initial commit

attn_gates.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3c487bdc5ccd9bf4d3f9307a79471e32b98e8abddf3c0059dca333f8b1a8d1ab
+size 503344034

config.json

@@ -0,0 +1,50 @@
+{
+  "_attn_implementation_autoset": true,
+  "_name_or_path": "models/SeerAttention/deepseek-ai/DeepSeek-R1-Distill-Llama-70B_Qavg_Kmaxminavg_lr1e-3_maxlen65536_warmup20_bs16_steps500_gatelossscale10.0",
+  "architectures": [
+    "SeerAttnLlamaForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "base_model": "deepseek-ai/DeepSeek-R1-Distill-Llama-70B",
+  "bos_token_id": 128000,
+  "eos_token_id": [
+    128001,
+    128008,
+    128009
+  ],
+  "head_dim": 128,
+  "hidden_act": "silu",
+  "hidden_size": 8192,
+  "initializer_range": 0.02,
+  "intermediate_size": 28672,
+  "max_position_embeddings": 131072,
+  "mlp_bias": false,
+  "model_type": "llama",
+  "num_attention_heads": 64,
+  "num_hidden_layers": 80,
+  "num_key_value_heads": 8,
+  "pretraining_tp": 1,
+  "rms_norm_eps": 1e-05,
+  "rope_scaling": {
+    "factor": 8.0,
+    "high_freq_factor": 4.0,
+    "low_freq_factor": 1.0,
+    "original_max_position_embeddings": 8192,
+    "rope_type": "llama3"
+  },
+  "rope_theta": 500000.0,
+  "seerattn_gate_block_size": 64,
+  "seerattn_gate_force_double": false,
+  "seerattn_gate_hidden_size": 128,
+  "seerattn_gate_type": "Qavg_Kmaxminavg",
+  "seerattn_last_block_dense": true,
+  "seerattn_nz_ratio": 1.0,
+  "seerattn_sparsity_method": "threshold",
+  "seerattn_threshold": 0.0,
+  "tie_word_embeddings": false,
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.49.0",
+  "use_cache": true,
+  "vocab_size": 128257
+}