Instructions to use CATIE-AQ/FAT5-small-flan-en with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use CATIE-AQ/FAT5-small-flan-en with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("feature-extraction", model="CATIE-AQ/FAT5-small-flan-en", trust_remote_code=True)# Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("CATIE-AQ/FAT5-small-flan-en", trust_remote_code=True, device_map="auto") - Notebooks
- Google Colab
- Kaggle
Loïck commited on
Update config.json
Browse files- config.json +18 -11
config.json
CHANGED
|
@@ -4,13 +4,18 @@
|
|
| 4 |
"FlashT5ForConditionalGeneration"
|
| 5 |
],
|
| 6 |
"attention_dropout_rate": 0.0,
|
|
|
|
|
|
|
| 7 |
"auto_map": {
|
| 8 |
"AutoConfig": "configuration_flash_t5.FlashT5Config",
|
| 9 |
"AutoModel": "modeling_flash_t5.FlashT5ForConditionalGeneration",
|
| 10 |
-
"
|
|
|
|
|
|
|
|
|
|
| 11 |
},
|
| 12 |
"classifier_dropout": 0.0,
|
| 13 |
-
"d_ff":
|
| 14 |
"d_kv": 64,
|
| 15 |
"d_model": 512,
|
| 16 |
"decoder_start_token_id": 0,
|
|
@@ -18,16 +23,17 @@
|
|
| 18 |
"dropout_rate": 0.0,
|
| 19 |
"eos_token_id": 1,
|
| 20 |
"feed_forward_proj": "relu",
|
|
|
|
| 21 |
"initializer_factor": 1.0,
|
| 22 |
"is_encoder_decoder": false,
|
| 23 |
"is_gated_act": false,
|
| 24 |
-
"label_smoothing": 0.
|
| 25 |
"layer_norm_epsilon": 1e-06,
|
| 26 |
"max_sequence_length": 1024,
|
| 27 |
"model_type": "flash_t5",
|
| 28 |
-
"num_decoder_layers":
|
| 29 |
-
"num_heads":
|
| 30 |
-
"num_layers":
|
| 31 |
"pad_token_id": 0,
|
| 32 |
"position_encoding_type": "t5",
|
| 33 |
"relative_attention_max_distance": 128,
|
|
@@ -38,16 +44,17 @@
|
|
| 38 |
"rotary_scale_base": null,
|
| 39 |
"tie_word_embeddings": false,
|
| 40 |
"torch_dtype": "float32",
|
| 41 |
-
"transformers_version": "4.
|
| 42 |
"use_cache": true,
|
| 43 |
-
"use_flash_attention": "
|
| 44 |
"use_full_bias_size": false,
|
| 45 |
"use_gelu_act": true,
|
| 46 |
"use_glu_mlp": true,
|
|
|
|
| 47 |
"use_randomized_position_encoding": false,
|
| 48 |
-
"use_triton_crossentropy":
|
| 49 |
"use_triton_gated_mlp": false,
|
| 50 |
"use_triton_layernorm": false,
|
| 51 |
-
"vocab_size":
|
| 52 |
"z_loss": 0.0001
|
| 53 |
-
}
|
|
|
|
| 4 |
"FlashT5ForConditionalGeneration"
|
| 5 |
],
|
| 6 |
"attention_dropout_rate": 0.0,
|
| 7 |
+
"attention_scale": 1.0,
|
| 8 |
+
"attention_type": "ref",
|
| 9 |
"auto_map": {
|
| 10 |
"AutoConfig": "configuration_flash_t5.FlashT5Config",
|
| 11 |
"AutoModel": "modeling_flash_t5.FlashT5ForConditionalGeneration",
|
| 12 |
+
"AutoModelForQuestionAnswering": "custom_heads_flash_t5.FlashT5ForQuestionAnswering",
|
| 13 |
+
"AutoModelForSeq2SeqLM": "modeling_flash_t5.FlashT5ForConditionalGeneration",
|
| 14 |
+
"AutoModelForSequenceClassification": "custom_heads_flash_t5.FlashT5ForSequenceClassification",
|
| 15 |
+
"AutoModelForTokenClassification": "custom_heads_flash_t5.FlashT5ForTokenClassification"
|
| 16 |
},
|
| 17 |
"classifier_dropout": 0.0,
|
| 18 |
+
"d_ff": 1024,
|
| 19 |
"d_kv": 64,
|
| 20 |
"d_model": 512,
|
| 21 |
"decoder_start_token_id": 0,
|
|
|
|
| 23 |
"dropout_rate": 0.0,
|
| 24 |
"eos_token_id": 1,
|
| 25 |
"feed_forward_proj": "relu",
|
| 26 |
+
"fire_mlp_width": 32,
|
| 27 |
"initializer_factor": 1.0,
|
| 28 |
"is_encoder_decoder": false,
|
| 29 |
"is_gated_act": false,
|
| 30 |
+
"label_smoothing": 0.1,
|
| 31 |
"layer_norm_epsilon": 1e-06,
|
| 32 |
"max_sequence_length": 1024,
|
| 33 |
"model_type": "flash_t5",
|
| 34 |
+
"num_decoder_layers": 8,
|
| 35 |
+
"num_heads": 6,
|
| 36 |
+
"num_layers": 8,
|
| 37 |
"pad_token_id": 0,
|
| 38 |
"position_encoding_type": "t5",
|
| 39 |
"relative_attention_max_distance": 128,
|
|
|
|
| 44 |
"rotary_scale_base": null,
|
| 45 |
"tie_word_embeddings": false,
|
| 46 |
"torch_dtype": "float32",
|
| 47 |
+
"transformers_version": "4.39.3",
|
| 48 |
"use_cache": true,
|
| 49 |
+
"use_flash_attention": "triton",
|
| 50 |
"use_full_bias_size": false,
|
| 51 |
"use_gelu_act": true,
|
| 52 |
"use_glu_mlp": true,
|
| 53 |
+
"use_masking": false,
|
| 54 |
"use_randomized_position_encoding": false,
|
| 55 |
+
"use_triton_crossentropy": false,
|
| 56 |
"use_triton_gated_mlp": false,
|
| 57 |
"use_triton_layernorm": false,
|
| 58 |
+
"vocab_size": 32128,
|
| 59 |
"z_loss": 0.0001
|
| 60 |
+
}
|