o wi@s,ddlmZmZddlZddlmZddlmZddl m Z ddl m Z ddl mZdd lmZmZd d lmZmZmZmZd d lmZd dlmZd dlmZeeZdZ GdddeZ!Gddde Z"GdddeZ#GdddeeZ$GdddeZ%GdddeZ&GdddeZ'gdZ(dS) )OptionalUnionN)Cache)FlashAttentionKwargs)GradientCheckpointingLayer)CausalLMOutputWithPast)Unpack) LossKwargslogging) GlmAttentionGlmForCausalLMGlmForSequenceClassificationGlmForTokenClassification)Phi3MLP) Glm4Config) Glm4RMSNormzTHUDM/GLM-4-9B-0414c@ eZdZdS)Glm4MLPN__name__ __module__ __qualname__rrb/home/ubuntu/sommelier/.venv/lib/python3.10/site-packages/transformers/models/glm4/modular_glm4.pyr%rcseZdZdedeffdd Z       ddejdeejd eej d ee d ee d ee d eej dee ejejfde ede ejee ejejfffddZZS)Glm4DecoderLayerconfig layer_idxcsvt|j|_t||d|_t||_t|j|jd|_ t|j|jd|_ t|j|jd|_ t|j|jd|_ dS)N)rr )eps) super__init__ hidden_size Glm4Attention self_attnrmlpr rms_norm_epsinput_layernormpost_attention_layernormpost_self_attn_layernormpost_mlp_layernorm)selfrr  __class__rrr#*s  zGlm4DecoderLayer.__init__NF hidden_statesattention_mask position_idspast_key_valueoutput_attentions use_cachecache_positionposition_embeddingskwargsreturnc Ks|} ||}|jd||||||||d| \}} ||}| |}|} ||}||}||}| |}|f} |rB| | f7} | S)N)r0r1r2r3r4r5r6r7r)r)r&r+r*r'r,) r-r0r1r2r3r4r5r6r7r8residualself_attn_weightsoutputsrrrforward5s2      zGlm4DecoderLayer.forward)NNNFFNN)rrrrintr#torchTensorr LongTensorrbooltupler r FloatTensorr= __classcell__rrr.rr)s<   rc@r)r%Nrrrrrr%crr%c@r)KwargsForCausalLMNrrrrrrFgs rFcs2eZdZdeedeeefffdd ZZ S)Glm4ForCausalLM super_kwargsr9c stjdi|S)ah labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*): Labels for computing the masked language modeling loss. Indices should either be in `[0, ..., config.vocab_size]` or -100 (see `input_ids` docstring). Tokens with indices set to `-100` are ignored (masked), the loss is only computed for the tokens with labels in `[0, ..., config.vocab_size]`. Example: ```python >>> from transformers import AutoTokenizer, Glm4ForCausalLM >>> model = Glm4ForCausalLM.from_pretrained("THUDM/GLM-4-9B-0414") >>> tokenizer = AutoTokenizer.from_pretrained("THUDM/GLM-4-9B-0414") >>> prompt = "Hey, are you conscious? Can you talk to me?" >>> inputs = tokenizer(prompt, return_tensors="pt") >>> # Generate >>> generate_ids = model.generate(inputs.input_ids, max_length=30) >>> tokenizer.batch_decode(generate_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False)[0] "Hey, are you conscious? Can you talk to me?\nI'm not conscious, but I can talk to you." ```Nr)r"r=)r-rHr.rrr=kszGlm4ForCausalLM.forward) rrrr rFrrCrr=rErrr.rrGjs  rGc@r)Glm4ForSequenceClassificationNrrrrrrIrrIc@r)Glm4ForTokenClassificationNrrrrrrJrrJ)Glm4PreTrainedModel Glm4ModelrGrIrJ))typingrrtorch.utils.checkpointr? cache_utilsrmodeling_flash_attention_utilsrmodeling_layersrmodeling_outputsrprocessing_utilsr utilsr r glm.modeling_glmr rrrphi3.modeling_phi3rconfiguration_glm4r modeling_glm4r get_loggerrlogger_CHECKPOINT_FOR_DOCrrr%rFrGrIrJ__all__rrrrs,         :