fix: gptattentionplugin onnxparser compatability

[jl749]

Issue

#2685

Changes made

Added missing declaration of the fields for tensorrt.OnnxParser conversion api.
I also removed the in_flight_batching plugin field as I couldn't see it was directly used anywhere.

# relevant part of the attached ONNX compile log
[01/23/2025-03:27:56] [TRT] [W] onnxOpImporters.cpp:6507: Attribute in_flight_batching not found in plugin node! Ensure that the plugin creator has a default value defined or the engine may fail to build.

TensorRT-LLM/tensorrt_llm/functional.py

Lines 5350 to 5370 in d93a2dd

	pfc = trt.PluginFieldCollection([
	layer_idx, nheads, vision_start, vision_length, num_kv_heads,
	layer_idx_in_cache_pool, head_size, unidirectional, q_scaling,
	attn_logit_softcapping_scale, position_embedding_type,
	rotary_embedding_dim, rotary_embedding_base,
	rotary_embedding_scale_type, rotary_embedding_scale,
	rotary_embedding_short_m_scale, rotary_embedding_long_m_scale,
	rotary_embedding_max_positions, rotary_embedding_original_max_positions,
	tp_size, tp_rank, unfuse_qkv_gemm, context_fmha_type,
	kv_cache_quant_mode_field, remove_input_padding, mask_type_filed,
	block_sparse_block_size, block_sparse_homo_head_pattern,
	block_sparse_num_local_blocks, block_sparse_vertical_stride,
	paged_kv_cache, tokens_per_block, pf_type, max_context_length,
	qkv_bias_enabled, do_cross_attention_field, max_distance,
	pos_shift_enabled, dense_context_fmha, use_paged_context_fmha_field,
	use_fp8_context_fmha_field, has_full_attention_mask_field, use_cache_pf,
	is_spec_decoding_enabled, spec_decoding_is_generation_length_variable,
	spec_decoding_max_generation_length, is_mla_enabled, q_lora_rank,
	kv_lora_rank, qk_nope_head_dim, qk_rope_head_dim, v_head_dim,
	skip_attn_pf, cp_size, cp_rank, cp_group, use_logn_scaling
	])

To test

1. Download this ONNX file
2. Use tensorrt.OnnxParser api with the strongly_typed flag
3. Check that engine file is generated without an error

ONNX compile log

[01/23/2025-03:27:48] [TRT] [I] [MemUsageChange] Init CUDA: CPU +0, GPU +0, now: CPU 217, GPU 423 (MiB)
[01/23/2025-03:27:56] [TRT] [I] [MemUsageChange] Init builder kernel library: CPU +2038, GPU +374, now: CPU 2395, GPU 797 (MiB)
config.default_device_type = DeviceType.GPU
config.max_aux_streams = -1
config.plugins_to_serialize = []
config.get_memory_pool_limit(trt.MemoryPoolType.WORKSPACE) = 42298834944 Byte (39.4 GiB)
[01/23/2025-03:27:56] [TRT] [I] ----------------------------------------------------------------
[01/23/2025-03:27:56] [TRT] [I] Input filename:   ./model.onnx
[01/23/2025-03:27:56] [TRT] [I] ONNX IR version:  0.0.8
[01/23/2025-03:27:56] [TRT] [I] Opset version:    17
[01/23/2025-03:27:56] [TRT] [I] Producer name:    pytorch
[01/23/2025-03:27:56] [TRT] [I] Producer version: 2.4.1
[01/23/2025-03:27:56] [TRT] [I] Domain:           
[01/23/2025-03:27:56] [TRT] [I] Model version:    0
[01/23/2025-03:27:56] [TRT] [I] Doc string:       
[01/23/2025-03:27:56] [TRT] [I] ----------------------------------------------------------------
[01/23/2025-03:27:56] [TRT] [W] ModelImporter.cpp:459: Make sure input input_ids has Int64 binding.
[01/23/2025-03:27:56] [TRT] [W] ModelImporter.cpp:459: Make sure input host_kv_cache_pool_pointers has Int64 binding.
[01/23/2025-03:27:56] [TRT] [W] ModelImporter.cpp:459: Make sure input host_runtime_perf_knobs has Int64 binding.
[01/23/2025-03:27:56] [TRT] [W] ModelImporter.cpp:459: Make sure input host_context_progress has Int64 binding.
[01/23/2025-03:27:56] [TRT] [I] No checker registered for op: GPTAttention. Attempting to check as plugin.
[01/23/2025-03:27:56] [TRT] [I] No importer registered for op: GPTAttention. Attempting to import as plugin.
[01/23/2025-03:27:56] [TRT] [I] Searching for plugin: GPTAttention, plugin_version: 1, plugin_namespace: tensorrt_llm
[01/23/2025-03:27:56] [TRT] [W] onnxOpImporters.cpp:6507: Attribute in_flight_batching not found in plugin node! Ensure that the plugin creator has a default value defined or the engine may fail to build.
[01/23/2025-03:27:56] [TRT] [I] Successfully created plugin: GPTAttention
[01/23/2025-03:27:56] [TRT] [W] Unused Input: position_ids
[01/23/2025-03:27:56] [TRT] [W] [RemoveDeadLayers] Input Tensor position_ids is unused or used only at compile-time, but is not being removed.
[01/23/2025-03:27:56] [TRT] [I] Global timing cache in use. Profiling results in this builder pass will be stored.
[01/23/2025-03:27:56] [TRT] [I] Compiler backend is used during engine build.
[01/23/2025-03:28:12] [TRT] [I] Detected 16 inputs and 1 output network tensors.
[01/23/2025-03:28:14] [TRT] [I] Total Host Persistent Memory: 1792 bytes
[01/23/2025-03:28:14] [TRT] [I] Total Device Persistent Memory: 0 bytes
[01/23/2025-03:28:14] [TRT] [I] Max Scratch Memory: 33555072 bytes
[01/23/2025-03:28:14] [TRT] [I] [BlockAssignment] Started assigning block shifts. This will take 6 steps to complete.
[01/23/2025-03:28:14] [TRT] [I] [BlockAssignment] Algorithm ShiftNTopDown took 0.024757ms to assign 4 blocks to 6 nodes requiring 40895488 bytes.
[01/23/2025-03:28:14] [TRT] [I] Total Activation Memory: 40895488 bytes
[01/23/2025-03:28:17] [TRT] [I] Detected 16 inputs and 1 output network tensors.
[01/23/2025-03:28:17] [TRT] [I] Total Host Persistent Memory: 1792 bytes
[01/23/2025-03:28:17] [TRT] [I] Total Device Persistent Memory: 0 bytes
[01/23/2025-03:28:17] [TRT] [I] Max Scratch Memory: 33555072 bytes
[01/23/2025-03:28:17] [TRT] [I] [BlockAssignment] Started assigning block shifts. This will take 6 steps to complete.
[01/23/2025-03:28:17] [TRT] [I] [BlockAssignment] Algorithm ShiftNTopDown took 0.024627ms to assign 4 blocks to 6 nodes requiring 33562624 bytes.
[01/23/2025-03:28:17] [TRT] [I] Total Activation Memory: 33562624 bytes
[01/23/2025-03:28:18] [TRT] [I] Total Weights Memory: 6339328 bytes
[01/23/2025-03:28:18] [TRT] [I] Compiler backend is used during engine execution.
[01/23/2025-03:28:18] [TRT] [I] Engine generation completed in 21.9013 seconds.
[01/23/2025-03:28:18] [TRT] [I] [MemUsageStats] Peak memory usage of TRT CPU/GPU memory allocators: CPU 0 MiB, GPU 6 MiB
[01/23/2025-03:28:19] [TRT] [I] Loaded engine size: 7 MiB
TRT Engine uses: 40895488 bytes of Memory