Upload 4 files

config.json

@@ -0,0 +1,27 @@
+{
+  "model_type": "privacy_filter",
+  "inference_contract_version": 1,
+  "encoding": "o200k_base",
+  "num_hidden_layers": 8,
+  "num_experts": 128,
+  "experts_per_token": 4,
+  "vocab_size": 200064,
+  "num_labels": 33,
+  "hidden_size": 640,
+  "intermediate_size": 640,
+  "head_dim": 64,
+  "num_attention_heads": 14,
+  "num_key_value_heads": 2,
+  "sliding_window": 257,
+  "bidirectional_context": true,
+  "bidirectional_left_context": 128,
+  "bidirectional_right_context": 128,
+  "initial_context_length": 4096,
+  "max_position_embeddings": 131072,
+  "default_n_ctx": 128000,
+  "rope_theta": 150000,
+  "rope_scaling_factor": 32.0,
+  "rope_ntk_alpha": 1.0,
+  "rope_ntk_beta": 32.0,
+  "param_dtype": "bfloat16"
+}

dtypes.json

@@ -0,0 +1,92 @@
+{
+  "embedding.weight": "torch.bfloat16",
+  "block.0.attn.qkv.weight": "torch.bfloat16",
+  "block.0.attn.qkv.bias": "torch.bfloat16",
+  "block.0.attn.sinks": "torch.float32",
+  "block.0.attn.out.weight": "torch.bfloat16",
+  "block.0.attn.out.bias": "torch.bfloat16",
+  "block.0.mlp.gate.weight": "torch.bfloat16",
+  "block.0.mlp.gate.bias": "torch.bfloat16",
+  "block.0.mlp.swiglu.weight": "torch.bfloat16",
+  "block.0.mlp.swiglu.bias": "torch.bfloat16",
+  "block.0.mlp.out.weight": "torch.bfloat16",
+  "block.0.mlp.out.bias": "torch.bfloat16",
+  "block.1.attn.qkv.weight": "torch.bfloat16",
+  "block.1.attn.qkv.bias": "torch.bfloat16",
+  "block.1.attn.sinks": "torch.float32",
+  "block.1.attn.out.weight": "torch.bfloat16",
+  "block.1.attn.out.bias": "torch.bfloat16",
+  "block.1.mlp.gate.weight": "torch.bfloat16",
+  "block.1.mlp.gate.bias": "torch.bfloat16",
+  "block.1.mlp.swiglu.weight": "torch.bfloat16",
+  "block.1.mlp.swiglu.bias": "torch.bfloat16",
+  "block.1.mlp.out.weight": "torch.bfloat16",
+  "block.1.mlp.out.bias": "torch.bfloat16",
+  "block.2.attn.qkv.weight": "torch.bfloat16",
+  "block.2.attn.qkv.bias": "torch.bfloat16",
+  "block.2.attn.sinks": "torch.float32",
+  "block.2.attn.out.weight": "torch.bfloat16",
+  "block.2.attn.out.bias": "torch.bfloat16",
+  "block.2.mlp.gate.weight": "torch.bfloat16",
+  "block.2.mlp.gate.bias": "torch.bfloat16",
+  "block.2.mlp.swiglu.weight": "torch.bfloat16",
+  "block.2.mlp.swiglu.bias": "torch.bfloat16",
+  "block.2.mlp.out.weight": "torch.bfloat16",
+  "block.2.mlp.out.bias": "torch.bfloat16",
+  "block.3.attn.qkv.weight": "torch.bfloat16",
+  "block.3.attn.qkv.bias": "torch.bfloat16",
+  "block.3.attn.sinks": "torch.float32",
+  "block.3.attn.out.weight": "torch.bfloat16",
+  "block.3.attn.out.bias": "torch.bfloat16",
+  "block.3.mlp.gate.weight": "torch.bfloat16",
+  "block.3.mlp.gate.bias": "torch.bfloat16",
+  "block.3.mlp.swiglu.weight": "torch.bfloat16",
+  "block.3.mlp.swiglu.bias": "torch.bfloat16",
+  "block.3.mlp.out.weight": "torch.bfloat16",
+  "block.3.mlp.out.bias": "torch.bfloat16",
+  "block.4.attn.qkv.weight": "torch.bfloat16",
+  "block.4.attn.qkv.bias": "torch.bfloat16",
+  "block.4.attn.sinks": "torch.float32",
+  "block.4.attn.out.weight": "torch.bfloat16",
+  "block.4.attn.out.bias": "torch.bfloat16",
+  "block.4.mlp.gate.weight": "torch.bfloat16",
+  "block.4.mlp.gate.bias": "torch.bfloat16",
+  "block.4.mlp.swiglu.weight": "torch.bfloat16",
+  "block.4.mlp.swiglu.bias": "torch.bfloat16",
+  "block.4.mlp.out.weight": "torch.bfloat16",
+  "block.4.mlp.out.bias": "torch.bfloat16",
+  "block.5.attn.qkv.weight": "torch.bfloat16",
+  "block.5.attn.qkv.bias": "torch.bfloat16",
+  "block.5.attn.sinks": "torch.float32",
+  "block.5.attn.out.weight": "torch.bfloat16",
+  "block.5.attn.out.bias": "torch.bfloat16",
+  "block.5.mlp.gate.weight": "torch.bfloat16",
+  "block.5.mlp.gate.bias": "torch.bfloat16",
+  "block.5.mlp.swiglu.weight": "torch.bfloat16",
+  "block.5.mlp.swiglu.bias": "torch.bfloat16",
+  "block.5.mlp.out.weight": "torch.bfloat16",
+  "block.5.mlp.out.bias": "torch.bfloat16",
+  "block.6.attn.qkv.weight": "torch.bfloat16",
+  "block.6.attn.qkv.bias": "torch.bfloat16",
+  "block.6.attn.sinks": "torch.float32",
+  "block.6.attn.out.weight": "torch.bfloat16",
+  "block.6.attn.out.bias": "torch.bfloat16",
+  "block.6.mlp.gate.weight": "torch.bfloat16",
+  "block.6.mlp.gate.bias": "torch.bfloat16",
+  "block.6.mlp.swiglu.weight": "torch.bfloat16",
+  "block.6.mlp.swiglu.bias": "torch.bfloat16",
+  "block.6.mlp.out.weight": "torch.bfloat16",
+  "block.6.mlp.out.bias": "torch.bfloat16",
+  "block.7.attn.qkv.weight": "torch.bfloat16",
+  "block.7.attn.qkv.bias": "torch.bfloat16",
+  "block.7.attn.sinks": "torch.float32",
+  "block.7.attn.out.weight": "torch.bfloat16",
+  "block.7.attn.out.bias": "torch.bfloat16",
+  "block.7.mlp.gate.weight": "torch.bfloat16",
+  "block.7.mlp.gate.bias": "torch.bfloat16",
+  "block.7.mlp.swiglu.weight": "torch.bfloat16",
+  "block.7.mlp.swiglu.bias": "torch.bfloat16",
+  "block.7.mlp.out.weight": "torch.bfloat16",
+  "block.7.mlp.out.bias": "torch.bfloat16",
+  "unembedding.weight": "torch.bfloat16"
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9c262cbe68a0c8a50590a648ef8341a2b7d3be1fa11dfb79893fe0b03ce57b5c
+size 2798984088

viterbi_calibration.json

@@ -0,0 +1,14 @@
+{
+  "operating_points": {
+    "default": {
+      "biases": {
+        "transition_bias_background_stay": 0.0,
+        "transition_bias_background_to_start": 0.0,
+        "transition_bias_end_to_background": 0.0,
+        "transition_bias_end_to_start": 0.0,
+        "transition_bias_inside_to_continue": 0.0,
+        "transition_bias_inside_to_end": 0.0
+      }
+    }
+  }
+}