a h'@svdZddlmZddlZddlmZddlmZGdd d ej j j Z Gd d d ej j j Z Gd d d ej j j ZdS)a Generic interface to various configurations of the Perceiver Resampler, that simply takes in a series of (potentially time-indexed) contextual embeddings, and "resamples" (compresses) them down to a pre-specified number of latents! Note that the Perceiver in general resamples based solely off the *long-range* context; there's a nice opportunity here to prime the Perceiver Resampler with say a single layer's worth of language embeddings (the target domain), and use that to softly "retrieve & compress" what we need --> this would be a novel contribution we should explore. References: - DeepMind's Flamingo: https://www.deepmind.com/blog/tackling-multiple-tasks-with-a-single-visual-language-model - Code borrowed w/ love from: https://github.com/lucidrains/flamingo-pytorch )OptionalN) shape_list) IdeficsConfigcsNeZdZeeeeeeddfdd ZfddZejejddd Z Z S) TFIdeficsPerceiverResamplerN)config embed_dimdepthn_headshead_dim n_latentsreturnc stjfi|||||f\|_|_|_|_|jj|_t |j dsN|jdn |j jd|_ g|_ t |D]F}|j t|j|j|j|jd|ddt|j |d|ddgqjtjjjddd |_d S) ao Instantiates a Perceiver Resampler that operates over a sequence of embeddings (say from a ResNet or ViT or MAE) of a given dimension, performs `depth` blocks of cross-attention with a fixed `n_latents` inputs, then returns a Tensor of shape [bsz, n_latents, embed_dim]. :param embed_dim: Dimensionality of embeddings being fed to the Perceiver Resampler (also dimensionality of latent embeddings *returned* by the Perceiver Resampler. Could be e.g., VIT embed_dim, ResNet pool dim, and so on. Args: config (`IdeficsConfig`): config object embed_dim (`int`): The size of each embedding vector depth (`int`): Depth of the Perceiver Resampler (Transformer w/ cross attention). Should be shallow (< 3). n_heads (`int`): Number of heads in each Transformer block (for multi-headed self-attention). head_dim (`int`): Dimensionality of each head projection in the Transformer block. n_latents (`int`): Number of latent embeddings to resample ("compress") the input sequence to (usually < 128). r zblocks.z.0namez.1h㈵> layer_normepsilonrN)super__init__r r r r Zperceiver_configZqk_layer_norms_perceiverqk_layer_normshasattr vision_configZintermediate_dimblocksrangeappendTFIdeficsPerceiverAttention TFIdeficsMLPtfkeraslayersLayerNormalizationr) selfrr r r r r kwargsi __class__d/var/www/html/assistant/venv/lib/python3.9/site-packages/transformers/models/idefics/perceiver_tf.pyr1s"     z$TFIdeficsPerceiverResampler.__init__cs,|j|j|jfdddd|_t|dS)NZ random_normalTlatents)shapeZ initializerZ trainabler)Z add_weightr r r+rbuild)r$Z input_shaper'r)r*r-\sz!TFIdeficsPerceiverResampler.build)contextrcCs`tj|jdd}t|t|dddg}|jD]"\}}||||}|||}q2||S)zWResample arbitrary length context & *compress* down to self.n_latents latent embeddingsraxisr)r Z expand_dimsr+Ztiler,rr)r$r.r+attnffr)r)r*callcs z TFIdeficsPerceiverResampler.call) __name__ __module__ __qualname__rintrr-r Tensorr3 __classcell__r)r)r'r*r0s  + rcsBeZdZeeeeddfdd ZejejejdddZZ S)rN)r r r rrc stjfi|||||_|_|_||_tjjj ddd|_ tjjj ddd|_ |jrtjjj ddd|_ tjjj ddd|_ |jd|_tjjj|j|jdd d |_tjjj|j|jdd d |_tjjj|j|jdd d |_tjjj|dd d |_dS)ziPerceiver Cross-Attention Module --> let long-form inputs be `context`, resampled embeddings be `latents`rcontext_layer_normrlatents_layer_norm q_layer_norm k_layer_normgFq_projZuse_biasrk_projv_proj output_projN)rrr r r rr r!r"r#r:r;r<r=qk_scaleDenser>r@rArB)r$r r r rr%r'r)r*rps z$TFIdeficsPerceiverAttention.__init__)r.r+rc s |}|}t|\}}|}tj||gdd}tj||gdd}fdd|||fD\}}}jr |} |}t d|j |}|tj |ddd} tjj| dd} t d | |} ttj| gd d djjfS) a= Runs Perceiver Self-Attention, with special (context, latents) appended along the `seq` dimension! Args: context (`tf.Tensor`): Tensor of shape `[bsz, seq, embed_dim]` representing long-form context to resample. latents (`tf.Tensor`): Tensor of shape `[bsz, n_latents, embed_dim]` representing fixed length latents to compress to. Returns: `tf.Tensor`: Tensor of shape `[bsz, n_latents, embed_dim]` representing attention over latents w/ cross from context. r/c s8g|]0}tjt||jdjjfgddqS)rrrrperm)r transposereshaper,r r ).0xZ batch_sizer$r)r* sz4TFIdeficsPerceiverAttention.call..z... i d, ... j d -> ... i jT)r0Zkeepdimsz... i j, ... j d -> ... i drFrH)r:r;rr>r@r concatrArr<r=ZeinsumrCZ reduce_maxnnZsoftmaxrBrKrJr r ) r$r.r+Z seq_lengthr qkvZscoresZstabilized_scoresr1Z resampledr)rNr*r3s&       &z TFIdeficsPerceiverAttention.call) r4r5r6r7boolrr r8r3r9r)r)r'r*rosrcs>eZdZedfdd ZeeejejdddZ Z S)r)rc sttjfi||jj|_tjjjddd|_tjjj |ddd|_ tjjj dd|_ tjjj |jdd d|_ d S) z:Simple MLP block with intermediate_size and embedding sizerlnrFfcr?actrc_projN)rrrr r r!r"r#rWrDrXZReLUrYrZ)r$Zintermediate_sizerr%r'r)r*rs  zTFIdeficsMLP.__init__) hidden_statesrcCs,||}||}||}||}|S)N)rWrXrYrZ)r$r[r)r)r*r3s     zTFIdeficsMLP.call) r4r5r6rrrtupler r8r3r9r)r)r'r*rs r)__doc__typingrZ tensorflowr Zmodeling_tf_utilsrZconfiguration_ideficsrr!r"ZLayerrrrr)r)r)r*s   ?D