token_id": 1024,
-  "bos_token_id": 1,
   "ctc_loss_reduction": "mean",
   "ctc_zero_infinity": true,
+  "dtype": "bfloat16",
   "encoder_config": {
     "activation_dropout": 0.1,
-    "activation_function": "silu",
-    "architectures": [
-      "FastConformerModel"
-    ],
+    "attention_bias": true,
     "attention_dropout": 0.1,
-    "attention_probs_dropout_prob": 0.1,
     "conv_kernel_size": 9,
-    "d_model": 1024,
     "dropout": 0.1,
-    "dropout_emb": 0.0,
-    "encoder_attention_heads": 8,
-    "encoder_ffn_dim": 4096,
-    "encoder_layerdrop": 0.1,
-    "encoder_layers": 42,
+    "dropout_positions": 0.0,
     "hidden_act": "silu",
-    "hidden_dropout_prob": 0.1,
     "hidden_size": 1024,
     "initializer_range": 0.02,
     "intermediate_size": 4096,
-    "model_type": "fastconformer",
-    "nemo_decoder_type": "nemo.collections.asr.modules.ConvASRDecoder",
-    "nemo_encoder_type": "nemo.collections.asr.modules.ConformerEncoder",
-    "nemo_model_type": "parakeet_ctc",
+    "layerdrop": 0.1,
+    "max_position_embeddings": 5000,
+    "model_type": "parakeet_encoder",
     "num_attention_heads": 8,
     "num_hidden_layers": 42,
+    "num_key_value_heads": 8,
     "num_mel_bins": 80,
+    "scale_input": true,
     "subsampling_conv_channels": 256,
-    "subsampling_factor": 8,
-    "use_bias": true,
-    "use_cache": false,
-    "vocab_size": 1024,
-    "xscaling": true
+    "subsampling_conv_kernel_size": 3,
+    "subsampling_conv_stride": 2,
+    "subsampling_factor": 8
   },
-  "eos_token_id": 2,
+  "initializer_range": 0.02,
   "model_type": "parakeet_ctc",
-  "pad_token_id": 0,
-  "torch_dtype": "float32",
-  "transformers_version": "4.54.0.dev0",
+  "pad_token_id": 1024,
+  "transformers_version": "4.57.0.dev0",
   "vocab_size": 1025
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ec43cbc00c2763555c0265948176889224d04fe0b4f688e532b6665a53843792
-size 4250700828
+oid sha256:57e0bc26772f3360b7ae0c087f184364179906674d08fc8b71d48a54d4f52145
+size 4250698604

preprocessor_config.json

@@ -1,20 +1,13 @@
 {
-  "f_max": 8000,
-  "f_min": 0,
-  "feature_extractor_type": "FastConformerFeatureExtractor",
+  "feature_extractor_type": "ParakeetFeatureExtractor",
   "feature_size": 80,
   "hop_length": 160,
-  "mag_power": 2.0,
-  "mel_scale": "slaney",
   "n_fft": 512,
-  "n_mels": 80,
-  "normalize": "per_feature",
   "padding_side": "right",
   "padding_value": 0.0,
   "preemphasis": 0.97,
+  "processor_class": "ParakeetProcessor",
   "return_attention_mask": true,
   "sampling_rate": 16000,
-  "win_length": 400,
-  "window_size": 0.025,
-  "window_stride": 0.01
+  "win_length": 400
 }

special_tokens_map.json

@@ -0,0 +1,16 @@
+{
+  "pad_token": {
+    "content": "<pad>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "unk_token": {
+    "content": "<unk>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

tokenizer_config.json

@@ -7,13 +7,21 @@
       "rstrip": false,
       "single_word": false,
       "special": true
+    },
+    "1024": {
+      "content": "<pad>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
     }
   },
-  "blank_token_id": 1024,
   "clean_up_tokenization_spaces": false,
-  "do_lower_case": false,
   "extra_special_tokens": {},
   "model_max_length": 1000000000000000019884624838656,
-  "tokenizer_class": "ParakeetCTCTokenizer",
+  "pad_token": "<pad>",
+  "processor_class": "ParakeetProcessor",
+  "tokenizer_class": "ParakeetTokenizerFast",
   "unk_token": "<unk>"
 }