update with dtype in the config

README.md

@@ -7,4 +7,4 @@ To build the model:
 
 ```bash
 optimum-cli export neuron --model hf-internal-testing/tiny-random-t5 --task text2text-generation --batch_size 1 --sequence_length 18 --num_beams 4 tiny_random_t5_neuronx/
-```
+```

config.json

@@ -1,5 +1,5 @@
 {
-  "_name_or_path": "/tmp/tmp1v7faz4n/encoder/config.json",
+  "_attn_implementation_autoset": true,
   "bos_token_id": 0,
   "classifier_dropout": 0.0,
   "d_ff": 37,
@@ -16,84 +16,14 @@
   "is_gated_act": false,
   "layer_norm_epsilon": 1e-06,
   "model_type": "t5",
-  "neuron": {
-    "auto_cast": "matmul",
-    "auto_cast_type": "bf16",
-    "compiler_type": "neuronx-cc",
-    "compiler_version": "2.11.0.34+c5231f848",
-    "decoder_input_names": [
-      "decoder_input_ids",
-      "decoder_attention_mask",
-      "encoder_hidden_states",
-      "attention_mask",
-      "beam_idx",
-      "beam_scores"
-    ],
-    "decoder_output_names": [
-      "next_tokens",
-      "past.0.self.key",
-      "past.1.self.key",
-      "past.2.self.key",
-      "past.3.self.key",
-      "past.4.self.key",
-      "past.0.self.value",
-      "past.1.self.value",
-      "past.2.self.value",
-      "past.3.self.value",
-      "past.4.self.value",
-      "past.0.cross.key",
-      "past.1.cross.key",
-      "past.2.cross.key",
-      "past.3.cross.key",
-      "past.4.cross.key",
-      "past.0.cross.value",
-      "past.1.cross.value",
-      "past.2.cross.value",
-      "past.3.cross.value",
-      "past.4.cross.value"
-    ],
-    "disable_fallback": false,
-    "disable_fast_relayout": false,
-    "dynamic_batch_size": false,
-    "encoder_input_names": [
-      "input_ids",
-      "attention_mask"
-    ],
-    "encoder_output_names": [
-      "present.0.self.key",
-      "present.1.self.key",
-      "present.2.self.key",
-      "present.3.self.key",
-      "present.4.self.key",
-      "present.0.self.value",
-      "present.1.self.value",
-      "present.2.self.value",
-      "present.3.self.value",
-      "present.4.self.value",
-      "present.0.cross.key",
-      "present.1.cross.key",
-      "present.2.cross.key",
-      "present.3.cross.key",
-      "present.4.cross.key",
-      "present.0.cross.value",
-      "present.1.cross.value",
-      "present.2.cross.value",
-      "present.3.cross.value",
-      "present.4.cross.value"
-    ],
-    "static_batch_size": 1,
-    "static_num_beams": 1,
-    "static_sequence_length": 64
-  },
   "num_decoder_layers": 5,
   "num_heads": 4,
   "num_layers": 5,
   "pad_token_id": 0,
   "relative_attention_max_distance": 128,
   "relative_attention_num_buckets": 8,
-  "task": "text2text-generation",
-  "torchscript": true,
-  "transformers_version": "4.35.0",
+  "torch_dtype": "float32",
+  "transformers_version": "4.51.0",
   "use_cache": true,
   "vocab_size": 1103
 }

decoder/config.json

@@ -1,6 +1,5 @@
 {
-  "_commit_hash": "2f582cd79ed5795b71539951d237945bc1c5ac7e",
-  "_name_or_path": "hf-internal-testing/tiny-random-t5",
+  "_attn_implementation_autoset": true,
   "bos_token_id": 0,
   "classifier_dropout": 0.0,
   "d_ff": 37,
@@ -21,8 +20,10 @@
     "auto_cast": null,
     "auto_cast_type": null,
     "compiler_type": "neuronx-cc",
-    "compiler_version": "2.12.54.0+f631c2365",
+    "compiler_version": "2.19.8089.0+8ab9f450",
     "dynamic_batch_size": false,
+    "float_dtype": "fp32",
+    "inline_weights_to_neff": false,
     "input_names": [
       "decoder_input_ids",
       "decoder_attention_mask",
@@ -31,6 +32,7 @@
       "beam_idx",
       "beam_scores"
     ],
+    "int_dtype": "int64",
     "model_type": "t5-decoder",
     "optlevel": "2",
     "output_attentions": false,
@@ -62,7 +64,9 @@
     ],
     "static_batch_size": 1,
     "static_num_beams": 4,
-    "static_sequence_length": 18
+    "static_sequence_length": 18,
+    "task": "text2text-generation",
+    "tensor_parallel_size": 1
   },
   "num_decoder_layers": 5,
   "num_heads": 4,
@@ -70,9 +74,9 @@
   "pad_token_id": 0,
   "relative_attention_max_distance": 128,
   "relative_attention_num_buckets": 8,
-  "task": "text2text-generation",
+  "torch_dtype": "float32",
   "torchscript": true,
-  "transformers_version": "4.11.0.dev0",
+  "transformers_version": "4.51.0",
   "use_cache": true,
   "vocab_size": 1103
 }

decoder/model.neuron

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:16fc7412146e0cf515099112ee32d3399c5bf8890fc57b77216a8edb3118c488
-size 783754
+oid sha256:29bba6b4ab3e0d3a51f998c8e35887f7083a7d49c88322d2eb1d75930af66000
+size 936999

encoder/config.json

@@ -1,6 +1,5 @@
 {
-  "_commit_hash": "2f582cd79ed5795b71539951d237945bc1c5ac7e",
-  "_name_or_path": "hf-internal-testing/tiny-random-t5",
+  "_attn_implementation_autoset": true,
   "bos_token_id": 0,
   "classifier_dropout": 0.0,
   "d_ff": 37,
@@ -21,12 +20,15 @@
     "auto_cast": null,
     "auto_cast_type": null,
     "compiler_type": "neuronx-cc",
-    "compiler_version": "2.12.54.0+f631c2365",
+    "compiler_version": "2.19.8089.0+8ab9f450",
     "dynamic_batch_size": false,
+    "float_dtype": "fp32",
+    "inline_weights_to_neff": false,
     "input_names": [
       "input_ids",
       "attention_mask"
     ],
+    "int_dtype": "int64",
     "model_type": "t5-encoder",
     "optlevel": "2",
     "output_attentions": false,
@@ -55,7 +57,9 @@
     ],
     "static_batch_size": 1,
     "static_num_beams": 4,
-    "static_sequence_length": 18
+    "static_sequence_length": 18,
+    "task": "text2text-generation",
+    "tensor_parallel_size": 1
   },
   "num_decoder_layers": 5,
   "num_heads": 4,
@@ -63,9 +67,9 @@
   "pad_token_id": 0,
   "relative_attention_max_distance": 128,
   "relative_attention_num_buckets": 8,
-  "task": "text2text-generation",
+  "torch_dtype": "float32",
   "torchscript": true,
-  "transformers_version": "4.11.0.dev0",
+  "transformers_version": "4.51.0",
   "use_cache": true,
   "vocab_size": 1103
 }

encoder/model.neuron

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:54d269f5f2f9f40833871e9e9d310c29351b1a7e05b360d7ddeb1addc151f227
-size 404354
+oid sha256:d06985bf02c880acb6fe126cf49167fb3e630b47c38fcba767fcd7c1677435e0
+size 485543

generation_config.json

@@ -4,5 +4,5 @@
   "decoder_start_token_id": 0,
   "eos_token_id": 1,
   "pad_token_id": 0,
-  "transformers_version": "4.35.0"
+  "transformers_version": "4.51.0"
 }

special_tokens_map.json

@@ -101,7 +101,25 @@
     "<extra_id_98>",
     "<extra_id_99>"
   ],
-  "eos_token": "</s>",
-  "pad_token": "<pad>",
-  "unk_token": "<unk>"
+  "eos_token": {
+    "content": "</s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "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.json

@@ -944,7 +944,8 @@
       {
         "type": "Metaspace",
         "replacement": "▁",
-        "add_prefix_space": true
+        "prepend_scheme": "always",
+        "split": true
       }
     ]
   },
@@ -1005,7 +1006,8 @@
   "decoder": {
     "type": "Metaspace",
     "replacement": "▁",
-    "add_prefix_space": true
+    "prepend_scheme": "always",
+    "split": true
   },
   "model": {
     "type": "Unigram",

tokenizer_config.json

@@ -1,4 +1,5 @@
 {
+  "add_prefix_space": null,
   "added_tokens_decoder": {
     "0": {
       "content": "<pad>",
@@ -927,11 +928,12 @@
     "<extra_id_98>",
     "<extra_id_99>"
   ],
-  "clean_up_tokenization_spaces": true,
+  "clean_up_tokenization_spaces": false,
   "eos_token": "</s>",
   "extra_ids": 100,
+  "extra_special_tokens": {},
   "model_max_length": 1024,
   "pad_token": "<pad>",
-  "tokenizer_class": "T5Tokenizer",
+  "tokenizer_class": "T5TokenizerFast",
   "unk_token": "<unk>"
 }