first commit

dev/online_quant_configs/MiniMax-M2.5-w4a8/config.json

@@ -0,0 +1,146 @@
+{
+  "architectures": [
+    "MiniMaxM2ForCausalLM"
+  ],
+  "attn_type_list": [
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1
+  ],
+  "auto_map": {
+    "AutoConfig": "configuration_minimax_m2.MiniMaxM2Config",
+    "AutoModelForCausalLM": "modeling_minimax_m2.MiniMaxM2ForCausalLM"
+  },
+  "head_dim": 128,
+  "hidden_act": "silu",
+  "hidden_size": 3072,
+  "intermediate_size": 1536,
+  "max_position_embeddings": 196608,
+  "model_type": "minimax_m2",
+  "mtp_transformer_layers": 1,
+  "num_attention_heads": 48,
+  "num_experts_per_tok": 8,
+  "num_hidden_layers": 62,
+  "num_key_value_heads": 8,
+  "num_local_experts": 256,
+  "num_mtp_modules": 3,
+  "qk_norm_type": "per_layer",
+  "rms_norm_eps": 1e-06,
+  "rope_theta": 5000000,
+  "rotary_dim": 64,
+  "scoring_func": "sigmoid",
+  "shared_intermediate_size": 0,
+  "tie_word_embeddings": false,
+  "transformers_version": "4.46.1",
+  "use_cache": true,
+  "use_mtp": true,
+  "use_qk_norm": true,
+  "use_routing_bias": true,
+  "vocab_size": 200064,
+  "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": [
+      "model.embed_tokens.weight",
+      "re:model.layers.\\d+.input_layernorm.weight",
+      "re:model.layers.\\d+.self_attn.*",
+      "re:model.layers.\\d+.post_attention_layernorm.weight",
+      "re:model.layers.\\d+.block_sparse_moe.gate.weight",
+      "re:model.layers.\\d+.block_sparse_moe.e_score_correction_bias",
+      "model.norm.weight",
+      "lm_head.weight"
+    ],
+    "kv_cache_scheme": null,
+    "quant_method": "compressed-tensors",
+    "quantization_status": "frozen"
+  }
+}