Automatic push from sapienzanlp

config.json

@@ -0,0 +1,30 @@
+{
+  "_name_or_path": "riccorl/e5-base-v2-blink-1M-32words-windows",
+  "architectures": [
+    "GoldenRetrieverModel"
+  ],
+  "attention_probs_dropout_prob": 0.1,
+  "auto_map": {
+    "AutoModel": "hf.GoldenRetrieverModel"
+  },
+  "classifier_dropout": null,
+  "gradient_checkpointing": false,
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 768,
+  "initializer_range": 0.02,
+  "intermediate_size": 3072,
+  "layer_norm_eps": 1e-12,
+  "max_position_embeddings": 512,
+  "model_type": "bert",
+  "num_attention_heads": 12,
+  "num_hidden_layers": 12,
+  "pad_token_id": 0,
+  "position_embedding_type": "absolute",
+  "projection_dim": null,
+  "torch_dtype": "float32",
+  "transformers_version": "4.33.3",
+  "type_vocab_size": 2,
+  "use_cache": true,
+  "vocab_size": 30522
+}

hf.py

@@ -0,0 +1,99 @@
+from typing import Tuple, Union
+
+import torch
+from transformers import PretrainedConfig
+from transformers.modeling_outputs import BaseModelOutputWithPoolingAndCrossAttentions
+from transformers.models.bert.modeling_bert import BertModel
+
+
+class GoldenRetrieverConfig(PretrainedConfig):
+    model_type = "bert"
+
+    def __init__(
+        self,
+        vocab_size=30522,
+        hidden_size=768,
+        num_hidden_layers=12,
+        num_attention_heads=12,
+        intermediate_size=3072,
+        hidden_act="gelu",
+        hidden_dropout_prob=0.1,
+        attention_probs_dropout_prob=0.1,
+        max_position_embeddings=512,
+        type_vocab_size=2,
+        initializer_range=0.02,
+        layer_norm_eps=1e-12,
+        pad_token_id=0,
+        position_embedding_type="absolute",
+        use_cache=True,
+        classifier_dropout=None,
+        projection_dim=None,
+        **kwargs,
+    ):
+        super().__init__(pad_token_id=pad_token_id, **kwargs)
+
+        self.vocab_size = vocab_size
+        self.hidden_size = hidden_size
+        self.num_hidden_layers = num_hidden_layers
+        self.num_attention_heads = num_attention_heads
+        self.hidden_act = hidden_act
+        self.intermediate_size = intermediate_size
+        self.hidden_dropout_prob = hidden_dropout_prob
+        self.attention_probs_dropout_prob = attention_probs_dropout_prob
+        self.max_position_embeddings = max_position_embeddings
+        self.type_vocab_size = type_vocab_size
+        self.initializer_range = initializer_range
+        self.layer_norm_eps = layer_norm_eps
+        self.position_embedding_type = position_embedding_type
+        self.use_cache = use_cache
+        self.classifier_dropout = classifier_dropout
+        self.projection_dim = projection_dim
+
+
+class GoldenRetrieverModel(BertModel):
+    config_class = GoldenRetrieverConfig
+
+    def __init__(self, config, *args, **kwargs):
+        super().__init__(config)
+        self.layer_norm_layer = torch.nn.LayerNorm(
+            config.hidden_size, eps=config.layer_norm_eps
+        )
+        self.projection: torch.nn.Module | None = None
+        if config.projection_dim is not None:
+            self.projection = torch.nn.Sequential(
+                torch.nn.Linear(config.hidden_size, config.projection_dim),
+                torch.nn.LayerNorm(config.projection_dim),
+            )
+
+    def forward(
+        self, **kwargs
+    ) -> Union[Tuple[torch.Tensor], BaseModelOutputWithPoolingAndCrossAttentions]:
+        attention_mask = kwargs.get("attention_mask", None)
+        model_outputs = super().forward(**kwargs)
+        if attention_mask is None:
+            pooler_output = model_outputs.pooler_output
+        else:
+            token_embeddings = model_outputs.last_hidden_state
+            input_mask_expanded = (
+                attention_mask.unsqueeze(-1).expand(token_embeddings.size()).float()
+            )
+            pooler_output = torch.sum(
+                token_embeddings * input_mask_expanded, 1
+            ) / torch.clamp(input_mask_expanded.sum(1), min=1e-9)
+
+        pooler_output = self.layer_norm_layer(pooler_output)
+
+        if self.projection is not None:
+            pooler_output = self.projection(pooler_output)
+
+        if not kwargs.get("return_dict", True):
+            return (model_outputs[0], pooler_output) + model_outputs[2:]
+
+        return BaseModelOutputWithPoolingAndCrossAttentions(
+            last_hidden_state=model_outputs.last_hidden_state,
+            pooler_output=pooler_output,
+            past_key_values=model_outputs.past_key_values,
+            hidden_states=model_outputs.hidden_states,
+            attentions=model_outputs.attentions,
+            cross_attentions=model_outputs.cross_attentions,
+        )

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6a8336a720e8e5259782fe255e1b797841fef88ad523c85883d84fcd8decf98c
+size 438002926

special_tokens_map.json

@@ -0,0 +1,7 @@
+{
+  "cls_token": "[CLS]",
+  "mask_token": "[MASK]",
+  "pad_token": "[PAD]",
+  "sep_token": "[SEP]",
+  "unk_token": "[UNK]"
+}

tokenizer_config.json

@@ -0,0 +1,17 @@
+{
+  "clean_up_tokenization_spaces": true,
+  "cls_token": "[CLS]",
+  "do_lower_case": true,
+  "mask_token": "[MASK]",
+  "max_length": 64,
+  "model_max_length": 512,
+  "pad_token": "[PAD]",
+  "sep_token": "[SEP]",
+  "stride": 0,
+  "strip_accents": null,
+  "tokenize_chinese_chars": true,
+  "tokenizer_class": "BertTokenizer",
+  "truncation_side": "right",
+  "truncation_strategy": "longest_first",
+  "unk_token": "[UNK]"
+}