initial commit

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+*.engine filter=lfs diff=lfs merge=lfs -text

tp2/config.json

@@ -0,0 +1,50 @@
+{
+  "builder_config": {
+    "apply_query_key_layer_scaling": false,
+    "fp8": false,
+    "gather_all_token_logits": false,
+    "hidden_act": "gelu",
+    "hidden_size": 6144,
+    "int8": false,
+    "max_batch_size": 256,
+    "max_input_len": 200,
+    "max_num_tokens": null,
+    "max_output_len": 200,
+    "max_position_embeddings": 8192,
+    "multi_query_mode": true,
+    "name": "gpt",
+    "num_heads": 48,
+    "num_kv_heads": 1,
+    "num_layers": 40,
+    "parallel_build": true,
+    "precision": "float16",
+    "tensor_parallel": 2,
+    "use_parallel_embedding": false,
+    "use_prompt_tuning": false,
+    "use_refit": false,
+    "vocab_size": 49152
+  },
+  "plugin_config": {
+    "attention_qk_half_accumulation": false,
+    "bert_attention_plugin": false,
+    "context_fmha_type": 1,
+    "gemm_plugin": "float16",
+    "gpt_attention_plugin": "float16",
+    "identity_plugin": false,
+    "layernorm_plugin": false,
+    "layernorm_quantization_plugin": false,
+    "lookup_plugin": false,
+    "nccl_plugin": "float16",
+    "paged_kv_cache": true,
+    "quantize_per_token_plugin": false,
+    "quantize_tensor_plugin": false,
+    "remove_input_padding": true,
+    "rmsnorm_plugin": false,
+    "rmsnorm_quantization_plugin": false,
+    "smooth_quant_gemm_plugin": false,
+    "tokens_per_block": 64,
+    "use_custom_all_reduce": false,
+    "weight_only_groupwise_quant_matmul_plugin": false,
+    "weight_only_quant_matmul_plugin": false
+  }
+}

tp2/gpt_float16_tp2_rank0.engine

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9e19b45e0994ccf015f847eeca3c75a13f5fffcc3f5571d1c6fbd2e1f0fdfe7d
+size 16243853196

tp2/gpt_float16_tp2_rank1.engine

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8fd21b88a1b4353fb2de111c4c0a1df800a01f230910bffe24100191d93d19f3
+size 16243853196