Upload compiled-model/config.json with huggingface_hub
Browse files- compiled-model/config.json +134 -0
compiled-model/config.json
ADDED
|
@@ -0,0 +1,134 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"version": "0.9.0.dev2024031900",
|
| 3 |
+
"pretrained_config": {
|
| 4 |
+
"architecture": "MistralForCausalLM",
|
| 5 |
+
"dtype": "float16",
|
| 6 |
+
"logits_dtype": "float32",
|
| 7 |
+
"vocab_size": 32000,
|
| 8 |
+
"max_position_embeddings": 32768,
|
| 9 |
+
"hidden_size": 4096,
|
| 10 |
+
"num_hidden_layers": 32,
|
| 11 |
+
"num_attention_heads": 32,
|
| 12 |
+
"num_key_value_heads": 8,
|
| 13 |
+
"head_size": 128,
|
| 14 |
+
"hidden_act": "silu",
|
| 15 |
+
"intermediate_size": 14336,
|
| 16 |
+
"norm_epsilon": 1e-05,
|
| 17 |
+
"position_embedding_type": "rope_gpt_neox",
|
| 18 |
+
"use_parallel_embedding": false,
|
| 19 |
+
"embedding_sharding_dim": 0,
|
| 20 |
+
"share_embedding_table": false,
|
| 21 |
+
"mapping": {
|
| 22 |
+
"world_size": 1,
|
| 23 |
+
"tp_size": 1,
|
| 24 |
+
"pp_size": 1
|
| 25 |
+
},
|
| 26 |
+
"quantization": {
|
| 27 |
+
"quant_algo": null,
|
| 28 |
+
"kv_cache_quant_algo": null,
|
| 29 |
+
"group_size": 128,
|
| 30 |
+
"has_zero_point": false,
|
| 31 |
+
"pre_quant_scale": false,
|
| 32 |
+
"exclude_modules": [
|
| 33 |
+
"lm_head"
|
| 34 |
+
]
|
| 35 |
+
},
|
| 36 |
+
"kv_dtype": "float16",
|
| 37 |
+
"rotary_scaling": null,
|
| 38 |
+
"moe_normalization_mode": null,
|
| 39 |
+
"rotary_base": 1000000.0,
|
| 40 |
+
"moe_num_experts": 0,
|
| 41 |
+
"moe_top_k": 0,
|
| 42 |
+
"moe_tp_mode": 2,
|
| 43 |
+
"attn_bias": false,
|
| 44 |
+
"disable_weight_only_quant_plugin": false,
|
| 45 |
+
"mlp_bias": false,
|
| 46 |
+
"use_prompt_tuning": false
|
| 47 |
+
},
|
| 48 |
+
"build_config": {
|
| 49 |
+
"max_input_len": 32256,
|
| 50 |
+
"max_output_len": 1024,
|
| 51 |
+
"max_batch_size": 1,
|
| 52 |
+
"max_beam_width": 1,
|
| 53 |
+
"max_num_tokens": 32256,
|
| 54 |
+
"max_prompt_embedding_table_size": 0,
|
| 55 |
+
"gather_context_logits": false,
|
| 56 |
+
"gather_generation_logits": false,
|
| 57 |
+
"strongly_typed": false,
|
| 58 |
+
"builder_opt": null,
|
| 59 |
+
"profiling_verbosity": "layer_names_only",
|
| 60 |
+
"enable_debug_output": false,
|
| 61 |
+
"max_draft_len": 0,
|
| 62 |
+
"use_refit": false,
|
| 63 |
+
"input_timing_cache": null,
|
| 64 |
+
"output_timing_cache": "model.cache",
|
| 65 |
+
"lora_config": {
|
| 66 |
+
"lora_dir": [],
|
| 67 |
+
"lora_ckpt_source": "hf",
|
| 68 |
+
"max_lora_rank": 64,
|
| 69 |
+
"lora_target_modules": [],
|
| 70 |
+
"trtllm_modules_to_hf_modules": {}
|
| 71 |
+
},
|
| 72 |
+
"auto_parallel_config": {
|
| 73 |
+
"world_size": 1,
|
| 74 |
+
"gpus_per_node": 8,
|
| 75 |
+
"cluster_key": "A10G",
|
| 76 |
+
"cluster_info": null,
|
| 77 |
+
"sharding_cost_model": "alpha_beta",
|
| 78 |
+
"comm_cost_model": "alpha_beta",
|
| 79 |
+
"enable_pipeline_parallelism": false,
|
| 80 |
+
"enable_shard_unbalanced_shape": false,
|
| 81 |
+
"enable_shard_dynamic_shape": false,
|
| 82 |
+
"enable_reduce_scatter": true,
|
| 83 |
+
"builder_flags": null,
|
| 84 |
+
"debug_mode": false,
|
| 85 |
+
"infer_shape": true,
|
| 86 |
+
"validation_mode": false,
|
| 87 |
+
"same_buffer_io": {
|
| 88 |
+
"past_key_value_(\\d+)": "present_key_value_\\1"
|
| 89 |
+
},
|
| 90 |
+
"same_spec_io": {},
|
| 91 |
+
"sharded_io_allowlist": [
|
| 92 |
+
"past_key_value_\\d+",
|
| 93 |
+
"present_key_value_\\d*"
|
| 94 |
+
],
|
| 95 |
+
"fast_reduce": true,
|
| 96 |
+
"fill_weights": false,
|
| 97 |
+
"parallel_config_cache": null,
|
| 98 |
+
"profile_cache": null,
|
| 99 |
+
"dump_path": null,
|
| 100 |
+
"debug_outputs": []
|
| 101 |
+
},
|
| 102 |
+
"plugin_config": {
|
| 103 |
+
"bert_attention_plugin": "float16",
|
| 104 |
+
"gpt_attention_plugin": "float16",
|
| 105 |
+
"gemm_plugin": "float16",
|
| 106 |
+
"smooth_quant_gemm_plugin": null,
|
| 107 |
+
"identity_plugin": null,
|
| 108 |
+
"layernorm_quantization_plugin": null,
|
| 109 |
+
"rmsnorm_quantization_plugin": null,
|
| 110 |
+
"nccl_plugin": null,
|
| 111 |
+
"lookup_plugin": null,
|
| 112 |
+
"lora_plugin": null,
|
| 113 |
+
"weight_only_groupwise_quant_matmul_plugin": null,
|
| 114 |
+
"weight_only_quant_matmul_plugin": null,
|
| 115 |
+
"quantize_per_token_plugin": false,
|
| 116 |
+
"quantize_tensor_plugin": false,
|
| 117 |
+
"moe_plugin": "float16",
|
| 118 |
+
"context_fmha": true,
|
| 119 |
+
"context_fmha_fp32_acc": false,
|
| 120 |
+
"paged_kv_cache": true,
|
| 121 |
+
"remove_input_padding": true,
|
| 122 |
+
"use_custom_all_reduce": true,
|
| 123 |
+
"multi_block_mode": false,
|
| 124 |
+
"enable_xqa": true,
|
| 125 |
+
"attention_qk_half_accumulation": false,
|
| 126 |
+
"tokens_per_block": 128,
|
| 127 |
+
"use_paged_context_fmha": false,
|
| 128 |
+
"use_context_fmha_for_generation": false,
|
| 129 |
+
"dense_context_fmha": false,
|
| 130 |
+
"pos_shift": false,
|
| 131 |
+
"multiple_profiles": false
|
| 132 |
+
}
|
| 133 |
+
}
|
| 134 |
+
}
|