| { |
| "method": "pytorch_int8", |
| "quant_bits": 8, |
| "family": "lfm2vl", |
| "model_id": "LiquidAI/LFM2-VL-3B", |
| "n_linear_quantized": 167, |
| "skip_vision_quant": true, |
| "int8_layers": { |
| "model.language_model.layers.0.conv.in_proj": { |
| "in_features": 2048, |
| "out_features": 6144, |
| "has_bias": false |
| }, |
| "model.language_model.layers.0.conv.out_proj": { |
| "in_features": 2048, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.0.feed_forward.w1": { |
| "in_features": 2048, |
| "out_features": 10752, |
| "has_bias": false |
| }, |
| "model.language_model.layers.0.feed_forward.w3": { |
| "in_features": 2048, |
| "out_features": 10752, |
| "has_bias": false |
| }, |
| "model.language_model.layers.0.feed_forward.w2": { |
| "in_features": 10752, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.1.conv.in_proj": { |
| "in_features": 2048, |
| "out_features": 6144, |
| "has_bias": false |
| }, |
| "model.language_model.layers.1.conv.out_proj": { |
| "in_features": 2048, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.1.feed_forward.w1": { |
| "in_features": 2048, |
| "out_features": 10752, |
| "has_bias": false |
| }, |
| "model.language_model.layers.1.feed_forward.w3": { |
| "in_features": 2048, |
| "out_features": 10752, |
| "has_bias": false |
| }, |
| "model.language_model.layers.1.feed_forward.w2": { |
| "in_features": 10752, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.2.self_attn.q_proj": { |
| "in_features": 2048, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.2.self_attn.k_proj": { |
| "in_features": 2048, |
| "out_features": 512, |
| "has_bias": false |
| }, |
| "model.language_model.layers.2.self_attn.v_proj": { |
| "in_features": 2048, |
| "out_features": 512, |
| "has_bias": false |
| }, |
| "model.language_model.layers.2.self_attn.out_proj": { |
| "in_features": 2048, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.2.feed_forward.w1": { |
| "in_features": 2048, |
| "out_features": 10752, |
| "has_bias": false |
| }, |
| "model.language_model.layers.2.feed_forward.w3": { |
| "in_features": 2048, |
| "out_features": 10752, |
| "has_bias": false |
| }, |
| "model.language_model.layers.2.feed_forward.w2": { |
| "in_features": 10752, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.3.conv.in_proj": { |
| "in_features": 2048, |
| "out_features": 6144, |
| "has_bias": false |
| }, |
| "model.language_model.layers.3.conv.out_proj": { |
| "in_features": 2048, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.3.feed_forward.w1": { |
| "in_features": 2048, |
| "out_features": 10752, |
| "has_bias": false |
| }, |
| "model.language_model.layers.3.feed_forward.w3": { |
| "in_features": 2048, |
| "out_features": 10752, |
| "has_bias": false |
| }, |
| "model.language_model.layers.3.feed_forward.w2": { |
| "in_features": 10752, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.4.conv.in_proj": { |
| "in_features": 2048, |
| "out_features": 6144, |
| "has_bias": false |
| }, |
| "model.language_model.layers.4.conv.out_proj": { |
| "in_features": 2048, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.4.feed_forward.w1": { |
| "in_features": 2048, |
| "out_features": 10752, |
| "has_bias": false |
| }, |
| "model.language_model.layers.4.feed_forward.w3": { |
| "in_features": 2048, |
| "out_features": 10752, |
| "has_bias": false |
| }, |
| "model.language_model.layers.4.feed_forward.w2": { |
| "in_features": 10752, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.5.self_attn.q_proj": { |
| "in_features": 2048, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.5.self_attn.k_proj": { |
| "in_features": 2048, |
| "out_features": 512, |
| "has_bias": false |
| }, |
| "model.language_model.layers.5.self_attn.v_proj": { |
| "in_features": 2048, |
| "out_features": 512, |
| "has_bias": false |
| }, |
| "model.language_model.layers.5.self_attn.out_proj": { |
| "in_features": 2048, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.5.feed_forward.w1": { |
| "in_features": 2048, |
| "out_features": 10752, |
| "has_bias": false |
| }, |
| "model.language_model.layers.5.feed_forward.w3": { |
| "in_features": 2048, |
| "out_features": 10752, |
| "has_bias": false |
| }, |
| "model.language_model.layers.5.feed_forward.w2": { |
| "in_features": 10752, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.6.conv.in_proj": { |
| "in_features": 2048, |
| "out_features": 6144, |
| "has_bias": false |
| }, |
| "model.language_model.layers.6.conv.out_proj": { |
| "in_features": 2048, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.6.feed_forward.w1": { |
| "in_features": 2048, |
| "out_features": 10752, |
| "has_bias": false |
| }, |
| "model.language_model.layers.6.feed_forward.w3": { |
| "in_features": 2048, |
| "out_features": 10752, |
| "has_bias": false |
| }, |
| "model.language_model.layers.6.feed_forward.w2": { |
| "in_features": 10752, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.7.conv.in_proj": { |
| "in_features": 2048, |
| "out_features": 6144, |
| "has_bias": false |
| }, |
| "model.language_model.layers.7.conv.out_proj": { |
| "in_features": 2048, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.7.feed_forward.w1": { |
| "in_features": 2048, |
| "out_features": 10752, |
| "has_bias": false |
| }, |
| "model.language_model.layers.7.feed_forward.w3": { |
| "in_features": 2048, |
| "out_features": 10752, |
| "has_bias": false |
| }, |
| "model.language_model.layers.7.feed_forward.w2": { |
| "in_features": 10752, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.8.conv.in_proj": { |
| "in_features": 2048, |
| "out_features": 6144, |
| "has_bias": false |
| }, |
| "model.language_model.layers.8.conv.out_proj": { |
| "in_features": 2048, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.8.feed_forward.w1": { |
| "in_features": 2048, |
| "out_features": 10752, |
| "has_bias": false |
| }, |
| "model.language_model.layers.8.feed_forward.w3": { |
| "in_features": 2048, |
| "out_features": 10752, |
| "has_bias": false |
| }, |
| "model.language_model.layers.8.feed_forward.w2": { |
| "in_features": 10752, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.9.self_attn.q_proj": { |
| "in_features": 2048, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.9.self_attn.k_proj": { |
| "in_features": 2048, |
| "out_features": 512, |
| "has_bias": false |
| }, |
| "model.language_model.layers.9.self_attn.v_proj": { |
| "in_features": 2048, |
| "out_features": 512, |
| "has_bias": false |
| }, |
| "model.language_model.layers.9.self_attn.out_proj": { |
| "in_features": 2048, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.9.feed_forward.w1": { |
| "in_features": 2048, |
| "out_features": 10752, |
| "has_bias": false |
| }, |
| "model.language_model.layers.9.feed_forward.w3": { |
| "in_features": 2048, |
| "out_features": 10752, |
| "has_bias": false |
| }, |
| "model.language_model.layers.9.feed_forward.w2": { |
| "in_features": 10752, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.10.conv.in_proj": { |
| "in_features": 2048, |
| "out_features": 6144, |
| "has_bias": false |
| }, |
| "model.language_model.layers.10.conv.out_proj": { |
| "in_features": 2048, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.10.feed_forward.w1": { |
| "in_features": 2048, |
| "out_features": 10752, |
| "has_bias": false |
| }, |
| "model.language_model.layers.10.feed_forward.w3": { |
| "in_features": 2048, |
| "out_features": 10752, |
| "has_bias": false |
| }, |
| "model.language_model.layers.10.feed_forward.w2": { |
| "in_features": 10752, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.11.conv.in_proj": { |
| "in_features": 2048, |
| "out_features": 6144, |
| "has_bias": false |
| }, |
| "model.language_model.layers.11.conv.out_proj": { |
| "in_features": 2048, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.11.feed_forward.w1": { |
| "in_features": 2048, |
| "out_features": 10752, |
| "has_bias": false |
| }, |
| "model.language_model.layers.11.feed_forward.w3": { |
| "in_features": 2048, |
| "out_features": 10752, |
| "has_bias": false |
| }, |
| "model.language_model.layers.11.feed_forward.w2": { |
| "in_features": 10752, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.12.conv.in_proj": { |
| "in_features": 2048, |
| "out_features": 6144, |
| "has_bias": false |
| }, |
| "model.language_model.layers.12.conv.out_proj": { |
| "in_features": 2048, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.12.feed_forward.w1": { |
| "in_features": 2048, |
| "out_features": 10752, |
| "has_bias": false |
| }, |
| "model.language_model.layers.12.feed_forward.w3": { |
| "in_features": 2048, |
| "out_features": 10752, |
| "has_bias": false |
| }, |
| "model.language_model.layers.12.feed_forward.w2": { |
| "in_features": 10752, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.13.self_attn.q_proj": { |
| "in_features": 2048, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.13.self_attn.k_proj": { |
| "in_features": 2048, |
| "out_features": 512, |
| "has_bias": false |
| }, |
| "model.language_model.layers.13.self_attn.v_proj": { |
| "in_features": 2048, |
| "out_features": 512, |
| "has_bias": false |
| }, |
| "model.language_model.layers.13.self_attn.out_proj": { |
| "in_features": 2048, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.13.feed_forward.w1": { |
| "in_features": 2048, |
| "out_features": 10752, |
| "has_bias": false |
| }, |
| "model.language_model.layers.13.feed_forward.w3": { |
| "in_features": 2048, |
| "out_features": 10752, |
| "has_bias": false |
| }, |
| "model.language_model.layers.13.feed_forward.w2": { |
| "in_features": 10752, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.14.conv.in_proj": { |
| "in_features": 2048, |
| "out_features": 6144, |
| "has_bias": false |
| }, |
| "model.language_model.layers.14.conv.out_proj": { |
| "in_features": 2048, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.14.feed_forward.w1": { |
| "in_features": 2048, |
| "out_features": 10752, |
| "has_bias": false |
| }, |
| "model.language_model.layers.14.feed_forward.w3": { |
| "in_features": 2048, |
| "out_features": 10752, |
| "has_bias": false |
| }, |
| "model.language_model.layers.14.feed_forward.w2": { |
| "in_features": 10752, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.15.conv.in_proj": { |
| "in_features": 2048, |
| "out_features": 6144, |
| "has_bias": false |
| }, |
| "model.language_model.layers.15.conv.out_proj": { |
| "in_features": 2048, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.15.feed_forward.w1": { |
| "in_features": 2048, |
| "out_features": 10752, |
| "has_bias": false |
| }, |
| "model.language_model.layers.15.feed_forward.w3": { |
| "in_features": 2048, |
| "out_features": 10752, |
| "has_bias": false |
| }, |
| "model.language_model.layers.15.feed_forward.w2": { |
| "in_features": 10752, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.16.conv.in_proj": { |
| "in_features": 2048, |
| "out_features": 6144, |
| "has_bias": false |
| }, |
| "model.language_model.layers.16.conv.out_proj": { |
| "in_features": 2048, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.16.feed_forward.w1": { |
| "in_features": 2048, |
| "out_features": 10752, |
| "has_bias": false |
| }, |
| "model.language_model.layers.16.feed_forward.w3": { |
| "in_features": 2048, |
| "out_features": 10752, |
| "has_bias": false |
| }, |
| "model.language_model.layers.16.feed_forward.w2": { |
| "in_features": 10752, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.17.self_attn.q_proj": { |
| "in_features": 2048, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.17.self_attn.k_proj": { |
| "in_features": 2048, |
| "out_features": 512, |
| "has_bias": false |
| }, |
| "model.language_model.layers.17.self_attn.v_proj": { |
| "in_features": 2048, |
| "out_features": 512, |
| "has_bias": false |
| }, |
| "model.language_model.layers.17.self_attn.out_proj": { |
| "in_features": 2048, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.17.feed_forward.w1": { |
| "in_features": 2048, |
| "out_features": 10752, |
| "has_bias": false |
| }, |
| "model.language_model.layers.17.feed_forward.w3": { |
| "in_features": 2048, |
| "out_features": 10752, |
| "has_bias": false |
| }, |
| "model.language_model.layers.17.feed_forward.w2": { |
| "in_features": 10752, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.18.conv.in_proj": { |
| "in_features": 2048, |
| "out_features": 6144, |
| "has_bias": false |
| }, |
| "model.language_model.layers.18.conv.out_proj": { |
| "in_features": 2048, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.18.feed_forward.w1": { |
| "in_features": 2048, |
| "out_features": 10752, |
| "has_bias": false |
| }, |
| "model.language_model.layers.18.feed_forward.w3": { |
| "in_features": 2048, |
| "out_features": 10752, |
| "has_bias": false |
| }, |
| "model.language_model.layers.18.feed_forward.w2": { |
| "in_features": 10752, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.19.conv.in_proj": { |
| "in_features": 2048, |
| "out_features": 6144, |
| "has_bias": false |
| }, |
| "model.language_model.layers.19.conv.out_proj": { |
| "in_features": 2048, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.19.feed_forward.w1": { |
| "in_features": 2048, |
| "out_features": 10752, |
| "has_bias": false |
| }, |
| "model.language_model.layers.19.feed_forward.w3": { |
| "in_features": 2048, |
| "out_features": 10752, |
| "has_bias": false |
| }, |
| "model.language_model.layers.19.feed_forward.w2": { |
| "in_features": 10752, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.20.conv.in_proj": { |
| "in_features": 2048, |
| "out_features": 6144, |
| "has_bias": false |
| }, |
| "model.language_model.layers.20.conv.out_proj": { |
| "in_features": 2048, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.20.feed_forward.w1": { |
| "in_features": 2048, |
| "out_features": 10752, |
| "has_bias": false |
| }, |
| "model.language_model.layers.20.feed_forward.w3": { |
| "in_features": 2048, |
| "out_features": 10752, |
| "has_bias": false |
| }, |
| "model.language_model.layers.20.feed_forward.w2": { |
| "in_features": 10752, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.21.self_attn.q_proj": { |
| "in_features": 2048, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.21.self_attn.k_proj": { |
| "in_features": 2048, |
| "out_features": 512, |
| "has_bias": false |
| }, |
| "model.language_model.layers.21.self_attn.v_proj": { |
| "in_features": 2048, |
| "out_features": 512, |
| "has_bias": false |
| }, |
| "model.language_model.layers.21.self_attn.out_proj": { |
| "in_features": 2048, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.21.feed_forward.w1": { |
| "in_features": 2048, |
| "out_features": 10752, |
| "has_bias": false |
| }, |
| "model.language_model.layers.21.feed_forward.w3": { |
| "in_features": 2048, |
| "out_features": 10752, |
| "has_bias": false |
| }, |
| "model.language_model.layers.21.feed_forward.w2": { |
| "in_features": 10752, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.22.conv.in_proj": { |
| "in_features": 2048, |
| "out_features": 6144, |
| "has_bias": false |
| }, |
| "model.language_model.layers.22.conv.out_proj": { |
| "in_features": 2048, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.22.feed_forward.w1": { |
| "in_features": 2048, |
| "out_features": 10752, |
| "has_bias": false |
| }, |
| "model.language_model.layers.22.feed_forward.w3": { |
| "in_features": 2048, |
| "out_features": 10752, |
| "has_bias": false |
| }, |
| "model.language_model.layers.22.feed_forward.w2": { |
| "in_features": 10752, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.23.conv.in_proj": { |
| "in_features": 2048, |
| "out_features": 6144, |
| "has_bias": false |
| }, |
| "model.language_model.layers.23.conv.out_proj": { |
| "in_features": 2048, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.23.feed_forward.w1": { |
| "in_features": 2048, |
| "out_features": 10752, |
| "has_bias": false |
| }, |
| "model.language_model.layers.23.feed_forward.w3": { |
| "in_features": 2048, |
| "out_features": 10752, |
| "has_bias": false |
| }, |
| "model.language_model.layers.23.feed_forward.w2": { |
| "in_features": 10752, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.24.self_attn.q_proj": { |
| "in_features": 2048, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.24.self_attn.k_proj": { |
| "in_features": 2048, |
| "out_features": 512, |
| "has_bias": false |
| }, |
| "model.language_model.layers.24.self_attn.v_proj": { |
| "in_features": 2048, |
| "out_features": 512, |
| "has_bias": false |
| }, |
| "model.language_model.layers.24.self_attn.out_proj": { |
| "in_features": 2048, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.24.feed_forward.w1": { |
| "in_features": 2048, |
| "out_features": 10752, |
| "has_bias": false |
| }, |
| "model.language_model.layers.24.feed_forward.w3": { |
| "in_features": 2048, |
| "out_features": 10752, |
| "has_bias": false |
| }, |
| "model.language_model.layers.24.feed_forward.w2": { |
| "in_features": 10752, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.25.conv.in_proj": { |
| "in_features": 2048, |
| "out_features": 6144, |
| "has_bias": false |
| }, |
| "model.language_model.layers.25.conv.out_proj": { |
| "in_features": 2048, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.25.feed_forward.w1": { |
| "in_features": 2048, |
| "out_features": 10752, |
| "has_bias": false |
| }, |
| "model.language_model.layers.25.feed_forward.w3": { |
| "in_features": 2048, |
| "out_features": 10752, |
| "has_bias": false |
| }, |
| "model.language_model.layers.25.feed_forward.w2": { |
| "in_features": 10752, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.26.conv.in_proj": { |
| "in_features": 2048, |
| "out_features": 6144, |
| "has_bias": false |
| }, |
| "model.language_model.layers.26.conv.out_proj": { |
| "in_features": 2048, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.26.feed_forward.w1": { |
| "in_features": 2048, |
| "out_features": 10752, |
| "has_bias": false |
| }, |
| "model.language_model.layers.26.feed_forward.w3": { |
| "in_features": 2048, |
| "out_features": 10752, |
| "has_bias": false |
| }, |
| "model.language_model.layers.26.feed_forward.w2": { |
| "in_features": 10752, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.27.self_attn.q_proj": { |
| "in_features": 2048, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.27.self_attn.k_proj": { |
| "in_features": 2048, |
| "out_features": 512, |
| "has_bias": false |
| }, |
| "model.language_model.layers.27.self_attn.v_proj": { |
| "in_features": 2048, |
| "out_features": 512, |
| "has_bias": false |
| }, |
| "model.language_model.layers.27.self_attn.out_proj": { |
| "in_features": 2048, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.27.feed_forward.w1": { |
| "in_features": 2048, |
| "out_features": 10752, |
| "has_bias": false |
| }, |
| "model.language_model.layers.27.feed_forward.w3": { |
| "in_features": 2048, |
| "out_features": 10752, |
| "has_bias": false |
| }, |
| "model.language_model.layers.27.feed_forward.w2": { |
| "in_features": 10752, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.28.conv.in_proj": { |
| "in_features": 2048, |
| "out_features": 6144, |
| "has_bias": false |
| }, |
| "model.language_model.layers.28.conv.out_proj": { |
| "in_features": 2048, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.28.feed_forward.w1": { |
| "in_features": 2048, |
| "out_features": 10752, |
| "has_bias": false |
| }, |
| "model.language_model.layers.28.feed_forward.w3": { |
| "in_features": 2048, |
| "out_features": 10752, |
| "has_bias": false |
| }, |
| "model.language_model.layers.28.feed_forward.w2": { |
| "in_features": 10752, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.29.conv.in_proj": { |
| "in_features": 2048, |
| "out_features": 6144, |
| "has_bias": false |
| }, |
| "model.language_model.layers.29.conv.out_proj": { |
| "in_features": 2048, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "model.language_model.layers.29.feed_forward.w1": { |
| "in_features": 2048, |
| "out_features": 10752, |
| "has_bias": false |
| }, |
| "model.language_model.layers.29.feed_forward.w3": { |
| "in_features": 2048, |
| "out_features": 10752, |
| "has_bias": false |
| }, |
| "model.language_model.layers.29.feed_forward.w2": { |
| "in_features": 10752, |
| "out_features": 2048, |
| "has_bias": false |
| }, |
| "lm_head": { |
| "in_features": 2048, |
| "out_features": 65536, |
| "has_bias": false |
| } |
| } |
| } |