first commit
This commit is contained in:
362
dev/online_quant_configs/Step-3.5-Flash-w8a8/config.json
Normal file
362
dev/online_quant_configs/Step-3.5-Flash-w8a8/config.json
Normal file
@@ -0,0 +1,362 @@
|
||||
{
|
||||
"architectures": [
|
||||
"Step3p5ForCausalLM"
|
||||
],
|
||||
"model_type": "step3p5",
|
||||
"auto_map": {
|
||||
"AutoConfig": "configuration_step3p5.Step3p5Config",
|
||||
"AutoModelForCausalLM": "modeling_step3p5.Step3p5ForCausalLM"
|
||||
},
|
||||
"rope_scaling": {
|
||||
"rope_type": "llama3",
|
||||
"factor": 2.0,
|
||||
"original_max_position_embeddings": 131072,
|
||||
"low_freq_factor": 1.0,
|
||||
"high_freq_factor": 32.0
|
||||
},
|
||||
"yarn_only_types": [
|
||||
"full_attention"
|
||||
],
|
||||
"hidden_size": 4096,
|
||||
"intermediate_size": 11264,
|
||||
"num_hidden_layers": 45,
|
||||
"max_seq_len": 262144,
|
||||
"vocab_size": 128896,
|
||||
"torch_dtype": "bfloat16",
|
||||
"use_qk_norm": true,
|
||||
"moe_layers_enum": "3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44",
|
||||
"num_attention_heads": 64,
|
||||
"num_attention_groups": 8,
|
||||
"head_dim": 128,
|
||||
"use_moe": true,
|
||||
"moe_num_experts": 288,
|
||||
"moe_top_k": 8,
|
||||
"moe_intermediate_size": 1280,
|
||||
"share_expert_dim": 1280,
|
||||
"moe_layer_offset": 0,
|
||||
"moe_every_n_layer": 1,
|
||||
"norm_expert_weight": true,
|
||||
"moe_router_activation": "sigmoid",
|
||||
"moe_router_scaling_factor": 3.0,
|
||||
"att_impl_type": "GQA",
|
||||
"tie_word_embeddings": false,
|
||||
"rope_theta": [
|
||||
5000000.0,
|
||||
10000.0,
|
||||
10000.0,
|
||||
10000.0,
|
||||
5000000.0,
|
||||
10000.0,
|
||||
10000.0,
|
||||
10000.0,
|
||||
5000000.0,
|
||||
10000.0,
|
||||
10000.0,
|
||||
10000.0,
|
||||
5000000.0,
|
||||
10000.0,
|
||||
10000.0,
|
||||
10000.0,
|
||||
5000000.0,
|
||||
10000.0,
|
||||
10000.0,
|
||||
10000.0,
|
||||
5000000.0,
|
||||
10000.0,
|
||||
10000.0,
|
||||
10000.0,
|
||||
5000000.0,
|
||||
10000.0,
|
||||
10000.0,
|
||||
10000.0,
|
||||
5000000.0,
|
||||
10000.0,
|
||||
10000.0,
|
||||
10000.0,
|
||||
5000000.0,
|
||||
10000.0,
|
||||
10000.0,
|
||||
10000.0,
|
||||
5000000.0,
|
||||
10000.0,
|
||||
10000.0,
|
||||
10000.0,
|
||||
5000000.0,
|
||||
10000.0,
|
||||
10000.0,
|
||||
10000.0,
|
||||
5000000.0,
|
||||
10000.0,
|
||||
10000.0,
|
||||
10000.0
|
||||
],
|
||||
"use_head_wise_attn_gate": true,
|
||||
"sliding_window": 512,
|
||||
"use_moe_router_bias": true,
|
||||
"need_fp32_gate": true,
|
||||
"sink": false,
|
||||
"layer_types": [
|
||||
"full_attention",
|
||||
"sliding_attention",
|
||||
"sliding_attention",
|
||||
"sliding_attention",
|
||||
"full_attention",
|
||||
"sliding_attention",
|
||||
"sliding_attention",
|
||||
"sliding_attention",
|
||||
"full_attention",
|
||||
"sliding_attention",
|
||||
"sliding_attention",
|
||||
"sliding_attention",
|
||||
"full_attention",
|
||||
"sliding_attention",
|
||||
"sliding_attention",
|
||||
"sliding_attention",
|
||||
"full_attention",
|
||||
"sliding_attention",
|
||||
"sliding_attention",
|
||||
"sliding_attention",
|
||||
"full_attention",
|
||||
"sliding_attention",
|
||||
"sliding_attention",
|
||||
"sliding_attention",
|
||||
"full_attention",
|
||||
"sliding_attention",
|
||||
"sliding_attention",
|
||||
"sliding_attention",
|
||||
"full_attention",
|
||||
"sliding_attention",
|
||||
"sliding_attention",
|
||||
"sliding_attention",
|
||||
"full_attention",
|
||||
"sliding_attention",
|
||||
"sliding_attention",
|
||||
"sliding_attention",
|
||||
"full_attention",
|
||||
"sliding_attention",
|
||||
"sliding_attention",
|
||||
"sliding_attention",
|
||||
"full_attention",
|
||||
"sliding_attention",
|
||||
"sliding_attention",
|
||||
"sliding_attention",
|
||||
"full_attention",
|
||||
"sliding_attention",
|
||||
"sliding_attention",
|
||||
"sliding_attention"
|
||||
],
|
||||
"use_rope_layers": [],
|
||||
"num_nextn_predict_layers": 3,
|
||||
"partial_rotary_factors": [
|
||||
0.5,
|
||||
1.0,
|
||||
1.0,
|
||||
1.0,
|
||||
0.5,
|
||||
1.0,
|
||||
1.0,
|
||||
1.0,
|
||||
0.5,
|
||||
1.0,
|
||||
1.0,
|
||||
1.0,
|
||||
0.5,
|
||||
1.0,
|
||||
1.0,
|
||||
1.0,
|
||||
0.5,
|
||||
1.0,
|
||||
1.0,
|
||||
1.0,
|
||||
0.5,
|
||||
1.0,
|
||||
1.0,
|
||||
1.0,
|
||||
0.5,
|
||||
1.0,
|
||||
1.0,
|
||||
1.0,
|
||||
0.5,
|
||||
1.0,
|
||||
1.0,
|
||||
1.0,
|
||||
0.5,
|
||||
1.0,
|
||||
1.0,
|
||||
1.0,
|
||||
0.5,
|
||||
1.0,
|
||||
1.0,
|
||||
1.0,
|
||||
0.5,
|
||||
1.0,
|
||||
1.0,
|
||||
1.0,
|
||||
0.5,
|
||||
1.0,
|
||||
1.0,
|
||||
1.0
|
||||
],
|
||||
"eos_token_id": [
|
||||
1,
|
||||
2,
|
||||
128007
|
||||
],
|
||||
"bos_token_id": 0,
|
||||
"attention_other_setting": {
|
||||
"attention_type": "sliding_attention",
|
||||
"num_attention_heads": 96,
|
||||
"num_attention_groups": 8,
|
||||
"head_dim": 128,
|
||||
"true_head_dim": 128
|
||||
},
|
||||
"swiglu_limits": [
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
7,
|
||||
7,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0
|
||||
],
|
||||
"swiglu_limits_shared": [
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0,
|
||||
16,
|
||||
0.0,
|
||||
0.0,
|
||||
0.0
|
||||
],
|
||||
"zero_centered": true,
|
||||
"max_position_embeddings": 262144,
|
||||
"quantization_config": {
|
||||
"config_groups": {
|
||||
"group_0": {
|
||||
"input_activations": {
|
||||
"block_structure": null,
|
||||
"dynamic": true,
|
||||
"group_size": null,
|
||||
"num_bits": 8,
|
||||
"observer": "memoryless",
|
||||
"observer_kwargs": {},
|
||||
"strategy": "token",
|
||||
"symmetric": true,
|
||||
"type": "int"
|
||||
},
|
||||
"output_activations": null,
|
||||
"targets": [
|
||||
"Linear"
|
||||
],
|
||||
"weights": {
|
||||
"block_structure": null,
|
||||
"dynamic": false,
|
||||
"group_size": null,
|
||||
"num_bits": 8,
|
||||
"observer": "minmax",
|
||||
"observer_kwargs": {},
|
||||
"strategy": "channel",
|
||||
"symmetric": true,
|
||||
"type": "int"
|
||||
}
|
||||
}
|
||||
},
|
||||
"format": "int-quantized",
|
||||
"global_compression_ratio": 1.0,
|
||||
"ignore": [
|
||||
"re:.*norm.*",
|
||||
"re:.*bias.*",
|
||||
"re:.*moe.gate\\..*",
|
||||
"re:.*eh_proj.*",
|
||||
"re:.*transformer.*",
|
||||
"re:.*lm_head.*",
|
||||
"re:.*embed_tokens.*",
|
||||
"re:.*share_expert.*"
|
||||
],
|
||||
"kv_cache_scheme": null,
|
||||
"quant_method": "compressed-tensors",
|
||||
"quantization_status": "frozen"
|
||||
}
|
||||
}
|
||||
806
dev/online_quant_configs/Step-3.5-Flash-w8a8/quant_map.json
Normal file
806
dev/online_quant_configs/Step-3.5-Flash-w8a8/quant_map.json
Normal file
@@ -0,0 +1,806 @@
|
||||
{
|
||||
"lm_head.weight": "torch.bfloat16",
|
||||
"model.layers.24.self_attn.g_proj.weight": "int8-TN",
|
||||
"model.layers.24.self_attn.k_proj.weight": "int8-TN",
|
||||
"model.layers.24.self_attn.q_proj.weight": "int8-TN",
|
||||
"model.layers.24.self_attn.v_proj.weight": "int8-TN",
|
||||
"model.layers.24.share_expert.gate_proj.weight": "torch.bfloat16",
|
||||
"model.layers.24.share_expert.up_proj.weight": "torch.bfloat16",
|
||||
"model.layers.25.self_attn.g_proj.weight": "int8-TN",
|
||||
"model.layers.25.self_attn.k_proj.weight": "int8-TN",
|
||||
"model.layers.25.self_attn.q_proj.weight": "int8-TN",
|
||||
"model.layers.25.self_attn.v_proj.weight": "int8-TN",
|
||||
"model.layers.25.share_expert.gate_proj.weight": "torch.bfloat16",
|
||||
"model.layers.25.share_expert.up_proj.weight": "torch.bfloat16",
|
||||
"model.layers.26.self_attn.g_proj.weight": "int8-TN",
|
||||
"model.layers.26.self_attn.k_proj.weight": "int8-TN",
|
||||
"model.layers.26.self_attn.q_proj.weight": "int8-TN",
|
||||
"model.layers.26.self_attn.v_proj.weight": "int8-TN",
|
||||
"model.layers.26.share_expert.gate_proj.weight": "torch.bfloat16",
|
||||
"model.layers.26.share_expert.up_proj.weight": "torch.bfloat16",
|
||||
"model.layers.27.self_attn.g_proj.weight": "int8-TN",
|
||||
"model.layers.27.self_attn.k_proj.weight": "int8-TN",
|
||||
"model.layers.27.self_attn.q_proj.weight": "int8-TN",
|
||||
"model.layers.27.self_attn.v_proj.weight": "int8-TN",
|
||||
"model.layers.27.share_expert.gate_proj.weight": "torch.bfloat16",
|
||||
"model.layers.27.share_expert.up_proj.weight": "torch.bfloat16",
|
||||
"model.layers.28.self_attn.g_proj.weight": "int8-TN",
|
||||
"model.layers.28.self_attn.k_proj.weight": "int8-TN",
|
||||
"model.layers.28.self_attn.q_proj.weight": "int8-TN",
|
||||
"model.layers.28.self_attn.v_proj.weight": "int8-TN",
|
||||
"model.layers.28.share_expert.gate_proj.weight": "torch.bfloat16",
|
||||
"model.layers.28.share_expert.up_proj.weight": "torch.bfloat16",
|
||||
"model.layers.29.self_attn.g_proj.weight": "int8-TN",
|
||||
"model.layers.29.self_attn.k_proj.weight": "int8-TN",
|
||||
"model.layers.29.self_attn.q_proj.weight": "int8-TN",
|
||||
"model.layers.29.self_attn.v_proj.weight": "int8-TN",
|
||||
"model.layers.29.share_expert.gate_proj.weight": "torch.bfloat16",
|
||||
"model.layers.29.share_expert.up_proj.weight": "torch.bfloat16",
|
||||
"model.layers.30.self_attn.g_proj.weight": "int8-TN",
|
||||
"model.layers.30.self_attn.k_proj.weight": "int8-TN",
|
||||
"model.layers.30.self_attn.q_proj.weight": "int8-TN",
|
||||
"model.layers.30.self_attn.v_proj.weight": "int8-TN",
|
||||
"model.layers.30.share_expert.gate_proj.weight": "torch.bfloat16",
|
||||
"model.layers.30.share_expert.up_proj.weight": "torch.bfloat16",
|
||||
"model.layers.31.self_attn.g_proj.weight": "int8-TN",
|
||||
"model.layers.31.self_attn.k_proj.weight": "int8-TN",
|
||||
"model.layers.31.self_attn.q_proj.weight": "int8-TN",
|
||||
"model.layers.31.self_attn.v_proj.weight": "int8-TN",
|
||||
"model.layers.31.share_expert.gate_proj.weight": "torch.bfloat16",
|
||||
"model.layers.31.share_expert.up_proj.weight": "torch.bfloat16",
|
||||
"model.layers.32.self_attn.g_proj.weight": "int8-TN",
|
||||
"model.layers.32.self_attn.k_proj.weight": "int8-TN",
|
||||
"model.layers.32.self_attn.q_proj.weight": "int8-TN",
|
||||
"model.layers.32.self_attn.v_proj.weight": "int8-TN",
|
||||
"model.layers.32.share_expert.gate_proj.weight": "torch.bfloat16",
|
||||
"model.layers.32.share_expert.up_proj.weight": "torch.bfloat16",
|
||||
"model.layers.33.self_attn.g_proj.weight": "int8-TN",
|
||||
"model.layers.33.self_attn.k_proj.weight": "int8-TN",
|
||||
"model.layers.33.self_attn.q_proj.weight": "int8-TN",
|
||||
"model.layers.33.self_attn.v_proj.weight": "int8-TN",
|
||||
"model.layers.33.share_expert.gate_proj.weight": "torch.bfloat16",
|
||||
"model.layers.33.share_expert.up_proj.weight": "torch.bfloat16",
|
||||
"model.layers.34.self_attn.g_proj.weight": "int8-TN",
|
||||
"model.layers.34.self_attn.k_proj.weight": "int8-TN",
|
||||
"model.layers.34.self_attn.q_proj.weight": "int8-TN",
|
||||
"model.layers.34.self_attn.v_proj.weight": "int8-TN",
|
||||
"model.layers.34.share_expert.gate_proj.weight": "torch.bfloat16",
|
||||
"model.layers.34.share_expert.up_proj.weight": "torch.bfloat16",
|
||||
"model.layers.35.self_attn.g_proj.weight": "int8-TN",
|
||||
"model.layers.35.self_attn.k_proj.weight": "int8-TN",
|
||||
"model.layers.35.self_attn.q_proj.weight": "int8-TN",
|
||||
"model.layers.35.self_attn.v_proj.weight": "int8-TN",
|
||||
"model.layers.35.share_expert.gate_proj.weight": "torch.bfloat16",
|
||||
"model.layers.35.share_expert.up_proj.weight": "torch.bfloat16",
|
||||
"model.layers.36.self_attn.g_proj.weight": "int8-TN",
|
||||
"model.layers.36.self_attn.k_proj.weight": "int8-TN",
|
||||
"model.layers.36.self_attn.q_proj.weight": "int8-TN",
|
||||
"model.layers.36.self_attn.v_proj.weight": "int8-TN",
|
||||
"model.layers.36.share_expert.gate_proj.weight": "torch.bfloat16",
|
||||
"model.layers.36.share_expert.up_proj.weight": "torch.bfloat16",
|
||||
"model.layers.37.self_attn.g_proj.weight": "int8-TN",
|
||||
"model.layers.37.self_attn.k_proj.weight": "int8-TN",
|
||||
"model.layers.37.self_attn.q_proj.weight": "int8-TN",
|
||||
"model.layers.37.self_attn.v_proj.weight": "int8-TN",
|
||||
"model.layers.37.share_expert.gate_proj.weight": "torch.bfloat16",
|
||||
"model.layers.37.share_expert.up_proj.weight": "torch.bfloat16",
|
||||
"model.layers.38.self_attn.g_proj.weight": "int8-TN",
|
||||
"model.layers.38.self_attn.k_proj.weight": "int8-TN",
|
||||
"model.layers.38.self_attn.q_proj.weight": "int8-TN",
|
||||
"model.layers.38.self_attn.v_proj.weight": "int8-TN",
|
||||
"model.layers.38.share_expert.gate_proj.weight": "torch.bfloat16",
|
||||
"model.layers.38.share_expert.up_proj.weight": "torch.bfloat16",
|
||||
"model.layers.39.self_attn.g_proj.weight": "int8-TN",
|
||||
"model.layers.39.self_attn.k_proj.weight": "int8-TN",
|
||||
"model.layers.39.self_attn.q_proj.weight": "int8-TN",
|
||||
"model.layers.39.self_attn.v_proj.weight": "int8-TN",
|
||||
"model.layers.39.share_expert.gate_proj.weight": "torch.bfloat16",
|
||||
"model.layers.39.share_expert.up_proj.weight": "torch.bfloat16",
|
||||
"model.layers.40.self_attn.g_proj.weight": "int8-TN",
|
||||
"model.layers.40.self_attn.k_proj.weight": "int8-TN",
|
||||
"model.layers.40.self_attn.q_proj.weight": "int8-TN",
|
||||
"model.layers.40.self_attn.v_proj.weight": "int8-TN",
|
||||
"model.layers.40.share_expert.gate_proj.weight": "torch.bfloat16",
|
||||
"model.layers.40.share_expert.up_proj.weight": "torch.bfloat16",
|
||||
"model.layers.41.self_attn.g_proj.weight": "int8-TN",
|
||||
"model.layers.41.self_attn.k_proj.weight": "int8-TN",
|
||||
"model.layers.41.self_attn.q_proj.weight": "int8-TN",
|
||||
"model.layers.41.self_attn.v_proj.weight": "int8-TN",
|
||||
"model.layers.41.share_expert.gate_proj.weight": "torch.bfloat16",
|
||||
"model.layers.41.share_expert.up_proj.weight": "torch.bfloat16",
|
||||
"model.layers.42.self_attn.g_proj.weight": "int8-TN",
|
||||
"model.layers.42.self_attn.k_proj.weight": "int8-TN",
|
||||
"model.layers.42.self_attn.q_proj.weight": "int8-TN",
|
||||
"model.layers.42.self_attn.v_proj.weight": "int8-TN",
|
||||
"model.layers.42.share_expert.gate_proj.weight": "torch.bfloat16",
|
||||
"model.layers.42.share_expert.up_proj.weight": "torch.bfloat16",
|
||||
"model.layers.43.self_attn.g_proj.weight": "int8-TN",
|
||||
"model.layers.43.self_attn.k_proj.weight": "int8-TN",
|
||||
"model.layers.43.self_attn.q_proj.weight": "int8-TN",
|
||||
"model.layers.43.self_attn.v_proj.weight": "int8-TN",
|
||||
"model.layers.43.share_expert.gate_proj.weight": "torch.bfloat16",
|
||||
"model.layers.43.share_expert.up_proj.weight": "torch.bfloat16",
|
||||
"model.layers.44.self_attn.g_proj.weight": "int8-TN",
|
||||
"model.layers.44.self_attn.k_proj.weight": "int8-TN",
|
||||
"model.layers.44.self_attn.q_proj.weight": "int8-TN",
|
||||
"model.layers.44.self_attn.v_proj.weight": "int8-TN",
|
||||
"model.layers.44.share_expert.gate_proj.weight": "torch.bfloat16",
|
||||
"model.layers.44.share_expert.up_proj.weight": "torch.bfloat16",
|
||||
"model.layers.45.eh_proj.weight": "torch.bfloat16",
|
||||
"model.layers.45.enorm.weight": "torch.bfloat16",
|
||||
"model.layers.45.hnorm.weight": "torch.bfloat16",
|
||||
"model.layers.45.input_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.45.mlp.down_proj.weight": "int8-TN",
|
||||
"model.layers.45.mlp.gate_proj.weight": "int8-TN",
|
||||
"model.layers.45.mlp.up_proj.weight": "int8-TN",
|
||||
"model.layers.45.post_attention_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.45.self_attn.g_proj.weight": "int8-TN",
|
||||
"model.layers.45.self_attn.k_norm.weight": "torch.bfloat16",
|
||||
"model.layers.45.self_attn.k_proj.weight": "int8-TN",
|
||||
"model.layers.45.self_attn.o_proj.weight": "int8-TN",
|
||||
"model.layers.45.self_attn.q_norm.weight": "torch.bfloat16",
|
||||
"model.layers.45.self_attn.q_proj.weight": "int8-TN",
|
||||
"model.layers.45.self_attn.v_proj.weight": "int8-TN",
|
||||
"model.layers.45.transformer.shared_head.output.weight": "torch.bfloat16",
|
||||
"model.layers.46.eh_proj.weight": "torch.bfloat16",
|
||||
"model.layers.46.enorm.weight": "torch.bfloat16",
|
||||
"model.layers.46.hnorm.weight": "torch.bfloat16",
|
||||
"model.layers.46.input_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.46.mlp.down_proj.weight": "int8-TN",
|
||||
"model.layers.46.mlp.gate_proj.weight": "int8-TN",
|
||||
"model.layers.46.mlp.up_proj.weight": "int8-TN",
|
||||
"model.layers.46.post_attention_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.46.self_attn.g_proj.weight": "int8-TN",
|
||||
"model.layers.46.self_attn.k_norm.weight": "torch.bfloat16",
|
||||
"model.layers.46.self_attn.k_proj.weight": "int8-TN",
|
||||
"model.layers.46.self_attn.o_proj.weight": "int8-TN",
|
||||
"model.layers.46.self_attn.q_norm.weight": "torch.bfloat16",
|
||||
"model.layers.46.self_attn.q_proj.weight": "int8-TN",
|
||||
"model.layers.46.self_attn.v_proj.weight": "int8-TN",
|
||||
"model.layers.46.transformer.shared_head.norm.weight": "torch.bfloat16",
|
||||
"model.layers.46.transformer.shared_head.output.weight": "torch.bfloat16",
|
||||
"model.layers.47.eh_proj.weight": "torch.bfloat16",
|
||||
"model.layers.47.enorm.weight": "torch.bfloat16",
|
||||
"model.layers.47.hnorm.weight": "torch.bfloat16",
|
||||
"model.layers.47.input_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.47.mlp.down_proj.weight": "int8-TN",
|
||||
"model.layers.47.mlp.gate_proj.weight": "int8-TN",
|
||||
"model.layers.47.mlp.up_proj.weight": "int8-TN",
|
||||
"model.layers.47.post_attention_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.47.self_attn.g_proj.weight": "int8-TN",
|
||||
"model.layers.47.self_attn.k_norm.weight": "torch.bfloat16",
|
||||
"model.layers.47.self_attn.k_proj.weight": "int8-TN",
|
||||
"model.layers.47.self_attn.o_proj.weight": "int8-TN",
|
||||
"model.layers.47.self_attn.q_norm.weight": "torch.bfloat16",
|
||||
"model.layers.47.self_attn.q_proj.weight": "int8-TN",
|
||||
"model.layers.47.self_attn.v_proj.weight": "int8-TN",
|
||||
"model.layers.47.transformer.shared_head.norm.weight": "torch.bfloat16",
|
||||
"model.layers.47.transformer.shared_head.output.weight": "torch.bfloat16",
|
||||
"model.norm.weight": "torch.bfloat16",
|
||||
"model.layers.10.moe.router_bias": "torch.float32",
|
||||
"model.layers.11.moe.router_bias": "torch.float32",
|
||||
"model.layers.12.moe.router_bias": "torch.float32",
|
||||
"model.layers.13.moe.router_bias": "torch.float32",
|
||||
"model.layers.14.moe.router_bias": "torch.float32",
|
||||
"model.layers.15.moe.router_bias": "torch.float32",
|
||||
"model.layers.16.moe.router_bias": "torch.float32",
|
||||
"model.layers.17.moe.router_bias": "torch.float32",
|
||||
"model.layers.18.moe.router_bias": "torch.float32",
|
||||
"model.layers.19.moe.router_bias": "torch.float32",
|
||||
"model.layers.20.moe.router_bias": "torch.float32",
|
||||
"model.layers.21.moe.router_bias": "torch.float32",
|
||||
"model.layers.22.moe.router_bias": "torch.float32",
|
||||
"model.layers.23.moe.router_bias": "torch.float32",
|
||||
"model.layers.24.moe.router_bias": "torch.float32",
|
||||
"model.layers.25.moe.router_bias": "torch.float32",
|
||||
"model.layers.26.moe.router_bias": "torch.float32",
|
||||
"model.layers.27.moe.router_bias": "torch.float32",
|
||||
"model.layers.28.moe.router_bias": "torch.float32",
|
||||
"model.layers.29.moe.router_bias": "torch.float32",
|
||||
"model.layers.3.moe.router_bias": "torch.float32",
|
||||
"model.layers.30.moe.router_bias": "torch.float32",
|
||||
"model.layers.31.moe.router_bias": "torch.float32",
|
||||
"model.layers.32.moe.router_bias": "torch.float32",
|
||||
"model.layers.33.moe.router_bias": "torch.float32",
|
||||
"model.layers.34.moe.router_bias": "torch.float32",
|
||||
"model.layers.35.moe.router_bias": "torch.float32",
|
||||
"model.layers.36.moe.router_bias": "torch.float32",
|
||||
"model.layers.37.moe.router_bias": "torch.float32",
|
||||
"model.layers.38.moe.router_bias": "torch.float32",
|
||||
"model.layers.39.moe.router_bias": "torch.float32",
|
||||
"model.layers.4.moe.router_bias": "torch.float32",
|
||||
"model.layers.40.moe.router_bias": "torch.float32",
|
||||
"model.layers.41.moe.router_bias": "torch.float32",
|
||||
"model.layers.42.moe.router_bias": "torch.float32",
|
||||
"model.layers.43.moe.router_bias": "torch.float32",
|
||||
"model.layers.44.moe.router_bias": "torch.float32",
|
||||
"model.layers.5.moe.router_bias": "torch.float32",
|
||||
"model.layers.6.moe.router_bias": "torch.float32",
|
||||
"model.layers.7.moe.router_bias": "torch.float32",
|
||||
"model.layers.8.moe.router_bias": "torch.float32",
|
||||
"model.layers.9.moe.router_bias": "torch.float32",
|
||||
"model.embed_tokens.weight": "torch.bfloat16",
|
||||
"model.layers.0.input_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.0.mlp.down_proj.weight": "int8-TN",
|
||||
"model.layers.0.mlp.gate_proj.weight": "int8-TN",
|
||||
"model.layers.0.mlp.up_proj.weight": "int8-TN",
|
||||
"model.layers.0.post_attention_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.0.self_attn.g_proj.weight": "int8-TN",
|
||||
"model.layers.0.self_attn.k_norm.weight": "torch.bfloat16",
|
||||
"model.layers.0.self_attn.k_proj.weight": "int8-TN",
|
||||
"model.layers.0.self_attn.o_proj.weight": "int8-TN",
|
||||
"model.layers.0.self_attn.q_norm.weight": "torch.bfloat16",
|
||||
"model.layers.0.self_attn.q_proj.weight": "int8-TN",
|
||||
"model.layers.0.self_attn.v_proj.weight": "int8-TN",
|
||||
"model.layers.1.input_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.1.mlp.down_proj.weight": "int8-TN",
|
||||
"model.layers.1.mlp.gate_proj.weight": "int8-TN",
|
||||
"model.layers.1.mlp.up_proj.weight": "int8-TN",
|
||||
"model.layers.1.post_attention_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.1.self_attn.g_proj.weight": "int8-TN",
|
||||
"model.layers.1.self_attn.k_norm.weight": "torch.bfloat16",
|
||||
"model.layers.1.self_attn.k_proj.weight": "int8-TN",
|
||||
"model.layers.1.self_attn.o_proj.weight": "int8-TN",
|
||||
"model.layers.1.self_attn.q_norm.weight": "torch.bfloat16",
|
||||
"model.layers.1.self_attn.q_proj.weight": "int8-TN",
|
||||
"model.layers.1.self_attn.v_proj.weight": "int8-TN",
|
||||
"model.layers.10.input_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.10.moe.gate.weight": "torch.bfloat16",
|
||||
"model.layers.10.post_attention_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.10.self_attn.g_proj.weight": "int8-TN",
|
||||
"model.layers.10.self_attn.k_norm.weight": "torch.bfloat16",
|
||||
"model.layers.10.self_attn.k_proj.weight": "int8-TN",
|
||||
"model.layers.10.self_attn.o_proj.weight": "int8-TN",
|
||||
"model.layers.10.self_attn.q_norm.weight": "torch.bfloat16",
|
||||
"model.layers.10.self_attn.q_proj.weight": "int8-TN",
|
||||
"model.layers.10.self_attn.v_proj.weight": "int8-TN",
|
||||
"model.layers.10.share_expert.down_proj.weight": "torch.bfloat16",
|
||||
"model.layers.10.share_expert.gate_proj.weight": "torch.bfloat16",
|
||||
"model.layers.10.share_expert.up_proj.weight": "torch.bfloat16",
|
||||
"model.layers.11.input_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.11.moe.gate.weight": "torch.bfloat16",
|
||||
"model.layers.11.post_attention_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.11.self_attn.g_proj.weight": "int8-TN",
|
||||
"model.layers.11.self_attn.k_norm.weight": "torch.bfloat16",
|
||||
"model.layers.11.self_attn.k_proj.weight": "int8-TN",
|
||||
"model.layers.11.self_attn.o_proj.weight": "int8-TN",
|
||||
"model.layers.11.self_attn.q_norm.weight": "torch.bfloat16",
|
||||
"model.layers.11.self_attn.q_proj.weight": "int8-TN",
|
||||
"model.layers.11.self_attn.v_proj.weight": "int8-TN",
|
||||
"model.layers.11.share_expert.down_proj.weight": "torch.bfloat16",
|
||||
"model.layers.11.share_expert.gate_proj.weight": "torch.bfloat16",
|
||||
"model.layers.11.share_expert.up_proj.weight": "torch.bfloat16",
|
||||
"model.layers.12.input_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.12.moe.gate.weight": "torch.bfloat16",
|
||||
"model.layers.12.post_attention_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.12.self_attn.g_proj.weight": "int8-TN",
|
||||
"model.layers.12.self_attn.k_norm.weight": "torch.bfloat16",
|
||||
"model.layers.12.self_attn.k_proj.weight": "int8-TN",
|
||||
"model.layers.12.self_attn.o_proj.weight": "int8-TN",
|
||||
"model.layers.12.self_attn.q_norm.weight": "torch.bfloat16",
|
||||
"model.layers.12.self_attn.q_proj.weight": "int8-TN",
|
||||
"model.layers.12.self_attn.v_proj.weight": "int8-TN",
|
||||
"model.layers.12.share_expert.down_proj.weight": "torch.bfloat16",
|
||||
"model.layers.12.share_expert.gate_proj.weight": "torch.bfloat16",
|
||||
"model.layers.12.share_expert.up_proj.weight": "torch.bfloat16",
|
||||
"model.layers.13.input_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.13.moe.gate.weight": "torch.bfloat16",
|
||||
"model.layers.13.post_attention_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.13.self_attn.g_proj.weight": "int8-TN",
|
||||
"model.layers.13.self_attn.k_norm.weight": "torch.bfloat16",
|
||||
"model.layers.13.self_attn.k_proj.weight": "int8-TN",
|
||||
"model.layers.13.self_attn.o_proj.weight": "int8-TN",
|
||||
"model.layers.13.self_attn.q_norm.weight": "torch.bfloat16",
|
||||
"model.layers.13.self_attn.q_proj.weight": "int8-TN",
|
||||
"model.layers.13.self_attn.v_proj.weight": "int8-TN",
|
||||
"model.layers.13.share_expert.down_proj.weight": "torch.bfloat16",
|
||||
"model.layers.13.share_expert.gate_proj.weight": "torch.bfloat16",
|
||||
"model.layers.13.share_expert.up_proj.weight": "torch.bfloat16",
|
||||
"model.layers.14.input_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.14.moe.gate.weight": "torch.bfloat16",
|
||||
"model.layers.14.post_attention_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.14.self_attn.g_proj.weight": "int8-TN",
|
||||
"model.layers.14.self_attn.k_norm.weight": "torch.bfloat16",
|
||||
"model.layers.14.self_attn.k_proj.weight": "int8-TN",
|
||||
"model.layers.14.self_attn.o_proj.weight": "int8-TN",
|
||||
"model.layers.14.self_attn.q_norm.weight": "torch.bfloat16",
|
||||
"model.layers.14.self_attn.q_proj.weight": "int8-TN",
|
||||
"model.layers.14.self_attn.v_proj.weight": "int8-TN",
|
||||
"model.layers.14.share_expert.down_proj.weight": "torch.bfloat16",
|
||||
"model.layers.14.share_expert.gate_proj.weight": "torch.bfloat16",
|
||||
"model.layers.14.share_expert.up_proj.weight": "torch.bfloat16",
|
||||
"model.layers.15.input_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.15.moe.gate.weight": "torch.bfloat16",
|
||||
"model.layers.15.post_attention_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.15.self_attn.g_proj.weight": "int8-TN",
|
||||
"model.layers.15.self_attn.k_norm.weight": "torch.bfloat16",
|
||||
"model.layers.15.self_attn.k_proj.weight": "int8-TN",
|
||||
"model.layers.15.self_attn.o_proj.weight": "int8-TN",
|
||||
"model.layers.15.self_attn.q_norm.weight": "torch.bfloat16",
|
||||
"model.layers.15.self_attn.q_proj.weight": "int8-TN",
|
||||
"model.layers.15.self_attn.v_proj.weight": "int8-TN",
|
||||
"model.layers.15.share_expert.down_proj.weight": "torch.bfloat16",
|
||||
"model.layers.15.share_expert.gate_proj.weight": "torch.bfloat16",
|
||||
"model.layers.15.share_expert.up_proj.weight": "torch.bfloat16",
|
||||
"model.layers.16.input_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.16.moe.gate.weight": "torch.bfloat16",
|
||||
"model.layers.16.post_attention_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.16.self_attn.g_proj.weight": "int8-TN",
|
||||
"model.layers.16.self_attn.k_norm.weight": "torch.bfloat16",
|
||||
"model.layers.16.self_attn.k_proj.weight": "int8-TN",
|
||||
"model.layers.16.self_attn.o_proj.weight": "int8-TN",
|
||||
"model.layers.16.self_attn.q_norm.weight": "torch.bfloat16",
|
||||
"model.layers.16.self_attn.q_proj.weight": "int8-TN",
|
||||
"model.layers.16.self_attn.v_proj.weight": "int8-TN",
|
||||
"model.layers.16.share_expert.down_proj.weight": "torch.bfloat16",
|
||||
"model.layers.16.share_expert.gate_proj.weight": "torch.bfloat16",
|
||||
"model.layers.16.share_expert.up_proj.weight": "torch.bfloat16",
|
||||
"model.layers.17.input_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.17.moe.gate.weight": "torch.bfloat16",
|
||||
"model.layers.17.post_attention_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.17.self_attn.g_proj.weight": "int8-TN",
|
||||
"model.layers.17.self_attn.k_norm.weight": "torch.bfloat16",
|
||||
"model.layers.17.self_attn.k_proj.weight": "int8-TN",
|
||||
"model.layers.17.self_attn.o_proj.weight": "int8-TN",
|
||||
"model.layers.17.self_attn.q_norm.weight": "torch.bfloat16",
|
||||
"model.layers.17.self_attn.q_proj.weight": "int8-TN",
|
||||
"model.layers.17.self_attn.v_proj.weight": "int8-TN",
|
||||
"model.layers.17.share_expert.down_proj.weight": "torch.bfloat16",
|
||||
"model.layers.17.share_expert.gate_proj.weight": "torch.bfloat16",
|
||||
"model.layers.17.share_expert.up_proj.weight": "torch.bfloat16",
|
||||
"model.layers.18.input_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.18.moe.gate.weight": "torch.bfloat16",
|
||||
"model.layers.18.post_attention_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.18.self_attn.g_proj.weight": "int8-TN",
|
||||
"model.layers.18.self_attn.k_norm.weight": "torch.bfloat16",
|
||||
"model.layers.18.self_attn.k_proj.weight": "int8-TN",
|
||||
"model.layers.18.self_attn.o_proj.weight": "int8-TN",
|
||||
"model.layers.18.self_attn.q_norm.weight": "torch.bfloat16",
|
||||
"model.layers.18.self_attn.q_proj.weight": "int8-TN",
|
||||
"model.layers.18.self_attn.v_proj.weight": "int8-TN",
|
||||
"model.layers.18.share_expert.down_proj.weight": "torch.bfloat16",
|
||||
"model.layers.18.share_expert.gate_proj.weight": "torch.bfloat16",
|
||||
"model.layers.18.share_expert.up_proj.weight": "torch.bfloat16",
|
||||
"model.layers.19.input_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.19.moe.gate.weight": "torch.bfloat16",
|
||||
"model.layers.19.post_attention_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.19.self_attn.g_proj.weight": "int8-TN",
|
||||
"model.layers.19.self_attn.k_norm.weight": "torch.bfloat16",
|
||||
"model.layers.19.self_attn.k_proj.weight": "int8-TN",
|
||||
"model.layers.19.self_attn.o_proj.weight": "int8-TN",
|
||||
"model.layers.19.self_attn.q_norm.weight": "torch.bfloat16",
|
||||
"model.layers.19.self_attn.q_proj.weight": "int8-TN",
|
||||
"model.layers.19.self_attn.v_proj.weight": "int8-TN",
|
||||
"model.layers.19.share_expert.down_proj.weight": "torch.bfloat16",
|
||||
"model.layers.19.share_expert.gate_proj.weight": "torch.bfloat16",
|
||||
"model.layers.19.share_expert.up_proj.weight": "torch.bfloat16",
|
||||
"model.layers.2.input_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.2.mlp.down_proj.weight": "int8-TN",
|
||||
"model.layers.2.mlp.gate_proj.weight": "int8-TN",
|
||||
"model.layers.2.mlp.up_proj.weight": "int8-TN",
|
||||
"model.layers.2.post_attention_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.2.self_attn.g_proj.weight": "int8-TN",
|
||||
"model.layers.2.self_attn.k_norm.weight": "torch.bfloat16",
|
||||
"model.layers.2.self_attn.k_proj.weight": "int8-TN",
|
||||
"model.layers.2.self_attn.o_proj.weight": "int8-TN",
|
||||
"model.layers.2.self_attn.q_norm.weight": "torch.bfloat16",
|
||||
"model.layers.2.self_attn.q_proj.weight": "int8-TN",
|
||||
"model.layers.2.self_attn.v_proj.weight": "int8-TN",
|
||||
"model.layers.20.input_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.20.moe.gate.weight": "torch.bfloat16",
|
||||
"model.layers.20.post_attention_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.20.self_attn.g_proj.weight": "int8-TN",
|
||||
"model.layers.20.self_attn.k_norm.weight": "torch.bfloat16",
|
||||
"model.layers.20.self_attn.k_proj.weight": "int8-TN",
|
||||
"model.layers.20.self_attn.o_proj.weight": "int8-TN",
|
||||
"model.layers.20.self_attn.q_norm.weight": "torch.bfloat16",
|
||||
"model.layers.20.self_attn.q_proj.weight": "int8-TN",
|
||||
"model.layers.20.self_attn.v_proj.weight": "int8-TN",
|
||||
"model.layers.20.share_expert.down_proj.weight": "torch.bfloat16",
|
||||
"model.layers.20.share_expert.gate_proj.weight": "torch.bfloat16",
|
||||
"model.layers.20.share_expert.up_proj.weight": "torch.bfloat16",
|
||||
"model.layers.21.input_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.21.moe.gate.weight": "torch.bfloat16",
|
||||
"model.layers.21.post_attention_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.21.self_attn.g_proj.weight": "int8-TN",
|
||||
"model.layers.21.self_attn.k_norm.weight": "torch.bfloat16",
|
||||
"model.layers.21.self_attn.k_proj.weight": "int8-TN",
|
||||
"model.layers.21.self_attn.o_proj.weight": "int8-TN",
|
||||
"model.layers.21.self_attn.q_norm.weight": "torch.bfloat16",
|
||||
"model.layers.21.self_attn.q_proj.weight": "int8-TN",
|
||||
"model.layers.21.self_attn.v_proj.weight": "int8-TN",
|
||||
"model.layers.21.share_expert.down_proj.weight": "torch.bfloat16",
|
||||
"model.layers.21.share_expert.gate_proj.weight": "torch.bfloat16",
|
||||
"model.layers.21.share_expert.up_proj.weight": "torch.bfloat16",
|
||||
"model.layers.22.input_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.22.moe.gate.weight": "torch.bfloat16",
|
||||
"model.layers.22.post_attention_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.22.self_attn.g_proj.weight": "int8-TN",
|
||||
"model.layers.22.self_attn.k_norm.weight": "torch.bfloat16",
|
||||
"model.layers.22.self_attn.k_proj.weight": "int8-TN",
|
||||
"model.layers.22.self_attn.o_proj.weight": "int8-TN",
|
||||
"model.layers.22.self_attn.q_norm.weight": "torch.bfloat16",
|
||||
"model.layers.22.self_attn.q_proj.weight": "int8-TN",
|
||||
"model.layers.22.self_attn.v_proj.weight": "int8-TN",
|
||||
"model.layers.22.share_expert.down_proj.weight": "torch.bfloat16",
|
||||
"model.layers.22.share_expert.gate_proj.weight": "torch.bfloat16",
|
||||
"model.layers.22.share_expert.up_proj.weight": "torch.bfloat16",
|
||||
"model.layers.23.input_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.23.moe.gate.weight": "torch.bfloat16",
|
||||
"model.layers.23.post_attention_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.23.self_attn.g_proj.weight": "int8-TN",
|
||||
"model.layers.23.self_attn.k_norm.weight": "torch.bfloat16",
|
||||
"model.layers.23.self_attn.k_proj.weight": "int8-TN",
|
||||
"model.layers.23.self_attn.o_proj.weight": "int8-TN",
|
||||
"model.layers.23.self_attn.q_norm.weight": "torch.bfloat16",
|
||||
"model.layers.23.self_attn.q_proj.weight": "int8-TN",
|
||||
"model.layers.23.self_attn.v_proj.weight": "int8-TN",
|
||||
"model.layers.23.share_expert.down_proj.weight": "torch.bfloat16",
|
||||
"model.layers.23.share_expert.gate_proj.weight": "torch.bfloat16",
|
||||
"model.layers.23.share_expert.up_proj.weight": "torch.bfloat16",
|
||||
"model.layers.24.input_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.24.moe.gate.weight": "torch.bfloat16",
|
||||
"model.layers.24.post_attention_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.24.self_attn.k_norm.weight": "torch.bfloat16",
|
||||
"model.layers.24.self_attn.o_proj.weight": "int8-TN",
|
||||
"model.layers.24.self_attn.q_norm.weight": "torch.bfloat16",
|
||||
"model.layers.24.share_expert.down_proj.weight": "torch.bfloat16",
|
||||
"model.layers.25.input_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.25.moe.gate.weight": "torch.bfloat16",
|
||||
"model.layers.25.post_attention_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.25.self_attn.k_norm.weight": "torch.bfloat16",
|
||||
"model.layers.25.self_attn.o_proj.weight": "int8-TN",
|
||||
"model.layers.25.self_attn.q_norm.weight": "torch.bfloat16",
|
||||
"model.layers.25.share_expert.down_proj.weight": "torch.bfloat16",
|
||||
"model.layers.26.input_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.26.moe.gate.weight": "torch.bfloat16",
|
||||
"model.layers.26.post_attention_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.26.self_attn.k_norm.weight": "torch.bfloat16",
|
||||
"model.layers.26.self_attn.o_proj.weight": "int8-TN",
|
||||
"model.layers.26.self_attn.q_norm.weight": "torch.bfloat16",
|
||||
"model.layers.26.share_expert.down_proj.weight": "torch.bfloat16",
|
||||
"model.layers.27.input_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.27.moe.gate.weight": "torch.bfloat16",
|
||||
"model.layers.27.post_attention_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.27.self_attn.k_norm.weight": "torch.bfloat16",
|
||||
"model.layers.27.self_attn.o_proj.weight": "int8-TN",
|
||||
"model.layers.27.self_attn.q_norm.weight": "torch.bfloat16",
|
||||
"model.layers.27.share_expert.down_proj.weight": "torch.bfloat16",
|
||||
"model.layers.28.input_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.28.moe.gate.weight": "torch.bfloat16",
|
||||
"model.layers.28.post_attention_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.28.self_attn.k_norm.weight": "torch.bfloat16",
|
||||
"model.layers.28.self_attn.o_proj.weight": "int8-TN",
|
||||
"model.layers.28.self_attn.q_norm.weight": "torch.bfloat16",
|
||||
"model.layers.28.share_expert.down_proj.weight": "torch.bfloat16",
|
||||
"model.layers.29.input_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.29.moe.gate.weight": "torch.bfloat16",
|
||||
"model.layers.29.post_attention_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.29.self_attn.k_norm.weight": "torch.bfloat16",
|
||||
"model.layers.29.self_attn.o_proj.weight": "int8-TN",
|
||||
"model.layers.29.self_attn.q_norm.weight": "torch.bfloat16",
|
||||
"model.layers.29.share_expert.down_proj.weight": "torch.bfloat16",
|
||||
"model.layers.3.input_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.3.moe.gate.weight": "torch.bfloat16",
|
||||
"model.layers.3.post_attention_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.3.self_attn.g_proj.weight": "int8-TN",
|
||||
"model.layers.3.self_attn.k_norm.weight": "torch.bfloat16",
|
||||
"model.layers.3.self_attn.k_proj.weight": "int8-TN",
|
||||
"model.layers.3.self_attn.o_proj.weight": "int8-TN",
|
||||
"model.layers.3.self_attn.q_norm.weight": "torch.bfloat16",
|
||||
"model.layers.3.self_attn.q_proj.weight": "int8-TN",
|
||||
"model.layers.3.self_attn.v_proj.weight": "int8-TN",
|
||||
"model.layers.3.share_expert.down_proj.weight": "torch.bfloat16",
|
||||
"model.layers.3.share_expert.gate_proj.weight": "torch.bfloat16",
|
||||
"model.layers.3.share_expert.up_proj.weight": "torch.bfloat16",
|
||||
"model.layers.30.input_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.30.moe.gate.weight": "torch.bfloat16",
|
||||
"model.layers.30.post_attention_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.30.self_attn.k_norm.weight": "torch.bfloat16",
|
||||
"model.layers.30.self_attn.o_proj.weight": "int8-TN",
|
||||
"model.layers.30.self_attn.q_norm.weight": "torch.bfloat16",
|
||||
"model.layers.30.share_expert.down_proj.weight": "torch.bfloat16",
|
||||
"model.layers.31.input_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.31.moe.gate.weight": "torch.bfloat16",
|
||||
"model.layers.31.post_attention_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.31.self_attn.k_norm.weight": "torch.bfloat16",
|
||||
"model.layers.31.self_attn.o_proj.weight": "int8-TN",
|
||||
"model.layers.31.self_attn.q_norm.weight": "torch.bfloat16",
|
||||
"model.layers.31.share_expert.down_proj.weight": "torch.bfloat16",
|
||||
"model.layers.32.input_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.32.moe.gate.weight": "torch.bfloat16",
|
||||
"model.layers.32.post_attention_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.32.self_attn.k_norm.weight": "torch.bfloat16",
|
||||
"model.layers.32.self_attn.o_proj.weight": "int8-TN",
|
||||
"model.layers.32.self_attn.q_norm.weight": "torch.bfloat16",
|
||||
"model.layers.32.share_expert.down_proj.weight": "torch.bfloat16",
|
||||
"model.layers.33.input_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.33.moe.gate.weight": "torch.bfloat16",
|
||||
"model.layers.33.post_attention_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.33.self_attn.k_norm.weight": "torch.bfloat16",
|
||||
"model.layers.33.self_attn.o_proj.weight": "int8-TN",
|
||||
"model.layers.33.self_attn.q_norm.weight": "torch.bfloat16",
|
||||
"model.layers.33.share_expert.down_proj.weight": "torch.bfloat16",
|
||||
"model.layers.34.input_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.34.moe.gate.weight": "torch.bfloat16",
|
||||
"model.layers.34.post_attention_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.34.self_attn.k_norm.weight": "torch.bfloat16",
|
||||
"model.layers.34.self_attn.o_proj.weight": "int8-TN",
|
||||
"model.layers.34.self_attn.q_norm.weight": "torch.bfloat16",
|
||||
"model.layers.34.share_expert.down_proj.weight": "torch.bfloat16",
|
||||
"model.layers.35.input_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.35.moe.gate.weight": "torch.bfloat16",
|
||||
"model.layers.35.post_attention_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.35.self_attn.k_norm.weight": "torch.bfloat16",
|
||||
"model.layers.35.self_attn.o_proj.weight": "int8-TN",
|
||||
"model.layers.35.self_attn.q_norm.weight": "torch.bfloat16",
|
||||
"model.layers.35.share_expert.down_proj.weight": "torch.bfloat16",
|
||||
"model.layers.36.input_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.36.moe.gate.weight": "torch.bfloat16",
|
||||
"model.layers.36.post_attention_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.36.self_attn.k_norm.weight": "torch.bfloat16",
|
||||
"model.layers.36.self_attn.o_proj.weight": "int8-TN",
|
||||
"model.layers.36.self_attn.q_norm.weight": "torch.bfloat16",
|
||||
"model.layers.36.share_expert.down_proj.weight": "torch.bfloat16",
|
||||
"model.layers.37.input_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.37.moe.gate.weight": "torch.bfloat16",
|
||||
"model.layers.37.post_attention_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.37.self_attn.k_norm.weight": "torch.bfloat16",
|
||||
"model.layers.37.self_attn.o_proj.weight": "int8-TN",
|
||||
"model.layers.37.self_attn.q_norm.weight": "torch.bfloat16",
|
||||
"model.layers.37.share_expert.down_proj.weight": "torch.bfloat16",
|
||||
"model.layers.38.input_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.38.moe.gate.weight": "torch.bfloat16",
|
||||
"model.layers.38.post_attention_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.38.self_attn.k_norm.weight": "torch.bfloat16",
|
||||
"model.layers.38.self_attn.o_proj.weight": "int8-TN",
|
||||
"model.layers.38.self_attn.q_norm.weight": "torch.bfloat16",
|
||||
"model.layers.38.share_expert.down_proj.weight": "torch.bfloat16",
|
||||
"model.layers.39.input_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.39.moe.gate.weight": "torch.bfloat16",
|
||||
"model.layers.39.post_attention_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.39.self_attn.k_norm.weight": "torch.bfloat16",
|
||||
"model.layers.39.self_attn.o_proj.weight": "int8-TN",
|
||||
"model.layers.39.self_attn.q_norm.weight": "torch.bfloat16",
|
||||
"model.layers.39.share_expert.down_proj.weight": "torch.bfloat16",
|
||||
"model.layers.4.input_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.4.moe.gate.weight": "torch.bfloat16",
|
||||
"model.layers.4.post_attention_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.4.self_attn.g_proj.weight": "int8-TN",
|
||||
"model.layers.4.self_attn.k_norm.weight": "torch.bfloat16",
|
||||
"model.layers.4.self_attn.k_proj.weight": "int8-TN",
|
||||
"model.layers.4.self_attn.o_proj.weight": "int8-TN",
|
||||
"model.layers.4.self_attn.q_norm.weight": "torch.bfloat16",
|
||||
"model.layers.4.self_attn.q_proj.weight": "int8-TN",
|
||||
"model.layers.4.self_attn.v_proj.weight": "int8-TN",
|
||||
"model.layers.4.share_expert.down_proj.weight": "torch.bfloat16",
|
||||
"model.layers.4.share_expert.gate_proj.weight": "torch.bfloat16",
|
||||
"model.layers.4.share_expert.up_proj.weight": "torch.bfloat16",
|
||||
"model.layers.40.input_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.40.moe.gate.weight": "torch.bfloat16",
|
||||
"model.layers.40.post_attention_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.40.self_attn.k_norm.weight": "torch.bfloat16",
|
||||
"model.layers.40.self_attn.o_proj.weight": "int8-TN",
|
||||
"model.layers.40.self_attn.q_norm.weight": "torch.bfloat16",
|
||||
"model.layers.40.share_expert.down_proj.weight": "torch.bfloat16",
|
||||
"model.layers.41.input_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.41.moe.gate.weight": "torch.bfloat16",
|
||||
"model.layers.41.post_attention_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.41.self_attn.k_norm.weight": "torch.bfloat16",
|
||||
"model.layers.41.self_attn.o_proj.weight": "int8-TN",
|
||||
"model.layers.41.self_attn.q_norm.weight": "torch.bfloat16",
|
||||
"model.layers.41.share_expert.down_proj.weight": "torch.bfloat16",
|
||||
"model.layers.42.input_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.42.moe.gate.weight": "torch.bfloat16",
|
||||
"model.layers.42.post_attention_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.42.self_attn.k_norm.weight": "torch.bfloat16",
|
||||
"model.layers.42.self_attn.o_proj.weight": "int8-TN",
|
||||
"model.layers.42.self_attn.q_norm.weight": "torch.bfloat16",
|
||||
"model.layers.42.share_expert.down_proj.weight": "torch.bfloat16",
|
||||
"model.layers.43.input_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.43.moe.gate.weight": "torch.bfloat16",
|
||||
"model.layers.43.post_attention_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.43.self_attn.k_norm.weight": "torch.bfloat16",
|
||||
"model.layers.43.self_attn.o_proj.weight": "int8-TN",
|
||||
"model.layers.43.self_attn.q_norm.weight": "torch.bfloat16",
|
||||
"model.layers.43.share_expert.down_proj.weight": "torch.bfloat16",
|
||||
"model.layers.44.input_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.44.moe.gate.weight": "torch.bfloat16",
|
||||
"model.layers.44.post_attention_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.44.self_attn.k_norm.weight": "torch.bfloat16",
|
||||
"model.layers.44.self_attn.o_proj.weight": "int8-TN",
|
||||
"model.layers.44.self_attn.q_norm.weight": "torch.bfloat16",
|
||||
"model.layers.44.share_expert.down_proj.weight": "torch.bfloat16",
|
||||
"model.layers.45.transformer.shared_head.norm.weight": "torch.bfloat16",
|
||||
"model.layers.5.input_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.5.moe.gate.weight": "torch.bfloat16",
|
||||
"model.layers.5.post_attention_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.5.self_attn.g_proj.weight": "int8-TN",
|
||||
"model.layers.5.self_attn.k_norm.weight": "torch.bfloat16",
|
||||
"model.layers.5.self_attn.k_proj.weight": "int8-TN",
|
||||
"model.layers.5.self_attn.o_proj.weight": "int8-TN",
|
||||
"model.layers.5.self_attn.q_norm.weight": "torch.bfloat16",
|
||||
"model.layers.5.self_attn.q_proj.weight": "int8-TN",
|
||||
"model.layers.5.self_attn.v_proj.weight": "int8-TN",
|
||||
"model.layers.5.share_expert.down_proj.weight": "torch.bfloat16",
|
||||
"model.layers.5.share_expert.gate_proj.weight": "torch.bfloat16",
|
||||
"model.layers.5.share_expert.up_proj.weight": "torch.bfloat16",
|
||||
"model.layers.6.input_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.6.moe.gate.weight": "torch.bfloat16",
|
||||
"model.layers.6.post_attention_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.6.self_attn.g_proj.weight": "int8-TN",
|
||||
"model.layers.6.self_attn.k_norm.weight": "torch.bfloat16",
|
||||
"model.layers.6.self_attn.k_proj.weight": "int8-TN",
|
||||
"model.layers.6.self_attn.o_proj.weight": "int8-TN",
|
||||
"model.layers.6.self_attn.q_norm.weight": "torch.bfloat16",
|
||||
"model.layers.6.self_attn.q_proj.weight": "int8-TN",
|
||||
"model.layers.6.self_attn.v_proj.weight": "int8-TN",
|
||||
"model.layers.6.share_expert.down_proj.weight": "torch.bfloat16",
|
||||
"model.layers.6.share_expert.gate_proj.weight": "torch.bfloat16",
|
||||
"model.layers.6.share_expert.up_proj.weight": "torch.bfloat16",
|
||||
"model.layers.7.input_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.7.moe.gate.weight": "torch.bfloat16",
|
||||
"model.layers.7.post_attention_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.7.self_attn.g_proj.weight": "int8-TN",
|
||||
"model.layers.7.self_attn.k_norm.weight": "torch.bfloat16",
|
||||
"model.layers.7.self_attn.k_proj.weight": "int8-TN",
|
||||
"model.layers.7.self_attn.o_proj.weight": "int8-TN",
|
||||
"model.layers.7.self_attn.q_norm.weight": "torch.bfloat16",
|
||||
"model.layers.7.self_attn.q_proj.weight": "int8-TN",
|
||||
"model.layers.7.self_attn.v_proj.weight": "int8-TN",
|
||||
"model.layers.7.share_expert.down_proj.weight": "torch.bfloat16",
|
||||
"model.layers.7.share_expert.gate_proj.weight": "torch.bfloat16",
|
||||
"model.layers.7.share_expert.up_proj.weight": "torch.bfloat16",
|
||||
"model.layers.8.input_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.8.moe.gate.weight": "torch.bfloat16",
|
||||
"model.layers.8.post_attention_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.8.self_attn.g_proj.weight": "int8-TN",
|
||||
"model.layers.8.self_attn.k_norm.weight": "torch.bfloat16",
|
||||
"model.layers.8.self_attn.k_proj.weight": "int8-TN",
|
||||
"model.layers.8.self_attn.o_proj.weight": "int8-TN",
|
||||
"model.layers.8.self_attn.q_norm.weight": "torch.bfloat16",
|
||||
"model.layers.8.self_attn.q_proj.weight": "int8-TN",
|
||||
"model.layers.8.self_attn.v_proj.weight": "int8-TN",
|
||||
"model.layers.8.share_expert.down_proj.weight": "torch.bfloat16",
|
||||
"model.layers.8.share_expert.gate_proj.weight": "torch.bfloat16",
|
||||
"model.layers.8.share_expert.up_proj.weight": "torch.bfloat16",
|
||||
"model.layers.9.input_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.9.moe.gate.weight": "torch.bfloat16",
|
||||
"model.layers.9.post_attention_layernorm.weight": "torch.bfloat16",
|
||||
"model.layers.9.self_attn.g_proj.weight": "int8-TN",
|
||||
"model.layers.9.self_attn.k_norm.weight": "torch.bfloat16",
|
||||
"model.layers.9.self_attn.k_proj.weight": "int8-TN",
|
||||
"model.layers.9.self_attn.o_proj.weight": "int8-TN",
|
||||
"model.layers.9.self_attn.q_norm.weight": "torch.bfloat16",
|
||||
"model.layers.9.self_attn.q_proj.weight": "int8-TN",
|
||||
"model.layers.9.self_attn.v_proj.weight": "int8-TN",
|
||||
"model.layers.9.share_expert.down_proj.weight": "torch.bfloat16",
|
||||
"model.layers.9.share_expert.gate_proj.weight": "torch.bfloat16",
|
||||
"model.layers.9.share_expert.up_proj.weight": "torch.bfloat16",
|
||||
"model.layers.5.moe.down_proj.weight": "int8-TN",
|
||||
"model.layers.5.moe.gate_proj.weight": "int8-TN",
|
||||
"model.layers.5.moe.up_proj.weight": "int8-TN",
|
||||
"model.layers.10.moe.down_proj.weight": "int8-TN",
|
||||
"model.layers.10.moe.gate_proj.weight": "int8-TN",
|
||||
"model.layers.10.moe.up_proj.weight": "int8-TN",
|
||||
"model.layers.14.moe.down_proj.weight": "int8-TN",
|
||||
"model.layers.14.moe.gate_proj.weight": "int8-TN",
|
||||
"model.layers.14.moe.up_proj.weight": "int8-TN",
|
||||
"model.layers.4.moe.down_proj.weight": "int8-TN",
|
||||
"model.layers.4.moe.gate_proj.weight": "int8-TN",
|
||||
"model.layers.4.moe.up_proj.weight": "int8-TN",
|
||||
"model.layers.6.moe.down_proj.weight": "int8-TN",
|
||||
"model.layers.6.moe.gate_proj.weight": "int8-TN",
|
||||
"model.layers.6.moe.up_proj.weight": "int8-TN",
|
||||
"model.layers.7.moe.down_proj.weight": "int8-TN",
|
||||
"model.layers.7.moe.gate_proj.weight": "int8-TN",
|
||||
"model.layers.7.moe.up_proj.weight": "int8-TN",
|
||||
"model.layers.11.moe.down_proj.weight": "int8-TN",
|
||||
"model.layers.11.moe.gate_proj.weight": "int8-TN",
|
||||
"model.layers.11.moe.up_proj.weight": "int8-TN",
|
||||
"model.layers.13.moe.down_proj.weight": "int8-TN",
|
||||
"model.layers.13.moe.gate_proj.weight": "int8-TN",
|
||||
"model.layers.13.moe.up_proj.weight": "int8-TN",
|
||||
"model.layers.12.moe.down_proj.weight": "int8-TN",
|
||||
"model.layers.12.moe.gate_proj.weight": "int8-TN",
|
||||
"model.layers.12.moe.up_proj.weight": "int8-TN",
|
||||
"model.layers.9.moe.down_proj.weight": "int8-TN",
|
||||
"model.layers.9.moe.gate_proj.weight": "int8-TN",
|
||||
"model.layers.9.moe.up_proj.weight": "int8-TN",
|
||||
"model.layers.8.moe.down_proj.weight": "int8-TN",
|
||||
"model.layers.8.moe.gate_proj.weight": "int8-TN",
|
||||
"model.layers.8.moe.up_proj.weight": "int8-TN",
|
||||
"model.layers.16.moe.down_proj.weight": "int8-TN",
|
||||
"model.layers.16.moe.gate_proj.weight": "int8-TN",
|
||||
"model.layers.16.moe.up_proj.weight": "int8-TN",
|
||||
"model.layers.3.moe.down_proj.weight": "int8-TN",
|
||||
"model.layers.3.moe.gate_proj.weight": "int8-TN",
|
||||
"model.layers.3.moe.up_proj.weight": "int8-TN",
|
||||
"model.layers.15.moe.down_proj.weight": "int8-TN",
|
||||
"model.layers.15.moe.gate_proj.weight": "int8-TN",
|
||||
"model.layers.15.moe.up_proj.weight": "int8-TN",
|
||||
"model.layers.17.moe.down_proj.weight": "int8-TN",
|
||||
"model.layers.17.moe.gate_proj.weight": "int8-TN",
|
||||
"model.layers.17.moe.up_proj.weight": "int8-TN",
|
||||
"model.layers.18.moe.down_proj.weight": "int8-TN",
|
||||
"model.layers.18.moe.gate_proj.weight": "int8-TN",
|
||||
"model.layers.18.moe.up_proj.weight": "int8-TN",
|
||||
"model.layers.19.moe.down_proj.weight": "int8-TN",
|
||||
"model.layers.19.moe.gate_proj.weight": "int8-TN",
|
||||
"model.layers.19.moe.up_proj.weight": "int8-TN",
|
||||
"model.layers.20.moe.down_proj.weight": "int8-TN",
|
||||
"model.layers.20.moe.gate_proj.weight": "int8-TN",
|
||||
"model.layers.20.moe.up_proj.weight": "int8-TN",
|
||||
"model.layers.21.moe.down_proj.weight": "int8-TN",
|
||||
"model.layers.21.moe.gate_proj.weight": "int8-TN",
|
||||
"model.layers.21.moe.up_proj.weight": "int8-TN",
|
||||
"model.layers.22.moe.down_proj.weight": "int8-TN",
|
||||
"model.layers.22.moe.gate_proj.weight": "int8-TN",
|
||||
"model.layers.22.moe.up_proj.weight": "int8-TN",
|
||||
"model.layers.23.moe.down_proj.weight": "int8-TN",
|
||||
"model.layers.23.moe.gate_proj.weight": "int8-TN",
|
||||
"model.layers.23.moe.up_proj.weight": "int8-TN",
|
||||
"model.layers.24.moe.down_proj.weight": "int8-TN",
|
||||
"model.layers.24.moe.gate_proj.weight": "int8-TN",
|
||||
"model.layers.24.moe.up_proj.weight": "int8-TN",
|
||||
"model.layers.25.moe.down_proj.weight": "int8-TN",
|
||||
"model.layers.25.moe.gate_proj.weight": "int8-TN",
|
||||
"model.layers.25.moe.up_proj.weight": "int8-TN",
|
||||
"model.layers.26.moe.down_proj.weight": "int8-TN",
|
||||
"model.layers.26.moe.gate_proj.weight": "int8-TN",
|
||||
"model.layers.26.moe.up_proj.weight": "int8-TN",
|
||||
"model.layers.27.moe.down_proj.weight": "int8-TN",
|
||||
"model.layers.27.moe.gate_proj.weight": "int8-TN",
|
||||
"model.layers.27.moe.up_proj.weight": "int8-TN",
|
||||
"model.layers.28.moe.down_proj.weight": "int8-TN",
|
||||
"model.layers.28.moe.gate_proj.weight": "int8-TN",
|
||||
"model.layers.28.moe.up_proj.weight": "int8-TN",
|
||||
"model.layers.29.moe.down_proj.weight": "int8-TN",
|
||||
"model.layers.29.moe.gate_proj.weight": "int8-TN",
|
||||
"model.layers.29.moe.up_proj.weight": "int8-TN",
|
||||
"model.layers.30.moe.down_proj.weight": "int8-TN",
|
||||
"model.layers.30.moe.gate_proj.weight": "int8-TN",
|
||||
"model.layers.30.moe.up_proj.weight": "int8-TN",
|
||||
"model.layers.31.moe.down_proj.weight": "int8-TN",
|
||||
"model.layers.31.moe.gate_proj.weight": "int8-TN",
|
||||
"model.layers.31.moe.up_proj.weight": "int8-TN",
|
||||
"model.layers.32.moe.down_proj.weight": "int8-TN",
|
||||
"model.layers.32.moe.gate_proj.weight": "int8-TN",
|
||||
"model.layers.32.moe.up_proj.weight": "int8-TN",
|
||||
"model.layers.33.moe.down_proj.weight": "int8-TN",
|
||||
"model.layers.33.moe.gate_proj.weight": "int8-TN",
|
||||
"model.layers.33.moe.up_proj.weight": "int8-TN",
|
||||
"model.layers.34.moe.down_proj.weight": "int8-TN",
|
||||
"model.layers.34.moe.gate_proj.weight": "int8-TN",
|
||||
"model.layers.34.moe.up_proj.weight": "int8-TN",
|
||||
"model.layers.35.moe.down_proj.weight": "int8-TN",
|
||||
"model.layers.35.moe.gate_proj.weight": "int8-TN",
|
||||
"model.layers.35.moe.up_proj.weight": "int8-TN",
|
||||
"model.layers.36.moe.down_proj.weight": "int8-TN",
|
||||
"model.layers.36.moe.gate_proj.weight": "int8-TN",
|
||||
"model.layers.36.moe.up_proj.weight": "int8-TN",
|
||||
"model.layers.37.moe.down_proj.weight": "int8-TN",
|
||||
"model.layers.37.moe.gate_proj.weight": "int8-TN",
|
||||
"model.layers.37.moe.up_proj.weight": "int8-TN",
|
||||
"model.layers.38.moe.down_proj.weight": "int8-TN",
|
||||
"model.layers.38.moe.gate_proj.weight": "int8-TN",
|
||||
"model.layers.38.moe.up_proj.weight": "int8-TN",
|
||||
"model.layers.39.moe.down_proj.weight": "int8-TN",
|
||||
"model.layers.39.moe.gate_proj.weight": "int8-TN",
|
||||
"model.layers.39.moe.up_proj.weight": "int8-TN",
|
||||
"model.layers.40.moe.down_proj.weight": "int8-TN",
|
||||
"model.layers.40.moe.gate_proj.weight": "int8-TN",
|
||||
"model.layers.40.moe.up_proj.weight": "int8-TN",
|
||||
"model.layers.41.moe.down_proj.weight": "int8-TN",
|
||||
"model.layers.41.moe.gate_proj.weight": "int8-TN",
|
||||
"model.layers.41.moe.up_proj.weight": "int8-TN",
|
||||
"model.layers.42.moe.down_proj.weight": "int8-TN",
|
||||
"model.layers.42.moe.gate_proj.weight": "int8-TN",
|
||||
"model.layers.42.moe.up_proj.weight": "int8-TN",
|
||||
"model.layers.43.moe.down_proj.weight": "int8-TN",
|
||||
"model.layers.43.moe.gate_proj.weight": "int8-TN",
|
||||
"model.layers.43.moe.up_proj.weight": "int8-TN",
|
||||
"model.layers.44.moe.down_proj.weight": "int8-TN",
|
||||
"model.layers.44.moe.gate_proj.weight": "int8-TN",
|
||||
"model.layers.44.moe.up_proj.weight": "int8-TN"
|
||||
}
|
||||
Reference in New Issue
Block a user