# RCO per-tensor GGUF allocation # Repository: ISTA-DASLab/DeepSeek-V4.1-Flash-1017-GSQ-RCO-GGUF # GGUF: DeepSeek-V4.1-Flash-1017-GSQ-RCO-IQ2_XXS-00001-of-00006.gguf (first of 6 shards; this dump covers all of them) # Source: https://huggingface.co/ISTA-DASLab/DeepSeek-V4.1-Flash-1017-GSQ-RCO-GGUF/resolve/main/DeepSeek-V4.1-Flash-1017-GSQ-RCO-IQ2_XXS-00001-of-00006.gguf # Source: https://huggingface.co/ISTA-DASLab/DeepSeek-V4.1-Flash-1017-GSQ-RCO-GGUF/resolve/main/DeepSeek-V4.1-Flash-1017-GSQ-RCO-IQ2_XXS-00002-of-00006.gguf # Source: https://huggingface.co/ISTA-DASLab/DeepSeek-V4.1-Flash-1017-GSQ-RCO-GGUF/resolve/main/DeepSeek-V4.1-Flash-1017-GSQ-RCO-IQ2_XXS-00003-of-00006.gguf # Source: https://huggingface.co/ISTA-DASLab/DeepSeek-V4.1-Flash-1017-GSQ-RCO-GGUF/resolve/main/DeepSeek-V4.1-Flash-1017-GSQ-RCO-IQ2_XXS-00004-of-00006.gguf # Source: https://huggingface.co/ISTA-DASLab/DeepSeek-V4.1-Flash-1017-GSQ-RCO-GGUF/resolve/main/DeepSeek-V4.1-Flash-1017-GSQ-RCO-IQ2_XXS-00005-of-00006.gguf # Source: https://huggingface.co/ISTA-DASLab/DeepSeek-V4.1-Flash-1017-GSQ-RCO-GGUF/resolve/main/DeepSeek-V4.1-Flash-Engram-IQ4_NL.gguf # Target bpw over the backbone (all tensors except the Engram table): 2.00 (measured: 2.0038) # The Engram table (per_layer_token_embd, shard 6 = the shared Engram file) is IQ4_NL in every build and not part of the search. # GGUF version: 3 # GGUF metadata entries: 77 # Tensor count: 1043 # Quant-type counts: BF16=67, F32=531, IQ1_M=60, IQ2_XS=18, IQ2_XXS=42, IQ4_NL=1, Q5_K=244, Q8_0=80 # Format: : # Includes fixed non-RCO tensors (e.g. F32/BF16) so the released GGUF is fully reproducible. # output.weight: Q5_K output_norm.weight: F32 per_layer_token_embd.weight: IQ4_NL token_embd.weight: Q5_K blk.0.attn_kv.weight: Q8_0 blk.0.attn_kv_a_norm.weight: F32 blk.0.attn_norm.weight: F32 blk.0.attn_output_a.weight: Q5_K blk.0.attn_output_b.weight: Q5_K blk.0.attn_q_a.weight: Q8_0 blk.0.attn_q_a_norm.weight: F32 blk.0.attn_q_b.weight: Q5_K blk.0.attn_sinks.weight: F32 blk.0.exp_probs_b.bias: F32 blk.0.exp_probs_b_vl.bias: F32 blk.0.ffn_down_exps.weight: IQ1_M blk.0.ffn_down_shexp.weight: Q5_K blk.0.ffn_gate_exps.weight: IQ1_M blk.0.ffn_gate_inp.weight: BF16 blk.0.ffn_gate_shexp.weight: Q5_K blk.0.ffn_norm.weight: F32 blk.0.ffn_up_exps.weight: IQ1_M blk.0.ffn_up_shexp.weight: Q5_K blk.0.hc_attn_base.weight: F32 blk.0.hc_attn_fn.weight: F32 blk.0.hc_attn_scale.weight: F32 blk.0.hc_ffn_base.weight: F32 blk.0.hc_ffn_fn.weight: F32 blk.0.hc_ffn_scale.weight: F32 blk.1.attn_kv.weight: Q8_0 blk.1.attn_kv_a_norm.weight: F32 blk.1.attn_norm.weight: F32 blk.1.attn_output_a.weight: Q5_K blk.1.attn_output_b.weight: Q5_K blk.1.attn_q_a.weight: Q8_0 blk.1.attn_q_a_norm.weight: F32 blk.1.attn_q_b.weight: Q5_K blk.1.attn_sinks.weight: F32 blk.1.engram_gate.weight: F32 blk.1.engram_kv.weight: Q5_K blk.1.exp_probs_b.bias: F32 blk.1.exp_probs_b_vl.bias: F32 blk.1.ffn_down_exps.weight: IQ1_M blk.1.ffn_down_shexp.weight: Q5_K blk.1.ffn_gate_exps.weight: IQ1_M blk.1.ffn_gate_inp.weight: BF16 blk.1.ffn_gate_shexp.weight: Q5_K blk.1.ffn_norm.weight: F32 blk.1.ffn_up_exps.weight: IQ1_M blk.1.ffn_up_shexp.weight: Q5_K blk.1.hc_attn_base.weight: F32 blk.1.hc_attn_fn.weight: F32 blk.1.hc_attn_scale.weight: F32 blk.1.hc_ffn_base.weight: F32 blk.1.hc_ffn_fn.weight: F32 blk.1.hc_ffn_scale.weight: F32 blk.2.attn_compressor_gate.weight: BF16 blk.2.attn_compressor_kv.weight: BF16 blk.2.attn_compressor_norm.weight: F32 blk.2.attn_kv.weight: Q8_0 blk.2.attn_kv_a_norm.weight: F32 blk.2.attn_norm.weight: F32 blk.2.attn_output_a.weight: Q5_K blk.2.attn_output_b.weight: Q5_K blk.2.attn_q_a.weight: Q8_0 blk.2.attn_q_a_norm.weight: F32 blk.2.attn_q_b.weight: Q5_K blk.2.attn_sinks.weight: F32 blk.2.exp_probs_b.bias: F32 blk.2.exp_probs_b_vl.bias: F32 blk.2.ffn_down_exps.weight: IQ1_M blk.2.ffn_down_shexp.weight: Q5_K blk.2.ffn_gate_exps.weight: IQ1_M blk.2.ffn_gate_inp.weight: BF16 blk.2.ffn_gate_shexp.weight: Q5_K blk.2.ffn_norm.weight: F32 blk.2.ffn_up_exps.weight: IQ1_M blk.2.ffn_up_shexp.weight: Q5_K blk.2.hc_attn_base.weight: F32 blk.2.hc_attn_fn.weight: F32 blk.2.hc_attn_scale.weight: F32 blk.2.hc_ffn_base.weight: F32 blk.2.hc_ffn_fn.weight: F32 blk.2.hc_ffn_scale.weight: F32 blk.2.indexer.attn_k.weight: BF16 blk.2.indexer.attn_q_b.weight: BF16 blk.2.indexer.k_norm.weight: F32 blk.2.indexer.proj.weight: BF16 blk.3.attn_kv.weight: Q8_0 blk.3.attn_kv_a_norm.weight: F32 blk.3.attn_norm.weight: F32 blk.3.attn_output_a.weight: Q5_K blk.3.attn_output_b.weight: Q5_K blk.3.attn_q_a.weight: Q8_0 blk.3.attn_q_a_norm.weight: F32 blk.3.attn_q_b.weight: Q5_K blk.3.attn_sinks.weight: F32 blk.3.exp_probs_b.bias: F32 blk.3.exp_probs_b_vl.bias: F32 blk.3.ffn_down_exps.weight: IQ1_M blk.3.ffn_down_shexp.weight: Q5_K blk.3.ffn_gate_exps.weight: IQ1_M blk.3.ffn_gate_inp.weight: BF16 blk.3.ffn_gate_shexp.weight: Q5_K blk.3.ffn_norm.weight: F32 blk.3.ffn_up_exps.weight: IQ1_M blk.3.ffn_up_shexp.weight: Q5_K blk.3.hc_attn_base.weight: F32 blk.3.hc_attn_fn.weight: F32 blk.3.hc_attn_scale.weight: F32 blk.3.hc_ffn_base.weight: F32 blk.3.hc_ffn_fn.weight: F32 blk.3.hc_ffn_scale.weight: F32 blk.4.attn_kv.weight: Q8_0 blk.4.attn_kv_a_norm.weight: F32 blk.4.attn_norm.weight: F32 blk.4.attn_output_a.weight: Q5_K blk.4.attn_output_b.weight: Q5_K blk.4.attn_q_a.weight: Q8_0 blk.4.attn_q_a_norm.weight: F32 blk.4.attn_q_b.weight: Q5_K blk.4.attn_sinks.weight: F32 blk.4.exp_probs_b.bias: F32 blk.4.exp_probs_b_vl.bias: F32 blk.4.ffn_down_exps.weight: IQ1_M blk.4.ffn_down_shexp.weight: Q5_K blk.4.ffn_gate_exps.weight: IQ1_M blk.4.ffn_gate_inp.weight: BF16 blk.4.ffn_gate_shexp.weight: Q5_K blk.4.ffn_norm.weight: F32 blk.4.ffn_up_exps.weight: IQ1_M blk.4.ffn_up_shexp.weight: Q5_K blk.4.hc_attn_base.weight: F32 blk.4.hc_attn_fn.weight: F32 blk.4.hc_attn_scale.weight: F32 blk.4.hc_ffn_base.weight: F32 blk.4.hc_ffn_fn.weight: F32 blk.4.hc_ffn_scale.weight: F32 blk.5.attn_kv.weight: Q8_0 blk.5.attn_kv_a_norm.weight: F32 blk.5.attn_norm.weight: F32 blk.5.attn_output_a.weight: Q5_K blk.5.attn_output_b.weight: Q5_K blk.5.attn_q_a.weight: Q8_0 blk.5.attn_q_a_norm.weight: F32 blk.5.attn_q_b.weight: Q5_K blk.5.attn_sinks.weight: F32 blk.5.exp_probs_b.bias: F32 blk.5.exp_probs_b_vl.bias: F32 blk.5.ffn_down_exps.weight: IQ1_M blk.5.ffn_down_shexp.weight: Q5_K blk.5.ffn_gate_exps.weight: IQ1_M blk.5.ffn_gate_inp.weight: BF16 blk.5.ffn_gate_shexp.weight: Q5_K blk.5.ffn_norm.weight: F32 blk.5.ffn_up_exps.weight: IQ1_M blk.5.ffn_up_shexp.weight: Q5_K blk.5.hc_attn_base.weight: F32 blk.5.hc_attn_fn.weight: F32 blk.5.hc_attn_scale.weight: F32 blk.5.hc_ffn_base.weight: F32 blk.5.hc_ffn_fn.weight: F32 blk.5.hc_ffn_scale.weight: F32 blk.6.attn_kv.weight: Q8_0 blk.6.attn_kv_a_norm.weight: F32 blk.6.attn_norm.weight: F32 blk.6.attn_output_a.weight: Q5_K blk.6.attn_output_b.weight: Q5_K blk.6.attn_q_a.weight: Q8_0 blk.6.attn_q_a_norm.weight: F32 blk.6.attn_q_b.weight: Q5_K blk.6.attn_sinks.weight: F32 blk.6.exp_probs_b.bias: F32 blk.6.exp_probs_b_vl.bias: F32 blk.6.ffn_down_exps.weight: IQ1_M blk.6.ffn_down_shexp.weight: Q5_K blk.6.ffn_gate_exps.weight: IQ1_M blk.6.ffn_gate_inp.weight: BF16 blk.6.ffn_gate_shexp.weight: Q5_K blk.6.ffn_norm.weight: F32 blk.6.ffn_up_exps.weight: IQ1_M blk.6.ffn_up_shexp.weight: Q5_K blk.6.hc_attn_base.weight: F32 blk.6.hc_attn_fn.weight: F32 blk.6.hc_attn_scale.weight: F32 blk.6.hc_ffn_base.weight: F32 blk.6.hc_ffn_fn.weight: F32 blk.6.hc_ffn_scale.weight: F32 blk.7.attn_kv.weight: Q8_0 blk.7.attn_kv_a_norm.weight: F32 blk.7.attn_norm.weight: F32 blk.7.attn_output_a.weight: Q5_K blk.7.attn_output_b.weight: Q5_K blk.7.attn_q_a.weight: Q8_0 blk.7.attn_q_a_norm.weight: F32 blk.7.attn_q_b.weight: Q5_K blk.7.attn_sinks.weight: F32 blk.7.exp_probs_b.bias: F32 blk.7.exp_probs_b_vl.bias: F32 blk.7.ffn_down_exps.weight: IQ2_XXS blk.7.ffn_down_shexp.weight: Q5_K blk.7.ffn_gate_exps.weight: IQ1_M blk.7.ffn_gate_inp.weight: BF16 blk.7.ffn_gate_shexp.weight: Q5_K blk.7.ffn_norm.weight: F32 blk.7.ffn_up_exps.weight: IQ1_M blk.7.ffn_up_shexp.weight: Q5_K blk.7.hc_attn_base.weight: F32 blk.7.hc_attn_fn.weight: F32 blk.7.hc_attn_scale.weight: F32 blk.7.hc_ffn_base.weight: F32 blk.7.hc_ffn_fn.weight: F32 blk.7.hc_ffn_scale.weight: F32 blk.8.attn_compressor_gate.weight: BF16 blk.8.attn_compressor_kv.weight: BF16 blk.8.attn_compressor_norm.weight: F32 blk.8.attn_kv.weight: Q8_0 blk.8.attn_kv_a_norm.weight: F32 blk.8.attn_norm.weight: F32 blk.8.attn_output_a.weight: Q5_K blk.8.attn_output_b.weight: Q5_K blk.8.attn_q_a.weight: Q8_0 blk.8.attn_q_a_norm.weight: F32 blk.8.attn_q_b.weight: Q5_K blk.8.attn_sinks.weight: F32 blk.8.exp_probs_b.bias: F32 blk.8.exp_probs_b_vl.bias: F32 blk.8.ffn_down_exps.weight: IQ2_XXS blk.8.ffn_down_shexp.weight: Q5_K blk.8.ffn_gate_exps.weight: IQ1_M blk.8.ffn_gate_inp.weight: BF16 blk.8.ffn_gate_shexp.weight: Q5_K blk.8.ffn_norm.weight: F32 blk.8.ffn_up_exps.weight: IQ1_M blk.8.ffn_up_shexp.weight: Q5_K blk.8.hc_attn_base.weight: F32 blk.8.hc_attn_fn.weight: F32 blk.8.hc_attn_scale.weight: F32 blk.8.hc_ffn_base.weight: F32 blk.8.hc_ffn_fn.weight: F32 blk.8.hc_ffn_scale.weight: F32 blk.8.indexer.attn_k.weight: BF16 blk.8.indexer.attn_q_b.weight: BF16 blk.8.indexer.k_norm.weight: F32 blk.8.indexer.proj.weight: BF16 blk.9.attn_kv.weight: Q8_0 blk.9.attn_kv_a_norm.weight: F32 blk.9.attn_norm.weight: F32 blk.9.attn_output_a.weight: Q5_K blk.9.attn_output_b.weight: Q5_K blk.9.attn_q_a.weight: Q8_0 blk.9.attn_q_a_norm.weight: F32 blk.9.attn_q_b.weight: Q5_K blk.9.attn_sinks.weight: F32 blk.9.exp_probs_b.bias: F32 blk.9.exp_probs_b_vl.bias: F32 blk.9.ffn_down_exps.weight: IQ1_M blk.9.ffn_down_shexp.weight: Q5_K blk.9.ffn_gate_exps.weight: IQ1_M blk.9.ffn_gate_inp.weight: BF16 blk.9.ffn_gate_shexp.weight: Q5_K blk.9.ffn_norm.weight: F32 blk.9.ffn_up_exps.weight: IQ1_M blk.9.ffn_up_shexp.weight: Q5_K blk.9.hc_attn_base.weight: F32 blk.9.hc_attn_fn.weight: F32 blk.9.hc_attn_scale.weight: F32 blk.9.hc_ffn_base.weight: F32 blk.9.hc_ffn_fn.weight: F32 blk.9.hc_ffn_scale.weight: F32 blk.10.attn_kv.weight: Q8_0 blk.10.attn_kv_a_norm.weight: F32 blk.10.attn_norm.weight: F32 blk.10.attn_output_a.weight: Q5_K blk.10.attn_output_b.weight: Q5_K blk.10.attn_q_a.weight: Q8_0 blk.10.attn_q_a_norm.weight: F32 blk.10.attn_q_b.weight: Q5_K blk.10.attn_sinks.weight: F32 blk.10.exp_probs_b.bias: F32 blk.10.exp_probs_b_vl.bias: F32 blk.10.ffn_down_exps.weight: IQ2_XXS blk.10.ffn_down_shexp.weight: Q5_K blk.10.ffn_gate_exps.weight: IQ2_XXS blk.10.ffn_gate_inp.weight: BF16 blk.10.ffn_gate_shexp.weight: Q5_K blk.10.ffn_norm.weight: F32 blk.10.ffn_up_exps.weight: IQ2_XXS blk.10.ffn_up_shexp.weight: Q5_K blk.10.hc_attn_base.weight: F32 blk.10.hc_attn_fn.weight: F32 blk.10.hc_attn_scale.weight: F32 blk.10.hc_ffn_base.weight: F32 blk.10.hc_ffn_fn.weight: F32 blk.10.hc_ffn_scale.weight: F32 blk.11.attn_kv.weight: Q8_0 blk.11.attn_kv_a_norm.weight: F32 blk.11.attn_norm.weight: F32 blk.11.attn_output_a.weight: Q5_K blk.11.attn_output_b.weight: Q5_K blk.11.attn_q_a.weight: Q8_0 blk.11.attn_q_a_norm.weight: F32 blk.11.attn_q_b.weight: Q5_K blk.11.attn_sinks.weight: F32 blk.11.exp_probs_b.bias: F32 blk.11.exp_probs_b_vl.bias: F32 blk.11.ffn_down_exps.weight: IQ1_M blk.11.ffn_down_shexp.weight: Q5_K blk.11.ffn_gate_exps.weight: IQ2_XXS blk.11.ffn_gate_inp.weight: BF16 blk.11.ffn_gate_shexp.weight: Q5_K blk.11.ffn_norm.weight: F32 blk.11.ffn_up_exps.weight: IQ2_XXS blk.11.ffn_up_shexp.weight: Q5_K blk.11.hc_attn_base.weight: F32 blk.11.hc_attn_fn.weight: F32 blk.11.hc_attn_scale.weight: F32 blk.11.hc_ffn_base.weight: F32 blk.11.hc_ffn_fn.weight: F32 blk.11.hc_ffn_scale.weight: F32 blk.12.attn_kv.weight: Q8_0 blk.12.attn_kv_a_norm.weight: F32 blk.12.attn_norm.weight: F32 blk.12.attn_output_a.weight: Q5_K blk.12.attn_output_b.weight: Q5_K blk.12.attn_q_a.weight: Q8_0 blk.12.attn_q_a_norm.weight: F32 blk.12.attn_q_b.weight: Q5_K blk.12.attn_sinks.weight: F32 blk.12.exp_probs_b.bias: F32 blk.12.exp_probs_b_vl.bias: F32 blk.12.ffn_down_exps.weight: IQ1_M blk.12.ffn_down_shexp.weight: Q5_K blk.12.ffn_gate_exps.weight: IQ2_XXS blk.12.ffn_gate_inp.weight: BF16 blk.12.ffn_gate_shexp.weight: Q5_K blk.12.ffn_norm.weight: F32 blk.12.ffn_up_exps.weight: IQ2_XXS blk.12.ffn_up_shexp.weight: Q5_K blk.12.hc_attn_base.weight: F32 blk.12.hc_attn_fn.weight: F32 blk.12.hc_attn_scale.weight: F32 blk.12.hc_ffn_base.weight: F32 blk.12.hc_ffn_fn.weight: F32 blk.12.hc_ffn_scale.weight: F32 blk.13.attn_kv.weight: Q8_0 blk.13.attn_kv_a_norm.weight: F32 blk.13.attn_norm.weight: F32 blk.13.attn_output_a.weight: Q5_K blk.13.attn_output_b.weight: Q5_K blk.13.attn_q_a.weight: Q8_0 blk.13.attn_q_a_norm.weight: F32 blk.13.attn_q_b.weight: Q5_K blk.13.attn_sinks.weight: F32 blk.13.exp_probs_b.bias: F32 blk.13.exp_probs_b_vl.bias: F32 blk.13.ffn_down_exps.weight: IQ1_M blk.13.ffn_down_shexp.weight: Q5_K blk.13.ffn_gate_exps.weight: IQ1_M blk.13.ffn_gate_inp.weight: BF16 blk.13.ffn_gate_shexp.weight: Q5_K blk.13.ffn_norm.weight: F32 blk.13.ffn_up_exps.weight: IQ1_M blk.13.ffn_up_shexp.weight: Q5_K blk.13.hc_attn_base.weight: F32 blk.13.hc_attn_fn.weight: F32 blk.13.hc_attn_scale.weight: F32 blk.13.hc_ffn_base.weight: F32 blk.13.hc_ffn_fn.weight: F32 blk.13.hc_ffn_scale.weight: F32 blk.14.attn_compressor_gate.weight: BF16 blk.14.attn_compressor_kv.weight: BF16 blk.14.attn_compressor_norm.weight: F32 blk.14.attn_kv.weight: Q8_0 blk.14.attn_kv_a_norm.weight: F32 blk.14.attn_norm.weight: F32 blk.14.attn_output_a.weight: Q5_K blk.14.attn_output_b.weight: Q5_K blk.14.attn_q_a.weight: Q8_0 blk.14.attn_q_a_norm.weight: F32 blk.14.attn_q_b.weight: Q5_K blk.14.attn_sinks.weight: F32 blk.14.engram_gate.weight: F32 blk.14.engram_kv.weight: Q5_K blk.14.exp_probs_b.bias: F32 blk.14.exp_probs_b_vl.bias: F32 blk.14.ffn_down_exps.weight: IQ1_M blk.14.ffn_down_shexp.weight: Q5_K blk.14.ffn_gate_exps.weight: IQ1_M blk.14.ffn_gate_inp.weight: BF16 blk.14.ffn_gate_shexp.weight: Q5_K blk.14.ffn_norm.weight: F32 blk.14.ffn_up_exps.weight: IQ1_M blk.14.ffn_up_shexp.weight: Q5_K blk.14.hc_attn_base.weight: F32 blk.14.hc_attn_fn.weight: F32 blk.14.hc_attn_scale.weight: F32 blk.14.hc_ffn_base.weight: F32 blk.14.hc_ffn_fn.weight: F32 blk.14.hc_ffn_scale.weight: F32 blk.14.indexer.attn_k.weight: BF16 blk.14.indexer.attn_q_b.weight: BF16 blk.14.indexer.k_norm.weight: F32 blk.14.indexer.proj.weight: BF16 blk.15.attn_kv.weight: Q8_0 blk.15.attn_kv_a_norm.weight: F32 blk.15.attn_norm.weight: F32 blk.15.attn_output_a.weight: Q5_K blk.15.attn_output_b.weight: Q5_K blk.15.attn_q_a.weight: Q8_0 blk.15.attn_q_a_norm.weight: F32 blk.15.attn_q_b.weight: Q5_K blk.15.attn_sinks.weight: F32 blk.15.exp_probs_b.bias: F32 blk.15.exp_probs_b_vl.bias: F32 blk.15.ffn_down_exps.weight: IQ1_M blk.15.ffn_down_shexp.weight: Q5_K blk.15.ffn_gate_exps.weight: IQ2_XXS blk.15.ffn_gate_inp.weight: BF16 blk.15.ffn_gate_shexp.weight: Q5_K blk.15.ffn_norm.weight: F32 blk.15.ffn_up_exps.weight: IQ2_XXS blk.15.ffn_up_shexp.weight: Q5_K blk.15.hc_attn_base.weight: F32 blk.15.hc_attn_fn.weight: F32 blk.15.hc_attn_scale.weight: F32 blk.15.hc_ffn_base.weight: F32 blk.15.hc_ffn_fn.weight: F32 blk.15.hc_ffn_scale.weight: F32 blk.16.attn_kv.weight: Q8_0 blk.16.attn_kv_a_norm.weight: F32 blk.16.attn_norm.weight: F32 blk.16.attn_output_a.weight: Q5_K blk.16.attn_output_b.weight: Q5_K blk.16.attn_q_a.weight: Q8_0 blk.16.attn_q_a_norm.weight: F32 blk.16.attn_q_b.weight: Q5_K blk.16.attn_sinks.weight: F32 blk.16.exp_probs_b.bias: F32 blk.16.exp_probs_b_vl.bias: F32 blk.16.ffn_down_exps.weight: IQ1_M blk.16.ffn_down_shexp.weight: Q5_K blk.16.ffn_gate_exps.weight: IQ1_M blk.16.ffn_gate_inp.weight: BF16 blk.16.ffn_gate_shexp.weight: Q5_K blk.16.ffn_norm.weight: F32 blk.16.ffn_up_exps.weight: IQ1_M blk.16.ffn_up_shexp.weight: Q5_K blk.16.hc_attn_base.weight: F32 blk.16.hc_attn_fn.weight: F32 blk.16.hc_attn_scale.weight: F32 blk.16.hc_ffn_base.weight: F32 blk.16.hc_ffn_fn.weight: F32 blk.16.hc_ffn_scale.weight: F32 blk.17.attn_kv.weight: Q8_0 blk.17.attn_kv_a_norm.weight: F32 blk.17.attn_norm.weight: F32 blk.17.attn_output_a.weight: Q5_K blk.17.attn_output_b.weight: Q5_K blk.17.attn_q_a.weight: Q8_0 blk.17.attn_q_a_norm.weight: F32 blk.17.attn_q_b.weight: Q5_K blk.17.attn_sinks.weight: F32 blk.17.exp_probs_b.bias: F32 blk.17.exp_probs_b_vl.bias: F32 blk.17.ffn_down_exps.weight: IQ1_M blk.17.ffn_down_shexp.weight: Q5_K blk.17.ffn_gate_exps.weight: IQ2_XS blk.17.ffn_gate_inp.weight: BF16 blk.17.ffn_gate_shexp.weight: Q5_K blk.17.ffn_norm.weight: F32 blk.17.ffn_up_exps.weight: IQ2_XS blk.17.ffn_up_shexp.weight: Q5_K blk.17.hc_attn_base.weight: F32 blk.17.hc_attn_fn.weight: F32 blk.17.hc_attn_scale.weight: F32 blk.17.hc_ffn_base.weight: F32 blk.17.hc_ffn_fn.weight: F32 blk.17.hc_ffn_scale.weight: F32 blk.18.attn_kv.weight: Q8_0 blk.18.attn_kv_a_norm.weight: F32 blk.18.attn_norm.weight: F32 blk.18.attn_output_a.weight: Q5_K blk.18.attn_output_b.weight: Q5_K blk.18.attn_q_a.weight: Q8_0 blk.18.attn_q_a_norm.weight: F32 blk.18.attn_q_b.weight: Q5_K blk.18.attn_sinks.weight: F32 blk.18.exp_probs_b.bias: F32 blk.18.exp_probs_b_vl.bias: F32 blk.18.ffn_down_exps.weight: IQ2_XXS blk.18.ffn_down_shexp.weight: Q5_K blk.18.ffn_gate_exps.weight: IQ2_XXS blk.18.ffn_gate_inp.weight: BF16 blk.18.ffn_gate_shexp.weight: Q5_K blk.18.ffn_norm.weight: F32 blk.18.ffn_up_exps.weight: IQ2_XXS blk.18.ffn_up_shexp.weight: Q5_K blk.18.hc_attn_base.weight: F32 blk.18.hc_attn_fn.weight: F32 blk.18.hc_attn_scale.weight: F32 blk.18.hc_ffn_base.weight: F32 blk.18.hc_ffn_fn.weight: F32 blk.18.hc_ffn_scale.weight: F32 blk.19.attn_kv.weight: Q8_0 blk.19.attn_kv_a_norm.weight: F32 blk.19.attn_norm.weight: F32 blk.19.attn_output_a.weight: Q5_K blk.19.attn_output_b.weight: Q5_K blk.19.attn_q_a.weight: Q8_0 blk.19.attn_q_a_norm.weight: F32 blk.19.attn_q_b.weight: Q5_K blk.19.attn_sinks.weight: F32 blk.19.exp_probs_b.bias: F32 blk.19.exp_probs_b_vl.bias: F32 blk.19.ffn_down_exps.weight: IQ1_M blk.19.ffn_down_shexp.weight: Q5_K blk.19.ffn_gate_exps.weight: IQ2_XS blk.19.ffn_gate_inp.weight: BF16 blk.19.ffn_gate_shexp.weight: Q5_K blk.19.ffn_norm.weight: F32 blk.19.ffn_up_exps.weight: IQ2_XS blk.19.ffn_up_shexp.weight: Q5_K blk.19.hc_attn_base.weight: F32 blk.19.hc_attn_fn.weight: F32 blk.19.hc_attn_scale.weight: F32 blk.19.hc_ffn_base.weight: F32 blk.19.hc_ffn_fn.weight: F32 blk.19.hc_ffn_scale.weight: F32 blk.20.attn_compressor_kv.weight: BF16 blk.20.attn_compressor_norm.weight: F32 blk.20.attn_kv.weight: Q8_0 blk.20.attn_kv_a_norm.weight: F32 blk.20.attn_norm.weight: F32 blk.20.attn_output_a.weight: Q5_K blk.20.attn_output_b.weight: Q5_K blk.20.attn_q_a.weight: Q8_0 blk.20.attn_q_a_norm.weight: F32 blk.20.attn_q_b.weight: Q5_K blk.20.attn_sinks.weight: F32 blk.20.exp_probs_b.bias: F32 blk.20.exp_probs_b_vl.bias: F32 blk.20.ffn_down_exps.weight: IQ1_M blk.20.ffn_down_shexp.weight: Q5_K blk.20.ffn_gate_exps.weight: IQ1_M blk.20.ffn_gate_inp.weight: BF16 blk.20.ffn_gate_shexp.weight: Q5_K blk.20.ffn_norm.weight: F32 blk.20.ffn_up_exps.weight: IQ1_M blk.20.ffn_up_shexp.weight: Q5_K blk.20.hc_attn_base.weight: F32 blk.20.hc_attn_fn.weight: F32 blk.20.hc_attn_scale.weight: F32 blk.20.hc_ffn_base.weight: F32 blk.20.hc_ffn_fn.weight: F32 blk.20.hc_ffn_scale.weight: F32 blk.20.indexer.attn_k.weight: BF16 blk.20.indexer.attn_q_b.weight: BF16 blk.20.indexer.k_norm.weight: F32 blk.20.indexer.proj.weight: BF16 blk.21.attn_kv.weight: Q8_0 blk.21.attn_kv_a_norm.weight: F32 blk.21.attn_norm.weight: F32 blk.21.attn_output_a.weight: Q5_K blk.21.attn_output_b.weight: Q5_K blk.21.attn_q_a.weight: Q8_0 blk.21.attn_q_a_norm.weight: F32 blk.21.attn_q_b.weight: Q5_K blk.21.attn_sinks.weight: F32 blk.21.exp_probs_b.bias: F32 blk.21.exp_probs_b_vl.bias: F32 blk.21.ffn_down_exps.weight: IQ1_M blk.21.ffn_down_shexp.weight: Q5_K blk.21.ffn_gate_exps.weight: IQ1_M blk.21.ffn_gate_inp.weight: BF16 blk.21.ffn_gate_shexp.weight: Q5_K blk.21.ffn_norm.weight: F32 blk.21.ffn_up_exps.weight: IQ1_M blk.21.ffn_up_shexp.weight: Q5_K blk.21.hc_attn_base.weight: F32 blk.21.hc_attn_fn.weight: F32 blk.21.hc_attn_scale.weight: F32 blk.21.hc_ffn_base.weight: F32 blk.21.hc_ffn_fn.weight: F32 blk.21.hc_ffn_scale.weight: F32 blk.22.attn_kv.weight: Q8_0 blk.22.attn_kv_a_norm.weight: F32 blk.22.attn_norm.weight: F32 blk.22.attn_output_a.weight: Q5_K blk.22.attn_output_b.weight: Q5_K blk.22.attn_q_a.weight: Q8_0 blk.22.attn_q_a_norm.weight: F32 blk.22.attn_q_b.weight: Q5_K blk.22.attn_sinks.weight: F32 blk.22.exp_probs_b.bias: F32 blk.22.exp_probs_b_vl.bias: F32 blk.22.ffn_down_exps.weight: IQ2_XXS blk.22.ffn_down_shexp.weight: Q5_K blk.22.ffn_gate_exps.weight: IQ2_XXS blk.22.ffn_gate_inp.weight: BF16 blk.22.ffn_gate_shexp.weight: Q5_K blk.22.ffn_norm.weight: F32 blk.22.ffn_up_exps.weight: IQ2_XXS blk.22.ffn_up_shexp.weight: Q5_K blk.22.hc_attn_base.weight: F32 blk.22.hc_attn_fn.weight: F32 blk.22.hc_attn_scale.weight: F32 blk.22.hc_ffn_base.weight: F32 blk.22.hc_ffn_fn.weight: F32 blk.22.hc_ffn_scale.weight: F32 blk.23.attn_kv.weight: Q8_0 blk.23.attn_kv_a_norm.weight: F32 blk.23.attn_norm.weight: F32 blk.23.attn_output_a.weight: Q5_K blk.23.attn_output_b.weight: Q5_K blk.23.attn_q_a.weight: Q8_0 blk.23.attn_q_a_norm.weight: F32 blk.23.attn_q_b.weight: Q5_K blk.23.attn_sinks.weight: F32 blk.23.exp_probs_b.bias: F32 blk.23.exp_probs_b_vl.bias: F32 blk.23.ffn_down_exps.weight: IQ2_XS blk.23.ffn_down_shexp.weight: Q5_K blk.23.ffn_gate_exps.weight: IQ2_XXS blk.23.ffn_gate_inp.weight: BF16 blk.23.ffn_gate_shexp.weight: Q5_K blk.23.ffn_norm.weight: F32 blk.23.ffn_up_exps.weight: IQ2_XXS blk.23.ffn_up_shexp.weight: Q5_K blk.23.hc_attn_base.weight: F32 blk.23.hc_attn_fn.weight: F32 blk.23.hc_attn_scale.weight: F32 blk.23.hc_ffn_base.weight: F32 blk.23.hc_ffn_fn.weight: F32 blk.23.hc_ffn_scale.weight: F32 blk.24.attn_kv.weight: Q8_0 blk.24.attn_kv_a_norm.weight: F32 blk.24.attn_norm.weight: F32 blk.24.attn_output_a.weight: Q5_K blk.24.attn_output_b.weight: Q5_K blk.24.attn_q_a.weight: Q8_0 blk.24.attn_q_a_norm.weight: F32 blk.24.attn_q_b.weight: Q5_K blk.24.attn_sinks.weight: F32 blk.24.exp_probs_b.bias: F32 blk.24.exp_probs_b_vl.bias: F32 blk.24.ffn_down_exps.weight: IQ1_M blk.24.ffn_down_shexp.weight: Q5_K blk.24.ffn_gate_exps.weight: IQ2_XXS blk.24.ffn_gate_inp.weight: BF16 blk.24.ffn_gate_shexp.weight: Q5_K blk.24.ffn_norm.weight: F32 blk.24.ffn_up_exps.weight: IQ2_XXS blk.24.ffn_up_shexp.weight: Q5_K blk.24.hc_attn_base.weight: F32 blk.24.hc_attn_fn.weight: F32 blk.24.hc_attn_scale.weight: F32 blk.24.hc_ffn_base.weight: F32 blk.24.hc_ffn_fn.weight: F32 blk.24.hc_ffn_scale.weight: F32 blk.24.indexer.attn_q_b.weight: BF16 blk.24.indexer.proj.weight: BF16 blk.25.attn_kv.weight: Q8_0 blk.25.attn_kv_a_norm.weight: F32 blk.25.attn_norm.weight: F32 blk.25.attn_output_a.weight: Q5_K blk.25.attn_output_b.weight: Q5_K blk.25.attn_q_a.weight: Q8_0 blk.25.attn_q_a_norm.weight: F32 blk.25.attn_q_b.weight: Q5_K blk.25.attn_sinks.weight: F32 blk.25.exp_probs_b.bias: F32 blk.25.exp_probs_b_vl.bias: F32 blk.25.ffn_down_exps.weight: IQ2_XXS blk.25.ffn_down_shexp.weight: Q5_K blk.25.ffn_gate_exps.weight: IQ2_XXS blk.25.ffn_gate_inp.weight: BF16 blk.25.ffn_gate_shexp.weight: Q5_K blk.25.ffn_norm.weight: F32 blk.25.ffn_up_exps.weight: IQ2_XXS blk.25.ffn_up_shexp.weight: Q5_K blk.25.hc_attn_base.weight: F32 blk.25.hc_attn_fn.weight: F32 blk.25.hc_attn_scale.weight: F32 blk.25.hc_ffn_base.weight: F32 blk.25.hc_ffn_fn.weight: F32 blk.25.hc_ffn_scale.weight: F32 blk.26.attn_kv.weight: Q8_0 blk.26.attn_kv_a_norm.weight: F32 blk.26.attn_norm.weight: F32 blk.26.attn_output_a.weight: Q5_K blk.26.attn_output_b.weight: Q5_K blk.26.attn_q_a.weight: Q8_0 blk.26.attn_q_a_norm.weight: F32 blk.26.attn_q_b.weight: Q5_K blk.26.attn_sinks.weight: F32 blk.26.exp_probs_b.bias: F32 blk.26.exp_probs_b_vl.bias: F32 blk.26.ffn_down_exps.weight: IQ2_XS blk.26.ffn_down_shexp.weight: Q5_K blk.26.ffn_gate_exps.weight: IQ2_XXS blk.26.ffn_gate_inp.weight: BF16 blk.26.ffn_gate_shexp.weight: Q5_K blk.26.ffn_norm.weight: F32 blk.26.ffn_up_exps.weight: IQ2_XXS blk.26.ffn_up_shexp.weight: Q5_K blk.26.hc_attn_base.weight: F32 blk.26.hc_attn_fn.weight: F32 blk.26.hc_attn_scale.weight: F32 blk.26.hc_ffn_base.weight: F32 blk.26.hc_ffn_fn.weight: F32 blk.26.hc_ffn_scale.weight: F32 blk.27.attn_kv.weight: Q8_0 blk.27.attn_kv_a_norm.weight: F32 blk.27.attn_norm.weight: F32 blk.27.attn_output_a.weight: Q5_K blk.27.attn_output_b.weight: Q5_K blk.27.attn_q_a.weight: Q8_0 blk.27.attn_q_a_norm.weight: F32 blk.27.attn_q_b.weight: Q5_K blk.27.attn_sinks.weight: F32 blk.27.exp_probs_b.bias: F32 blk.27.exp_probs_b_vl.bias: F32 blk.27.ffn_down_exps.weight: IQ1_M blk.27.ffn_down_shexp.weight: Q5_K blk.27.ffn_gate_exps.weight: IQ2_XS blk.27.ffn_gate_inp.weight: BF16 blk.27.ffn_gate_shexp.weight: Q5_K blk.27.ffn_norm.weight: F32 blk.27.ffn_up_exps.weight: IQ2_XS blk.27.ffn_up_shexp.weight: Q5_K blk.27.hc_attn_base.weight: F32 blk.27.hc_attn_fn.weight: F32 blk.27.hc_attn_scale.weight: F32 blk.27.hc_ffn_base.weight: F32 blk.27.hc_ffn_fn.weight: F32 blk.27.hc_ffn_scale.weight: F32 blk.28.attn_kv.weight: Q8_0 blk.28.attn_kv_a_norm.weight: F32 blk.28.attn_norm.weight: F32 blk.28.attn_output_a.weight: Q5_K blk.28.attn_output_b.weight: Q5_K blk.28.attn_q_a.weight: Q8_0 blk.28.attn_q_a_norm.weight: F32 blk.28.attn_q_b.weight: Q5_K blk.28.attn_sinks.weight: F32 blk.28.exp_probs_b.bias: F32 blk.28.exp_probs_b_vl.bias: F32 blk.28.ffn_down_exps.weight: IQ2_XXS blk.28.ffn_down_shexp.weight: Q5_K blk.28.ffn_gate_exps.weight: IQ2_XXS blk.28.ffn_gate_inp.weight: BF16 blk.28.ffn_gate_shexp.weight: Q5_K blk.28.ffn_norm.weight: F32 blk.28.ffn_up_exps.weight: IQ2_XXS blk.28.ffn_up_shexp.weight: Q5_K blk.28.hc_attn_base.weight: F32 blk.28.hc_attn_fn.weight: F32 blk.28.hc_attn_scale.weight: F32 blk.28.hc_ffn_base.weight: F32 blk.28.hc_ffn_fn.weight: F32 blk.28.hc_ffn_scale.weight: F32 blk.28.indexer.attn_q_b.weight: BF16 blk.28.indexer.proj.weight: BF16 blk.29.attn_kv.weight: Q8_0 blk.29.attn_kv_a_norm.weight: F32 blk.29.attn_norm.weight: F32 blk.29.attn_output_a.weight: Q5_K blk.29.attn_output_b.weight: Q5_K blk.29.attn_q_a.weight: Q8_0 blk.29.attn_q_a_norm.weight: F32 blk.29.attn_q_b.weight: Q5_K blk.29.attn_sinks.weight: F32 blk.29.exp_probs_b.bias: F32 blk.29.exp_probs_b_vl.bias: F32 blk.29.ffn_down_exps.weight: IQ1_M blk.29.ffn_down_shexp.weight: Q5_K blk.29.ffn_gate_exps.weight: IQ2_XXS blk.29.ffn_gate_inp.weight: BF16 blk.29.ffn_gate_shexp.weight: Q5_K blk.29.ffn_norm.weight: F32 blk.29.ffn_up_exps.weight: IQ2_XXS blk.29.ffn_up_shexp.weight: Q5_K blk.29.hc_attn_base.weight: F32 blk.29.hc_attn_fn.weight: F32 blk.29.hc_attn_scale.weight: F32 blk.29.hc_ffn_base.weight: F32 blk.29.hc_ffn_fn.weight: F32 blk.29.hc_ffn_scale.weight: F32 blk.30.attn_kv.weight: Q8_0 blk.30.attn_kv_a_norm.weight: F32 blk.30.attn_norm.weight: F32 blk.30.attn_output_a.weight: Q5_K blk.30.attn_output_b.weight: Q5_K blk.30.attn_q_a.weight: Q8_0 blk.30.attn_q_a_norm.weight: F32 blk.30.attn_q_b.weight: Q5_K blk.30.attn_sinks.weight: F32 blk.30.exp_probs_b.bias: F32 blk.30.exp_probs_b_vl.bias: F32 blk.30.ffn_down_exps.weight: IQ1_M blk.30.ffn_down_shexp.weight: Q5_K blk.30.ffn_gate_exps.weight: IQ1_M blk.30.ffn_gate_inp.weight: BF16 blk.30.ffn_gate_shexp.weight: Q5_K blk.30.ffn_norm.weight: F32 blk.30.ffn_up_exps.weight: IQ1_M blk.30.ffn_up_shexp.weight: Q5_K blk.30.hc_attn_base.weight: F32 blk.30.hc_attn_fn.weight: F32 blk.30.hc_attn_scale.weight: F32 blk.30.hc_ffn_base.weight: F32 blk.30.hc_ffn_fn.weight: F32 blk.30.hc_ffn_scale.weight: F32 blk.31.attn_kv.weight: Q8_0 blk.31.attn_kv_a_norm.weight: F32 blk.31.attn_norm.weight: F32 blk.31.attn_output_a.weight: Q5_K blk.31.attn_output_b.weight: Q5_K blk.31.attn_q_a.weight: Q8_0 blk.31.attn_q_a_norm.weight: F32 blk.31.attn_q_b.weight: Q5_K blk.31.attn_sinks.weight: F32 blk.31.exp_probs_b.bias: F32 blk.31.exp_probs_b_vl.bias: F32 blk.31.ffn_down_exps.weight: IQ1_M blk.31.ffn_down_shexp.weight: Q5_K blk.31.ffn_gate_exps.weight: IQ2_XXS blk.31.ffn_gate_inp.weight: BF16 blk.31.ffn_gate_shexp.weight: Q5_K blk.31.ffn_norm.weight: F32 blk.31.ffn_up_exps.weight: IQ2_XXS blk.31.ffn_up_shexp.weight: Q5_K blk.31.hc_attn_base.weight: F32 blk.31.hc_attn_fn.weight: F32 blk.31.hc_attn_scale.weight: F32 blk.31.hc_ffn_base.weight: F32 blk.31.hc_ffn_fn.weight: F32 blk.31.hc_ffn_scale.weight: F32 blk.32.attn_kv.weight: Q8_0 blk.32.attn_kv_a_norm.weight: F32 blk.32.attn_norm.weight: F32 blk.32.attn_output_a.weight: Q5_K blk.32.attn_output_b.weight: Q5_K blk.32.attn_q_a.weight: Q8_0 blk.32.attn_q_a_norm.weight: F32 blk.32.attn_q_b.weight: Q5_K blk.32.attn_sinks.weight: F32 blk.32.exp_probs_b.bias: F32 blk.32.exp_probs_b_vl.bias: F32 blk.32.ffn_down_exps.weight: IQ1_M blk.32.ffn_down_shexp.weight: Q5_K blk.32.ffn_gate_exps.weight: IQ2_XXS blk.32.ffn_gate_inp.weight: BF16 blk.32.ffn_gate_shexp.weight: Q5_K blk.32.ffn_norm.weight: F32 blk.32.ffn_up_exps.weight: IQ2_XXS blk.32.ffn_up_shexp.weight: Q5_K blk.32.hc_attn_base.weight: F32 blk.32.hc_attn_fn.weight: F32 blk.32.hc_attn_scale.weight: F32 blk.32.hc_ffn_base.weight: F32 blk.32.hc_ffn_fn.weight: F32 blk.32.hc_ffn_scale.weight: F32 blk.32.indexer.attn_q_b.weight: BF16 blk.32.indexer.proj.weight: BF16 blk.33.attn_kv.weight: Q8_0 blk.33.attn_kv_a_norm.weight: F32 blk.33.attn_norm.weight: F32 blk.33.attn_output_a.weight: Q5_K blk.33.attn_output_b.weight: Q5_K blk.33.attn_q_a.weight: Q8_0 blk.33.attn_q_a_norm.weight: F32 blk.33.attn_q_b.weight: Q5_K blk.33.attn_sinks.weight: F32 blk.33.exp_probs_b.bias: F32 blk.33.exp_probs_b_vl.bias: F32 blk.33.ffn_down_exps.weight: IQ2_XS blk.33.ffn_down_shexp.weight: Q5_K blk.33.ffn_gate_exps.weight: IQ2_XXS blk.33.ffn_gate_inp.weight: BF16 blk.33.ffn_gate_shexp.weight: Q5_K blk.33.ffn_norm.weight: F32 blk.33.ffn_up_exps.weight: IQ2_XXS blk.33.ffn_up_shexp.weight: Q5_K blk.33.hc_attn_base.weight: F32 blk.33.hc_attn_fn.weight: F32 blk.33.hc_attn_scale.weight: F32 blk.33.hc_ffn_base.weight: F32 blk.33.hc_ffn_fn.weight: F32 blk.33.hc_ffn_scale.weight: F32 blk.34.attn_kv.weight: Q8_0 blk.34.attn_kv_a_norm.weight: F32 blk.34.attn_norm.weight: F32 blk.34.attn_output_a.weight: Q5_K blk.34.attn_output_b.weight: Q5_K blk.34.attn_q_a.weight: Q8_0 blk.34.attn_q_a_norm.weight: F32 blk.34.attn_q_b.weight: Q5_K blk.34.attn_sinks.weight: F32 blk.34.exp_probs_b.bias: F32 blk.34.exp_probs_b_vl.bias: F32 blk.34.ffn_down_exps.weight: IQ1_M blk.34.ffn_down_shexp.weight: Q5_K blk.34.ffn_gate_exps.weight: IQ2_XS blk.34.ffn_gate_inp.weight: BF16 blk.34.ffn_gate_shexp.weight: Q5_K blk.34.ffn_norm.weight: F32 blk.34.ffn_up_exps.weight: IQ2_XS blk.34.ffn_up_shexp.weight: Q5_K blk.34.hc_attn_base.weight: F32 blk.34.hc_attn_fn.weight: F32 blk.34.hc_attn_scale.weight: F32 blk.34.hc_ffn_base.weight: F32 blk.34.hc_ffn_fn.weight: F32 blk.34.hc_ffn_scale.weight: F32 blk.35.attn_kv.weight: Q8_0 blk.35.attn_kv_a_norm.weight: F32 blk.35.attn_norm.weight: F32 blk.35.attn_output_a.weight: Q5_K blk.35.attn_output_b.weight: Q5_K blk.35.attn_q_a.weight: Q8_0 blk.35.attn_q_a_norm.weight: F32 blk.35.attn_q_b.weight: Q5_K blk.35.attn_sinks.weight: F32 blk.35.exp_probs_b.bias: F32 blk.35.exp_probs_b_vl.bias: F32 blk.35.ffn_down_exps.weight: IQ1_M blk.35.ffn_down_shexp.weight: Q5_K blk.35.ffn_gate_exps.weight: IQ2_XS blk.35.ffn_gate_inp.weight: BF16 blk.35.ffn_gate_shexp.weight: Q5_K blk.35.ffn_norm.weight: F32 blk.35.ffn_up_exps.weight: IQ2_XS blk.35.ffn_up_shexp.weight: Q5_K blk.35.hc_attn_base.weight: F32 blk.35.hc_attn_fn.weight: F32 blk.35.hc_attn_scale.weight: F32 blk.35.hc_ffn_base.weight: F32 blk.35.hc_ffn_fn.weight: F32 blk.35.hc_ffn_scale.weight: F32 blk.36.attn_kv.weight: Q8_0 blk.36.attn_kv_a_norm.weight: F32 blk.36.attn_norm.weight: F32 blk.36.attn_output_a.weight: Q5_K blk.36.attn_output_b.weight: Q5_K blk.36.attn_q_a.weight: Q8_0 blk.36.attn_q_a_norm.weight: F32 blk.36.attn_q_b.weight: Q5_K blk.36.attn_sinks.weight: F32 blk.36.exp_probs_b.bias: F32 blk.36.exp_probs_b_vl.bias: F32 blk.36.ffn_down_exps.weight: IQ1_M blk.36.ffn_down_shexp.weight: Q5_K blk.36.ffn_gate_exps.weight: IQ2_XXS blk.36.ffn_gate_inp.weight: BF16 blk.36.ffn_gate_shexp.weight: Q5_K blk.36.ffn_norm.weight: F32 blk.36.ffn_up_exps.weight: IQ2_XXS blk.36.ffn_up_shexp.weight: Q5_K blk.36.hc_attn_base.weight: F32 blk.36.hc_attn_fn.weight: F32 blk.36.hc_attn_scale.weight: F32 blk.36.hc_ffn_base.weight: F32 blk.36.hc_ffn_fn.weight: F32 blk.36.hc_ffn_scale.weight: F32 blk.36.indexer.attn_q_b.weight: BF16 blk.36.indexer.proj.weight: BF16 blk.37.attn_kv.weight: Q8_0 blk.37.attn_kv_a_norm.weight: F32 blk.37.attn_norm.weight: F32 blk.37.attn_output_a.weight: Q5_K blk.37.attn_output_b.weight: Q5_K blk.37.attn_q_a.weight: Q8_0 blk.37.attn_q_a_norm.weight: F32 blk.37.attn_q_b.weight: Q5_K blk.37.attn_sinks.weight: F32 blk.37.exp_probs_b.bias: F32 blk.37.exp_probs_b_vl.bias: F32 blk.37.ffn_down_exps.weight: IQ1_M blk.37.ffn_down_shexp.weight: Q5_K blk.37.ffn_gate_exps.weight: IQ2_XXS blk.37.ffn_gate_inp.weight: BF16 blk.37.ffn_gate_shexp.weight: Q5_K blk.37.ffn_norm.weight: F32 blk.37.ffn_up_exps.weight: IQ2_XXS blk.37.ffn_up_shexp.weight: Q5_K blk.37.hc_attn_base.weight: F32 blk.37.hc_attn_fn.weight: F32 blk.37.hc_attn_scale.weight: F32 blk.37.hc_ffn_base.weight: F32 blk.37.hc_ffn_fn.weight: F32 blk.37.hc_ffn_scale.weight: F32 blk.38.attn_kv.weight: Q8_0 blk.38.attn_kv_a_norm.weight: F32 blk.38.attn_norm.weight: F32 blk.38.attn_output_a.weight: Q5_K blk.38.attn_output_b.weight: Q5_K blk.38.attn_q_a.weight: Q8_0 blk.38.attn_q_a_norm.weight: F32 blk.38.attn_q_b.weight: Q5_K blk.38.attn_sinks.weight: F32 blk.38.exp_probs_b.bias: F32 blk.38.exp_probs_b_vl.bias: F32 blk.38.ffn_down_exps.weight: IQ2_XXS blk.38.ffn_down_shexp.weight: Q5_K blk.38.ffn_gate_exps.weight: IQ2_XS blk.38.ffn_gate_inp.weight: BF16 blk.38.ffn_gate_shexp.weight: Q5_K blk.38.ffn_norm.weight: F32 blk.38.ffn_up_exps.weight: IQ2_XS blk.38.ffn_up_shexp.weight: Q5_K blk.38.hc_attn_base.weight: F32 blk.38.hc_attn_fn.weight: F32 blk.38.hc_attn_scale.weight: F32 blk.38.hc_ffn_base.weight: F32 blk.38.hc_ffn_fn.weight: F32 blk.38.hc_ffn_scale.weight: F32 blk.39.attn_kv.weight: Q8_0 blk.39.attn_kv_a_norm.weight: F32 blk.39.attn_norm.weight: F32 blk.39.attn_output_a.weight: Q5_K blk.39.attn_output_b.weight: Q5_K blk.39.attn_q_a.weight: Q8_0 blk.39.attn_q_a_norm.weight: F32 blk.39.attn_q_b.weight: Q5_K blk.39.attn_sinks.weight: F32 blk.39.exp_probs_b.bias: F32 blk.39.exp_probs_b_vl.bias: F32 blk.39.ffn_down_exps.weight: IQ2_XS blk.39.ffn_down_shexp.weight: Q5_K blk.39.ffn_gate_exps.weight: IQ2_XS blk.39.ffn_gate_inp.weight: BF16 blk.39.ffn_gate_shexp.weight: Q5_K blk.39.ffn_norm.weight: F32 blk.39.ffn_up_exps.weight: IQ2_XS blk.39.ffn_up_shexp.weight: Q5_K blk.39.hc_attn_base.weight: F32 blk.39.hc_attn_fn.weight: F32 blk.39.hc_attn_scale.weight: F32 blk.39.hc_ffn_base.weight: F32 blk.39.hc_ffn_fn.weight: F32 blk.39.hc_ffn_scale.weight: F32