upload bf16

config.json

@@ -0,0 +1,41 @@
+{
+  "_name_or_path": "../fairseq-moe/fairseq-moe-15b",
+  "activation_function": "gelu",
+  "architectures": [
+    "FSGPTMoEForCausalLM"
+  ],
+  "attention_dropout": 0.1,
+  "attention_heads": 12,
+  "batch_prioritized_routing": false,
+  "bos_token_id": 50257,
+  "classifier_dropout": 0.1,
+  "d_model": 768,
+  "embed_dropout": 0.0,
+  "eos_token_id": 50259,
+  "expert_capacity": null,
+  "ffn_dim": 3072,
+  "hidden_size": 768,
+  "intermediate_size": 3072,
+  "max_position_embeddings": 2048,
+  "moe_eval_capacity_token_fraction": -1.0,
+  "moe_token_dropout": 0.2,
+  "no_scale_embedding": false,
+  "normalize_router_prob_before_dropping": false,
+  "num_experts": 512,
+  "num_heads": 12,
+  "num_layers": 12,
+  "output_router_logits": true,
+  "pad_token_id": 50258,
+  "resid_dropout": 0.1,
+  "router_bias": false,
+  "router_dtype": "float32",
+  "router_ignore_padding_tokens": true,
+  "second_expert_policy": "all",
+  "sparse_step": 2,
+  "tokenizer_class": "GPT2Tokenizer",
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.33.2",
+  "unk_token_id": 3,
+  "use_cache": true,
+  "vocab_size": 51200
+}

generation_config.json

@@ -0,0 +1,7 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 50257,
+  "eos_token_id": 50259,
+  "pad_token_id": 50258,
+  "transformers_version": "4.33.2"
+}

pytorch_model-00001-of-00003.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c6041b590f6ce3df211c5f738b42df13b5fa7682f6dc9db0812da0f6addfebd1
+size 9999674523

pytorch_model-00002-of-00003.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cd34a9d107666427d70a9b9850453a1db3af82c05ee255fb0590e3e961203e8c
+size 10000519407

pytorch_model-00003-of-00003.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:82c4946f3cdebb9bebe8b5683f7fba19f17c7a84c1e85c0572a406fe1da1f47d
+size 9215670223