Deepseek V3

config.json

@@ -9,6 +9,7 @@
     "AutoModel": "modeling_deepseek.DeepseekV3Model",
     "AutoModelForCausalLM": "modeling_deepseek.DeepseekV3ForCausalLM"
   },
+  "aux_loss_alpha": 0.001,
   "bos_token_id": 0,
   "eos_token_id": 1,
   "ep_size": 1,
@@ -31,6 +32,7 @@
   "num_hidden_layers": 61,
   "num_key_value_heads": 128,
   "num_nextn_predict_layers": 1,
+  "pretraining_tp": 1,
   "q_lora_rank": 1536,
   "qk_nope_head_dim": 128,
   "qk_rope_head_dim": 64,
@@ -56,6 +58,7 @@
   "rope_theta": 10000,
   "routed_scaling_factor": 2.5,
   "scoring_func": "sigmoid",
+  "seq_aux": true,
   "tie_word_embeddings": false,
   "topk_group": 4,
   "topk_method": "noaux_tc",

configuration_deepseek.py

@@ -82,6 +82,11 @@ class DeepseekV3Config(PretrainedConfig):
             Beginning of stream token id.
         eos_token_id (`int`, *optional*, defaults to 2):
             End of stream token id.
+        pretraining_tp (`int`, *optional*, defaults to 1):
+            Experimental feature. Tensor parallelism rank used during pretraining. Please refer to [this
+            document](https://huggingface.co/docs/transformers/parallelism) to understand more about it. This value is
+            necessary to ensure exact reproducibility of the pretraining results. Please refer to [this
+            issue](https://github.com/pytorch/pytorch/issues/76232).
         tie_word_embeddings (`bool`, *optional*, defaults to `False`):
             Whether to tie weight embeddings
         rope_theta (`float`, *optional*, defaults to 10000.0):
@@ -136,6 +141,8 @@ class DeepseekV3Config(PretrainedConfig):
         first_k_dense_replace = 3,
         norm_topk_prob = True,
         scoring_func = 'sigmoid',
+        aux_loss_alpha = 0.001,
+        seq_aux = True,
         hidden_act="silu",
         max_position_embeddings=4096,
         initializer_range=0.02,
@@ -144,6 +151,7 @@ class DeepseekV3Config(PretrainedConfig):
         pad_token_id=None,
         bos_token_id=0,
         eos_token_id=1,
+        pretraining_tp=1,
         tie_word_embeddings=False,
         rope_theta=10000.0,
         rope_scaling=None,
@@ -176,6 +184,8 @@ class DeepseekV3Config(PretrainedConfig):
         self.first_k_dense_replace = first_k_dense_replace
         self.norm_topk_prob = norm_topk_prob
         self.scoring_func = scoring_func
+        self.aux_loss_alpha = aux_loss_alpha
+        self.seq_aux = seq_aux
         # for backward compatibility
         if num_key_value_heads is None:
             num_key_value_heads = num_attention_heads
@@ -184,6 +194,7 @@ class DeepseekV3Config(PretrainedConfig):
         self.hidden_act = hidden_act
         self.initializer_range = initializer_range
         self.rms_norm_eps = rms_norm_eps
+        self.pretraining_tp = pretraining_tp
         self.use_cache = use_cache
         self.rope_theta = rope_theta
         self.rope_scaling = rope_scaling

modeling_deepseek.py

@@ -398,6 +398,7 @@ class MoEGate(nn.Module):
         self.n_routed_experts = config.n_routed_experts
         self.routed_scaling_factor = config.routed_scaling_factor
         self.scoring_func = config.scoring_func
+        self.seq_aux = config.seq_aux
         self.topk_method = config.topk_method
         self.n_group = config.n_group
         self.topk_group = config.topk_group
@@ -454,7 +455,7 @@ class MoEGate(nn.Module):
                 )
                 .reshape(bsz * seq_len, -1)
             )  # [n, e]
-            tmp_scores = scores_for_choice.masked_fill(~score_mask.bool(), float("-inf"))  # [n, e]
+            tmp_scores = scores_for_choice.masked_fill(~score_mask.bool(), 0.0)  # [n, e]
             _, topk_idx = torch.topk(
                 tmp_scores, k=self.top_k, dim=-1, sorted=False
             )