uploading v75 bundle for qwen3 0.8B

.gitattributes

@@ -33,5 +33,7 @@ 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
+v75/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+v79/tokenizer.json filter=lfs diff=lfs merge=lfs -text
 v81/tokenizer.json filter=lfs diff=lfs merge=lfs -text
-v81-w8/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+v81-w8/tokenizer.json filter=lfs diff=lfs merge=lfs -text

v75/qwen3.5-0.8b-1024.json

@@ -0,0 +1,65 @@
+{
+  "schema_version": 1,
+  "_comment": "Qwen3.5-0.8B GatedDeltaNet hybrid for Galaxy S24 (Hexagon v75). W8 decode + fp16 lm-head + fp16 embed; decode-over-prompt via qwen3_5_generate. Same flat bundle layout as v81-w8 on HF.",
+  "model": {
+    "name": "qwen3.5-0.8b-1024",
+    "family": "llm",
+    "dsp_arch": "v75",
+    "tokenizer_pre": "qwen2"
+  },
+  "params": {
+    "hidden": 1024,
+    "vocab": 248320,
+    "n_layers": 24,
+    "max_ctx": 1024,
+    "kv_dim": 512,
+    "head_dim": 256,
+    "rope_theta": 10000000.0,
+    "eos_token_id": 248044,
+    "gated_delta": {
+      "q_heads": 8,
+      "kv_heads": 2,
+      "lin_k_heads": 16,
+      "lin_v_heads": 16,
+      "lin_k_dim": 128,
+      "lin_v_dim": 128,
+      "conv_kernel": 4,
+      "full_every": 4,
+      "rot_dim": 64,
+      "attn_gate": true,
+      "full_attn_idx": [
+        3,
+        7,
+        11,
+        15,
+        19,
+        23
+      ]
+    }
+  },
+  "artifacts": {
+    "contexts": {
+      "decode": {
+        "bin": "qwen3508b_decode_w8.bin"
+      },
+      "lmhead": {
+        "bin": "qwen3508b_lmhead_f16.bin"
+      }
+    },
+    "embed": "qwen3508b_embed_f16.bin",
+    "tokenizer": "tokenizer.json"
+  },
+  "plan": {
+    "steps": [
+      {
+        "host": "qwen3_5_generate",
+        "params": {
+          "decode": "qwen3508b_decode_w8",
+          "lmhead": "qwen3508b_lmhead_f16",
+          "max_new": 64,
+          "neg": -50000
+        }
+      }
+    ]
+  }
+}

v75/qwen3508b_decode_w8.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6d9f86c6d14339cc931a377ecb592a314530a511e7fd0942ff8b6020b0c122fc
+size 512131072

v75/qwen3508b_embed_f16.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:24375b3c4ceac14cb5a66efa63545eae9ca0f37a6b57ea7a713797784249771b
+size 508559360

v75/qwen3508b_lmhead_f16.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6f086350a96bd4cf16b943ee88f8309fe032c1c2f006cd0d176afbb7ae64d11b
+size 509870080

v75/tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:06b9509352d2af50381ab2247e083b80d32d5c0aba91c272ca9ff729b6a0e523
+size 19989325