diff --git "a/model.safetensors.index.json" "b/model.safetensors.index.json" new file mode 100644--- /dev/null +++ "b/model.safetensors.index.json" @@ -0,0 +1,1421 @@ +{ + "metadata": { + "total_size": 75401035776 + }, + "weight_map": { + "head.weight": "model-00001-of-00016.safetensors", + "model.blocks.0.attention.gate.weight": "model-00001-of-00016.safetensors", + "model.blocks.0.attention.key.weight": "model-00001-of-00016.safetensors", + "model.blocks.0.attention.ln_x.bias": "model-00001-of-00016.safetensors", + "model.blocks.0.attention.ln_x.weight": "model-00001-of-00016.safetensors", + "model.blocks.0.attention.output.weight": "model-00001-of-00016.safetensors", + "model.blocks.0.attention.receptance.weight": "model-00001-of-00016.safetensors", + "model.blocks.0.attention.time_decay": "model-00001-of-00016.safetensors", + "model.blocks.0.attention.time_decay_w1": "model-00001-of-00016.safetensors", + "model.blocks.0.attention.time_decay_w2": "model-00001-of-00016.safetensors", + "model.blocks.0.attention.time_faaaa": "model-00001-of-00016.safetensors", + "model.blocks.0.attention.time_maa_g": "model-00001-of-00016.safetensors", + "model.blocks.0.attention.time_maa_k": "model-00001-of-00016.safetensors", + "model.blocks.0.attention.time_maa_r": "model-00001-of-00016.safetensors", + "model.blocks.0.attention.time_maa_v": "model-00001-of-00016.safetensors", + "model.blocks.0.attention.time_maa_w": "model-00001-of-00016.safetensors", + "model.blocks.0.attention.time_maa_w1": "model-00001-of-00016.safetensors", + "model.blocks.0.attention.time_maa_w2": "model-00001-of-00016.safetensors", + "model.blocks.0.attention.time_maa_x": "model-00001-of-00016.safetensors", + "model.blocks.0.attention.value.weight": "model-00001-of-00016.safetensors", + "model.blocks.0.feed_forward.experts.0.key.weight": "model-00001-of-00016.safetensors", + "model.blocks.0.feed_forward.experts.0.value.weight": "model-00001-of-00016.safetensors", + "model.blocks.0.feed_forward.experts.1.key.weight": "model-00001-of-00016.safetensors", + "model.blocks.0.feed_forward.experts.1.value.weight": "model-00001-of-00016.safetensors", + "model.blocks.0.feed_forward.experts.2.key.weight": "model-00001-of-00016.safetensors", + "model.blocks.0.feed_forward.experts.2.value.weight": "model-00001-of-00016.safetensors", + "model.blocks.0.feed_forward.experts.3.key.weight": "model-00001-of-00016.safetensors", + "model.blocks.0.feed_forward.experts.3.value.weight": "model-00001-of-00016.safetensors", + "model.blocks.0.feed_forward.experts.4.key.weight": "model-00001-of-00016.safetensors", + "model.blocks.0.feed_forward.experts.4.value.weight": "model-00001-of-00016.safetensors", + "model.blocks.0.feed_forward.experts.5.key.weight": "model-00001-of-00016.safetensors", + "model.blocks.0.feed_forward.experts.5.value.weight": "model-00001-of-00016.safetensors", + "model.blocks.0.feed_forward.experts.6.key.weight": "model-00001-of-00016.safetensors", + "model.blocks.0.feed_forward.experts.6.value.weight": "model-00001-of-00016.safetensors", + "model.blocks.0.feed_forward.experts.7.key.weight": "model-00001-of-00016.safetensors", + "model.blocks.0.feed_forward.experts.7.value.weight": "model-00001-of-00016.safetensors", + "model.blocks.0.feed_forward.receptance.weight": "model-00001-of-00016.safetensors", + "model.blocks.0.feed_forward.shared_expert.key.weight": "model-00001-of-00016.safetensors", + "model.blocks.0.feed_forward.shared_expert.value.weight": "model-00001-of-00016.safetensors", + "model.blocks.0.feed_forward.time_maa_k": "model-00001-of-00016.safetensors", + "model.blocks.0.feed_forward.time_maa_r": "model-00001-of-00016.safetensors", + "model.blocks.0.ln1.bias": "model-00001-of-00016.safetensors", + "model.blocks.0.ln1.weight": "model-00001-of-00016.safetensors", + "model.blocks.0.ln2.bias": "model-00001-of-00016.safetensors", + "model.blocks.0.ln2.weight": "model-00001-of-00016.safetensors", + "model.blocks.0.pre_ln.bias": "model-00001-of-00016.safetensors", + "model.blocks.0.pre_ln.weight": "model-00001-of-00016.safetensors", + "model.blocks.1.attention.gate.weight": "model-00001-of-00016.safetensors", + "model.blocks.1.attention.key.weight": "model-00001-of-00016.safetensors", + "model.blocks.1.attention.ln_x.bias": "model-00001-of-00016.safetensors", + "model.blocks.1.attention.ln_x.weight": "model-00001-of-00016.safetensors", + "model.blocks.1.attention.output.weight": "model-00001-of-00016.safetensors", + "model.blocks.1.attention.receptance.weight": "model-00001-of-00016.safetensors", + "model.blocks.1.attention.time_decay": "model-00001-of-00016.safetensors", + "model.blocks.1.attention.time_decay_w1": "model-00001-of-00016.safetensors", + "model.blocks.1.attention.time_decay_w2": "model-00001-of-00016.safetensors", + "model.blocks.1.attention.time_faaaa": "model-00001-of-00016.safetensors", + "model.blocks.1.attention.time_maa_g": "model-00001-of-00016.safetensors", + "model.blocks.1.attention.time_maa_k": "model-00001-of-00016.safetensors", + "model.blocks.1.attention.time_maa_r": "model-00001-of-00016.safetensors", + "model.blocks.1.attention.time_maa_v": "model-00001-of-00016.safetensors", + "model.blocks.1.attention.time_maa_w": "model-00001-of-00016.safetensors", + "model.blocks.1.attention.time_maa_w1": "model-00001-of-00016.safetensors", + "model.blocks.1.attention.time_maa_w2": "model-00001-of-00016.safetensors", + "model.blocks.1.attention.time_maa_x": "model-00001-of-00016.safetensors", + "model.blocks.1.attention.value.weight": "model-00001-of-00016.safetensors", + "model.blocks.1.feed_forward.experts.0.key.weight": "model-00001-of-00016.safetensors", + "model.blocks.1.feed_forward.experts.0.value.weight": "model-00001-of-00016.safetensors", + "model.blocks.1.feed_forward.experts.1.key.weight": "model-00001-of-00016.safetensors", + "model.blocks.1.feed_forward.experts.1.value.weight": "model-00001-of-00016.safetensors", + "model.blocks.1.feed_forward.experts.2.key.weight": "model-00001-of-00016.safetensors", + "model.blocks.1.feed_forward.experts.2.value.weight": "model-00001-of-00016.safetensors", + "model.blocks.1.feed_forward.experts.3.key.weight": "model-00001-of-00016.safetensors", + "model.blocks.1.feed_forward.experts.3.value.weight": "model-00001-of-00016.safetensors", + "model.blocks.1.feed_forward.experts.4.key.weight": "model-00001-of-00016.safetensors", + "model.blocks.1.feed_forward.experts.4.value.weight": "model-00001-of-00016.safetensors", + "model.blocks.1.feed_forward.experts.5.key.weight": "model-00001-of-00016.safetensors", + "model.blocks.1.feed_forward.experts.5.value.weight": "model-00001-of-00016.safetensors", + "model.blocks.1.feed_forward.experts.6.key.weight": "model-00001-of-00016.safetensors", + "model.blocks.1.feed_forward.experts.6.value.weight": "model-00001-of-00016.safetensors", + "model.blocks.1.feed_forward.experts.7.key.weight": "model-00001-of-00016.safetensors", + "model.blocks.1.feed_forward.experts.7.value.weight": "model-00001-of-00016.safetensors", + "model.blocks.1.feed_forward.receptance.weight": "model-00001-of-00016.safetensors", + "model.blocks.1.feed_forward.shared_expert.key.weight": "model-00002-of-00016.safetensors", + "model.blocks.1.feed_forward.shared_expert.value.weight": "model-00002-of-00016.safetensors", + "model.blocks.1.feed_forward.time_maa_k": "model-00002-of-00016.safetensors", + "model.blocks.1.feed_forward.time_maa_r": "model-00002-of-00016.safetensors", + "model.blocks.1.ln1.bias": "model-00002-of-00016.safetensors", + "model.blocks.1.ln1.weight": "model-00002-of-00016.safetensors", + "model.blocks.1.ln2.bias": "model-00002-of-00016.safetensors", + "model.blocks.1.ln2.weight": "model-00002-of-00016.safetensors", + "model.blocks.10.attention.gate.weight": "model-00002-of-00016.safetensors", + "model.blocks.10.attention.key.weight": "model-00002-of-00016.safetensors", + "model.blocks.10.attention.ln_x.bias": "model-00002-of-00016.safetensors", + "model.blocks.10.attention.ln_x.weight": "model-00002-of-00016.safetensors", + "model.blocks.10.attention.output.weight": "model-00002-of-00016.safetensors", + "model.blocks.10.attention.receptance.weight": "model-00002-of-00016.safetensors", + "model.blocks.10.attention.time_decay": "model-00002-of-00016.safetensors", + "model.blocks.10.attention.time_decay_w1": "model-00002-of-00016.safetensors", + "model.blocks.10.attention.time_decay_w2": "model-00002-of-00016.safetensors", + "model.blocks.10.attention.time_faaaa": "model-00002-of-00016.safetensors", + "model.blocks.10.attention.time_maa_g": "model-00002-of-00016.safetensors", + "model.blocks.10.attention.time_maa_k": "model-00002-of-00016.safetensors", + "model.blocks.10.attention.time_maa_r": "model-00002-of-00016.safetensors", + "model.blocks.10.attention.time_maa_v": "model-00002-of-00016.safetensors", + "model.blocks.10.attention.time_maa_w": "model-00002-of-00016.safetensors", + "model.blocks.10.attention.time_maa_w1": "model-00002-of-00016.safetensors", + "model.blocks.10.attention.time_maa_w2": "model-00002-of-00016.safetensors", + "model.blocks.10.attention.time_maa_x": "model-00002-of-00016.safetensors", + "model.blocks.10.attention.value.weight": "model-00002-of-00016.safetensors", + "model.blocks.10.feed_forward.experts.0.key.weight": "model-00002-of-00016.safetensors", + "model.blocks.10.feed_forward.experts.0.value.weight": "model-00002-of-00016.safetensors", + "model.blocks.10.feed_forward.experts.1.key.weight": "model-00002-of-00016.safetensors", + "model.blocks.10.feed_forward.experts.1.value.weight": "model-00002-of-00016.safetensors", + "model.blocks.10.feed_forward.experts.2.key.weight": "model-00002-of-00016.safetensors", + "model.blocks.10.feed_forward.experts.2.value.weight": "model-00002-of-00016.safetensors", + "model.blocks.10.feed_forward.experts.3.key.weight": "model-00002-of-00016.safetensors", + "model.blocks.10.feed_forward.experts.3.value.weight": "model-00002-of-00016.safetensors", + "model.blocks.10.feed_forward.experts.4.key.weight": "model-00002-of-00016.safetensors", + "model.blocks.10.feed_forward.experts.4.value.weight": "model-00002-of-00016.safetensors", + "model.blocks.10.feed_forward.experts.5.key.weight": "model-00002-of-00016.safetensors", + "model.blocks.10.feed_forward.experts.5.value.weight": "model-00002-of-00016.safetensors", + "model.blocks.10.feed_forward.experts.6.key.weight": "model-00002-of-00016.safetensors", + "model.blocks.10.feed_forward.experts.6.value.weight": "model-00002-of-00016.safetensors", + "model.blocks.10.feed_forward.experts.7.key.weight": "model-00002-of-00016.safetensors", + "model.blocks.10.feed_forward.experts.7.value.weight": "model-00002-of-00016.safetensors", + "model.blocks.10.feed_forward.receptance.weight": "model-00002-of-00016.safetensors", + "model.blocks.10.feed_forward.shared_expert.key.weight": "model-00002-of-00016.safetensors", + "model.blocks.10.feed_forward.shared_expert.value.weight": "model-00002-of-00016.safetensors", + "model.blocks.10.feed_forward.time_maa_k": "model-00002-of-00016.safetensors", + "model.blocks.10.feed_forward.time_maa_r": "model-00002-of-00016.safetensors", + "model.blocks.10.ln1.bias": "model-00002-of-00016.safetensors", + "model.blocks.10.ln1.weight": "model-00002-of-00016.safetensors", + "model.blocks.10.ln2.bias": "model-00002-of-00016.safetensors", + "model.blocks.10.ln2.weight": "model-00002-of-00016.safetensors", + "model.blocks.11.attention.gate.weight": "model-00002-of-00016.safetensors", + "model.blocks.11.attention.key.weight": "model-00002-of-00016.safetensors", + "model.blocks.11.attention.ln_x.bias": "model-00002-of-00016.safetensors", + "model.blocks.11.attention.ln_x.weight": "model-00002-of-00016.safetensors", + "model.blocks.11.attention.output.weight": "model-00002-of-00016.safetensors", + "model.blocks.11.attention.receptance.weight": "model-00002-of-00016.safetensors", + "model.blocks.11.attention.time_decay": "model-00002-of-00016.safetensors", + "model.blocks.11.attention.time_decay_w1": "model-00002-of-00016.safetensors", + "model.blocks.11.attention.time_decay_w2": "model-00002-of-00016.safetensors", + "model.blocks.11.attention.time_faaaa": "model-00002-of-00016.safetensors", + "model.blocks.11.attention.time_maa_g": "model-00002-of-00016.safetensors", + "model.blocks.11.attention.time_maa_k": "model-00002-of-00016.safetensors", + "model.blocks.11.attention.time_maa_r": "model-00002-of-00016.safetensors", + "model.blocks.11.attention.time_maa_v": "model-00002-of-00016.safetensors", + "model.blocks.11.attention.time_maa_w": "model-00002-of-00016.safetensors", + "model.blocks.11.attention.time_maa_w1": "model-00002-of-00016.safetensors", + "model.blocks.11.attention.time_maa_w2": "model-00002-of-00016.safetensors", + "model.blocks.11.attention.time_maa_x": "model-00002-of-00016.safetensors", + "model.blocks.11.attention.value.weight": "model-00002-of-00016.safetensors", + "model.blocks.11.feed_forward.experts.0.key.weight": "model-00002-of-00016.safetensors", + "model.blocks.11.feed_forward.experts.0.value.weight": "model-00002-of-00016.safetensors", + "model.blocks.11.feed_forward.experts.1.key.weight": "model-00002-of-00016.safetensors", + "model.blocks.11.feed_forward.experts.1.value.weight": "model-00002-of-00016.safetensors", + "model.blocks.11.feed_forward.experts.2.key.weight": "model-00002-of-00016.safetensors", + "model.blocks.11.feed_forward.experts.2.value.weight": "model-00002-of-00016.safetensors", + "model.blocks.11.feed_forward.experts.3.key.weight": "model-00002-of-00016.safetensors", + "model.blocks.11.feed_forward.experts.3.value.weight": "model-00002-of-00016.safetensors", + "model.blocks.11.feed_forward.experts.4.key.weight": "model-00002-of-00016.safetensors", + "model.blocks.11.feed_forward.experts.4.value.weight": "model-00002-of-00016.safetensors", + "model.blocks.11.feed_forward.experts.5.key.weight": "model-00002-of-00016.safetensors", + "model.blocks.11.feed_forward.experts.5.value.weight": "model-00002-of-00016.safetensors", + "model.blocks.11.feed_forward.experts.6.key.weight": "model-00002-of-00016.safetensors", + "model.blocks.11.feed_forward.experts.6.value.weight": "model-00002-of-00016.safetensors", + "model.blocks.11.feed_forward.experts.7.key.weight": "model-00002-of-00016.safetensors", + "model.blocks.11.feed_forward.experts.7.value.weight": "model-00002-of-00016.safetensors", + "model.blocks.11.feed_forward.receptance.weight": "model-00002-of-00016.safetensors", + "model.blocks.11.feed_forward.shared_expert.key.weight": "model-00002-of-00016.safetensors", + "model.blocks.11.feed_forward.shared_expert.value.weight": "model-00002-of-00016.safetensors", + "model.blocks.11.feed_forward.time_maa_k": "model-00002-of-00016.safetensors", + "model.blocks.11.feed_forward.time_maa_r": "model-00002-of-00016.safetensors", + "model.blocks.11.ln1.bias": "model-00002-of-00016.safetensors", + "model.blocks.11.ln1.weight": "model-00002-of-00016.safetensors", + "model.blocks.11.ln2.bias": "model-00002-of-00016.safetensors", + "model.blocks.11.ln2.weight": "model-00002-of-00016.safetensors", + "model.blocks.12.attention.gate.weight": "model-00002-of-00016.safetensors", + "model.blocks.12.attention.key.weight": "model-00002-of-00016.safetensors", + "model.blocks.12.attention.ln_x.bias": "model-00002-of-00016.safetensors", + "model.blocks.12.attention.ln_x.weight": "model-00002-of-00016.safetensors", + "model.blocks.12.attention.output.weight": "model-00002-of-00016.safetensors", + "model.blocks.12.attention.receptance.weight": "model-00003-of-00016.safetensors", + "model.blocks.12.attention.time_decay": "model-00003-of-00016.safetensors", + "model.blocks.12.attention.time_decay_w1": "model-00003-of-00016.safetensors", + "model.blocks.12.attention.time_decay_w2": "model-00003-of-00016.safetensors", + "model.blocks.12.attention.time_faaaa": "model-00003-of-00016.safetensors", + "model.blocks.12.attention.time_maa_g": "model-00003-of-00016.safetensors", + "model.blocks.12.attention.time_maa_k": "model-00003-of-00016.safetensors", + "model.blocks.12.attention.time_maa_r": "model-00003-of-00016.safetensors", + "model.blocks.12.attention.time_maa_v": "model-00003-of-00016.safetensors", + "model.blocks.12.attention.time_maa_w": "model-00003-of-00016.safetensors", + "model.blocks.12.attention.time_maa_w1": "model-00003-of-00016.safetensors", + "model.blocks.12.attention.time_maa_w2": "model-00003-of-00016.safetensors", + "model.blocks.12.attention.time_maa_x": "model-00003-of-00016.safetensors", + "model.blocks.12.attention.value.weight": "model-00003-of-00016.safetensors", + "model.blocks.12.feed_forward.experts.0.key.weight": "model-00003-of-00016.safetensors", + "model.blocks.12.feed_forward.experts.0.value.weight": "model-00003-of-00016.safetensors", + "model.blocks.12.feed_forward.experts.1.key.weight": "model-00003-of-00016.safetensors", + "model.blocks.12.feed_forward.experts.1.value.weight": "model-00003-of-00016.safetensors", + "model.blocks.12.feed_forward.experts.2.key.weight": "model-00003-of-00016.safetensors", + "model.blocks.12.feed_forward.experts.2.value.weight": "model-00003-of-00016.safetensors", + "model.blocks.12.feed_forward.experts.3.key.weight": "model-00003-of-00016.safetensors", + "model.blocks.12.feed_forward.experts.3.value.weight": "model-00003-of-00016.safetensors", + "model.blocks.12.feed_forward.experts.4.key.weight": "model-00003-of-00016.safetensors", + "model.blocks.12.feed_forward.experts.4.value.weight": "model-00003-of-00016.safetensors", + "model.blocks.12.feed_forward.experts.5.key.weight": "model-00003-of-00016.safetensors", + "model.blocks.12.feed_forward.experts.5.value.weight": "model-00003-of-00016.safetensors", + "model.blocks.12.feed_forward.experts.6.key.weight": "model-00003-of-00016.safetensors", + "model.blocks.12.feed_forward.experts.6.value.weight": "model-00003-of-00016.safetensors", + "model.blocks.12.feed_forward.experts.7.key.weight": "model-00003-of-00016.safetensors", + "model.blocks.12.feed_forward.experts.7.value.weight": "model-00003-of-00016.safetensors", + "model.blocks.12.feed_forward.receptance.weight": "model-00003-of-00016.safetensors", + "model.blocks.12.feed_forward.shared_expert.key.weight": "model-00003-of-00016.safetensors", + "model.blocks.12.feed_forward.shared_expert.value.weight": "model-00003-of-00016.safetensors", + "model.blocks.12.feed_forward.time_maa_k": "model-00003-of-00016.safetensors", + "model.blocks.12.feed_forward.time_maa_r": "model-00003-of-00016.safetensors", + "model.blocks.12.ln1.bias": "model-00003-of-00016.safetensors", + "model.blocks.12.ln1.weight": "model-00003-of-00016.safetensors", + "model.blocks.12.ln2.bias": "model-00003-of-00016.safetensors", + "model.blocks.12.ln2.weight": "model-00003-of-00016.safetensors", + "model.blocks.13.attention.gate.weight": "model-00003-of-00016.safetensors", + "model.blocks.13.attention.key.weight": "model-00003-of-00016.safetensors", + "model.blocks.13.attention.ln_x.bias": "model-00003-of-00016.safetensors", + "model.blocks.13.attention.ln_x.weight": "model-00003-of-00016.safetensors", + "model.blocks.13.attention.output.weight": "model-00003-of-00016.safetensors", + "model.blocks.13.attention.receptance.weight": "model-00003-of-00016.safetensors", + "model.blocks.13.attention.time_decay": "model-00003-of-00016.safetensors", + "model.blocks.13.attention.time_decay_w1": "model-00003-of-00016.safetensors", + "model.blocks.13.attention.time_decay_w2": "model-00003-of-00016.safetensors", + "model.blocks.13.attention.time_faaaa": "model-00003-of-00016.safetensors", + "model.blocks.13.attention.time_maa_g": "model-00003-of-00016.safetensors", + "model.blocks.13.attention.time_maa_k": "model-00003-of-00016.safetensors", + "model.blocks.13.attention.time_maa_r": "model-00003-of-00016.safetensors", + "model.blocks.13.attention.time_maa_v": "model-00003-of-00016.safetensors", + "model.blocks.13.attention.time_maa_w": "model-00003-of-00016.safetensors", + "model.blocks.13.attention.time_maa_w1": "model-00003-of-00016.safetensors", + "model.blocks.13.attention.time_maa_w2": "model-00003-of-00016.safetensors", + "model.blocks.13.attention.time_maa_x": "model-00003-of-00016.safetensors", + "model.blocks.13.attention.value.weight": "model-00003-of-00016.safetensors", + "model.blocks.13.feed_forward.experts.0.key.weight": "model-00003-of-00016.safetensors", + "model.blocks.13.feed_forward.experts.0.value.weight": "model-00003-of-00016.safetensors", + "model.blocks.13.feed_forward.experts.1.key.weight": "model-00003-of-00016.safetensors", + "model.blocks.13.feed_forward.experts.1.value.weight": "model-00003-of-00016.safetensors", + "model.blocks.13.feed_forward.experts.2.key.weight": "model-00003-of-00016.safetensors", + "model.blocks.13.feed_forward.experts.2.value.weight": "model-00003-of-00016.safetensors", + "model.blocks.13.feed_forward.experts.3.key.weight": "model-00003-of-00016.safetensors", + "model.blocks.13.feed_forward.experts.3.value.weight": "model-00003-of-00016.safetensors", + "model.blocks.13.feed_forward.experts.4.key.weight": "model-00003-of-00016.safetensors", + "model.blocks.13.feed_forward.experts.4.value.weight": "model-00003-of-00016.safetensors", + "model.blocks.13.feed_forward.experts.5.key.weight": "model-00003-of-00016.safetensors", + "model.blocks.13.feed_forward.experts.5.value.weight": "model-00003-of-00016.safetensors", + "model.blocks.13.feed_forward.experts.6.key.weight": "model-00003-of-00016.safetensors", + "model.blocks.13.feed_forward.experts.6.value.weight": "model-00003-of-00016.safetensors", + "model.blocks.13.feed_forward.experts.7.key.weight": "model-00003-of-00016.safetensors", + "model.blocks.13.feed_forward.experts.7.value.weight": "model-00003-of-00016.safetensors", + "model.blocks.13.feed_forward.receptance.weight": "model-00003-of-00016.safetensors", + "model.blocks.13.feed_forward.shared_expert.key.weight": "model-00003-of-00016.safetensors", + "model.blocks.13.feed_forward.shared_expert.value.weight": "model-00003-of-00016.safetensors", + "model.blocks.13.feed_forward.time_maa_k": "model-00003-of-00016.safetensors", + "model.blocks.13.feed_forward.time_maa_r": "model-00003-of-00016.safetensors", + "model.blocks.13.ln1.bias": "model-00003-of-00016.safetensors", + "model.blocks.13.ln1.weight": "model-00003-of-00016.safetensors", + "model.blocks.13.ln2.bias": "model-00003-of-00016.safetensors", + "model.blocks.13.ln2.weight": "model-00003-of-00016.safetensors", + "model.blocks.14.attention.gate.weight": "model-00003-of-00016.safetensors", + "model.blocks.14.attention.key.weight": "model-00003-of-00016.safetensors", + "model.blocks.14.attention.ln_x.bias": "model-00003-of-00016.safetensors", + "model.blocks.14.attention.ln_x.weight": "model-00003-of-00016.safetensors", + "model.blocks.14.attention.output.weight": "model-00003-of-00016.safetensors", + "model.blocks.14.attention.receptance.weight": "model-00003-of-00016.safetensors", + "model.blocks.14.attention.time_decay": "model-00003-of-00016.safetensors", + "model.blocks.14.attention.time_decay_w1": "model-00003-of-00016.safetensors", + "model.blocks.14.attention.time_decay_w2": "model-00003-of-00016.safetensors", + "model.blocks.14.attention.time_faaaa": "model-00003-of-00016.safetensors", + "model.blocks.14.attention.time_maa_g": "model-00003-of-00016.safetensors", + "model.blocks.14.attention.time_maa_k": "model-00003-of-00016.safetensors", + "model.blocks.14.attention.time_maa_r": "model-00003-of-00016.safetensors", + "model.blocks.14.attention.time_maa_v": "model-00003-of-00016.safetensors", + "model.blocks.14.attention.time_maa_w": "model-00003-of-00016.safetensors", + "model.blocks.14.attention.time_maa_w1": "model-00003-of-00016.safetensors", + "model.blocks.14.attention.time_maa_w2": "model-00003-of-00016.safetensors", + "model.blocks.14.attention.time_maa_x": "model-00003-of-00016.safetensors", + "model.blocks.14.attention.value.weight": "model-00003-of-00016.safetensors", + "model.blocks.14.feed_forward.experts.0.key.weight": "model-00003-of-00016.safetensors", + "model.blocks.14.feed_forward.experts.0.value.weight": "model-00003-of-00016.safetensors", + "model.blocks.14.feed_forward.experts.1.key.weight": "model-00004-of-00016.safetensors", + "model.blocks.14.feed_forward.experts.1.value.weight": "model-00004-of-00016.safetensors", + "model.blocks.14.feed_forward.experts.2.key.weight": "model-00004-of-00016.safetensors", + "model.blocks.14.feed_forward.experts.2.value.weight": "model-00004-of-00016.safetensors", + "model.blocks.14.feed_forward.experts.3.key.weight": "model-00004-of-00016.safetensors", + "model.blocks.14.feed_forward.experts.3.value.weight": "model-00004-of-00016.safetensors", + "model.blocks.14.feed_forward.experts.4.key.weight": "model-00004-of-00016.safetensors", + "model.blocks.14.feed_forward.experts.4.value.weight": "model-00004-of-00016.safetensors", + "model.blocks.14.feed_forward.experts.5.key.weight": "model-00004-of-00016.safetensors", + "model.blocks.14.feed_forward.experts.5.value.weight": "model-00004-of-00016.safetensors", + "model.blocks.14.feed_forward.experts.6.key.weight": "model-00004-of-00016.safetensors", + "model.blocks.14.feed_forward.experts.6.value.weight": "model-00004-of-00016.safetensors", + "model.blocks.14.feed_forward.experts.7.key.weight": "model-00004-of-00016.safetensors", + "model.blocks.14.feed_forward.experts.7.value.weight": "model-00004-of-00016.safetensors", + "model.blocks.14.feed_forward.receptance.weight": "model-00004-of-00016.safetensors", + "model.blocks.14.feed_forward.shared_expert.key.weight": "model-00004-of-00016.safetensors", + "model.blocks.14.feed_forward.shared_expert.value.weight": "model-00004-of-00016.safetensors", + "model.blocks.14.feed_forward.time_maa_k": "model-00004-of-00016.safetensors", + "model.blocks.14.feed_forward.time_maa_r": "model-00004-of-00016.safetensors", + "model.blocks.14.ln1.bias": "model-00004-of-00016.safetensors", + "model.blocks.14.ln1.weight": "model-00004-of-00016.safetensors", + "model.blocks.14.ln2.bias": "model-00004-of-00016.safetensors", + "model.blocks.14.ln2.weight": "model-00004-of-00016.safetensors", + "model.blocks.15.attention.gate.weight": "model-00004-of-00016.safetensors", + "model.blocks.15.attention.key.weight": "model-00004-of-00016.safetensors", + "model.blocks.15.attention.ln_x.bias": "model-00004-of-00016.safetensors", + "model.blocks.15.attention.ln_x.weight": "model-00004-of-00016.safetensors", + "model.blocks.15.attention.output.weight": "model-00004-of-00016.safetensors", + "model.blocks.15.attention.receptance.weight": "model-00004-of-00016.safetensors", + "model.blocks.15.attention.time_decay": "model-00004-of-00016.safetensors", + "model.blocks.15.attention.time_decay_w1": "model-00004-of-00016.safetensors", + "model.blocks.15.attention.time_decay_w2": "model-00004-of-00016.safetensors", + "model.blocks.15.attention.time_faaaa": "model-00004-of-00016.safetensors", + "model.blocks.15.attention.time_maa_g": "model-00004-of-00016.safetensors", + "model.blocks.15.attention.time_maa_k": "model-00004-of-00016.safetensors", + "model.blocks.15.attention.time_maa_r": "model-00004-of-00016.safetensors", + "model.blocks.15.attention.time_maa_v": "model-00004-of-00016.safetensors", + "model.blocks.15.attention.time_maa_w": "model-00004-of-00016.safetensors", + "model.blocks.15.attention.time_maa_w1": "model-00004-of-00016.safetensors", + "model.blocks.15.attention.time_maa_w2": "model-00004-of-00016.safetensors", + "model.blocks.15.attention.time_maa_x": "model-00004-of-00016.safetensors", + "model.blocks.15.attention.value.weight": "model-00004-of-00016.safetensors", + "model.blocks.15.feed_forward.experts.0.key.weight": "model-00004-of-00016.safetensors", + "model.blocks.15.feed_forward.experts.0.value.weight": "model-00004-of-00016.safetensors", + "model.blocks.15.feed_forward.experts.1.key.weight": "model-00004-of-00016.safetensors", + "model.blocks.15.feed_forward.experts.1.value.weight": "model-00004-of-00016.safetensors", + "model.blocks.15.feed_forward.experts.2.key.weight": "model-00004-of-00016.safetensors", + "model.blocks.15.feed_forward.experts.2.value.weight": "model-00004-of-00016.safetensors", + "model.blocks.15.feed_forward.experts.3.key.weight": "model-00004-of-00016.safetensors", + "model.blocks.15.feed_forward.experts.3.value.weight": "model-00004-of-00016.safetensors", + "model.blocks.15.feed_forward.experts.4.key.weight": "model-00004-of-00016.safetensors", + "model.blocks.15.feed_forward.experts.4.value.weight": "model-00004-of-00016.safetensors", + "model.blocks.15.feed_forward.experts.5.key.weight": "model-00004-of-00016.safetensors", + "model.blocks.15.feed_forward.experts.5.value.weight": "model-00004-of-00016.safetensors", + "model.blocks.15.feed_forward.experts.6.key.weight": "model-00004-of-00016.safetensors", + "model.blocks.15.feed_forward.experts.6.value.weight": "model-00004-of-00016.safetensors", + "model.blocks.15.feed_forward.experts.7.key.weight": "model-00004-of-00016.safetensors", + "model.blocks.15.feed_forward.experts.7.value.weight": "model-00004-of-00016.safetensors", + "model.blocks.15.feed_forward.receptance.weight": "model-00004-of-00016.safetensors", + "model.blocks.15.feed_forward.shared_expert.key.weight": "model-00004-of-00016.safetensors", + "model.blocks.15.feed_forward.shared_expert.value.weight": "model-00004-of-00016.safetensors", + "model.blocks.15.feed_forward.time_maa_k": "model-00004-of-00016.safetensors", + "model.blocks.15.feed_forward.time_maa_r": "model-00004-of-00016.safetensors", + "model.blocks.15.ln1.bias": "model-00004-of-00016.safetensors", + "model.blocks.15.ln1.weight": "model-00004-of-00016.safetensors", + "model.blocks.15.ln2.bias": "model-00004-of-00016.safetensors", + "model.blocks.15.ln2.weight": "model-00004-of-00016.safetensors", + "model.blocks.16.attention.gate.weight": "model-00004-of-00016.safetensors", + "model.blocks.16.attention.key.weight": "model-00004-of-00016.safetensors", + "model.blocks.16.attention.ln_x.bias": "model-00004-of-00016.safetensors", + "model.blocks.16.attention.ln_x.weight": "model-00004-of-00016.safetensors", + "model.blocks.16.attention.output.weight": "model-00004-of-00016.safetensors", + "model.blocks.16.attention.receptance.weight": "model-00004-of-00016.safetensors", + "model.blocks.16.attention.time_decay": "model-00004-of-00016.safetensors", + "model.blocks.16.attention.time_decay_w1": "model-00004-of-00016.safetensors", + "model.blocks.16.attention.time_decay_w2": "model-00004-of-00016.safetensors", + "model.blocks.16.attention.time_faaaa": "model-00004-of-00016.safetensors", + "model.blocks.16.attention.time_maa_g": "model-00004-of-00016.safetensors", + "model.blocks.16.attention.time_maa_k": "model-00004-of-00016.safetensors", + "model.blocks.16.attention.time_maa_r": "model-00004-of-00016.safetensors", + "model.blocks.16.attention.time_maa_v": "model-00004-of-00016.safetensors", + "model.blocks.16.attention.time_maa_w": "model-00004-of-00016.safetensors", + "model.blocks.16.attention.time_maa_w1": "model-00004-of-00016.safetensors", + "model.blocks.16.attention.time_maa_w2": "model-00004-of-00016.safetensors", + "model.blocks.16.attention.time_maa_x": "model-00004-of-00016.safetensors", + "model.blocks.16.attention.value.weight": "model-00004-of-00016.safetensors", + "model.blocks.16.feed_forward.experts.0.key.weight": "model-00004-of-00016.safetensors", + "model.blocks.16.feed_forward.experts.0.value.weight": "model-00004-of-00016.safetensors", + "model.blocks.16.feed_forward.experts.1.key.weight": "model-00004-of-00016.safetensors", + "model.blocks.16.feed_forward.experts.1.value.weight": "model-00004-of-00016.safetensors", + "model.blocks.16.feed_forward.experts.2.key.weight": "model-00004-of-00016.safetensors", + "model.blocks.16.feed_forward.experts.2.value.weight": "model-00005-of-00016.safetensors", + "model.blocks.16.feed_forward.experts.3.key.weight": "model-00005-of-00016.safetensors", + "model.blocks.16.feed_forward.experts.3.value.weight": "model-00005-of-00016.safetensors", + "model.blocks.16.feed_forward.experts.4.key.weight": "model-00005-of-00016.safetensors", + "model.blocks.16.feed_forward.experts.4.value.weight": "model-00005-of-00016.safetensors", + "model.blocks.16.feed_forward.experts.5.key.weight": "model-00005-of-00016.safetensors", + "model.blocks.16.feed_forward.experts.5.value.weight": "model-00005-of-00016.safetensors", + "model.blocks.16.feed_forward.experts.6.key.weight": "model-00005-of-00016.safetensors", + "model.blocks.16.feed_forward.experts.6.value.weight": "model-00005-of-00016.safetensors", + "model.blocks.16.feed_forward.experts.7.key.weight": "model-00005-of-00016.safetensors", + "model.blocks.16.feed_forward.experts.7.value.weight": "model-00005-of-00016.safetensors", + "model.blocks.16.feed_forward.receptance.weight": "model-00005-of-00016.safetensors", + "model.blocks.16.feed_forward.shared_expert.key.weight": "model-00005-of-00016.safetensors", + "model.blocks.16.feed_forward.shared_expert.value.weight": "model-00005-of-00016.safetensors", + "model.blocks.16.feed_forward.time_maa_k": "model-00005-of-00016.safetensors", + "model.blocks.16.feed_forward.time_maa_r": "model-00005-of-00016.safetensors", + "model.blocks.16.ln1.bias": "model-00005-of-00016.safetensors", + "model.blocks.16.ln1.weight": "model-00005-of-00016.safetensors", + "model.blocks.16.ln2.bias": "model-00005-of-00016.safetensors", + "model.blocks.16.ln2.weight": "model-00005-of-00016.safetensors", + "model.blocks.17.attention.gate.weight": "model-00005-of-00016.safetensors", + "model.blocks.17.attention.key.weight": "model-00005-of-00016.safetensors", + "model.blocks.17.attention.ln_x.bias": "model-00005-of-00016.safetensors", + "model.blocks.17.attention.ln_x.weight": "model-00005-of-00016.safetensors", + "model.blocks.17.attention.output.weight": "model-00005-of-00016.safetensors", + "model.blocks.17.attention.receptance.weight": "model-00005-of-00016.safetensors", + "model.blocks.17.attention.time_decay": "model-00005-of-00016.safetensors", + "model.blocks.17.attention.time_decay_w1": "model-00005-of-00016.safetensors", + "model.blocks.17.attention.time_decay_w2": "model-00005-of-00016.safetensors", + "model.blocks.17.attention.time_faaaa": "model-00005-of-00016.safetensors", + "model.blocks.17.attention.time_maa_g": "model-00005-of-00016.safetensors", + "model.blocks.17.attention.time_maa_k": "model-00005-of-00016.safetensors", + "model.blocks.17.attention.time_maa_r": "model-00005-of-00016.safetensors", + "model.blocks.17.attention.time_maa_v": "model-00005-of-00016.safetensors", + "model.blocks.17.attention.time_maa_w": "model-00005-of-00016.safetensors", + "model.blocks.17.attention.time_maa_w1": "model-00005-of-00016.safetensors", + "model.blocks.17.attention.time_maa_w2": "model-00005-of-00016.safetensors", + "model.blocks.17.attention.time_maa_x": "model-00005-of-00016.safetensors", + "model.blocks.17.attention.value.weight": "model-00005-of-00016.safetensors", + "model.blocks.17.feed_forward.experts.0.key.weight": "model-00005-of-00016.safetensors", + "model.blocks.17.feed_forward.experts.0.value.weight": "model-00005-of-00016.safetensors", + "model.blocks.17.feed_forward.experts.1.key.weight": "model-00005-of-00016.safetensors", + "model.blocks.17.feed_forward.experts.1.value.weight": "model-00005-of-00016.safetensors", + "model.blocks.17.feed_forward.experts.2.key.weight": "model-00005-of-00016.safetensors", + "model.blocks.17.feed_forward.experts.2.value.weight": "model-00005-of-00016.safetensors", + "model.blocks.17.feed_forward.experts.3.key.weight": "model-00005-of-00016.safetensors", + "model.blocks.17.feed_forward.experts.3.value.weight": "model-00005-of-00016.safetensors", + "model.blocks.17.feed_forward.experts.4.key.weight": "model-00005-of-00016.safetensors", + "model.blocks.17.feed_forward.experts.4.value.weight": "model-00005-of-00016.safetensors", + "model.blocks.17.feed_forward.experts.5.key.weight": "model-00005-of-00016.safetensors", + "model.blocks.17.feed_forward.experts.5.value.weight": "model-00005-of-00016.safetensors", + "model.blocks.17.feed_forward.experts.6.key.weight": "model-00005-of-00016.safetensors", + "model.blocks.17.feed_forward.experts.6.value.weight": "model-00005-of-00016.safetensors", + "model.blocks.17.feed_forward.experts.7.key.weight": "model-00005-of-00016.safetensors", + "model.blocks.17.feed_forward.experts.7.value.weight": "model-00005-of-00016.safetensors", + "model.blocks.17.feed_forward.receptance.weight": "model-00005-of-00016.safetensors", + "model.blocks.17.feed_forward.shared_expert.key.weight": "model-00005-of-00016.safetensors", + "model.blocks.17.feed_forward.shared_expert.value.weight": "model-00005-of-00016.safetensors", + "model.blocks.17.feed_forward.time_maa_k": "model-00005-of-00016.safetensors", + "model.blocks.17.feed_forward.time_maa_r": "model-00005-of-00016.safetensors", + "model.blocks.17.ln1.bias": "model-00005-of-00016.safetensors", + "model.blocks.17.ln1.weight": "model-00005-of-00016.safetensors", + "model.blocks.17.ln2.bias": "model-00005-of-00016.safetensors", + "model.blocks.17.ln2.weight": "model-00005-of-00016.safetensors", + "model.blocks.18.attention.gate.weight": "model-00005-of-00016.safetensors", + "model.blocks.18.attention.key.weight": "model-00005-of-00016.safetensors", + "model.blocks.18.attention.ln_x.bias": "model-00005-of-00016.safetensors", + "model.blocks.18.attention.ln_x.weight": "model-00005-of-00016.safetensors", + "model.blocks.18.attention.output.weight": "model-00005-of-00016.safetensors", + "model.blocks.18.attention.receptance.weight": "model-00005-of-00016.safetensors", + "model.blocks.18.attention.time_decay": "model-00005-of-00016.safetensors", + "model.blocks.18.attention.time_decay_w1": "model-00005-of-00016.safetensors", + "model.blocks.18.attention.time_decay_w2": "model-00005-of-00016.safetensors", + "model.blocks.18.attention.time_faaaa": "model-00005-of-00016.safetensors", + "model.blocks.18.attention.time_maa_g": "model-00005-of-00016.safetensors", + "model.blocks.18.attention.time_maa_k": "model-00005-of-00016.safetensors", + "model.blocks.18.attention.time_maa_r": "model-00005-of-00016.safetensors", + "model.blocks.18.attention.time_maa_v": "model-00005-of-00016.safetensors", + "model.blocks.18.attention.time_maa_w": "model-00005-of-00016.safetensors", + "model.blocks.18.attention.time_maa_w1": "model-00005-of-00016.safetensors", + "model.blocks.18.attention.time_maa_w2": "model-00005-of-00016.safetensors", + "model.blocks.18.attention.time_maa_x": "model-00005-of-00016.safetensors", + "model.blocks.18.attention.value.weight": "model-00005-of-00016.safetensors", + "model.blocks.18.feed_forward.experts.0.key.weight": "model-00005-of-00016.safetensors", + "model.blocks.18.feed_forward.experts.0.value.weight": "model-00005-of-00016.safetensors", + "model.blocks.18.feed_forward.experts.1.key.weight": "model-00005-of-00016.safetensors", + "model.blocks.18.feed_forward.experts.1.value.weight": "model-00005-of-00016.safetensors", + "model.blocks.18.feed_forward.experts.2.key.weight": "model-00005-of-00016.safetensors", + "model.blocks.18.feed_forward.experts.2.value.weight": "model-00005-of-00016.safetensors", + "model.blocks.18.feed_forward.experts.3.key.weight": "model-00005-of-00016.safetensors", + "model.blocks.18.feed_forward.experts.3.value.weight": "model-00005-of-00016.safetensors", + "model.blocks.18.feed_forward.experts.4.key.weight": "model-00006-of-00016.safetensors", + "model.blocks.18.feed_forward.experts.4.value.weight": "model-00006-of-00016.safetensors", + "model.blocks.18.feed_forward.experts.5.key.weight": "model-00006-of-00016.safetensors", + "model.blocks.18.feed_forward.experts.5.value.weight": "model-00006-of-00016.safetensors", + "model.blocks.18.feed_forward.experts.6.key.weight": "model-00006-of-00016.safetensors", + "model.blocks.18.feed_forward.experts.6.value.weight": "model-00006-of-00016.safetensors", + "model.blocks.18.feed_forward.experts.7.key.weight": "model-00006-of-00016.safetensors", + "model.blocks.18.feed_forward.experts.7.value.weight": "model-00006-of-00016.safetensors", + "model.blocks.18.feed_forward.receptance.weight": "model-00006-of-00016.safetensors", + "model.blocks.18.feed_forward.shared_expert.key.weight": "model-00006-of-00016.safetensors", + "model.blocks.18.feed_forward.shared_expert.value.weight": "model-00006-of-00016.safetensors", + "model.blocks.18.feed_forward.time_maa_k": "model-00006-of-00016.safetensors", + "model.blocks.18.feed_forward.time_maa_r": "model-00006-of-00016.safetensors", + "model.blocks.18.ln1.bias": "model-00006-of-00016.safetensors", + "model.blocks.18.ln1.weight": "model-00006-of-00016.safetensors", + "model.blocks.18.ln2.bias": "model-00006-of-00016.safetensors", + "model.blocks.18.ln2.weight": "model-00006-of-00016.safetensors", + "model.blocks.19.attention.gate.weight": "model-00006-of-00016.safetensors", + "model.blocks.19.attention.key.weight": "model-00006-of-00016.safetensors", + "model.blocks.19.attention.ln_x.bias": "model-00006-of-00016.safetensors", + "model.blocks.19.attention.ln_x.weight": "model-00006-of-00016.safetensors", + "model.blocks.19.attention.output.weight": "model-00006-of-00016.safetensors", + "model.blocks.19.attention.receptance.weight": "model-00006-of-00016.safetensors", + "model.blocks.19.attention.time_decay": "model-00006-of-00016.safetensors", + "model.blocks.19.attention.time_decay_w1": "model-00006-of-00016.safetensors", + "model.blocks.19.attention.time_decay_w2": "model-00006-of-00016.safetensors", + "model.blocks.19.attention.time_faaaa": "model-00006-of-00016.safetensors", + "model.blocks.19.attention.time_maa_g": "model-00006-of-00016.safetensors", + "model.blocks.19.attention.time_maa_k": "model-00006-of-00016.safetensors", + "model.blocks.19.attention.time_maa_r": "model-00006-of-00016.safetensors", + "model.blocks.19.attention.time_maa_v": "model-00006-of-00016.safetensors", + "model.blocks.19.attention.time_maa_w": "model-00006-of-00016.safetensors", + "model.blocks.19.attention.time_maa_w1": "model-00006-of-00016.safetensors", + "model.blocks.19.attention.time_maa_w2": "model-00006-of-00016.safetensors", + "model.blocks.19.attention.time_maa_x": "model-00006-of-00016.safetensors", + "model.blocks.19.attention.value.weight": "model-00006-of-00016.safetensors", + "model.blocks.19.feed_forward.experts.0.key.weight": "model-00006-of-00016.safetensors", + "model.blocks.19.feed_forward.experts.0.value.weight": "model-00006-of-00016.safetensors", + "model.blocks.19.feed_forward.experts.1.key.weight": "model-00006-of-00016.safetensors", + "model.blocks.19.feed_forward.experts.1.value.weight": "model-00006-of-00016.safetensors", + "model.blocks.19.feed_forward.experts.2.key.weight": "model-00006-of-00016.safetensors", + "model.blocks.19.feed_forward.experts.2.value.weight": "model-00006-of-00016.safetensors", + "model.blocks.19.feed_forward.experts.3.key.weight": "model-00006-of-00016.safetensors", + "model.blocks.19.feed_forward.experts.3.value.weight": "model-00006-of-00016.safetensors", + "model.blocks.19.feed_forward.experts.4.key.weight": "model-00006-of-00016.safetensors", + "model.blocks.19.feed_forward.experts.4.value.weight": "model-00006-of-00016.safetensors", + "model.blocks.19.feed_forward.experts.5.key.weight": "model-00006-of-00016.safetensors", + "model.blocks.19.feed_forward.experts.5.value.weight": "model-00006-of-00016.safetensors", + "model.blocks.19.feed_forward.experts.6.key.weight": "model-00006-of-00016.safetensors", + "model.blocks.19.feed_forward.experts.6.value.weight": "model-00006-of-00016.safetensors", + "model.blocks.19.feed_forward.experts.7.key.weight": "model-00006-of-00016.safetensors", + "model.blocks.19.feed_forward.experts.7.value.weight": "model-00006-of-00016.safetensors", + "model.blocks.19.feed_forward.receptance.weight": "model-00006-of-00016.safetensors", + "model.blocks.19.feed_forward.shared_expert.key.weight": "model-00006-of-00016.safetensors", + "model.blocks.19.feed_forward.shared_expert.value.weight": "model-00006-of-00016.safetensors", + "model.blocks.19.feed_forward.time_maa_k": "model-00006-of-00016.safetensors", + "model.blocks.19.feed_forward.time_maa_r": "model-00006-of-00016.safetensors", + "model.blocks.19.ln1.bias": "model-00006-of-00016.safetensors", + "model.blocks.19.ln1.weight": "model-00006-of-00016.safetensors", + "model.blocks.19.ln2.bias": "model-00006-of-00016.safetensors", + "model.blocks.19.ln2.weight": "model-00006-of-00016.safetensors", + "model.blocks.2.attention.gate.weight": "model-00006-of-00016.safetensors", + "model.blocks.2.attention.key.weight": "model-00006-of-00016.safetensors", + "model.blocks.2.attention.ln_x.bias": "model-00006-of-00016.safetensors", + "model.blocks.2.attention.ln_x.weight": "model-00006-of-00016.safetensors", + "model.blocks.2.attention.output.weight": "model-00006-of-00016.safetensors", + "model.blocks.2.attention.receptance.weight": "model-00006-of-00016.safetensors", + "model.blocks.2.attention.time_decay": "model-00006-of-00016.safetensors", + "model.blocks.2.attention.time_decay_w1": "model-00006-of-00016.safetensors", + "model.blocks.2.attention.time_decay_w2": "model-00006-of-00016.safetensors", + "model.blocks.2.attention.time_faaaa": "model-00006-of-00016.safetensors", + "model.blocks.2.attention.time_maa_g": "model-00006-of-00016.safetensors", + "model.blocks.2.attention.time_maa_k": "model-00006-of-00016.safetensors", + "model.blocks.2.attention.time_maa_r": "model-00006-of-00016.safetensors", + "model.blocks.2.attention.time_maa_v": "model-00006-of-00016.safetensors", + "model.blocks.2.attention.time_maa_w": "model-00006-of-00016.safetensors", + "model.blocks.2.attention.time_maa_w1": "model-00006-of-00016.safetensors", + "model.blocks.2.attention.time_maa_w2": "model-00006-of-00016.safetensors", + "model.blocks.2.attention.time_maa_x": "model-00006-of-00016.safetensors", + "model.blocks.2.attention.value.weight": "model-00006-of-00016.safetensors", + "model.blocks.2.feed_forward.experts.0.key.weight": "model-00006-of-00016.safetensors", + "model.blocks.2.feed_forward.experts.0.value.weight": "model-00006-of-00016.safetensors", + "model.blocks.2.feed_forward.experts.1.key.weight": "model-00006-of-00016.safetensors", + "model.blocks.2.feed_forward.experts.1.value.weight": "model-00006-of-00016.safetensors", + "model.blocks.2.feed_forward.experts.2.key.weight": "model-00006-of-00016.safetensors", + "model.blocks.2.feed_forward.experts.2.value.weight": "model-00006-of-00016.safetensors", + "model.blocks.2.feed_forward.experts.3.key.weight": "model-00006-of-00016.safetensors", + "model.blocks.2.feed_forward.experts.3.value.weight": "model-00006-of-00016.safetensors", + "model.blocks.2.feed_forward.experts.4.key.weight": "model-00006-of-00016.safetensors", + "model.blocks.2.feed_forward.experts.4.value.weight": "model-00006-of-00016.safetensors", + "model.blocks.2.feed_forward.experts.5.key.weight": "model-00006-of-00016.safetensors", + "model.blocks.2.feed_forward.experts.5.value.weight": "model-00007-of-00016.safetensors", + "model.blocks.2.feed_forward.experts.6.key.weight": "model-00007-of-00016.safetensors", + "model.blocks.2.feed_forward.experts.6.value.weight": "model-00007-of-00016.safetensors", + "model.blocks.2.feed_forward.experts.7.key.weight": "model-00007-of-00016.safetensors", + "model.blocks.2.feed_forward.experts.7.value.weight": "model-00007-of-00016.safetensors", + "model.blocks.2.feed_forward.receptance.weight": "model-00007-of-00016.safetensors", + "model.blocks.2.feed_forward.shared_expert.key.weight": "model-00007-of-00016.safetensors", + "model.blocks.2.feed_forward.shared_expert.value.weight": "model-00007-of-00016.safetensors", + "model.blocks.2.feed_forward.time_maa_k": "model-00007-of-00016.safetensors", + "model.blocks.2.feed_forward.time_maa_r": "model-00007-of-00016.safetensors", + "model.blocks.2.ln1.bias": "model-00007-of-00016.safetensors", + "model.blocks.2.ln1.weight": "model-00007-of-00016.safetensors", + "model.blocks.2.ln2.bias": "model-00007-of-00016.safetensors", + "model.blocks.2.ln2.weight": "model-00007-of-00016.safetensors", + "model.blocks.20.attention.gate.weight": "model-00007-of-00016.safetensors", + "model.blocks.20.attention.key.weight": "model-00007-of-00016.safetensors", + "model.blocks.20.attention.ln_x.bias": "model-00007-of-00016.safetensors", + "model.blocks.20.attention.ln_x.weight": "model-00007-of-00016.safetensors", + "model.blocks.20.attention.output.weight": "model-00007-of-00016.safetensors", + "model.blocks.20.attention.receptance.weight": "model-00007-of-00016.safetensors", + "model.blocks.20.attention.time_decay": "model-00007-of-00016.safetensors", + "model.blocks.20.attention.time_decay_w1": "model-00007-of-00016.safetensors", + "model.blocks.20.attention.time_decay_w2": "model-00007-of-00016.safetensors", + "model.blocks.20.attention.time_faaaa": "model-00007-of-00016.safetensors", + "model.blocks.20.attention.time_maa_g": "model-00007-of-00016.safetensors", + "model.blocks.20.attention.time_maa_k": "model-00007-of-00016.safetensors", + "model.blocks.20.attention.time_maa_r": "model-00007-of-00016.safetensors", + "model.blocks.20.attention.time_maa_v": "model-00007-of-00016.safetensors", + "model.blocks.20.attention.time_maa_w": "model-00007-of-00016.safetensors", + "model.blocks.20.attention.time_maa_w1": "model-00007-of-00016.safetensors", + "model.blocks.20.attention.time_maa_w2": "model-00007-of-00016.safetensors", + "model.blocks.20.attention.time_maa_x": "model-00007-of-00016.safetensors", + "model.blocks.20.attention.value.weight": "model-00007-of-00016.safetensors", + "model.blocks.20.feed_forward.experts.0.key.weight": "model-00007-of-00016.safetensors", + "model.blocks.20.feed_forward.experts.0.value.weight": "model-00007-of-00016.safetensors", + "model.blocks.20.feed_forward.experts.1.key.weight": "model-00007-of-00016.safetensors", + "model.blocks.20.feed_forward.experts.1.value.weight": "model-00007-of-00016.safetensors", + "model.blocks.20.feed_forward.experts.2.key.weight": "model-00007-of-00016.safetensors", + "model.blocks.20.feed_forward.experts.2.value.weight": "model-00007-of-00016.safetensors", + "model.blocks.20.feed_forward.experts.3.key.weight": "model-00007-of-00016.safetensors", + "model.blocks.20.feed_forward.experts.3.value.weight": "model-00007-of-00016.safetensors", + "model.blocks.20.feed_forward.experts.4.key.weight": "model-00007-of-00016.safetensors", + "model.blocks.20.feed_forward.experts.4.value.weight": "model-00007-of-00016.safetensors", + "model.blocks.20.feed_forward.experts.5.key.weight": "model-00007-of-00016.safetensors", + "model.blocks.20.feed_forward.experts.5.value.weight": "model-00007-of-00016.safetensors", + "model.blocks.20.feed_forward.experts.6.key.weight": "model-00007-of-00016.safetensors", + "model.blocks.20.feed_forward.experts.6.value.weight": "model-00007-of-00016.safetensors", + "model.blocks.20.feed_forward.experts.7.key.weight": "model-00007-of-00016.safetensors", + "model.blocks.20.feed_forward.experts.7.value.weight": "model-00007-of-00016.safetensors", + "model.blocks.20.feed_forward.receptance.weight": "model-00007-of-00016.safetensors", + "model.blocks.20.feed_forward.shared_expert.key.weight": "model-00007-of-00016.safetensors", + "model.blocks.20.feed_forward.shared_expert.value.weight": "model-00007-of-00016.safetensors", + "model.blocks.20.feed_forward.time_maa_k": "model-00007-of-00016.safetensors", + "model.blocks.20.feed_forward.time_maa_r": "model-00007-of-00016.safetensors", + "model.blocks.20.ln1.bias": "model-00007-of-00016.safetensors", + "model.blocks.20.ln1.weight": "model-00007-of-00016.safetensors", + "model.blocks.20.ln2.bias": "model-00007-of-00016.safetensors", + "model.blocks.20.ln2.weight": "model-00007-of-00016.safetensors", + "model.blocks.21.attention.gate.weight": "model-00007-of-00016.safetensors", + "model.blocks.21.attention.key.weight": "model-00007-of-00016.safetensors", + "model.blocks.21.attention.ln_x.bias": "model-00007-of-00016.safetensors", + "model.blocks.21.attention.ln_x.weight": "model-00007-of-00016.safetensors", + "model.blocks.21.attention.output.weight": "model-00007-of-00016.safetensors", + "model.blocks.21.attention.receptance.weight": "model-00007-of-00016.safetensors", + "model.blocks.21.attention.time_decay": "model-00007-of-00016.safetensors", + "model.blocks.21.attention.time_decay_w1": "model-00007-of-00016.safetensors", + "model.blocks.21.attention.time_decay_w2": "model-00007-of-00016.safetensors", + "model.blocks.21.attention.time_faaaa": "model-00007-of-00016.safetensors", + "model.blocks.21.attention.time_maa_g": "model-00007-of-00016.safetensors", + "model.blocks.21.attention.time_maa_k": "model-00007-of-00016.safetensors", + "model.blocks.21.attention.time_maa_r": "model-00007-of-00016.safetensors", + "model.blocks.21.attention.time_maa_v": "model-00007-of-00016.safetensors", + "model.blocks.21.attention.time_maa_w": "model-00007-of-00016.safetensors", + "model.blocks.21.attention.time_maa_w1": "model-00007-of-00016.safetensors", + "model.blocks.21.attention.time_maa_w2": "model-00007-of-00016.safetensors", + "model.blocks.21.attention.time_maa_x": "model-00007-of-00016.safetensors", + "model.blocks.21.attention.value.weight": "model-00007-of-00016.safetensors", + "model.blocks.21.feed_forward.experts.0.key.weight": "model-00007-of-00016.safetensors", + "model.blocks.21.feed_forward.experts.0.value.weight": "model-00007-of-00016.safetensors", + "model.blocks.21.feed_forward.experts.1.key.weight": "model-00007-of-00016.safetensors", + "model.blocks.21.feed_forward.experts.1.value.weight": "model-00007-of-00016.safetensors", + "model.blocks.21.feed_forward.experts.2.key.weight": "model-00007-of-00016.safetensors", + "model.blocks.21.feed_forward.experts.2.value.weight": "model-00007-of-00016.safetensors", + "model.blocks.21.feed_forward.experts.3.key.weight": "model-00007-of-00016.safetensors", + "model.blocks.21.feed_forward.experts.3.value.weight": "model-00007-of-00016.safetensors", + "model.blocks.21.feed_forward.experts.4.key.weight": "model-00007-of-00016.safetensors", + "model.blocks.21.feed_forward.experts.4.value.weight": "model-00007-of-00016.safetensors", + "model.blocks.21.feed_forward.experts.5.key.weight": "model-00007-of-00016.safetensors", + "model.blocks.21.feed_forward.experts.5.value.weight": "model-00007-of-00016.safetensors", + "model.blocks.21.feed_forward.experts.6.key.weight": "model-00007-of-00016.safetensors", + "model.blocks.21.feed_forward.experts.6.value.weight": "model-00007-of-00016.safetensors", + "model.blocks.21.feed_forward.experts.7.key.weight": "model-00008-of-00016.safetensors", + "model.blocks.21.feed_forward.experts.7.value.weight": "model-00008-of-00016.safetensors", + "model.blocks.21.feed_forward.receptance.weight": "model-00008-of-00016.safetensors", + "model.blocks.21.feed_forward.shared_expert.key.weight": "model-00008-of-00016.safetensors", + "model.blocks.21.feed_forward.shared_expert.value.weight": "model-00008-of-00016.safetensors", + "model.blocks.21.feed_forward.time_maa_k": "model-00008-of-00016.safetensors", + "model.blocks.21.feed_forward.time_maa_r": "model-00008-of-00016.safetensors", + "model.blocks.21.ln1.bias": "model-00008-of-00016.safetensors", + "model.blocks.21.ln1.weight": "model-00008-of-00016.safetensors", + "model.blocks.21.ln2.bias": "model-00008-of-00016.safetensors", + "model.blocks.21.ln2.weight": "model-00008-of-00016.safetensors", + "model.blocks.22.attention.gate.weight": "model-00008-of-00016.safetensors", + "model.blocks.22.attention.key.weight": "model-00008-of-00016.safetensors", + "model.blocks.22.attention.ln_x.bias": "model-00008-of-00016.safetensors", + "model.blocks.22.attention.ln_x.weight": "model-00008-of-00016.safetensors", + "model.blocks.22.attention.output.weight": "model-00008-of-00016.safetensors", + "model.blocks.22.attention.receptance.weight": "model-00008-of-00016.safetensors", + "model.blocks.22.attention.time_decay": "model-00008-of-00016.safetensors", + "model.blocks.22.attention.time_decay_w1": "model-00008-of-00016.safetensors", + "model.blocks.22.attention.time_decay_w2": "model-00008-of-00016.safetensors", + "model.blocks.22.attention.time_faaaa": "model-00008-of-00016.safetensors", + "model.blocks.22.attention.time_maa_g": "model-00008-of-00016.safetensors", + "model.blocks.22.attention.time_maa_k": "model-00008-of-00016.safetensors", + "model.blocks.22.attention.time_maa_r": "model-00008-of-00016.safetensors", + "model.blocks.22.attention.time_maa_v": "model-00008-of-00016.safetensors", + "model.blocks.22.attention.time_maa_w": "model-00008-of-00016.safetensors", + "model.blocks.22.attention.time_maa_w1": "model-00008-of-00016.safetensors", + "model.blocks.22.attention.time_maa_w2": "model-00008-of-00016.safetensors", + "model.blocks.22.attention.time_maa_x": "model-00008-of-00016.safetensors", + "model.blocks.22.attention.value.weight": "model-00008-of-00016.safetensors", + "model.blocks.22.feed_forward.experts.0.key.weight": "model-00008-of-00016.safetensors", + "model.blocks.22.feed_forward.experts.0.value.weight": "model-00008-of-00016.safetensors", + "model.blocks.22.feed_forward.experts.1.key.weight": "model-00008-of-00016.safetensors", + "model.blocks.22.feed_forward.experts.1.value.weight": "model-00008-of-00016.safetensors", + "model.blocks.22.feed_forward.experts.2.key.weight": "model-00008-of-00016.safetensors", + "model.blocks.22.feed_forward.experts.2.value.weight": "model-00008-of-00016.safetensors", + "model.blocks.22.feed_forward.experts.3.key.weight": "model-00008-of-00016.safetensors", + "model.blocks.22.feed_forward.experts.3.value.weight": "model-00008-of-00016.safetensors", + "model.blocks.22.feed_forward.experts.4.key.weight": "model-00008-of-00016.safetensors", + "model.blocks.22.feed_forward.experts.4.value.weight": "model-00008-of-00016.safetensors", + "model.blocks.22.feed_forward.experts.5.key.weight": "model-00008-of-00016.safetensors", + "model.blocks.22.feed_forward.experts.5.value.weight": "model-00008-of-00016.safetensors", + "model.blocks.22.feed_forward.experts.6.key.weight": "model-00008-of-00016.safetensors", + "model.blocks.22.feed_forward.experts.6.value.weight": "model-00008-of-00016.safetensors", + "model.blocks.22.feed_forward.experts.7.key.weight": "model-00008-of-00016.safetensors", + "model.blocks.22.feed_forward.experts.7.value.weight": "model-00008-of-00016.safetensors", + "model.blocks.22.feed_forward.receptance.weight": "model-00008-of-00016.safetensors", + "model.blocks.22.feed_forward.shared_expert.key.weight": "model-00008-of-00016.safetensors", + "model.blocks.22.feed_forward.shared_expert.value.weight": "model-00008-of-00016.safetensors", + "model.blocks.22.feed_forward.time_maa_k": "model-00008-of-00016.safetensors", + "model.blocks.22.feed_forward.time_maa_r": "model-00008-of-00016.safetensors", + "model.blocks.22.ln1.bias": "model-00008-of-00016.safetensors", + "model.blocks.22.ln1.weight": "model-00008-of-00016.safetensors", + "model.blocks.22.ln2.bias": "model-00008-of-00016.safetensors", + "model.blocks.22.ln2.weight": "model-00008-of-00016.safetensors", + "model.blocks.23.attention.gate.weight": "model-00008-of-00016.safetensors", + "model.blocks.23.attention.key.weight": "model-00008-of-00016.safetensors", + "model.blocks.23.attention.ln_x.bias": "model-00008-of-00016.safetensors", + "model.blocks.23.attention.ln_x.weight": "model-00008-of-00016.safetensors", + "model.blocks.23.attention.output.weight": "model-00008-of-00016.safetensors", + "model.blocks.23.attention.receptance.weight": "model-00008-of-00016.safetensors", + "model.blocks.23.attention.time_decay": "model-00008-of-00016.safetensors", + "model.blocks.23.attention.time_decay_w1": "model-00008-of-00016.safetensors", + "model.blocks.23.attention.time_decay_w2": "model-00008-of-00016.safetensors", + "model.blocks.23.attention.time_faaaa": "model-00008-of-00016.safetensors", + "model.blocks.23.attention.time_maa_g": "model-00008-of-00016.safetensors", + "model.blocks.23.attention.time_maa_k": "model-00008-of-00016.safetensors", + "model.blocks.23.attention.time_maa_r": "model-00008-of-00016.safetensors", + "model.blocks.23.attention.time_maa_v": "model-00008-of-00016.safetensors", + "model.blocks.23.attention.time_maa_w": "model-00008-of-00016.safetensors", + "model.blocks.23.attention.time_maa_w1": "model-00008-of-00016.safetensors", + "model.blocks.23.attention.time_maa_w2": "model-00008-of-00016.safetensors", + "model.blocks.23.attention.time_maa_x": "model-00008-of-00016.safetensors", + "model.blocks.23.attention.value.weight": "model-00008-of-00016.safetensors", + "model.blocks.23.feed_forward.experts.0.key.weight": "model-00008-of-00016.safetensors", + "model.blocks.23.feed_forward.experts.0.value.weight": "model-00008-of-00016.safetensors", + "model.blocks.23.feed_forward.experts.1.key.weight": "model-00008-of-00016.safetensors", + "model.blocks.23.feed_forward.experts.1.value.weight": "model-00008-of-00016.safetensors", + "model.blocks.23.feed_forward.experts.2.key.weight": "model-00008-of-00016.safetensors", + "model.blocks.23.feed_forward.experts.2.value.weight": "model-00008-of-00016.safetensors", + "model.blocks.23.feed_forward.experts.3.key.weight": "model-00008-of-00016.safetensors", + "model.blocks.23.feed_forward.experts.3.value.weight": "model-00008-of-00016.safetensors", + "model.blocks.23.feed_forward.experts.4.key.weight": "model-00008-of-00016.safetensors", + "model.blocks.23.feed_forward.experts.4.value.weight": "model-00008-of-00016.safetensors", + "model.blocks.23.feed_forward.experts.5.key.weight": "model-00008-of-00016.safetensors", + "model.blocks.23.feed_forward.experts.5.value.weight": "model-00008-of-00016.safetensors", + "model.blocks.23.feed_forward.experts.6.key.weight": "model-00008-of-00016.safetensors", + "model.blocks.23.feed_forward.experts.6.value.weight": "model-00008-of-00016.safetensors", + "model.blocks.23.feed_forward.experts.7.key.weight": "model-00008-of-00016.safetensors", + "model.blocks.23.feed_forward.experts.7.value.weight": "model-00008-of-00016.safetensors", + "model.blocks.23.feed_forward.receptance.weight": "model-00008-of-00016.safetensors", + "model.blocks.23.feed_forward.shared_expert.key.weight": "model-00009-of-00016.safetensors", + "model.blocks.23.feed_forward.shared_expert.value.weight": "model-00009-of-00016.safetensors", + "model.blocks.23.feed_forward.time_maa_k": "model-00009-of-00016.safetensors", + "model.blocks.23.feed_forward.time_maa_r": "model-00009-of-00016.safetensors", + "model.blocks.23.ln1.bias": "model-00009-of-00016.safetensors", + "model.blocks.23.ln1.weight": "model-00009-of-00016.safetensors", + "model.blocks.23.ln2.bias": "model-00009-of-00016.safetensors", + "model.blocks.23.ln2.weight": "model-00009-of-00016.safetensors", + "model.blocks.24.attention.gate.weight": "model-00009-of-00016.safetensors", + "model.blocks.24.attention.key.weight": "model-00009-of-00016.safetensors", + "model.blocks.24.attention.ln_x.bias": "model-00009-of-00016.safetensors", + "model.blocks.24.attention.ln_x.weight": "model-00009-of-00016.safetensors", + "model.blocks.24.attention.output.weight": "model-00009-of-00016.safetensors", + "model.blocks.24.attention.receptance.weight": "model-00009-of-00016.safetensors", + "model.blocks.24.attention.time_decay": "model-00009-of-00016.safetensors", + "model.blocks.24.attention.time_decay_w1": "model-00009-of-00016.safetensors", + "model.blocks.24.attention.time_decay_w2": "model-00009-of-00016.safetensors", + "model.blocks.24.attention.time_faaaa": "model-00009-of-00016.safetensors", + "model.blocks.24.attention.time_maa_g": "model-00009-of-00016.safetensors", + "model.blocks.24.attention.time_maa_k": "model-00009-of-00016.safetensors", + "model.blocks.24.attention.time_maa_r": "model-00009-of-00016.safetensors", + "model.blocks.24.attention.time_maa_v": "model-00009-of-00016.safetensors", + "model.blocks.24.attention.time_maa_w": "model-00009-of-00016.safetensors", + "model.blocks.24.attention.time_maa_w1": "model-00009-of-00016.safetensors", + "model.blocks.24.attention.time_maa_w2": "model-00009-of-00016.safetensors", + "model.blocks.24.attention.time_maa_x": "model-00009-of-00016.safetensors", + "model.blocks.24.attention.value.weight": "model-00009-of-00016.safetensors", + "model.blocks.24.feed_forward.experts.0.key.weight": "model-00009-of-00016.safetensors", + "model.blocks.24.feed_forward.experts.0.value.weight": "model-00009-of-00016.safetensors", + "model.blocks.24.feed_forward.experts.1.key.weight": "model-00009-of-00016.safetensors", + "model.blocks.24.feed_forward.experts.1.value.weight": "model-00009-of-00016.safetensors", + "model.blocks.24.feed_forward.experts.2.key.weight": "model-00009-of-00016.safetensors", + "model.blocks.24.feed_forward.experts.2.value.weight": "model-00009-of-00016.safetensors", + "model.blocks.24.feed_forward.experts.3.key.weight": "model-00009-of-00016.safetensors", + "model.blocks.24.feed_forward.experts.3.value.weight": "model-00009-of-00016.safetensors", + "model.blocks.24.feed_forward.experts.4.key.weight": "model-00009-of-00016.safetensors", + "model.blocks.24.feed_forward.experts.4.value.weight": "model-00009-of-00016.safetensors", + "model.blocks.24.feed_forward.experts.5.key.weight": "model-00009-of-00016.safetensors", + "model.blocks.24.feed_forward.experts.5.value.weight": "model-00009-of-00016.safetensors", + "model.blocks.24.feed_forward.experts.6.key.weight": "model-00009-of-00016.safetensors", + "model.blocks.24.feed_forward.experts.6.value.weight": "model-00009-of-00016.safetensors", + "model.blocks.24.feed_forward.experts.7.key.weight": "model-00009-of-00016.safetensors", + "model.blocks.24.feed_forward.experts.7.value.weight": "model-00009-of-00016.safetensors", + "model.blocks.24.feed_forward.receptance.weight": "model-00009-of-00016.safetensors", + "model.blocks.24.feed_forward.shared_expert.key.weight": "model-00009-of-00016.safetensors", + "model.blocks.24.feed_forward.shared_expert.value.weight": "model-00009-of-00016.safetensors", + "model.blocks.24.feed_forward.time_maa_k": "model-00009-of-00016.safetensors", + "model.blocks.24.feed_forward.time_maa_r": "model-00009-of-00016.safetensors", + "model.blocks.24.ln1.bias": "model-00009-of-00016.safetensors", + "model.blocks.24.ln1.weight": "model-00009-of-00016.safetensors", + "model.blocks.24.ln2.bias": "model-00009-of-00016.safetensors", + "model.blocks.24.ln2.weight": "model-00009-of-00016.safetensors", + "model.blocks.25.attention.gate.weight": "model-00009-of-00016.safetensors", + "model.blocks.25.attention.key.weight": "model-00009-of-00016.safetensors", + "model.blocks.25.attention.ln_x.bias": "model-00009-of-00016.safetensors", + "model.blocks.25.attention.ln_x.weight": "model-00009-of-00016.safetensors", + "model.blocks.25.attention.output.weight": "model-00009-of-00016.safetensors", + "model.blocks.25.attention.receptance.weight": "model-00009-of-00016.safetensors", + "model.blocks.25.attention.time_decay": "model-00009-of-00016.safetensors", + "model.blocks.25.attention.time_decay_w1": "model-00009-of-00016.safetensors", + "model.blocks.25.attention.time_decay_w2": "model-00009-of-00016.safetensors", + "model.blocks.25.attention.time_faaaa": "model-00009-of-00016.safetensors", + "model.blocks.25.attention.time_maa_g": "model-00009-of-00016.safetensors", + "model.blocks.25.attention.time_maa_k": "model-00009-of-00016.safetensors", + "model.blocks.25.attention.time_maa_r": "model-00009-of-00016.safetensors", + "model.blocks.25.attention.time_maa_v": "model-00009-of-00016.safetensors", + "model.blocks.25.attention.time_maa_w": "model-00009-of-00016.safetensors", + "model.blocks.25.attention.time_maa_w1": "model-00009-of-00016.safetensors", + "model.blocks.25.attention.time_maa_w2": "model-00009-of-00016.safetensors", + "model.blocks.25.attention.time_maa_x": "model-00009-of-00016.safetensors", + "model.blocks.25.attention.value.weight": "model-00009-of-00016.safetensors", + "model.blocks.25.feed_forward.experts.0.key.weight": "model-00009-of-00016.safetensors", + "model.blocks.25.feed_forward.experts.0.value.weight": "model-00009-of-00016.safetensors", + "model.blocks.25.feed_forward.experts.1.key.weight": "model-00009-of-00016.safetensors", + "model.blocks.25.feed_forward.experts.1.value.weight": "model-00009-of-00016.safetensors", + "model.blocks.25.feed_forward.experts.2.key.weight": "model-00009-of-00016.safetensors", + "model.blocks.25.feed_forward.experts.2.value.weight": "model-00009-of-00016.safetensors", + "model.blocks.25.feed_forward.experts.3.key.weight": "model-00009-of-00016.safetensors", + "model.blocks.25.feed_forward.experts.3.value.weight": "model-00009-of-00016.safetensors", + "model.blocks.25.feed_forward.experts.4.key.weight": "model-00009-of-00016.safetensors", + "model.blocks.25.feed_forward.experts.4.value.weight": "model-00009-of-00016.safetensors", + "model.blocks.25.feed_forward.experts.5.key.weight": "model-00009-of-00016.safetensors", + "model.blocks.25.feed_forward.experts.5.value.weight": "model-00009-of-00016.safetensors", + "model.blocks.25.feed_forward.experts.6.key.weight": "model-00009-of-00016.safetensors", + "model.blocks.25.feed_forward.experts.6.value.weight": "model-00009-of-00016.safetensors", + "model.blocks.25.feed_forward.experts.7.key.weight": "model-00009-of-00016.safetensors", + "model.blocks.25.feed_forward.experts.7.value.weight": "model-00009-of-00016.safetensors", + "model.blocks.25.feed_forward.receptance.weight": "model-00009-of-00016.safetensors", + "model.blocks.25.feed_forward.shared_expert.key.weight": "model-00009-of-00016.safetensors", + "model.blocks.25.feed_forward.shared_expert.value.weight": "model-00009-of-00016.safetensors", + "model.blocks.25.feed_forward.time_maa_k": "model-00009-of-00016.safetensors", + "model.blocks.25.feed_forward.time_maa_r": "model-00009-of-00016.safetensors", + "model.blocks.25.ln1.bias": "model-00009-of-00016.safetensors", + "model.blocks.25.ln1.weight": "model-00009-of-00016.safetensors", + "model.blocks.25.ln2.bias": "model-00009-of-00016.safetensors", + "model.blocks.25.ln2.weight": "model-00009-of-00016.safetensors", + "model.blocks.26.attention.gate.weight": "model-00009-of-00016.safetensors", + "model.blocks.26.attention.key.weight": "model-00009-of-00016.safetensors", + "model.blocks.26.attention.ln_x.bias": "model-00009-of-00016.safetensors", + "model.blocks.26.attention.ln_x.weight": "model-00009-of-00016.safetensors", + "model.blocks.26.attention.output.weight": "model-00009-of-00016.safetensors", + "model.blocks.26.attention.receptance.weight": "model-00010-of-00016.safetensors", + "model.blocks.26.attention.time_decay": "model-00010-of-00016.safetensors", + "model.blocks.26.attention.time_decay_w1": "model-00010-of-00016.safetensors", + "model.blocks.26.attention.time_decay_w2": "model-00010-of-00016.safetensors", + "model.blocks.26.attention.time_faaaa": "model-00010-of-00016.safetensors", + "model.blocks.26.attention.time_maa_g": "model-00010-of-00016.safetensors", + "model.blocks.26.attention.time_maa_k": "model-00010-of-00016.safetensors", + "model.blocks.26.attention.time_maa_r": "model-00010-of-00016.safetensors", + "model.blocks.26.attention.time_maa_v": "model-00010-of-00016.safetensors", + "model.blocks.26.attention.time_maa_w": "model-00010-of-00016.safetensors", + "model.blocks.26.attention.time_maa_w1": "model-00010-of-00016.safetensors", + "model.blocks.26.attention.time_maa_w2": "model-00010-of-00016.safetensors", + "model.blocks.26.attention.time_maa_x": "model-00010-of-00016.safetensors", + "model.blocks.26.attention.value.weight": "model-00010-of-00016.safetensors", + "model.blocks.26.feed_forward.experts.0.key.weight": "model-00010-of-00016.safetensors", + "model.blocks.26.feed_forward.experts.0.value.weight": "model-00010-of-00016.safetensors", + "model.blocks.26.feed_forward.experts.1.key.weight": "model-00010-of-00016.safetensors", + "model.blocks.26.feed_forward.experts.1.value.weight": "model-00010-of-00016.safetensors", + "model.blocks.26.feed_forward.experts.2.key.weight": "model-00010-of-00016.safetensors", + "model.blocks.26.feed_forward.experts.2.value.weight": "model-00010-of-00016.safetensors", + "model.blocks.26.feed_forward.experts.3.key.weight": "model-00010-of-00016.safetensors", + "model.blocks.26.feed_forward.experts.3.value.weight": "model-00010-of-00016.safetensors", + "model.blocks.26.feed_forward.experts.4.key.weight": "model-00010-of-00016.safetensors", + "model.blocks.26.feed_forward.experts.4.value.weight": "model-00010-of-00016.safetensors", + "model.blocks.26.feed_forward.experts.5.key.weight": "model-00010-of-00016.safetensors", + "model.blocks.26.feed_forward.experts.5.value.weight": "model-00010-of-00016.safetensors", + "model.blocks.26.feed_forward.experts.6.key.weight": "model-00010-of-00016.safetensors", + "model.blocks.26.feed_forward.experts.6.value.weight": "model-00010-of-00016.safetensors", + "model.blocks.26.feed_forward.experts.7.key.weight": "model-00010-of-00016.safetensors", + "model.blocks.26.feed_forward.experts.7.value.weight": "model-00010-of-00016.safetensors", + "model.blocks.26.feed_forward.receptance.weight": "model-00010-of-00016.safetensors", + "model.blocks.26.feed_forward.shared_expert.key.weight": "model-00010-of-00016.safetensors", + "model.blocks.26.feed_forward.shared_expert.value.weight": "model-00010-of-00016.safetensors", + "model.blocks.26.feed_forward.time_maa_k": "model-00010-of-00016.safetensors", + "model.blocks.26.feed_forward.time_maa_r": "model-00010-of-00016.safetensors", + "model.blocks.26.ln1.bias": "model-00010-of-00016.safetensors", + "model.blocks.26.ln1.weight": "model-00010-of-00016.safetensors", + "model.blocks.26.ln2.bias": "model-00010-of-00016.safetensors", + "model.blocks.26.ln2.weight": "model-00010-of-00016.safetensors", + "model.blocks.27.attention.gate.weight": "model-00010-of-00016.safetensors", + "model.blocks.27.attention.key.weight": "model-00010-of-00016.safetensors", + "model.blocks.27.attention.ln_x.bias": "model-00010-of-00016.safetensors", + "model.blocks.27.attention.ln_x.weight": "model-00010-of-00016.safetensors", + "model.blocks.27.attention.output.weight": "model-00010-of-00016.safetensors", + "model.blocks.27.attention.receptance.weight": "model-00010-of-00016.safetensors", + "model.blocks.27.attention.time_decay": "model-00010-of-00016.safetensors", + "model.blocks.27.attention.time_decay_w1": "model-00010-of-00016.safetensors", + "model.blocks.27.attention.time_decay_w2": "model-00010-of-00016.safetensors", + "model.blocks.27.attention.time_faaaa": "model-00010-of-00016.safetensors", + "model.blocks.27.attention.time_maa_g": "model-00010-of-00016.safetensors", + "model.blocks.27.attention.time_maa_k": "model-00010-of-00016.safetensors", + "model.blocks.27.attention.time_maa_r": "model-00010-of-00016.safetensors", + "model.blocks.27.attention.time_maa_v": "model-00010-of-00016.safetensors", + "model.blocks.27.attention.time_maa_w": "model-00010-of-00016.safetensors", + "model.blocks.27.attention.time_maa_w1": "model-00010-of-00016.safetensors", + "model.blocks.27.attention.time_maa_w2": "model-00010-of-00016.safetensors", + "model.blocks.27.attention.time_maa_x": "model-00010-of-00016.safetensors", + "model.blocks.27.attention.value.weight": "model-00010-of-00016.safetensors", + "model.blocks.27.feed_forward.experts.0.key.weight": "model-00010-of-00016.safetensors", + "model.blocks.27.feed_forward.experts.0.value.weight": "model-00010-of-00016.safetensors", + "model.blocks.27.feed_forward.experts.1.key.weight": "model-00010-of-00016.safetensors", + "model.blocks.27.feed_forward.experts.1.value.weight": "model-00010-of-00016.safetensors", + "model.blocks.27.feed_forward.experts.2.key.weight": "model-00010-of-00016.safetensors", + "model.blocks.27.feed_forward.experts.2.value.weight": "model-00010-of-00016.safetensors", + "model.blocks.27.feed_forward.experts.3.key.weight": "model-00010-of-00016.safetensors", + "model.blocks.27.feed_forward.experts.3.value.weight": "model-00010-of-00016.safetensors", + "model.blocks.27.feed_forward.experts.4.key.weight": "model-00010-of-00016.safetensors", + "model.blocks.27.feed_forward.experts.4.value.weight": "model-00010-of-00016.safetensors", + "model.blocks.27.feed_forward.experts.5.key.weight": "model-00010-of-00016.safetensors", + "model.blocks.27.feed_forward.experts.5.value.weight": "model-00010-of-00016.safetensors", + "model.blocks.27.feed_forward.experts.6.key.weight": "model-00010-of-00016.safetensors", + "model.blocks.27.feed_forward.experts.6.value.weight": "model-00010-of-00016.safetensors", + "model.blocks.27.feed_forward.experts.7.key.weight": "model-00010-of-00016.safetensors", + "model.blocks.27.feed_forward.experts.7.value.weight": "model-00010-of-00016.safetensors", + "model.blocks.27.feed_forward.receptance.weight": "model-00010-of-00016.safetensors", + "model.blocks.27.feed_forward.shared_expert.key.weight": "model-00010-of-00016.safetensors", + "model.blocks.27.feed_forward.shared_expert.value.weight": "model-00010-of-00016.safetensors", + "model.blocks.27.feed_forward.time_maa_k": "model-00010-of-00016.safetensors", + "model.blocks.27.feed_forward.time_maa_r": "model-00010-of-00016.safetensors", + "model.blocks.27.ln1.bias": "model-00010-of-00016.safetensors", + "model.blocks.27.ln1.weight": "model-00010-of-00016.safetensors", + "model.blocks.27.ln2.bias": "model-00010-of-00016.safetensors", + "model.blocks.27.ln2.weight": "model-00010-of-00016.safetensors", + "model.blocks.28.attention.gate.weight": "model-00010-of-00016.safetensors", + "model.blocks.28.attention.key.weight": "model-00010-of-00016.safetensors", + "model.blocks.28.attention.ln_x.bias": "model-00010-of-00016.safetensors", + "model.blocks.28.attention.ln_x.weight": "model-00010-of-00016.safetensors", + "model.blocks.28.attention.output.weight": "model-00010-of-00016.safetensors", + "model.blocks.28.attention.receptance.weight": "model-00010-of-00016.safetensors", + "model.blocks.28.attention.time_decay": "model-00010-of-00016.safetensors", + "model.blocks.28.attention.time_decay_w1": "model-00010-of-00016.safetensors", + "model.blocks.28.attention.time_decay_w2": "model-00010-of-00016.safetensors", + "model.blocks.28.attention.time_faaaa": "model-00010-of-00016.safetensors", + "model.blocks.28.attention.time_maa_g": "model-00010-of-00016.safetensors", + "model.blocks.28.attention.time_maa_k": "model-00010-of-00016.safetensors", + "model.blocks.28.attention.time_maa_r": "model-00010-of-00016.safetensors", + "model.blocks.28.attention.time_maa_v": "model-00010-of-00016.safetensors", + "model.blocks.28.attention.time_maa_w": "model-00010-of-00016.safetensors", + "model.blocks.28.attention.time_maa_w1": "model-00010-of-00016.safetensors", + "model.blocks.28.attention.time_maa_w2": "model-00010-of-00016.safetensors", + "model.blocks.28.attention.time_maa_x": "model-00010-of-00016.safetensors", + "model.blocks.28.attention.value.weight": "model-00010-of-00016.safetensors", + "model.blocks.28.feed_forward.experts.0.key.weight": "model-00010-of-00016.safetensors", + "model.blocks.28.feed_forward.experts.0.value.weight": "model-00010-of-00016.safetensors", + "model.blocks.28.feed_forward.experts.1.key.weight": "model-00011-of-00016.safetensors", + "model.blocks.28.feed_forward.experts.1.value.weight": "model-00011-of-00016.safetensors", + "model.blocks.28.feed_forward.experts.2.key.weight": "model-00011-of-00016.safetensors", + "model.blocks.28.feed_forward.experts.2.value.weight": "model-00011-of-00016.safetensors", + "model.blocks.28.feed_forward.experts.3.key.weight": "model-00011-of-00016.safetensors", + "model.blocks.28.feed_forward.experts.3.value.weight": "model-00011-of-00016.safetensors", + "model.blocks.28.feed_forward.experts.4.key.weight": "model-00011-of-00016.safetensors", + "model.blocks.28.feed_forward.experts.4.value.weight": "model-00011-of-00016.safetensors", + "model.blocks.28.feed_forward.experts.5.key.weight": "model-00011-of-00016.safetensors", + "model.blocks.28.feed_forward.experts.5.value.weight": "model-00011-of-00016.safetensors", + "model.blocks.28.feed_forward.experts.6.key.weight": "model-00011-of-00016.safetensors", + "model.blocks.28.feed_forward.experts.6.value.weight": "model-00011-of-00016.safetensors", + "model.blocks.28.feed_forward.experts.7.key.weight": "model-00011-of-00016.safetensors", + "model.blocks.28.feed_forward.experts.7.value.weight": "model-00011-of-00016.safetensors", + "model.blocks.28.feed_forward.receptance.weight": "model-00011-of-00016.safetensors", + "model.blocks.28.feed_forward.shared_expert.key.weight": "model-00011-of-00016.safetensors", + "model.blocks.28.feed_forward.shared_expert.value.weight": "model-00011-of-00016.safetensors", + "model.blocks.28.feed_forward.time_maa_k": "model-00011-of-00016.safetensors", + "model.blocks.28.feed_forward.time_maa_r": "model-00011-of-00016.safetensors", + "model.blocks.28.ln1.bias": "model-00011-of-00016.safetensors", + "model.blocks.28.ln1.weight": "model-00011-of-00016.safetensors", + "model.blocks.28.ln2.bias": "model-00011-of-00016.safetensors", + "model.blocks.28.ln2.weight": "model-00011-of-00016.safetensors", + "model.blocks.29.attention.gate.weight": "model-00011-of-00016.safetensors", + "model.blocks.29.attention.key.weight": "model-00011-of-00016.safetensors", + "model.blocks.29.attention.ln_x.bias": "model-00011-of-00016.safetensors", + "model.blocks.29.attention.ln_x.weight": "model-00011-of-00016.safetensors", + "model.blocks.29.attention.output.weight": "model-00011-of-00016.safetensors", + "model.blocks.29.attention.receptance.weight": "model-00011-of-00016.safetensors", + "model.blocks.29.attention.time_decay": "model-00011-of-00016.safetensors", + "model.blocks.29.attention.time_decay_w1": "model-00011-of-00016.safetensors", + "model.blocks.29.attention.time_decay_w2": "model-00011-of-00016.safetensors", + "model.blocks.29.attention.time_faaaa": "model-00011-of-00016.safetensors", + "model.blocks.29.attention.time_maa_g": "model-00011-of-00016.safetensors", + "model.blocks.29.attention.time_maa_k": "model-00011-of-00016.safetensors", + "model.blocks.29.attention.time_maa_r": "model-00011-of-00016.safetensors", + "model.blocks.29.attention.time_maa_v": "model-00011-of-00016.safetensors", + "model.blocks.29.attention.time_maa_w": "model-00011-of-00016.safetensors", + "model.blocks.29.attention.time_maa_w1": "model-00011-of-00016.safetensors", + "model.blocks.29.attention.time_maa_w2": "model-00011-of-00016.safetensors", + "model.blocks.29.attention.time_maa_x": "model-00011-of-00016.safetensors", + "model.blocks.29.attention.value.weight": "model-00011-of-00016.safetensors", + "model.blocks.29.feed_forward.experts.0.key.weight": "model-00011-of-00016.safetensors", + "model.blocks.29.feed_forward.experts.0.value.weight": "model-00011-of-00016.safetensors", + "model.blocks.29.feed_forward.experts.1.key.weight": "model-00011-of-00016.safetensors", + "model.blocks.29.feed_forward.experts.1.value.weight": "model-00011-of-00016.safetensors", + "model.blocks.29.feed_forward.experts.2.key.weight": "model-00011-of-00016.safetensors", + "model.blocks.29.feed_forward.experts.2.value.weight": "model-00011-of-00016.safetensors", + "model.blocks.29.feed_forward.experts.3.key.weight": "model-00011-of-00016.safetensors", + "model.blocks.29.feed_forward.experts.3.value.weight": "model-00011-of-00016.safetensors", + "model.blocks.29.feed_forward.experts.4.key.weight": "model-00011-of-00016.safetensors", + "model.blocks.29.feed_forward.experts.4.value.weight": "model-00011-of-00016.safetensors", + "model.blocks.29.feed_forward.experts.5.key.weight": "model-00011-of-00016.safetensors", + "model.blocks.29.feed_forward.experts.5.value.weight": "model-00011-of-00016.safetensors", + "model.blocks.29.feed_forward.experts.6.key.weight": "model-00011-of-00016.safetensors", + "model.blocks.29.feed_forward.experts.6.value.weight": "model-00011-of-00016.safetensors", + "model.blocks.29.feed_forward.experts.7.key.weight": "model-00011-of-00016.safetensors", + "model.blocks.29.feed_forward.experts.7.value.weight": "model-00011-of-00016.safetensors", + "model.blocks.29.feed_forward.receptance.weight": "model-00011-of-00016.safetensors", + "model.blocks.29.feed_forward.shared_expert.key.weight": "model-00011-of-00016.safetensors", + "model.blocks.29.feed_forward.shared_expert.value.weight": "model-00011-of-00016.safetensors", + "model.blocks.29.feed_forward.time_maa_k": "model-00011-of-00016.safetensors", + "model.blocks.29.feed_forward.time_maa_r": "model-00011-of-00016.safetensors", + "model.blocks.29.ln1.bias": "model-00011-of-00016.safetensors", + "model.blocks.29.ln1.weight": "model-00011-of-00016.safetensors", + "model.blocks.29.ln2.bias": "model-00011-of-00016.safetensors", + "model.blocks.29.ln2.weight": "model-00011-of-00016.safetensors", + "model.blocks.3.attention.gate.weight": "model-00011-of-00016.safetensors", + "model.blocks.3.attention.key.weight": "model-00011-of-00016.safetensors", + "model.blocks.3.attention.ln_x.bias": "model-00011-of-00016.safetensors", + "model.blocks.3.attention.ln_x.weight": "model-00011-of-00016.safetensors", + "model.blocks.3.attention.output.weight": "model-00011-of-00016.safetensors", + "model.blocks.3.attention.receptance.weight": "model-00011-of-00016.safetensors", + "model.blocks.3.attention.time_decay": "model-00011-of-00016.safetensors", + "model.blocks.3.attention.time_decay_w1": "model-00011-of-00016.safetensors", + "model.blocks.3.attention.time_decay_w2": "model-00011-of-00016.safetensors", + "model.blocks.3.attention.time_faaaa": "model-00011-of-00016.safetensors", + "model.blocks.3.attention.time_maa_g": "model-00011-of-00016.safetensors", + "model.blocks.3.attention.time_maa_k": "model-00011-of-00016.safetensors", + "model.blocks.3.attention.time_maa_r": "model-00011-of-00016.safetensors", + "model.blocks.3.attention.time_maa_v": "model-00011-of-00016.safetensors", + "model.blocks.3.attention.time_maa_w": "model-00011-of-00016.safetensors", + "model.blocks.3.attention.time_maa_w1": "model-00011-of-00016.safetensors", + "model.blocks.3.attention.time_maa_w2": "model-00011-of-00016.safetensors", + "model.blocks.3.attention.time_maa_x": "model-00011-of-00016.safetensors", + "model.blocks.3.attention.value.weight": "model-00011-of-00016.safetensors", + "model.blocks.3.feed_forward.experts.0.key.weight": "model-00011-of-00016.safetensors", + "model.blocks.3.feed_forward.experts.0.value.weight": "model-00011-of-00016.safetensors", + "model.blocks.3.feed_forward.experts.1.key.weight": "model-00011-of-00016.safetensors", + "model.blocks.3.feed_forward.experts.1.value.weight": "model-00011-of-00016.safetensors", + "model.blocks.3.feed_forward.experts.2.key.weight": "model-00011-of-00016.safetensors", + "model.blocks.3.feed_forward.experts.2.value.weight": "model-00012-of-00016.safetensors", + "model.blocks.3.feed_forward.experts.3.key.weight": "model-00012-of-00016.safetensors", + "model.blocks.3.feed_forward.experts.3.value.weight": "model-00012-of-00016.safetensors", + "model.blocks.3.feed_forward.experts.4.key.weight": "model-00012-of-00016.safetensors", + "model.blocks.3.feed_forward.experts.4.value.weight": "model-00012-of-00016.safetensors", + "model.blocks.3.feed_forward.experts.5.key.weight": "model-00012-of-00016.safetensors", + "model.blocks.3.feed_forward.experts.5.value.weight": "model-00012-of-00016.safetensors", + "model.blocks.3.feed_forward.experts.6.key.weight": "model-00012-of-00016.safetensors", + "model.blocks.3.feed_forward.experts.6.value.weight": "model-00012-of-00016.safetensors", + "model.blocks.3.feed_forward.experts.7.key.weight": "model-00012-of-00016.safetensors", + "model.blocks.3.feed_forward.experts.7.value.weight": "model-00012-of-00016.safetensors", + "model.blocks.3.feed_forward.receptance.weight": "model-00012-of-00016.safetensors", + "model.blocks.3.feed_forward.shared_expert.key.weight": "model-00012-of-00016.safetensors", + "model.blocks.3.feed_forward.shared_expert.value.weight": "model-00012-of-00016.safetensors", + "model.blocks.3.feed_forward.time_maa_k": "model-00012-of-00016.safetensors", + "model.blocks.3.feed_forward.time_maa_r": "model-00012-of-00016.safetensors", + "model.blocks.3.ln1.bias": "model-00012-of-00016.safetensors", + "model.blocks.3.ln1.weight": "model-00012-of-00016.safetensors", + "model.blocks.3.ln2.bias": "model-00012-of-00016.safetensors", + "model.blocks.3.ln2.weight": "model-00012-of-00016.safetensors", + "model.blocks.30.attention.gate.weight": "model-00012-of-00016.safetensors", + "model.blocks.30.attention.key.weight": "model-00012-of-00016.safetensors", + "model.blocks.30.attention.ln_x.bias": "model-00012-of-00016.safetensors", + "model.blocks.30.attention.ln_x.weight": "model-00012-of-00016.safetensors", + "model.blocks.30.attention.output.weight": "model-00012-of-00016.safetensors", + "model.blocks.30.attention.receptance.weight": "model-00012-of-00016.safetensors", + "model.blocks.30.attention.time_decay": "model-00012-of-00016.safetensors", + "model.blocks.30.attention.time_decay_w1": "model-00012-of-00016.safetensors", + "model.blocks.30.attention.time_decay_w2": "model-00012-of-00016.safetensors", + "model.blocks.30.attention.time_faaaa": "model-00012-of-00016.safetensors", + "model.blocks.30.attention.time_maa_g": "model-00012-of-00016.safetensors", + "model.blocks.30.attention.time_maa_k": "model-00012-of-00016.safetensors", + "model.blocks.30.attention.time_maa_r": "model-00012-of-00016.safetensors", + "model.blocks.30.attention.time_maa_v": "model-00012-of-00016.safetensors", + "model.blocks.30.attention.time_maa_w": "model-00012-of-00016.safetensors", + "model.blocks.30.attention.time_maa_w1": "model-00012-of-00016.safetensors", + "model.blocks.30.attention.time_maa_w2": "model-00012-of-00016.safetensors", + "model.blocks.30.attention.time_maa_x": "model-00012-of-00016.safetensors", + "model.blocks.30.attention.value.weight": "model-00012-of-00016.safetensors", + "model.blocks.30.feed_forward.experts.0.key.weight": "model-00012-of-00016.safetensors", + "model.blocks.30.feed_forward.experts.0.value.weight": "model-00012-of-00016.safetensors", + "model.blocks.30.feed_forward.experts.1.key.weight": "model-00012-of-00016.safetensors", + "model.blocks.30.feed_forward.experts.1.value.weight": "model-00012-of-00016.safetensors", + "model.blocks.30.feed_forward.experts.2.key.weight": "model-00012-of-00016.safetensors", + "model.blocks.30.feed_forward.experts.2.value.weight": "model-00012-of-00016.safetensors", + "model.blocks.30.feed_forward.experts.3.key.weight": "model-00012-of-00016.safetensors", + "model.blocks.30.feed_forward.experts.3.value.weight": "model-00012-of-00016.safetensors", + "model.blocks.30.feed_forward.experts.4.key.weight": "model-00012-of-00016.safetensors", + "model.blocks.30.feed_forward.experts.4.value.weight": "model-00012-of-00016.safetensors", + "model.blocks.30.feed_forward.experts.5.key.weight": "model-00012-of-00016.safetensors", + "model.blocks.30.feed_forward.experts.5.value.weight": "model-00012-of-00016.safetensors", + "model.blocks.30.feed_forward.experts.6.key.weight": "model-00012-of-00016.safetensors", + "model.blocks.30.feed_forward.experts.6.value.weight": "model-00012-of-00016.safetensors", + "model.blocks.30.feed_forward.experts.7.key.weight": "model-00012-of-00016.safetensors", + "model.blocks.30.feed_forward.experts.7.value.weight": "model-00012-of-00016.safetensors", + "model.blocks.30.feed_forward.receptance.weight": "model-00012-of-00016.safetensors", + "model.blocks.30.feed_forward.shared_expert.key.weight": "model-00012-of-00016.safetensors", + "model.blocks.30.feed_forward.shared_expert.value.weight": "model-00012-of-00016.safetensors", + "model.blocks.30.feed_forward.time_maa_k": "model-00012-of-00016.safetensors", + "model.blocks.30.feed_forward.time_maa_r": "model-00012-of-00016.safetensors", + "model.blocks.30.ln1.bias": "model-00012-of-00016.safetensors", + "model.blocks.30.ln1.weight": "model-00012-of-00016.safetensors", + "model.blocks.30.ln2.bias": "model-00012-of-00016.safetensors", + "model.blocks.30.ln2.weight": "model-00012-of-00016.safetensors", + "model.blocks.31.attention.gate.weight": "model-00012-of-00016.safetensors", + "model.blocks.31.attention.key.weight": "model-00012-of-00016.safetensors", + "model.blocks.31.attention.ln_x.bias": "model-00012-of-00016.safetensors", + "model.blocks.31.attention.ln_x.weight": "model-00012-of-00016.safetensors", + "model.blocks.31.attention.output.weight": "model-00012-of-00016.safetensors", + "model.blocks.31.attention.receptance.weight": "model-00012-of-00016.safetensors", + "model.blocks.31.attention.time_decay": "model-00012-of-00016.safetensors", + "model.blocks.31.attention.time_decay_w1": "model-00012-of-00016.safetensors", + "model.blocks.31.attention.time_decay_w2": "model-00012-of-00016.safetensors", + "model.blocks.31.attention.time_faaaa": "model-00012-of-00016.safetensors", + "model.blocks.31.attention.time_maa_g": "model-00012-of-00016.safetensors", + "model.blocks.31.attention.time_maa_k": "model-00012-of-00016.safetensors", + "model.blocks.31.attention.time_maa_r": "model-00012-of-00016.safetensors", + "model.blocks.31.attention.time_maa_v": "model-00012-of-00016.safetensors", + "model.blocks.31.attention.time_maa_w": "model-00012-of-00016.safetensors", + "model.blocks.31.attention.time_maa_w1": "model-00012-of-00016.safetensors", + "model.blocks.31.attention.time_maa_w2": "model-00012-of-00016.safetensors", + "model.blocks.31.attention.time_maa_x": "model-00012-of-00016.safetensors", + "model.blocks.31.attention.value.weight": "model-00012-of-00016.safetensors", + "model.blocks.31.feed_forward.experts.0.key.weight": "model-00012-of-00016.safetensors", + "model.blocks.31.feed_forward.experts.0.value.weight": "model-00012-of-00016.safetensors", + "model.blocks.31.feed_forward.experts.1.key.weight": "model-00012-of-00016.safetensors", + "model.blocks.31.feed_forward.experts.1.value.weight": "model-00012-of-00016.safetensors", + "model.blocks.31.feed_forward.experts.2.key.weight": "model-00012-of-00016.safetensors", + "model.blocks.31.feed_forward.experts.2.value.weight": "model-00012-of-00016.safetensors", + "model.blocks.31.feed_forward.experts.3.key.weight": "model-00012-of-00016.safetensors", + "model.blocks.31.feed_forward.experts.3.value.weight": "model-00012-of-00016.safetensors", + "model.blocks.31.feed_forward.experts.4.key.weight": "model-00013-of-00016.safetensors", + "model.blocks.31.feed_forward.experts.4.value.weight": "model-00013-of-00016.safetensors", + "model.blocks.31.feed_forward.experts.5.key.weight": "model-00013-of-00016.safetensors", + "model.blocks.31.feed_forward.experts.5.value.weight": "model-00013-of-00016.safetensors", + "model.blocks.31.feed_forward.experts.6.key.weight": "model-00013-of-00016.safetensors", + "model.blocks.31.feed_forward.experts.6.value.weight": "model-00013-of-00016.safetensors", + "model.blocks.31.feed_forward.experts.7.key.weight": "model-00013-of-00016.safetensors", + "model.blocks.31.feed_forward.experts.7.value.weight": "model-00013-of-00016.safetensors", + "model.blocks.31.feed_forward.receptance.weight": "model-00013-of-00016.safetensors", + "model.blocks.31.feed_forward.shared_expert.key.weight": "model-00013-of-00016.safetensors", + "model.blocks.31.feed_forward.shared_expert.value.weight": "model-00013-of-00016.safetensors", + "model.blocks.31.feed_forward.time_maa_k": "model-00013-of-00016.safetensors", + "model.blocks.31.feed_forward.time_maa_r": "model-00013-of-00016.safetensors", + "model.blocks.31.ln1.bias": "model-00013-of-00016.safetensors", + "model.blocks.31.ln1.weight": "model-00013-of-00016.safetensors", + "model.blocks.31.ln2.bias": "model-00013-of-00016.safetensors", + "model.blocks.31.ln2.weight": "model-00013-of-00016.safetensors", + "model.blocks.4.attention.gate.weight": "model-00013-of-00016.safetensors", + "model.blocks.4.attention.key.weight": "model-00013-of-00016.safetensors", + "model.blocks.4.attention.ln_x.bias": "model-00013-of-00016.safetensors", + "model.blocks.4.attention.ln_x.weight": "model-00013-of-00016.safetensors", + "model.blocks.4.attention.output.weight": "model-00013-of-00016.safetensors", + "model.blocks.4.attention.receptance.weight": "model-00013-of-00016.safetensors", + "model.blocks.4.attention.time_decay": "model-00013-of-00016.safetensors", + "model.blocks.4.attention.time_decay_w1": "model-00013-of-00016.safetensors", + "model.blocks.4.attention.time_decay_w2": "model-00013-of-00016.safetensors", + "model.blocks.4.attention.time_faaaa": "model-00013-of-00016.safetensors", + "model.blocks.4.attention.time_maa_g": "model-00013-of-00016.safetensors", + "model.blocks.4.attention.time_maa_k": "model-00013-of-00016.safetensors", + "model.blocks.4.attention.time_maa_r": "model-00013-of-00016.safetensors", + "model.blocks.4.attention.time_maa_v": "model-00013-of-00016.safetensors", + "model.blocks.4.attention.time_maa_w": "model-00013-of-00016.safetensors", + "model.blocks.4.attention.time_maa_w1": "model-00013-of-00016.safetensors", + "model.blocks.4.attention.time_maa_w2": "model-00013-of-00016.safetensors", + "model.blocks.4.attention.time_maa_x": "model-00013-of-00016.safetensors", + "model.blocks.4.attention.value.weight": "model-00013-of-00016.safetensors", + "model.blocks.4.feed_forward.experts.0.key.weight": "model-00013-of-00016.safetensors", + "model.blocks.4.feed_forward.experts.0.value.weight": "model-00013-of-00016.safetensors", + "model.blocks.4.feed_forward.experts.1.key.weight": "model-00013-of-00016.safetensors", + "model.blocks.4.feed_forward.experts.1.value.weight": "model-00013-of-00016.safetensors", + "model.blocks.4.feed_forward.experts.2.key.weight": "model-00013-of-00016.safetensors", + "model.blocks.4.feed_forward.experts.2.value.weight": "model-00013-of-00016.safetensors", + "model.blocks.4.feed_forward.experts.3.key.weight": "model-00013-of-00016.safetensors", + "model.blocks.4.feed_forward.experts.3.value.weight": "model-00013-of-00016.safetensors", + "model.blocks.4.feed_forward.experts.4.key.weight": "model-00013-of-00016.safetensors", + "model.blocks.4.feed_forward.experts.4.value.weight": "model-00013-of-00016.safetensors", + "model.blocks.4.feed_forward.experts.5.key.weight": "model-00013-of-00016.safetensors", + "model.blocks.4.feed_forward.experts.5.value.weight": "model-00013-of-00016.safetensors", + "model.blocks.4.feed_forward.experts.6.key.weight": "model-00013-of-00016.safetensors", + "model.blocks.4.feed_forward.experts.6.value.weight": "model-00013-of-00016.safetensors", + "model.blocks.4.feed_forward.experts.7.key.weight": "model-00013-of-00016.safetensors", + "model.blocks.4.feed_forward.experts.7.value.weight": "model-00013-of-00016.safetensors", + "model.blocks.4.feed_forward.receptance.weight": "model-00013-of-00016.safetensors", + "model.blocks.4.feed_forward.shared_expert.key.weight": "model-00013-of-00016.safetensors", + "model.blocks.4.feed_forward.shared_expert.value.weight": "model-00013-of-00016.safetensors", + "model.blocks.4.feed_forward.time_maa_k": "model-00013-of-00016.safetensors", + "model.blocks.4.feed_forward.time_maa_r": "model-00013-of-00016.safetensors", + "model.blocks.4.ln1.bias": "model-00013-of-00016.safetensors", + "model.blocks.4.ln1.weight": "model-00013-of-00016.safetensors", + "model.blocks.4.ln2.bias": "model-00013-of-00016.safetensors", + "model.blocks.4.ln2.weight": "model-00013-of-00016.safetensors", + "model.blocks.5.attention.gate.weight": "model-00013-of-00016.safetensors", + "model.blocks.5.attention.key.weight": "model-00013-of-00016.safetensors", + "model.blocks.5.attention.ln_x.bias": "model-00013-of-00016.safetensors", + "model.blocks.5.attention.ln_x.weight": "model-00013-of-00016.safetensors", + "model.blocks.5.attention.output.weight": "model-00013-of-00016.safetensors", + "model.blocks.5.attention.receptance.weight": "model-00013-of-00016.safetensors", + "model.blocks.5.attention.time_decay": "model-00013-of-00016.safetensors", + "model.blocks.5.attention.time_decay_w1": "model-00013-of-00016.safetensors", + "model.blocks.5.attention.time_decay_w2": "model-00013-of-00016.safetensors", + "model.blocks.5.attention.time_faaaa": "model-00013-of-00016.safetensors", + "model.blocks.5.attention.time_maa_g": "model-00013-of-00016.safetensors", + "model.blocks.5.attention.time_maa_k": "model-00013-of-00016.safetensors", + "model.blocks.5.attention.time_maa_r": "model-00013-of-00016.safetensors", + "model.blocks.5.attention.time_maa_v": "model-00013-of-00016.safetensors", + "model.blocks.5.attention.time_maa_w": "model-00013-of-00016.safetensors", + "model.blocks.5.attention.time_maa_w1": "model-00013-of-00016.safetensors", + "model.blocks.5.attention.time_maa_w2": "model-00013-of-00016.safetensors", + "model.blocks.5.attention.time_maa_x": "model-00013-of-00016.safetensors", + "model.blocks.5.attention.value.weight": "model-00013-of-00016.safetensors", + "model.blocks.5.feed_forward.experts.0.key.weight": "model-00013-of-00016.safetensors", + "model.blocks.5.feed_forward.experts.0.value.weight": "model-00013-of-00016.safetensors", + "model.blocks.5.feed_forward.experts.1.key.weight": "model-00013-of-00016.safetensors", + "model.blocks.5.feed_forward.experts.1.value.weight": "model-00013-of-00016.safetensors", + "model.blocks.5.feed_forward.experts.2.key.weight": "model-00013-of-00016.safetensors", + "model.blocks.5.feed_forward.experts.2.value.weight": "model-00013-of-00016.safetensors", + "model.blocks.5.feed_forward.experts.3.key.weight": "model-00013-of-00016.safetensors", + "model.blocks.5.feed_forward.experts.3.value.weight": "model-00013-of-00016.safetensors", + "model.blocks.5.feed_forward.experts.4.key.weight": "model-00013-of-00016.safetensors", + "model.blocks.5.feed_forward.experts.4.value.weight": "model-00013-of-00016.safetensors", + "model.blocks.5.feed_forward.experts.5.key.weight": "model-00013-of-00016.safetensors", + "model.blocks.5.feed_forward.experts.5.value.weight": "model-00014-of-00016.safetensors", + "model.blocks.5.feed_forward.experts.6.key.weight": "model-00014-of-00016.safetensors", + "model.blocks.5.feed_forward.experts.6.value.weight": "model-00014-of-00016.safetensors", + "model.blocks.5.feed_forward.experts.7.key.weight": "model-00014-of-00016.safetensors", + "model.blocks.5.feed_forward.experts.7.value.weight": "model-00014-of-00016.safetensors", + "model.blocks.5.feed_forward.receptance.weight": "model-00014-of-00016.safetensors", + "model.blocks.5.feed_forward.shared_expert.key.weight": "model-00014-of-00016.safetensors", + "model.blocks.5.feed_forward.shared_expert.value.weight": "model-00014-of-00016.safetensors", + "model.blocks.5.feed_forward.time_maa_k": "model-00014-of-00016.safetensors", + "model.blocks.5.feed_forward.time_maa_r": "model-00014-of-00016.safetensors", + "model.blocks.5.ln1.bias": "model-00014-of-00016.safetensors", + "model.blocks.5.ln1.weight": "model-00014-of-00016.safetensors", + "model.blocks.5.ln2.bias": "model-00014-of-00016.safetensors", + "model.blocks.5.ln2.weight": "model-00014-of-00016.safetensors", + "model.blocks.6.attention.gate.weight": "model-00014-of-00016.safetensors", + "model.blocks.6.attention.key.weight": "model-00014-of-00016.safetensors", + "model.blocks.6.attention.ln_x.bias": "model-00014-of-00016.safetensors", + "model.blocks.6.attention.ln_x.weight": "model-00014-of-00016.safetensors", + "model.blocks.6.attention.output.weight": "model-00014-of-00016.safetensors", + "model.blocks.6.attention.receptance.weight": "model-00014-of-00016.safetensors", + "model.blocks.6.attention.time_decay": "model-00014-of-00016.safetensors", + "model.blocks.6.attention.time_decay_w1": "model-00014-of-00016.safetensors", + "model.blocks.6.attention.time_decay_w2": "model-00014-of-00016.safetensors", + "model.blocks.6.attention.time_faaaa": "model-00014-of-00016.safetensors", + "model.blocks.6.attention.time_maa_g": "model-00014-of-00016.safetensors", + "model.blocks.6.attention.time_maa_k": "model-00014-of-00016.safetensors", + "model.blocks.6.attention.time_maa_r": "model-00014-of-00016.safetensors", + "model.blocks.6.attention.time_maa_v": "model-00014-of-00016.safetensors", + "model.blocks.6.attention.time_maa_w": "model-00014-of-00016.safetensors", + "model.blocks.6.attention.time_maa_w1": "model-00014-of-00016.safetensors", + "model.blocks.6.attention.time_maa_w2": "model-00014-of-00016.safetensors", + "model.blocks.6.attention.time_maa_x": "model-00014-of-00016.safetensors", + "model.blocks.6.attention.value.weight": "model-00014-of-00016.safetensors", + "model.blocks.6.feed_forward.experts.0.key.weight": "model-00014-of-00016.safetensors", + "model.blocks.6.feed_forward.experts.0.value.weight": "model-00014-of-00016.safetensors", + "model.blocks.6.feed_forward.experts.1.key.weight": "model-00014-of-00016.safetensors", + "model.blocks.6.feed_forward.experts.1.value.weight": "model-00014-of-00016.safetensors", + "model.blocks.6.feed_forward.experts.2.key.weight": "model-00014-of-00016.safetensors", + "model.blocks.6.feed_forward.experts.2.value.weight": "model-00014-of-00016.safetensors", + "model.blocks.6.feed_forward.experts.3.key.weight": "model-00014-of-00016.safetensors", + "model.blocks.6.feed_forward.experts.3.value.weight": "model-00014-of-00016.safetensors", + "model.blocks.6.feed_forward.experts.4.key.weight": "model-00014-of-00016.safetensors", + "model.blocks.6.feed_forward.experts.4.value.weight": "model-00014-of-00016.safetensors", + "model.blocks.6.feed_forward.experts.5.key.weight": "model-00014-of-00016.safetensors", + "model.blocks.6.feed_forward.experts.5.value.weight": "model-00014-of-00016.safetensors", + "model.blocks.6.feed_forward.experts.6.key.weight": "model-00014-of-00016.safetensors", + "model.blocks.6.feed_forward.experts.6.value.weight": "model-00014-of-00016.safetensors", + "model.blocks.6.feed_forward.experts.7.key.weight": "model-00014-of-00016.safetensors", + "model.blocks.6.feed_forward.experts.7.value.weight": "model-00014-of-00016.safetensors", + "model.blocks.6.feed_forward.receptance.weight": "model-00014-of-00016.safetensors", + "model.blocks.6.feed_forward.shared_expert.key.weight": "model-00014-of-00016.safetensors", + "model.blocks.6.feed_forward.shared_expert.value.weight": "model-00014-of-00016.safetensors", + "model.blocks.6.feed_forward.time_maa_k": "model-00014-of-00016.safetensors", + "model.blocks.6.feed_forward.time_maa_r": "model-00014-of-00016.safetensors", + "model.blocks.6.ln1.bias": "model-00014-of-00016.safetensors", + "model.blocks.6.ln1.weight": "model-00014-of-00016.safetensors", + "model.blocks.6.ln2.bias": "model-00014-of-00016.safetensors", + "model.blocks.6.ln2.weight": "model-00014-of-00016.safetensors", + "model.blocks.7.attention.gate.weight": "model-00014-of-00016.safetensors", + "model.blocks.7.attention.key.weight": "model-00014-of-00016.safetensors", + "model.blocks.7.attention.ln_x.bias": "model-00014-of-00016.safetensors", + "model.blocks.7.attention.ln_x.weight": "model-00014-of-00016.safetensors", + "model.blocks.7.attention.output.weight": "model-00014-of-00016.safetensors", + "model.blocks.7.attention.receptance.weight": "model-00014-of-00016.safetensors", + "model.blocks.7.attention.time_decay": "model-00014-of-00016.safetensors", + "model.blocks.7.attention.time_decay_w1": "model-00014-of-00016.safetensors", + "model.blocks.7.attention.time_decay_w2": "model-00014-of-00016.safetensors", + "model.blocks.7.attention.time_faaaa": "model-00014-of-00016.safetensors", + "model.blocks.7.attention.time_maa_g": "model-00014-of-00016.safetensors", + "model.blocks.7.attention.time_maa_k": "model-00014-of-00016.safetensors", + "model.blocks.7.attention.time_maa_r": "model-00014-of-00016.safetensors", + "model.blocks.7.attention.time_maa_v": "model-00014-of-00016.safetensors", + "model.blocks.7.attention.time_maa_w": "model-00014-of-00016.safetensors", + "model.blocks.7.attention.time_maa_w1": "model-00014-of-00016.safetensors", + "model.blocks.7.attention.time_maa_w2": "model-00014-of-00016.safetensors", + "model.blocks.7.attention.time_maa_x": "model-00014-of-00016.safetensors", + "model.blocks.7.attention.value.weight": "model-00014-of-00016.safetensors", + "model.blocks.7.feed_forward.experts.0.key.weight": "model-00014-of-00016.safetensors", + "model.blocks.7.feed_forward.experts.0.value.weight": "model-00014-of-00016.safetensors", + "model.blocks.7.feed_forward.experts.1.key.weight": "model-00014-of-00016.safetensors", + "model.blocks.7.feed_forward.experts.1.value.weight": "model-00014-of-00016.safetensors", + "model.blocks.7.feed_forward.experts.2.key.weight": "model-00014-of-00016.safetensors", + "model.blocks.7.feed_forward.experts.2.value.weight": "model-00014-of-00016.safetensors", + "model.blocks.7.feed_forward.experts.3.key.weight": "model-00014-of-00016.safetensors", + "model.blocks.7.feed_forward.experts.3.value.weight": "model-00014-of-00016.safetensors", + "model.blocks.7.feed_forward.experts.4.key.weight": "model-00014-of-00016.safetensors", + "model.blocks.7.feed_forward.experts.4.value.weight": "model-00014-of-00016.safetensors", + "model.blocks.7.feed_forward.experts.5.key.weight": "model-00014-of-00016.safetensors", + "model.blocks.7.feed_forward.experts.5.value.weight": "model-00014-of-00016.safetensors", + "model.blocks.7.feed_forward.experts.6.key.weight": "model-00014-of-00016.safetensors", + "model.blocks.7.feed_forward.experts.6.value.weight": "model-00014-of-00016.safetensors", + "model.blocks.7.feed_forward.experts.7.key.weight": "model-00015-of-00016.safetensors", + "model.blocks.7.feed_forward.experts.7.value.weight": "model-00015-of-00016.safetensors", + "model.blocks.7.feed_forward.receptance.weight": "model-00015-of-00016.safetensors", + "model.blocks.7.feed_forward.shared_expert.key.weight": "model-00015-of-00016.safetensors", + "model.blocks.7.feed_forward.shared_expert.value.weight": "model-00015-of-00016.safetensors", + "model.blocks.7.feed_forward.time_maa_k": "model-00015-of-00016.safetensors", + "model.blocks.7.feed_forward.time_maa_r": "model-00015-of-00016.safetensors", + "model.blocks.7.ln1.bias": "model-00015-of-00016.safetensors", + "model.blocks.7.ln1.weight": "model-00015-of-00016.safetensors", + "model.blocks.7.ln2.bias": "model-00015-of-00016.safetensors", + "model.blocks.7.ln2.weight": "model-00015-of-00016.safetensors", + "model.blocks.8.attention.gate.weight": "model-00015-of-00016.safetensors", + "model.blocks.8.attention.key.weight": "model-00015-of-00016.safetensors", + "model.blocks.8.attention.ln_x.bias": "model-00015-of-00016.safetensors", + "model.blocks.8.attention.ln_x.weight": "model-00015-of-00016.safetensors", + "model.blocks.8.attention.output.weight": "model-00015-of-00016.safetensors", + "model.blocks.8.attention.receptance.weight": "model-00015-of-00016.safetensors", + "model.blocks.8.attention.time_decay": "model-00015-of-00016.safetensors", + "model.blocks.8.attention.time_decay_w1": "model-00015-of-00016.safetensors", + "model.blocks.8.attention.time_decay_w2": "model-00015-of-00016.safetensors", + "model.blocks.8.attention.time_faaaa": "model-00015-of-00016.safetensors", + "model.blocks.8.attention.time_maa_g": "model-00015-of-00016.safetensors", + "model.blocks.8.attention.time_maa_k": "model-00015-of-00016.safetensors", + "model.blocks.8.attention.time_maa_r": "model-00015-of-00016.safetensors", + "model.blocks.8.attention.time_maa_v": "model-00015-of-00016.safetensors", + "model.blocks.8.attention.time_maa_w": "model-00015-of-00016.safetensors", + "model.blocks.8.attention.time_maa_w1": "model-00015-of-00016.safetensors", + "model.blocks.8.attention.time_maa_w2": "model-00015-of-00016.safetensors", + "model.blocks.8.attention.time_maa_x": "model-00015-of-00016.safetensors", + "model.blocks.8.attention.value.weight": "model-00015-of-00016.safetensors", + "model.blocks.8.feed_forward.experts.0.key.weight": "model-00015-of-00016.safetensors", + "model.blocks.8.feed_forward.experts.0.value.weight": "model-00015-of-00016.safetensors", + "model.blocks.8.feed_forward.experts.1.key.weight": "model-00015-of-00016.safetensors", + "model.blocks.8.feed_forward.experts.1.value.weight": "model-00015-of-00016.safetensors", + "model.blocks.8.feed_forward.experts.2.key.weight": "model-00015-of-00016.safetensors", + "model.blocks.8.feed_forward.experts.2.value.weight": "model-00015-of-00016.safetensors", + "model.blocks.8.feed_forward.experts.3.key.weight": "model-00015-of-00016.safetensors", + "model.blocks.8.feed_forward.experts.3.value.weight": "model-00015-of-00016.safetensors", + "model.blocks.8.feed_forward.experts.4.key.weight": "model-00015-of-00016.safetensors", + "model.blocks.8.feed_forward.experts.4.value.weight": "model-00015-of-00016.safetensors", + "model.blocks.8.feed_forward.experts.5.key.weight": "model-00015-of-00016.safetensors", + "model.blocks.8.feed_forward.experts.5.value.weight": "model-00015-of-00016.safetensors", + "model.blocks.8.feed_forward.experts.6.key.weight": "model-00015-of-00016.safetensors", + "model.blocks.8.feed_forward.experts.6.value.weight": "model-00015-of-00016.safetensors", + "model.blocks.8.feed_forward.experts.7.key.weight": "model-00015-of-00016.safetensors", + "model.blocks.8.feed_forward.experts.7.value.weight": "model-00015-of-00016.safetensors", + "model.blocks.8.feed_forward.receptance.weight": "model-00015-of-00016.safetensors", + "model.blocks.8.feed_forward.shared_expert.key.weight": "model-00015-of-00016.safetensors", + "model.blocks.8.feed_forward.shared_expert.value.weight": "model-00015-of-00016.safetensors", + "model.blocks.8.feed_forward.time_maa_k": "model-00015-of-00016.safetensors", + "model.blocks.8.feed_forward.time_maa_r": "model-00015-of-00016.safetensors", + "model.blocks.8.ln1.bias": "model-00015-of-00016.safetensors", + "model.blocks.8.ln1.weight": "model-00015-of-00016.safetensors", + "model.blocks.8.ln2.bias": "model-00015-of-00016.safetensors", + "model.blocks.8.ln2.weight": "model-00015-of-00016.safetensors", + "model.blocks.9.attention.gate.weight": "model-00015-of-00016.safetensors", + "model.blocks.9.attention.key.weight": "model-00015-of-00016.safetensors", + "model.blocks.9.attention.ln_x.bias": "model-00015-of-00016.safetensors", + "model.blocks.9.attention.ln_x.weight": "model-00015-of-00016.safetensors", + "model.blocks.9.attention.output.weight": "model-00015-of-00016.safetensors", + "model.blocks.9.attention.receptance.weight": "model-00015-of-00016.safetensors", + "model.blocks.9.attention.time_decay": "model-00015-of-00016.safetensors", + "model.blocks.9.attention.time_decay_w1": "model-00015-of-00016.safetensors", + "model.blocks.9.attention.time_decay_w2": "model-00015-of-00016.safetensors", + "model.blocks.9.attention.time_faaaa": "model-00015-of-00016.safetensors", + "model.blocks.9.attention.time_maa_g": "model-00015-of-00016.safetensors", + "model.blocks.9.attention.time_maa_k": "model-00015-of-00016.safetensors", + "model.blocks.9.attention.time_maa_r": "model-00015-of-00016.safetensors", + "model.blocks.9.attention.time_maa_v": "model-00015-of-00016.safetensors", + "model.blocks.9.attention.time_maa_w": "model-00015-of-00016.safetensors", + "model.blocks.9.attention.time_maa_w1": "model-00015-of-00016.safetensors", + "model.blocks.9.attention.time_maa_w2": "model-00015-of-00016.safetensors", + "model.blocks.9.attention.time_maa_x": "model-00015-of-00016.safetensors", + "model.blocks.9.attention.value.weight": "model-00015-of-00016.safetensors", + "model.blocks.9.feed_forward.experts.0.key.weight": "model-00015-of-00016.safetensors", + "model.blocks.9.feed_forward.experts.0.value.weight": "model-00015-of-00016.safetensors", + "model.blocks.9.feed_forward.experts.1.key.weight": "model-00015-of-00016.safetensors", + "model.blocks.9.feed_forward.experts.1.value.weight": "model-00015-of-00016.safetensors", + "model.blocks.9.feed_forward.experts.2.key.weight": "model-00015-of-00016.safetensors", + "model.blocks.9.feed_forward.experts.2.value.weight": "model-00015-of-00016.safetensors", + "model.blocks.9.feed_forward.experts.3.key.weight": "model-00015-of-00016.safetensors", + "model.blocks.9.feed_forward.experts.3.value.weight": "model-00015-of-00016.safetensors", + "model.blocks.9.feed_forward.experts.4.key.weight": "model-00015-of-00016.safetensors", + "model.blocks.9.feed_forward.experts.4.value.weight": "model-00015-of-00016.safetensors", + "model.blocks.9.feed_forward.experts.5.key.weight": "model-00015-of-00016.safetensors", + "model.blocks.9.feed_forward.experts.5.value.weight": "model-00015-of-00016.safetensors", + "model.blocks.9.feed_forward.experts.6.key.weight": "model-00015-of-00016.safetensors", + "model.blocks.9.feed_forward.experts.6.value.weight": "model-00015-of-00016.safetensors", + "model.blocks.9.feed_forward.experts.7.key.weight": "model-00015-of-00016.safetensors", + "model.blocks.9.feed_forward.experts.7.value.weight": "model-00015-of-00016.safetensors", + "model.blocks.9.feed_forward.receptance.weight": "model-00015-of-00016.safetensors", + "model.blocks.9.feed_forward.shared_expert.key.weight": "model-00016-of-00016.safetensors", + "model.blocks.9.feed_forward.shared_expert.value.weight": "model-00016-of-00016.safetensors", + "model.blocks.9.feed_forward.time_maa_k": "model-00016-of-00016.safetensors", + "model.blocks.9.feed_forward.time_maa_r": "model-00016-of-00016.safetensors", + "model.blocks.9.ln1.bias": "model-00016-of-00016.safetensors", + "model.blocks.9.ln1.weight": "model-00016-of-00016.safetensors", + "model.blocks.9.ln2.bias": "model-00016-of-00016.safetensors", + "model.blocks.9.ln2.weight": "model-00016-of-00016.safetensors", + "model.embeddings.weight": "model-00016-of-00016.safetensors", + "model.ln_out.bias": "model-00016-of-00016.safetensors", + "model.ln_out.weight": "model-00016-of-00016.safetensors" + } +}