Upload model

config.json

@@ -0,0 +1,32 @@
+{
+  "_name_or_path": "EleutherAI/pythia-70m",
+  "architectures": [
+    "CustomModel"
+  ],
+  "attention_dropout": 0.0,
+  "auto_map": {
+    "AutoModelForCausalLM": "modeling_custom.CustomModel"
+  },
+  "bos_token_id": 0,
+  "classifier_dropout": 0.1,
+  "eos_token_id": 0,
+  "hidden_act": "gelu",
+  "hidden_dropout": 0.0,
+  "hidden_size": 512,
+  "initializer_range": 0.02,
+  "intermediate_size": 2048,
+  "layer_norm_eps": 1e-05,
+  "max_position_embeddings": 2048,
+  "model_type": "gpt_neox",
+  "num_attention_heads": 8,
+  "num_hidden_layers": 6,
+  "rope_scaling": null,
+  "rotary_emb_base": 10000,
+  "rotary_pct": 0.25,
+  "tie_word_embeddings": false,
+  "torch_dtype": "float32",
+  "transformers_version": "4.31.0",
+  "use_cache": true,
+  "use_parallel_residual": true,
+  "vocab_size": 50304
+}

generation_config.json

@@ -0,0 +1,6 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 0,
+  "eos_token_id": 0,
+  "transformers_version": "4.31.0"
+}

modeling_custom.py

@@ -0,0 +1,25 @@
+# https://huggingface.co/docs/transformers/custom_models
+
+from transformers import PreTrainedModel, GPTNeoXForCausalLM, AutoModelForCausalLM
+from torch.nn.functional import log_softmax
+
+# In the example there is also config class but we'll just use the one from GPTNeoX
+# The norm is to import from PreTrainedModel but we'll take a shortcut
+class CustomModel(GPTNeoXForCausalLM):
+    def __init__(self, config):
+        super().__init__(config)
+
+    def forward(self, *args, **kwargs):
+        # See https://huggingface.co/docs/transformers/main_classes/output
+        out = super().forward(*args, **kwargs)
+        out.logits = log_softmax(out.logits, dim=-1)
+        return out
+
+    @classmethod
+    def copy_from_neox(cls, *args, **kwargs):
+        m0 = GPTNeoXForCausalLM.from_pretrained(*args, **kwargs)
+        m1 = cls(m0.config)
+        m1.load_state_dict(m0.state_dict())
+        return m1
+
+CustomModel.register_for_auto_class('AutoModelForCausalLM')

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3e4128420302e58c85fc79e4553c320c270cab3169bc72310bdf2b5b75032a6d
+size 281732711