diff --git a/deepseek-r1-1.5b-packed/config.json b/deepseek-r1-1.5b-packed/config.json new file mode 100644 index 0000000000000000000000000000000000000000..3e61a980288109cc87c1f932d0057933290f0520 --- /dev/null +++ b/deepseek-r1-1.5b-packed/config.json @@ -0,0 +1,13 @@ +{ + "hidden_size": 1536, + "intermediate_size": 8960, + "num_attention_heads": 12, + "num_key_value_heads": 2, + "num_hidden_layers": 28, + "vocab_size": 151936, + "head_dim": 128, + "rope_theta": 1000000.0, + "rms_norm_eps": 1e-06, + "n_levels": 7, + "quant_type": "packed_unary" +} \ No newline at end of file diff --git a/deepseek-r1-1.5b-packed/model_layers_0_self_attn_q_proj_weight.scales b/deepseek-r1-1.5b-packed/model_layers_0_self_attn_q_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..b94d3a26056ae13e003511a9842a1f021895be27 Binary files /dev/null and b/deepseek-r1-1.5b-packed/model_layers_0_self_attn_q_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-packed/model_layers_10_mlp_up_proj_weight.scales b/deepseek-r1-1.5b-packed/model_layers_10_mlp_up_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..97e85772f291873729de4a628593c8146d4529ad Binary files /dev/null and b/deepseek-r1-1.5b-packed/model_layers_10_mlp_up_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-packed/model_layers_10_self_attn_q_proj_bias.fp16 b/deepseek-r1-1.5b-packed/model_layers_10_self_attn_q_proj_bias.fp16 new file mode 100644 index 0000000000000000000000000000000000000000..37c91a27fa24debad38b68b4937718c96e4bfad4 Binary files /dev/null and b/deepseek-r1-1.5b-packed/model_layers_10_self_attn_q_proj_bias.fp16 differ diff --git a/deepseek-r1-1.5b-packed/model_layers_12_self_attn_k_proj_weight.scales b/deepseek-r1-1.5b-packed/model_layers_12_self_attn_k_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..3d5fb21e947304ce8698d78b41a7ad53225f45f1 Binary files /dev/null and b/deepseek-r1-1.5b-packed/model_layers_12_self_attn_k_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-packed/model_layers_12_self_attn_k_proj_weight.signs b/deepseek-r1-1.5b-packed/model_layers_12_self_attn_k_proj_weight.signs new file mode 100644 index 0000000000000000000000000000000000000000..5ddb67908a3e1b5a8fab7a2e821812244d7471c2 Binary files /dev/null and b/deepseek-r1-1.5b-packed/model_layers_12_self_attn_k_proj_weight.signs differ diff --git a/deepseek-r1-1.5b-packed/model_layers_12_self_attn_o_proj_weight.scales b/deepseek-r1-1.5b-packed/model_layers_12_self_attn_o_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..96918a7603808743e4e1c65c6e419b8a9660e535 Binary files /dev/null and b/deepseek-r1-1.5b-packed/model_layers_12_self_attn_o_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-packed/model_layers_14_self_attn_v_proj_weight.scales b/deepseek-r1-1.5b-packed/model_layers_14_self_attn_v_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..7a0b45d4580ae03c74559ab37da0d33c9a2546e6 Binary files /dev/null and b/deepseek-r1-1.5b-packed/model_layers_14_self_attn_v_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-packed/model_layers_15_mlp_gate_proj_weight.scales b/deepseek-r1-1.5b-packed/model_layers_15_mlp_gate_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..c2431977ce88f68fb0e625d42a26c6980ba2be85 Binary files /dev/null and b/deepseek-r1-1.5b-packed/model_layers_15_mlp_gate_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-packed/model_layers_15_self_attn_k_proj_weight.signs b/deepseek-r1-1.5b-packed/model_layers_15_self_attn_k_proj_weight.signs new file mode 100644 index 0000000000000000000000000000000000000000..b225cf5b31630407d56f21aeaa62e86fb6d246fb Binary files /dev/null and b/deepseek-r1-1.5b-packed/model_layers_15_self_attn_k_proj_weight.signs differ diff --git a/deepseek-r1-1.5b-packed/model_layers_15_self_attn_q_proj_weight.rmm b/deepseek-r1-1.5b-packed/model_layers_15_self_attn_q_proj_weight.rmm new file mode 100644 index 0000000000000000000000000000000000000000..c86839c59145849e367cc3aa33182b1c6c31a614 --- /dev/null +++ b/deepseek-r1-1.5b-packed/model_layers_15_self_attn_q_proj_weight.rmm @@ -0,0 +1 @@ + \ No newline at end of file diff --git a/deepseek-r1-1.5b-packed/model_layers_16_mlp_gate_proj_weight.scales b/deepseek-r1-1.5b-packed/model_layers_16_mlp_gate_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..f56ece90071e0e2a193307b0a81483ad2454f358 Binary files /dev/null and b/deepseek-r1-1.5b-packed/model_layers_16_mlp_gate_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-packed/model_layers_16_self_attn_q_proj_weight.scales b/deepseek-r1-1.5b-packed/model_layers_16_self_attn_q_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..8d8ea604df8d924f9dd685f262fda6b5d06b53fa Binary files /dev/null and b/deepseek-r1-1.5b-packed/model_layers_16_self_attn_q_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-packed/model_layers_17_input_layernorm_weight.fp16 b/deepseek-r1-1.5b-packed/model_layers_17_input_layernorm_weight.fp16 new file mode 100644 index 0000000000000000000000000000000000000000..f65011ae1846ecb6b63f5a65c0ddbbfc830647e7 Binary files /dev/null and b/deepseek-r1-1.5b-packed/model_layers_17_input_layernorm_weight.fp16 differ diff --git a/deepseek-r1-1.5b-packed/model_layers_17_mlp_down_proj_weight.rmm b/deepseek-r1-1.5b-packed/model_layers_17_mlp_down_proj_weight.rmm new file mode 100644 index 0000000000000000000000000000000000000000..c86839c59145849e367cc3aa33182b1c6c31a614 --- /dev/null +++ b/deepseek-r1-1.5b-packed/model_layers_17_mlp_down_proj_weight.rmm @@ -0,0 +1 @@ + \ No newline at end of file diff --git a/deepseek-r1-1.5b-packed/model_layers_17_mlp_up_proj_weight.rmm b/deepseek-r1-1.5b-packed/model_layers_17_mlp_up_proj_weight.rmm new file mode 100644 index 0000000000000000000000000000000000000000..3d5a76a5a0c43488828c17fc8a7002591a508098 --- /dev/null +++ b/deepseek-r1-1.5b-packed/model_layers_17_mlp_up_proj_weight.rmm @@ -0,0 +1 @@ + \ No newline at end of file diff --git a/deepseek-r1-1.5b-packed/model_layers_17_post_attention_layernorm_weight.fp16 b/deepseek-r1-1.5b-packed/model_layers_17_post_attention_layernorm_weight.fp16 new file mode 100644 index 0000000000000000000000000000000000000000..00f889b99c2652a77daf31ccc53d9ced249c3ba3 Binary files /dev/null and b/deepseek-r1-1.5b-packed/model_layers_17_post_attention_layernorm_weight.fp16 differ diff --git a/deepseek-r1-1.5b-packed/model_layers_17_self_attn_k_proj_weight.scales b/deepseek-r1-1.5b-packed/model_layers_17_self_attn_k_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..3c9eaca0825c289cec36699ad686980f5de86a47 Binary files /dev/null and b/deepseek-r1-1.5b-packed/model_layers_17_self_attn_k_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-packed/model_layers_17_self_attn_q_proj_bias.fp16 b/deepseek-r1-1.5b-packed/model_layers_17_self_attn_q_proj_bias.fp16 new file mode 100644 index 0000000000000000000000000000000000000000..879dbcd56affcfbd20af9b3450a827e8ac23ac7d Binary files /dev/null and b/deepseek-r1-1.5b-packed/model_layers_17_self_attn_q_proj_bias.fp16 differ diff --git a/deepseek-r1-1.5b-packed/model_layers_18_self_attn_o_proj_weight.scales b/deepseek-r1-1.5b-packed/model_layers_18_self_attn_o_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..b92a3d1aa059446f67f29273333c92a73000e9fb Binary files /dev/null and b/deepseek-r1-1.5b-packed/model_layers_18_self_attn_o_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-packed/model_layers_1_mlp_down_proj_weight.scales b/deepseek-r1-1.5b-packed/model_layers_1_mlp_down_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..0efa47d2f3091f4cb7c90bdef7320406ad4f6723 Binary files /dev/null and b/deepseek-r1-1.5b-packed/model_layers_1_mlp_down_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-packed/model_layers_1_mlp_up_proj_weight.rmm b/deepseek-r1-1.5b-packed/model_layers_1_mlp_up_proj_weight.rmm new file mode 100644 index 0000000000000000000000000000000000000000..3d5a76a5a0c43488828c17fc8a7002591a508098 --- /dev/null +++ b/deepseek-r1-1.5b-packed/model_layers_1_mlp_up_proj_weight.rmm @@ -0,0 +1 @@ + \ No newline at end of file diff --git a/deepseek-r1-1.5b-packed/model_layers_1_self_attn_k_proj_weight.scales b/deepseek-r1-1.5b-packed/model_layers_1_self_attn_k_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..5828613ee205b174dbfc1f4f632dd30a4aba5f85 Binary files /dev/null and b/deepseek-r1-1.5b-packed/model_layers_1_self_attn_k_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-packed/model_layers_20_input_layernorm_weight.fp16 b/deepseek-r1-1.5b-packed/model_layers_20_input_layernorm_weight.fp16 new file mode 100644 index 0000000000000000000000000000000000000000..8a042895077821880f716b03d5f483d8669cad26 Binary files /dev/null and b/deepseek-r1-1.5b-packed/model_layers_20_input_layernorm_weight.fp16 differ diff --git a/deepseek-r1-1.5b-packed/model_layers_20_self_attn_q_proj_weight.rmm b/deepseek-r1-1.5b-packed/model_layers_20_self_attn_q_proj_weight.rmm new file mode 100644 index 0000000000000000000000000000000000000000..c86839c59145849e367cc3aa33182b1c6c31a614 --- /dev/null +++ b/deepseek-r1-1.5b-packed/model_layers_20_self_attn_q_proj_weight.rmm @@ -0,0 +1 @@ + \ No newline at end of file diff --git a/deepseek-r1-1.5b-packed/model_layers_20_self_attn_v_proj_weight.rmm b/deepseek-r1-1.5b-packed/model_layers_20_self_attn_v_proj_weight.rmm new file mode 100644 index 0000000000000000000000000000000000000000..9401fc7bec41306420e2f625efdc88cd8c92717d --- /dev/null +++ b/deepseek-r1-1.5b-packed/model_layers_20_self_attn_v_proj_weight.rmm @@ -0,0 +1 @@ + \ No newline at end of file diff --git a/deepseek-r1-1.5b-packed/model_layers_21_self_attn_q_proj_weight.rmm b/deepseek-r1-1.5b-packed/model_layers_21_self_attn_q_proj_weight.rmm new file mode 100644 index 0000000000000000000000000000000000000000..c86839c59145849e367cc3aa33182b1c6c31a614 --- /dev/null +++ b/deepseek-r1-1.5b-packed/model_layers_21_self_attn_q_proj_weight.rmm @@ -0,0 +1 @@ + \ No newline at end of file diff --git a/deepseek-r1-1.5b-packed/model_layers_21_self_attn_q_proj_weight.scales b/deepseek-r1-1.5b-packed/model_layers_21_self_attn_q_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..7b24f3b68a4be50ecc2515725b7a46127de44851 Binary files /dev/null and b/deepseek-r1-1.5b-packed/model_layers_21_self_attn_q_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-packed/model_layers_22_mlp_down_proj_weight.rmm b/deepseek-r1-1.5b-packed/model_layers_22_mlp_down_proj_weight.rmm new file mode 100644 index 0000000000000000000000000000000000000000..c86839c59145849e367cc3aa33182b1c6c31a614 --- /dev/null +++ b/deepseek-r1-1.5b-packed/model_layers_22_mlp_down_proj_weight.rmm @@ -0,0 +1 @@ + \ No newline at end of file diff --git a/deepseek-r1-1.5b-packed/model_layers_22_mlp_down_proj_weight.scales b/deepseek-r1-1.5b-packed/model_layers_22_mlp_down_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..2d8bad17a13035c76a7c181d856e186dacfdd0c3 Binary files /dev/null and b/deepseek-r1-1.5b-packed/model_layers_22_mlp_down_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-packed/model_layers_22_mlp_up_proj_weight.rmm b/deepseek-r1-1.5b-packed/model_layers_22_mlp_up_proj_weight.rmm new file mode 100644 index 0000000000000000000000000000000000000000..3d5a76a5a0c43488828c17fc8a7002591a508098 --- /dev/null +++ b/deepseek-r1-1.5b-packed/model_layers_22_mlp_up_proj_weight.rmm @@ -0,0 +1 @@ + \ No newline at end of file diff --git a/deepseek-r1-1.5b-packed/model_layers_22_self_attn_k_proj_weight.signs b/deepseek-r1-1.5b-packed/model_layers_22_self_attn_k_proj_weight.signs new file mode 100644 index 0000000000000000000000000000000000000000..51103b9cea5888331cf394ba8948f1014a260a27 Binary files /dev/null and b/deepseek-r1-1.5b-packed/model_layers_22_self_attn_k_proj_weight.signs differ diff --git a/deepseek-r1-1.5b-packed/model_layers_23_mlp_down_proj_weight.rmm b/deepseek-r1-1.5b-packed/model_layers_23_mlp_down_proj_weight.rmm new file mode 100644 index 0000000000000000000000000000000000000000..c86839c59145849e367cc3aa33182b1c6c31a614 --- /dev/null +++ b/deepseek-r1-1.5b-packed/model_layers_23_mlp_down_proj_weight.rmm @@ -0,0 +1 @@ + \ No newline at end of file diff --git a/deepseek-r1-1.5b-packed/model_layers_23_mlp_gate_proj_weight.scales b/deepseek-r1-1.5b-packed/model_layers_23_mlp_gate_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..29791c352e74b53934fbfc242d29474b4c49db40 Binary files /dev/null and b/deepseek-r1-1.5b-packed/model_layers_23_mlp_gate_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-packed/model_layers_23_self_attn_o_proj_weight.scales b/deepseek-r1-1.5b-packed/model_layers_23_self_attn_o_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..72120be0ab4fa286df617efb1151549528705b6e Binary files /dev/null and b/deepseek-r1-1.5b-packed/model_layers_23_self_attn_o_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-packed/model_layers_24_mlp_gate_proj_weight.rmm b/deepseek-r1-1.5b-packed/model_layers_24_mlp_gate_proj_weight.rmm new file mode 100644 index 0000000000000000000000000000000000000000..3d5a76a5a0c43488828c17fc8a7002591a508098 --- /dev/null +++ b/deepseek-r1-1.5b-packed/model_layers_24_mlp_gate_proj_weight.rmm @@ -0,0 +1 @@ + \ No newline at end of file diff --git a/deepseek-r1-1.5b-packed/model_layers_25_self_attn_q_proj_bias.fp16 b/deepseek-r1-1.5b-packed/model_layers_25_self_attn_q_proj_bias.fp16 new file mode 100644 index 0000000000000000000000000000000000000000..410431ac742d3ccc6fd4615eb565773e2aa1f755 Binary files /dev/null and b/deepseek-r1-1.5b-packed/model_layers_25_self_attn_q_proj_bias.fp16 differ diff --git a/deepseek-r1-1.5b-packed/model_layers_26_self_attn_v_proj_bias.fp16 b/deepseek-r1-1.5b-packed/model_layers_26_self_attn_v_proj_bias.fp16 new file mode 100644 index 0000000000000000000000000000000000000000..989a0bba49e703f8f30ef4a088a69eccd1f5872a Binary files /dev/null and b/deepseek-r1-1.5b-packed/model_layers_26_self_attn_v_proj_bias.fp16 differ diff --git a/deepseek-r1-1.5b-packed/model_layers_27_self_attn_v_proj_weight.scales b/deepseek-r1-1.5b-packed/model_layers_27_self_attn_v_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..922e1c31b73aeda5ea1dd86e771cf724ac1df764 Binary files /dev/null and b/deepseek-r1-1.5b-packed/model_layers_27_self_attn_v_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-packed/model_layers_27_self_attn_v_proj_weight.signs b/deepseek-r1-1.5b-packed/model_layers_27_self_attn_v_proj_weight.signs new file mode 100644 index 0000000000000000000000000000000000000000..075911e4210586b9cb0d3a01902ec36c2d905586 Binary files /dev/null and b/deepseek-r1-1.5b-packed/model_layers_27_self_attn_v_proj_weight.signs differ diff --git a/deepseek-r1-1.5b-packed/model_layers_3_self_attn_q_proj_weight.rmm b/deepseek-r1-1.5b-packed/model_layers_3_self_attn_q_proj_weight.rmm new file mode 100644 index 0000000000000000000000000000000000000000..c86839c59145849e367cc3aa33182b1c6c31a614 --- /dev/null +++ b/deepseek-r1-1.5b-packed/model_layers_3_self_attn_q_proj_weight.rmm @@ -0,0 +1 @@ + \ No newline at end of file diff --git a/deepseek-r1-1.5b-packed/model_layers_3_self_attn_q_proj_weight.scales b/deepseek-r1-1.5b-packed/model_layers_3_self_attn_q_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..b9ed0be039e4f2c1aab8ac5c8660d4394fb1d0a9 Binary files /dev/null and b/deepseek-r1-1.5b-packed/model_layers_3_self_attn_q_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-packed/model_layers_3_self_attn_v_proj_weight.scales b/deepseek-r1-1.5b-packed/model_layers_3_self_attn_v_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..fc64201b6dc14b00ff0e2d9383d23ce1a95e7b0f Binary files /dev/null and b/deepseek-r1-1.5b-packed/model_layers_3_self_attn_v_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-packed/model_layers_4_self_attn_v_proj_weight.signs b/deepseek-r1-1.5b-packed/model_layers_4_self_attn_v_proj_weight.signs new file mode 100644 index 0000000000000000000000000000000000000000..5e9a333c25f3936527245b9f270b9c80343b135b Binary files /dev/null and b/deepseek-r1-1.5b-packed/model_layers_4_self_attn_v_proj_weight.signs differ diff --git a/deepseek-r1-1.5b-packed/model_layers_5_self_attn_k_proj_bias.fp16 b/deepseek-r1-1.5b-packed/model_layers_5_self_attn_k_proj_bias.fp16 new file mode 100644 index 0000000000000000000000000000000000000000..770c264cb5354136f1f8a94488d3d68584801d04 Binary files /dev/null and b/deepseek-r1-1.5b-packed/model_layers_5_self_attn_k_proj_bias.fp16 differ diff --git a/deepseek-r1-1.5b-packed/model_layers_5_self_attn_k_proj_weight.scales b/deepseek-r1-1.5b-packed/model_layers_5_self_attn_k_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..73f209462b595cae21608b0f7176940d808b6258 Binary files /dev/null and b/deepseek-r1-1.5b-packed/model_layers_5_self_attn_k_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-packed/model_layers_5_self_attn_o_proj_weight.rmm b/deepseek-r1-1.5b-packed/model_layers_5_self_attn_o_proj_weight.rmm new file mode 100644 index 0000000000000000000000000000000000000000..c86839c59145849e367cc3aa33182b1c6c31a614 --- /dev/null +++ b/deepseek-r1-1.5b-packed/model_layers_5_self_attn_o_proj_weight.rmm @@ -0,0 +1 @@ + \ No newline at end of file diff --git a/deepseek-r1-1.5b-packed/model_layers_5_self_attn_o_proj_weight.scales b/deepseek-r1-1.5b-packed/model_layers_5_self_attn_o_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..14a7a5670850d5831f031ed13229270b2f3bb9cc Binary files /dev/null and b/deepseek-r1-1.5b-packed/model_layers_5_self_attn_o_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-packed/model_layers_6_mlp_down_proj_weight.rmm b/deepseek-r1-1.5b-packed/model_layers_6_mlp_down_proj_weight.rmm new file mode 100644 index 0000000000000000000000000000000000000000..c86839c59145849e367cc3aa33182b1c6c31a614 --- /dev/null +++ b/deepseek-r1-1.5b-packed/model_layers_6_mlp_down_proj_weight.rmm @@ -0,0 +1 @@ + \ No newline at end of file diff --git a/deepseek-r1-1.5b-packed/model_layers_7_mlp_up_proj_weight.rmm b/deepseek-r1-1.5b-packed/model_layers_7_mlp_up_proj_weight.rmm new file mode 100644 index 0000000000000000000000000000000000000000..3d5a76a5a0c43488828c17fc8a7002591a508098 --- /dev/null +++ b/deepseek-r1-1.5b-packed/model_layers_7_mlp_up_proj_weight.rmm @@ -0,0 +1 @@ + \ No newline at end of file diff --git a/deepseek-r1-1.5b-packed/model_layers_7_self_attn_k_proj_weight.scales b/deepseek-r1-1.5b-packed/model_layers_7_self_attn_k_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..f547eab26f3f5a68e9709a358460a4716a9581bf Binary files /dev/null and b/deepseek-r1-1.5b-packed/model_layers_7_self_attn_k_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-packed/model_layers_7_self_attn_q_proj_weight.rmm b/deepseek-r1-1.5b-packed/model_layers_7_self_attn_q_proj_weight.rmm new file mode 100644 index 0000000000000000000000000000000000000000..c86839c59145849e367cc3aa33182b1c6c31a614 --- /dev/null +++ b/deepseek-r1-1.5b-packed/model_layers_7_self_attn_q_proj_weight.rmm @@ -0,0 +1 @@ + \ No newline at end of file diff --git a/deepseek-r1-1.5b-packed/model_layers_8_self_attn_q_proj_weight.scales b/deepseek-r1-1.5b-packed/model_layers_8_self_attn_q_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..31b65c06bb9ae8b0ec0e238a79bda6e97a8e3b55 Binary files /dev/null and b/deepseek-r1-1.5b-packed/model_layers_8_self_attn_q_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-packed/model_layers_9_self_attn_q_proj_bias.fp16 b/deepseek-r1-1.5b-packed/model_layers_9_self_attn_q_proj_bias.fp16 new file mode 100644 index 0000000000000000000000000000000000000000..481af2f5d19a52afd43711747fe2739a46864394 Binary files /dev/null and b/deepseek-r1-1.5b-packed/model_layers_9_self_attn_q_proj_bias.fp16 differ diff --git a/deepseek-r1-1.5b-packed/model_layers_9_self_attn_q_proj_weight.rmm b/deepseek-r1-1.5b-packed/model_layers_9_self_attn_q_proj_weight.rmm new file mode 100644 index 0000000000000000000000000000000000000000..c86839c59145849e367cc3aa33182b1c6c31a614 --- /dev/null +++ b/deepseek-r1-1.5b-packed/model_layers_9_self_attn_q_proj_weight.rmm @@ -0,0 +1 @@ + \ No newline at end of file diff --git a/deepseek-r1-1.5b-ternary/model_layers_0_mlp_up_proj_weight.scales b/deepseek-r1-1.5b-ternary/model_layers_0_mlp_up_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..fd067b43060516755f51d3b659a73da9e7461059 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_0_mlp_up_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_0_post_attention_layernorm_weight.fp16 b/deepseek-r1-1.5b-ternary/model_layers_0_post_attention_layernorm_weight.fp16 new file mode 100644 index 0000000000000000000000000000000000000000..4480c766ecab325c11ba7ca372dbd090f47e5b66 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_0_post_attention_layernorm_weight.fp16 differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_0_self_attn_q_proj_bias.fp16 b/deepseek-r1-1.5b-ternary/model_layers_0_self_attn_q_proj_bias.fp16 new file mode 100644 index 0000000000000000000000000000000000000000..2c352bc4acbcfe7d8e68b1208ac895328786874d Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_0_self_attn_q_proj_bias.fp16 differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_10_self_attn_k_proj_bias.fp16 b/deepseek-r1-1.5b-ternary/model_layers_10_self_attn_k_proj_bias.fp16 new file mode 100644 index 0000000000000000000000000000000000000000..f12798d185f3a26d7b58dca998da4ad94d1d0a4c Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_10_self_attn_k_proj_bias.fp16 differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_10_self_attn_o_proj_weight.scales b/deepseek-r1-1.5b-ternary/model_layers_10_self_attn_o_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..e1fc5c1ba8c6d4e63f5d0507e8c9529882f8a90c Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_10_self_attn_o_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_10_self_attn_q_proj_weight.scales b/deepseek-r1-1.5b-ternary/model_layers_10_self_attn_q_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..4d00b57f8ac629b56fd96673a6405354114be17c Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_10_self_attn_q_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_10_self_attn_v_proj_bias.fp16 b/deepseek-r1-1.5b-ternary/model_layers_10_self_attn_v_proj_bias.fp16 new file mode 100644 index 0000000000000000000000000000000000000000..1f3165213bff8155723b1ad8450afb9bab120399 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_10_self_attn_v_proj_bias.fp16 differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_10_self_attn_v_proj_weight.pos b/deepseek-r1-1.5b-ternary/model_layers_10_self_attn_v_proj_weight.pos new file mode 100644 index 0000000000000000000000000000000000000000..661a94cbf1127e76ee36194d8940f288eb996992 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_10_self_attn_v_proj_weight.pos differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_11_input_layernorm_weight.fp16 b/deepseek-r1-1.5b-ternary/model_layers_11_input_layernorm_weight.fp16 new file mode 100644 index 0000000000000000000000000000000000000000..eacc9d59d93c4c2cb731f52dbd3be00600d8ba57 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_11_input_layernorm_weight.fp16 differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_11_mlp_down_proj_weight.scales b/deepseek-r1-1.5b-ternary/model_layers_11_mlp_down_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..4ad50b780241af2d1352d8fa21b6339de1a6ff14 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_11_mlp_down_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_11_self_attn_v_proj_weight.pos b/deepseek-r1-1.5b-ternary/model_layers_11_self_attn_v_proj_weight.pos new file mode 100644 index 0000000000000000000000000000000000000000..71fcbfd249ed2411cc58ffc336ad832a07d4dae1 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_11_self_attn_v_proj_weight.pos differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_11_self_attn_v_proj_weight.scales b/deepseek-r1-1.5b-ternary/model_layers_11_self_attn_v_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..edac18e65d42880e15c93f9998459e86fcf09e0e Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_11_self_attn_v_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_12_mlp_down_proj_weight.scales b/deepseek-r1-1.5b-ternary/model_layers_12_mlp_down_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..8acaaa9739bae466e1392c67b0baa70e1afea093 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_12_mlp_down_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_12_mlp_up_proj_weight.scales b/deepseek-r1-1.5b-ternary/model_layers_12_mlp_up_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..87d32ceea14e66692e6c383b35bcb5cc9b96dea6 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_12_mlp_up_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_12_post_attention_layernorm_weight.fp16 b/deepseek-r1-1.5b-ternary/model_layers_12_post_attention_layernorm_weight.fp16 new file mode 100644 index 0000000000000000000000000000000000000000..0fb4e1e8dc6a981526052aec4e271aa819bd9dd0 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_12_post_attention_layernorm_weight.fp16 differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_12_self_attn_k_proj_weight.neg b/deepseek-r1-1.5b-ternary/model_layers_12_self_attn_k_proj_weight.neg new file mode 100644 index 0000000000000000000000000000000000000000..979c59a3a9032c789af9c162ff42d7584ba30114 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_12_self_attn_k_proj_weight.neg differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_12_self_attn_v_proj_weight.neg b/deepseek-r1-1.5b-ternary/model_layers_12_self_attn_v_proj_weight.neg new file mode 100644 index 0000000000000000000000000000000000000000..3a7134568c04b77fb75cbaeeb60f8ac03e35cd96 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_12_self_attn_v_proj_weight.neg differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_13_post_attention_layernorm_weight.fp16 b/deepseek-r1-1.5b-ternary/model_layers_13_post_attention_layernorm_weight.fp16 new file mode 100644 index 0000000000000000000000000000000000000000..408a4fe4330502360320954c9e4c94350380eafe Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_13_post_attention_layernorm_weight.fp16 differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_13_self_attn_k_proj_weight.neg b/deepseek-r1-1.5b-ternary/model_layers_13_self_attn_k_proj_weight.neg new file mode 100644 index 0000000000000000000000000000000000000000..66dc48ca9ef3bb16e132a2fb83d108a884eda1fa Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_13_self_attn_k_proj_weight.neg differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_13_self_attn_o_proj_weight.scales b/deepseek-r1-1.5b-ternary/model_layers_13_self_attn_o_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..3f5dbe8628df7df8e12c081644b3ca551a09a67c Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_13_self_attn_o_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_13_self_attn_q_proj_bias.fp16 b/deepseek-r1-1.5b-ternary/model_layers_13_self_attn_q_proj_bias.fp16 new file mode 100644 index 0000000000000000000000000000000000000000..3037fc4357b76a73522638553ed5df0a87ece153 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_13_self_attn_q_proj_bias.fp16 differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_13_self_attn_v_proj_bias.fp16 b/deepseek-r1-1.5b-ternary/model_layers_13_self_attn_v_proj_bias.fp16 new file mode 100644 index 0000000000000000000000000000000000000000..a3117acfad675e99118bcbf2cc87a0db9725a61b Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_13_self_attn_v_proj_bias.fp16 differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_13_self_attn_v_proj_weight.scales b/deepseek-r1-1.5b-ternary/model_layers_13_self_attn_v_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..00f00649b65a4bcc4149ed34c740b633279c4e2d Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_13_self_attn_v_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_14_input_layernorm_weight.fp16 b/deepseek-r1-1.5b-ternary/model_layers_14_input_layernorm_weight.fp16 new file mode 100644 index 0000000000000000000000000000000000000000..e6b9907c05e8517ba0c33c7c2bd8a1ec773f2920 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_14_input_layernorm_weight.fp16 differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_14_post_attention_layernorm_weight.fp16 b/deepseek-r1-1.5b-ternary/model_layers_14_post_attention_layernorm_weight.fp16 new file mode 100644 index 0000000000000000000000000000000000000000..90cf66bf7f23dbce533a9156464500f3ce2edc41 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_14_post_attention_layernorm_weight.fp16 differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_14_self_attn_k_proj_weight.pos b/deepseek-r1-1.5b-ternary/model_layers_14_self_attn_k_proj_weight.pos new file mode 100644 index 0000000000000000000000000000000000000000..255765c37b2a0533f0bc24274a461e61ec3a14c0 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_14_self_attn_k_proj_weight.pos differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_14_self_attn_k_proj_weight.scales b/deepseek-r1-1.5b-ternary/model_layers_14_self_attn_k_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..7a0d052195cbec3a51189a415b536bf87484ed42 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_14_self_attn_k_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_14_self_attn_q_proj_weight.scales b/deepseek-r1-1.5b-ternary/model_layers_14_self_attn_q_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..11fc3593ea9670d29d529e25ca5f6f4a54597dd5 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_14_self_attn_q_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_14_self_attn_v_proj_weight.pos b/deepseek-r1-1.5b-ternary/model_layers_14_self_attn_v_proj_weight.pos new file mode 100644 index 0000000000000000000000000000000000000000..4373f508561b11264c099c42eab72e5199c34e6a Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_14_self_attn_v_proj_weight.pos differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_15_post_attention_layernorm_weight.fp16 b/deepseek-r1-1.5b-ternary/model_layers_15_post_attention_layernorm_weight.fp16 new file mode 100644 index 0000000000000000000000000000000000000000..6a3082c006e0687aaff28354dbf020b4d36b5cde Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_15_post_attention_layernorm_weight.fp16 differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_15_self_attn_q_proj_weight.scales b/deepseek-r1-1.5b-ternary/model_layers_15_self_attn_q_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..eeb856d0ac85a8ae497cc5886d85137c71902375 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_15_self_attn_q_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_15_self_attn_v_proj_bias.fp16 b/deepseek-r1-1.5b-ternary/model_layers_15_self_attn_v_proj_bias.fp16 new file mode 100644 index 0000000000000000000000000000000000000000..bedd656cb61e1b16ec719cd5cefd43ad5eea50d2 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_15_self_attn_v_proj_bias.fp16 differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_15_self_attn_v_proj_weight.neg b/deepseek-r1-1.5b-ternary/model_layers_15_self_attn_v_proj_weight.neg new file mode 100644 index 0000000000000000000000000000000000000000..1995b3be6452c29fe818c5510175fb5e54298cc4 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_15_self_attn_v_proj_weight.neg differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_15_self_attn_v_proj_weight.pos b/deepseek-r1-1.5b-ternary/model_layers_15_self_attn_v_proj_weight.pos new file mode 100644 index 0000000000000000000000000000000000000000..565ebb370f58959d078d171bad1c137f7ec6e259 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_15_self_attn_v_proj_weight.pos differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_15_self_attn_v_proj_weight.scales b/deepseek-r1-1.5b-ternary/model_layers_15_self_attn_v_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..bd9cc07a57debc38ba247b5d249c622dfed65a09 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_15_self_attn_v_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_16_self_attn_k_proj_weight.pos b/deepseek-r1-1.5b-ternary/model_layers_16_self_attn_k_proj_weight.pos new file mode 100644 index 0000000000000000000000000000000000000000..ab16e6c5fa652e405823ba04333c87645d9f94d6 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_16_self_attn_k_proj_weight.pos differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_16_self_attn_o_proj_weight.scales b/deepseek-r1-1.5b-ternary/model_layers_16_self_attn_o_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..847f2e9e1d10ee92fa2ef884a20450f8a894c80e Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_16_self_attn_o_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_17_mlp_down_proj_weight.scales b/deepseek-r1-1.5b-ternary/model_layers_17_mlp_down_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..a4f49fa2f298e1fca9d362fe9e890f176a32c243 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_17_mlp_down_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_17_mlp_gate_proj_weight.scales b/deepseek-r1-1.5b-ternary/model_layers_17_mlp_gate_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..4dadfb0ea4863849fe70edf9690176dd9c670ae5 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_17_mlp_gate_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_17_self_attn_k_proj_bias.fp16 b/deepseek-r1-1.5b-ternary/model_layers_17_self_attn_k_proj_bias.fp16 new file mode 100644 index 0000000000000000000000000000000000000000..d44cca0dd8ea6a11bbd4a9bbdc6575b50c34f948 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_17_self_attn_k_proj_bias.fp16 differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_17_self_attn_k_proj_weight.pos b/deepseek-r1-1.5b-ternary/model_layers_17_self_attn_k_proj_weight.pos new file mode 100644 index 0000000000000000000000000000000000000000..2732aac03c3c4719aad35fdcbcb2e577d71bd0eb Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_17_self_attn_k_proj_weight.pos differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_17_self_attn_v_proj_weight.pos b/deepseek-r1-1.5b-ternary/model_layers_17_self_attn_v_proj_weight.pos new file mode 100644 index 0000000000000000000000000000000000000000..ac0dd6198b95148d0624b075ca8ed0556fa730c2 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_17_self_attn_v_proj_weight.pos differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_18_post_attention_layernorm_weight.fp16 b/deepseek-r1-1.5b-ternary/model_layers_18_post_attention_layernorm_weight.fp16 new file mode 100644 index 0000000000000000000000000000000000000000..ed1a9c348a0afdb0628a14416dd97e4f101469b7 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_18_post_attention_layernorm_weight.fp16 differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_18_self_attn_k_proj_weight.neg b/deepseek-r1-1.5b-ternary/model_layers_18_self_attn_k_proj_weight.neg new file mode 100644 index 0000000000000000000000000000000000000000..f9a80a006af781d757776d153319cd0b0a20e3b4 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_18_self_attn_k_proj_weight.neg differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_18_self_attn_v_proj_weight.pos b/deepseek-r1-1.5b-ternary/model_layers_18_self_attn_v_proj_weight.pos new file mode 100644 index 0000000000000000000000000000000000000000..ceadb3fda72bbd2fced7007e0c92212209070de8 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_18_self_attn_v_proj_weight.pos differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_19_mlp_down_proj_weight.scales b/deepseek-r1-1.5b-ternary/model_layers_19_mlp_down_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..6a5566acb3ca3ccd99d362a6c2fbc1f3a72b6102 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_19_mlp_down_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_19_mlp_up_proj_weight.scales b/deepseek-r1-1.5b-ternary/model_layers_19_mlp_up_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..1bd657340fdf6c92cc5b9094d04ada039b163ba4 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_19_mlp_up_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_19_self_attn_k_proj_bias.fp16 b/deepseek-r1-1.5b-ternary/model_layers_19_self_attn_k_proj_bias.fp16 new file mode 100644 index 0000000000000000000000000000000000000000..e9bfa94937c80e227e43b2591120196b57471c90 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_19_self_attn_k_proj_bias.fp16 differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_19_self_attn_k_proj_weight.scales b/deepseek-r1-1.5b-ternary/model_layers_19_self_attn_k_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..265a6e5d589b40d1b8eba21fecba809c421dc65b Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_19_self_attn_k_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_19_self_attn_q_proj_bias.fp16 b/deepseek-r1-1.5b-ternary/model_layers_19_self_attn_q_proj_bias.fp16 new file mode 100644 index 0000000000000000000000000000000000000000..0385ee4a6ec2b0e55b92dbd7e5ac8995d8a1d4b4 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_19_self_attn_q_proj_bias.fp16 differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_19_self_attn_q_proj_weight.scales b/deepseek-r1-1.5b-ternary/model_layers_19_self_attn_q_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..53bc1a3859f990804c284a963e60a46cf0a732ab Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_19_self_attn_q_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_19_self_attn_v_proj_bias.fp16 b/deepseek-r1-1.5b-ternary/model_layers_19_self_attn_v_proj_bias.fp16 new file mode 100644 index 0000000000000000000000000000000000000000..3a20a9baba7e7b5f11f0c545db427ef58d85c276 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_19_self_attn_v_proj_bias.fp16 differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_19_self_attn_v_proj_weight.neg b/deepseek-r1-1.5b-ternary/model_layers_19_self_attn_v_proj_weight.neg new file mode 100644 index 0000000000000000000000000000000000000000..c9d58ea62de5f9374c49978eec13ed807ffb7a5f Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_19_self_attn_v_proj_weight.neg differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_19_self_attn_v_proj_weight.scales b/deepseek-r1-1.5b-ternary/model_layers_19_self_attn_v_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..762888cf591b295632a64e9d89994d13b04891df Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_19_self_attn_v_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_1_input_layernorm_weight.fp16 b/deepseek-r1-1.5b-ternary/model_layers_1_input_layernorm_weight.fp16 new file mode 100644 index 0000000000000000000000000000000000000000..58cfcc62ae22017c2355f5abacc4896ccebb9722 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_1_input_layernorm_weight.fp16 differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_1_mlp_up_proj_weight.scales b/deepseek-r1-1.5b-ternary/model_layers_1_mlp_up_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..fcc9c3382989febb93207b8019c8171034cf96e0 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_1_mlp_up_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_1_post_attention_layernorm_weight.fp16 b/deepseek-r1-1.5b-ternary/model_layers_1_post_attention_layernorm_weight.fp16 new file mode 100644 index 0000000000000000000000000000000000000000..2d66603de162864edd6c64e2fa5a308470e3bd81 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_1_post_attention_layernorm_weight.fp16 differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_1_self_attn_k_proj_bias.fp16 b/deepseek-r1-1.5b-ternary/model_layers_1_self_attn_k_proj_bias.fp16 new file mode 100644 index 0000000000000000000000000000000000000000..c75ebdad2866070c227bef111584780a52937315 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_1_self_attn_k_proj_bias.fp16 differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_1_self_attn_k_proj_weight.pos b/deepseek-r1-1.5b-ternary/model_layers_1_self_attn_k_proj_weight.pos new file mode 100644 index 0000000000000000000000000000000000000000..21a20cd575c3fe0d8b4888777d077f605ddde24c Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_1_self_attn_k_proj_weight.pos differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_1_self_attn_v_proj_bias.fp16 b/deepseek-r1-1.5b-ternary/model_layers_1_self_attn_v_proj_bias.fp16 new file mode 100644 index 0000000000000000000000000000000000000000..b3c2a5838c67fa4a37e84df45a2ca5ca5edf8c05 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_1_self_attn_v_proj_bias.fp16 differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_20_mlp_gate_proj_weight.scales b/deepseek-r1-1.5b-ternary/model_layers_20_mlp_gate_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..1cf0b0a9f40a995d10eae061cc6c6c76c88e68a8 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_20_mlp_gate_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_20_mlp_up_proj_weight.scales b/deepseek-r1-1.5b-ternary/model_layers_20_mlp_up_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..1e1f10b4d1390b105e9026f89500c6e4eb276214 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_20_mlp_up_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_20_self_attn_o_proj_weight.scales b/deepseek-r1-1.5b-ternary/model_layers_20_self_attn_o_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..49455014539daf19284f186aeae0c796b9a74f21 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_20_self_attn_o_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_20_self_attn_v_proj_bias.fp16 b/deepseek-r1-1.5b-ternary/model_layers_20_self_attn_v_proj_bias.fp16 new file mode 100644 index 0000000000000000000000000000000000000000..e1a165cb2b8233c01da78474d8e7d4efaca539a8 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_20_self_attn_v_proj_bias.fp16 differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_20_self_attn_v_proj_weight.scales b/deepseek-r1-1.5b-ternary/model_layers_20_self_attn_v_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..8fb00b62d2d88fdd668a888c889d91199aa0c867 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_20_self_attn_v_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_21_input_layernorm_weight.fp16 b/deepseek-r1-1.5b-ternary/model_layers_21_input_layernorm_weight.fp16 new file mode 100644 index 0000000000000000000000000000000000000000..d5442af4c06c3d73058f742d5dcc76b443362488 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_21_input_layernorm_weight.fp16 differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_21_mlp_gate_proj_weight.scales b/deepseek-r1-1.5b-ternary/model_layers_21_mlp_gate_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..9f4d18f8eca18dd306dbf90a2fee29bf26e4f62d Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_21_mlp_gate_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_21_self_attn_k_proj_weight.scales b/deepseek-r1-1.5b-ternary/model_layers_21_self_attn_k_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..aa4e9176404625c6038330326adec3b9b1c7c8ec Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_21_self_attn_k_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_21_self_attn_o_proj_weight.scales b/deepseek-r1-1.5b-ternary/model_layers_21_self_attn_o_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..41f535e02ac9d69f391cb8794e12cc69cd924b9b Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_21_self_attn_o_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_21_self_attn_q_proj_bias.fp16 b/deepseek-r1-1.5b-ternary/model_layers_21_self_attn_q_proj_bias.fp16 new file mode 100644 index 0000000000000000000000000000000000000000..52070528d3cccbe104ed1b0c462add5e96f45f7f Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_21_self_attn_q_proj_bias.fp16 differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_21_self_attn_v_proj_weight.neg b/deepseek-r1-1.5b-ternary/model_layers_21_self_attn_v_proj_weight.neg new file mode 100644 index 0000000000000000000000000000000000000000..2a653c34367dca2440a397efb22c04bd80080a10 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_21_self_attn_v_proj_weight.neg differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_22_post_attention_layernorm_weight.fp16 b/deepseek-r1-1.5b-ternary/model_layers_22_post_attention_layernorm_weight.fp16 new file mode 100644 index 0000000000000000000000000000000000000000..e5cc5cafbd37ea08214564cf9ab44fc1a3b5e27e Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_22_post_attention_layernorm_weight.fp16 differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_22_self_attn_k_proj_weight.neg b/deepseek-r1-1.5b-ternary/model_layers_22_self_attn_k_proj_weight.neg new file mode 100644 index 0000000000000000000000000000000000000000..b7a4d6d959dd3cd0856f4f4b3f985b991d794ea5 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_22_self_attn_k_proj_weight.neg differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_22_self_attn_k_proj_weight.pos b/deepseek-r1-1.5b-ternary/model_layers_22_self_attn_k_proj_weight.pos new file mode 100644 index 0000000000000000000000000000000000000000..c5b99c0440698886212f5f33e52b8da67dbfe034 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_22_self_attn_k_proj_weight.pos differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_22_self_attn_k_proj_weight.scales b/deepseek-r1-1.5b-ternary/model_layers_22_self_attn_k_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..f510fccb68f3fa3e7b9f92553e05ba4e4a7f8861 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_22_self_attn_k_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_23_self_attn_k_proj_weight.neg b/deepseek-r1-1.5b-ternary/model_layers_23_self_attn_k_proj_weight.neg new file mode 100644 index 0000000000000000000000000000000000000000..e01f0d44c127ccbd3a0d76ffd8313051e0b31435 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_23_self_attn_k_proj_weight.neg differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_23_self_attn_k_proj_weight.scales b/deepseek-r1-1.5b-ternary/model_layers_23_self_attn_k_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..e0cc11e4d8fbb47a274fcb4b77fdd93b1e137c56 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_23_self_attn_k_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_23_self_attn_q_proj_weight.scales b/deepseek-r1-1.5b-ternary/model_layers_23_self_attn_q_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..fbfdfe45b074daeaf4b05779bcab8d0cac72deb1 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_23_self_attn_q_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_23_self_attn_v_proj_weight.scales b/deepseek-r1-1.5b-ternary/model_layers_23_self_attn_v_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..8921c4263dd9375d76931bb3d158373bea9873f6 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_23_self_attn_v_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_24_input_layernorm_weight.fp16 b/deepseek-r1-1.5b-ternary/model_layers_24_input_layernorm_weight.fp16 new file mode 100644 index 0000000000000000000000000000000000000000..93e70fde36a686eebd1aec80445a130a6481199c Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_24_input_layernorm_weight.fp16 differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_24_mlp_up_proj_weight.scales b/deepseek-r1-1.5b-ternary/model_layers_24_mlp_up_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..ca1a79ae67df1b208970b74086c605dd95b4271f Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_24_mlp_up_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_24_self_attn_k_proj_weight.neg b/deepseek-r1-1.5b-ternary/model_layers_24_self_attn_k_proj_weight.neg new file mode 100644 index 0000000000000000000000000000000000000000..dc70f3c536577b3ba39546d8e119982974c49134 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_24_self_attn_k_proj_weight.neg differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_24_self_attn_k_proj_weight.scales b/deepseek-r1-1.5b-ternary/model_layers_24_self_attn_k_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..c3476e47e38daa89bfef6b65a2a106545c0ba446 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_24_self_attn_k_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_24_self_attn_o_proj_weight.scales b/deepseek-r1-1.5b-ternary/model_layers_24_self_attn_o_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..cf469bc99e9ba275fe46128ad54fa7eace294374 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_24_self_attn_o_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_24_self_attn_q_proj_weight.scales b/deepseek-r1-1.5b-ternary/model_layers_24_self_attn_q_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..7a765ed9690ac05d5cb5ce19c7bda508a208a58a Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_24_self_attn_q_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_25_mlp_gate_proj_weight.scales b/deepseek-r1-1.5b-ternary/model_layers_25_mlp_gate_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..eb556b8fa285a6d58fc9b0caa412160596ea4c14 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_25_mlp_gate_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_25_self_attn_v_proj_bias.fp16 b/deepseek-r1-1.5b-ternary/model_layers_25_self_attn_v_proj_bias.fp16 new file mode 100644 index 0000000000000000000000000000000000000000..142ca0c79cb9915517d2f204fcfc606ac025a0a0 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_25_self_attn_v_proj_bias.fp16 differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_26_mlp_down_proj_weight.scales b/deepseek-r1-1.5b-ternary/model_layers_26_mlp_down_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..65fad430591db0c5c2753eaae9e2c55a058c3c1b Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_26_mlp_down_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_26_self_attn_k_proj_bias.fp16 b/deepseek-r1-1.5b-ternary/model_layers_26_self_attn_k_proj_bias.fp16 new file mode 100644 index 0000000000000000000000000000000000000000..dd99f004d6c98e411bf9172a770ad5f195453c7e Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_26_self_attn_k_proj_bias.fp16 differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_26_self_attn_k_proj_weight.scales b/deepseek-r1-1.5b-ternary/model_layers_26_self_attn_k_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..838cf92c6b3a9c8d4551dfaf374d94ccdd0b3c19 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_26_self_attn_k_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_26_self_attn_q_proj_bias.fp16 b/deepseek-r1-1.5b-ternary/model_layers_26_self_attn_q_proj_bias.fp16 new file mode 100644 index 0000000000000000000000000000000000000000..2c70b4025d2fcb30057f94c604e21dcd3d32c7f4 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_26_self_attn_q_proj_bias.fp16 differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_26_self_attn_v_proj_weight.neg b/deepseek-r1-1.5b-ternary/model_layers_26_self_attn_v_proj_weight.neg new file mode 100644 index 0000000000000000000000000000000000000000..61356aded89be9a5661af0c05a4f3e2ee47011b8 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_26_self_attn_v_proj_weight.neg differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_26_self_attn_v_proj_weight.pos b/deepseek-r1-1.5b-ternary/model_layers_26_self_attn_v_proj_weight.pos new file mode 100644 index 0000000000000000000000000000000000000000..9d149ec3ded76a6b20275f9ed6145ee63caa2628 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_26_self_attn_v_proj_weight.pos differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_27_input_layernorm_weight.fp16 b/deepseek-r1-1.5b-ternary/model_layers_27_input_layernorm_weight.fp16 new file mode 100644 index 0000000000000000000000000000000000000000..708e7b09ea3bdacf953c85c2016e72a4401f6a04 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_27_input_layernorm_weight.fp16 differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_27_mlp_down_proj_weight.scales b/deepseek-r1-1.5b-ternary/model_layers_27_mlp_down_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..45f2d068d5e64590760653ee34de39aa67276459 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_27_mlp_down_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_27_mlp_gate_proj_weight.scales b/deepseek-r1-1.5b-ternary/model_layers_27_mlp_gate_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..4c732699af4f8f207a195ea4b57695bb78f82dc9 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_27_mlp_gate_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_27_post_attention_layernorm_weight.fp16 b/deepseek-r1-1.5b-ternary/model_layers_27_post_attention_layernorm_weight.fp16 new file mode 100644 index 0000000000000000000000000000000000000000..556cc9f350ff7645f4505a070ddacdeef1913989 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_27_post_attention_layernorm_weight.fp16 differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_27_self_attn_k_proj_weight.neg b/deepseek-r1-1.5b-ternary/model_layers_27_self_attn_k_proj_weight.neg new file mode 100644 index 0000000000000000000000000000000000000000..0c4e97a7bbfb07f4f42cc4f2baa7aeff6531d876 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_27_self_attn_k_proj_weight.neg differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_27_self_attn_k_proj_weight.pos b/deepseek-r1-1.5b-ternary/model_layers_27_self_attn_k_proj_weight.pos new file mode 100644 index 0000000000000000000000000000000000000000..195f92de8d8e144482f4088f2697961160e8b2fe Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_27_self_attn_k_proj_weight.pos differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_27_self_attn_k_proj_weight.scales b/deepseek-r1-1.5b-ternary/model_layers_27_self_attn_k_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..2a83de0ab30ceaf15b846369738242a643780d3d Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_27_self_attn_k_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_27_self_attn_o_proj_weight.scales b/deepseek-r1-1.5b-ternary/model_layers_27_self_attn_o_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..25f676317ccb8a212e3c4c71bba9664cac90d4fd Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_27_self_attn_o_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_27_self_attn_v_proj_weight.neg b/deepseek-r1-1.5b-ternary/model_layers_27_self_attn_v_proj_weight.neg new file mode 100644 index 0000000000000000000000000000000000000000..f5674ae13a2071543aa82c9d38c95ebc7f88652e Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_27_self_attn_v_proj_weight.neg differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_2_input_layernorm_weight.fp16 b/deepseek-r1-1.5b-ternary/model_layers_2_input_layernorm_weight.fp16 new file mode 100644 index 0000000000000000000000000000000000000000..d9e2acdd45184532118543b69e8e3ab58c2968b5 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_2_input_layernorm_weight.fp16 differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_2_self_attn_k_proj_weight.neg b/deepseek-r1-1.5b-ternary/model_layers_2_self_attn_k_proj_weight.neg new file mode 100644 index 0000000000000000000000000000000000000000..238a2abaa4debd89526123e92a287df1a8b8eac7 Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_2_self_attn_k_proj_weight.neg differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_2_self_attn_k_proj_weight.scales b/deepseek-r1-1.5b-ternary/model_layers_2_self_attn_k_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..f5de7b38cb7558a779cd62edf5c17cb71eade78e Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_2_self_attn_k_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_2_self_attn_q_proj_weight.scales b/deepseek-r1-1.5b-ternary/model_layers_2_self_attn_q_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..bf28e965c809ad071b6b0d596559ddb49191649a Binary files /dev/null and b/deepseek-r1-1.5b-ternary/model_layers_2_self_attn_q_proj_weight.scales differ diff --git a/deepseek-r1-1.5b-ternary/model_layers_2_self_attn_v_proj_weight.scales b/deepseek-r1-1.5b-ternary/model_layers_2_self_attn_v_proj_weight.scales new file mode 100644 index 0000000000000000000000000000000000000000..00833730b5a447675519f6dc03d33b6976c2bc78 --- /dev/null +++ b/deepseek-r1-1.5b-ternary/model_layers_2_self_attn_v_proj_weight.scales @@ -0,0 +1,6 @@ + ===<===m=6=d!=@=====}=8@=q=%K===O=v=t=S =ڝ===6=k==i=i==+ ==0U=O=# ==.==O=UJ=(W=F?= 7=='=T==_=N]==C!=ɚ=7.==}===J=#=== =dt=*=P==H=j=w=_=Ơ=5=FT=7=== +===#==S ====N=UU== E=a==c=p==P=/==.=9=|R==D==H=h<=s=ۮ=R= =s===׮=G= ==m