Add pipeline_state + per-model padding_embedding for SA3 TRT pipeline

Required runtime support for the TRT engines:
- pipeline_state/tokenizer/ T5Gemma BPE tokenizer
- pipeline_state/num_conditioner.* seconds_total conditioner (768D)
- pipeline_state/dist_shift_config DistributionShift schedule warp (base_shift=0.5, max_shift=1.15)
- pipeline_state/{cond_routing,model_info}.json pipeline metadata
- sa3-{m,sm-music,sm-sfx}/padding_embedding.pt per-DiT learned padding embedding for T5 cross-attn

.gitattributes

@@ -35,3 +35,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
 Stable_Audio_3.0_Thumbnail_1x1.png filter=lfs diff=lfs merge=lfs -text
 *.trt filter=lfs diff=lfs merge=lfs -text
+tokenizer.json filter=lfs diff=lfs merge=lfs -text

tensorRT/pipeline_state/cond_routing.json

@@ -0,0 +1 @@
+{"cross_attn_cond_ids": ["prompt", "seconds_total"], "global_cond_ids": ["seconds_total"], "local_add_cond_ids": ["inpaint_mask", "inpaint_masked_input"]}

tensorRT/pipeline_state/dist_shift_config.json

@@ -0,0 +1 @@
+{"type": "DistributionShift", "base_shift": 0.5, "max_shift": 1.15, "max_length": 4096, "min_length": 256, "use_sine": false}

tensorRT/pipeline_state/model_info.json

@@ -0,0 +1 @@
+{"io_channels": 256, "sample_rate": 44100, "model_dtype": "torch.float32"}

tensorRT/pipeline_state/num_conditioner.npz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a26ca5ef3af62677d82a3199195aebd6b6a74c15ab4b303851557b6e493f60a4
+size 790094

tensorRT/pipeline_state/num_conditioner.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0f2aecdb2436073f3aedb37b4c129e42b5ad7dce5aeca16fabdbe100bbaf607e
+size 791304

tensorRT/pipeline_state/num_conditioner_config.json

@@ -0,0 +1 @@
+{"min_val": 0, "max_val": 384, "type": "NumberConditioner", "output_dim": 768}

tensorRT/pipeline_state/tokenizer/tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f7f12d987b3614408d037113c1fe9348bc1a1c4994807ad07c71b280e33abdbd
+size 34362304

tensorRT/pipeline_state/tokenizer/tokenizer_config.json

@@ -0,0 +1,16 @@
+{
+  "backend": "tokenizers",
+  "bos_token": "<bos>",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<eos>",
+  "is_local": false,
+  "mask_token": "<mask>",
+  "model_max_length": 1000000000000000019884624838656,
+  "pad_token": "<pad>",
+  "padding_side": "right",
+  "sp_model_kwargs": {},
+  "spaces_between_special_tokens": false,
+  "tokenizer_class": "GemmaTokenizer",
+  "unk_token": "<unk>",
+  "use_default_system_prompt": false
+}

tensorRT/sa3-m/padding_embedding.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:63aa7c591786ff515371f2ae0bcec3dea4d75fe22f7a5d88c786a222c8d9c62c
+size 4302

tensorRT/sa3-sm-music/padding_embedding.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9e9b6a46c5d3fb07761378a94d6ea9bfc6312d1b18f4e106875d5c14e46b6bbb
+size 4302

tensorRT/sa3-sm-sfx/padding_embedding.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3f3f55bbd085051b506e94f6c7e651e78eaddb3d444f493888e747f348bd6549
+size 4302