Upload DeepseekV3ForCausalLM

config.json

@@ -18,24 +18,24 @@
   "hidden_size": 128,
   "initializer_range": 0.02,
   "intermediate_size": 256,
-  "kv_lora_rank": 16,
+  "kv_lora_rank": 32,
   "max_position_embeddings": 2048,
   "model_type": "deepseek_v3",
   "moe_intermediate_size": 64,
   "moe_layer_freq": 1,
   "n_group": 2,
-  "n_routed_experts": 4,
+  "n_routed_experts": 8,
   "n_shared_experts": 1,
   "norm_topk_prob": true,
-  "num_attention_heads": 4,
+  "num_attention_heads": 8,
   "num_experts_per_tok": 2,
   "num_hidden_layers": 6,
-  "num_key_value_heads": 2,
+  "num_key_value_heads": 8,
   "num_nextn_predict_layers": 1,
   "pretraining_tp": 1,
   "q_lora_rank": 32,
-  "qk_head_dim": 32,
-  "qk_nope_head_dim": 24,
+  "qk_head_dim": 24,
+  "qk_nope_head_dim": 16,
   "qk_rope_head_dim": 8,
   "quantization_config": {
     "activation_scheme": "dynamic",
@@ -49,9 +49,9 @@
   "rms_norm_eps": 1e-06,
   "rope_interleave": true,
   "rope_scaling": {
-    "beta_fast": 32,
-    "beta_slow": 1,
-    "factor": 40,
+    "beta_fast": 32.0,
+    "beta_slow": 1.0,
+    "factor": 40.0,
     "mscale": 1.0,
     "mscale_all_dim": 1.0,
     "original_max_position_embeddings": 2048,
@@ -67,6 +67,6 @@
   "torch_dtype": "float32",
   "transformers_version": "4.51.3",
   "use_cache": true,
-  "v_head_dim": 32,
-  "vocab_size": 163840
+  "v_head_dim": 16,
+  "vocab_size": 129280
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e068d3a17bef93953848bfdbf2a11b644d1f85316c1c35a07470d690b86ddcd4
-size 171101328
+oid sha256:5070c89c812ace8f741f9ffb2cb89dec28766ea792a77634f15988afe306fa2d
+size 136714544