o
    'Zh*                      @   s   d Z ddlZddlmZ ddlZddlZG dd dZ	ddeee	f ded	ee
 d
ee	 fddZ	dde
de	de
d	ee
 d
eee	f f
ddZdee d
dfddZdd ZdS )zJThis is an educational implementation of the byte pair encoding algorithm.    N)Optionalc                   @   s   e Zd Zdedeeef ddfddZdded	ee de	e fd
dZ
de	e defddZde	e defddZde	e de	e fddZedededefddZedd ZdS )SimpleBytePairEncodingpat_strmergeable_ranksreturnNc                C   s0   || _ || _dd | D | _t|| _dS )zCreates an Encoding object.c                 S   s   i | ]\}}||qS  r   ).0token_bytestokenr   r   L/var/www/html/lang_env/lib/python3.10/site-packages/tiktoken/_educational.py
<dictcomp>       z3SimpleBytePairEncoding.__init__.<locals>.<dictcomp>N)r   r   items_decoderregexcompile_pat)selfr   r   r   r   r   __init__   s   zSimpleBytePairEncoding.__init__colourtext	visualisec                 C   sB   | j |}g }|D ]}|d}t| j||d}|| q
|S )z`Encodes a string into tokens.

        >>> enc.encode("hello world")
        [388, 372]
        utf-8)r   )r   findallencode
bpe_encoder   extend)r   r   r   wordstokenswordZ
word_bytesZword_tokensr   r   r   r      s   
zSimpleBytePairEncoding.encoder   c                    s   d  fdd|D S )znDecodes a list of tokens into bytes.

        >>> enc.decode_bytes([388, 372])
        b'hello world'
            c                 3   s    | ]} j | V  qd S Nr   r   r
   r   r   r   	<genexpr>+   s    z6SimpleBytePairEncoding.decode_bytes.<locals>.<genexpr>)joinr   r   r   r$   r   decode_bytes%   s   z#SimpleBytePairEncoding.decode_bytesc                 C   s   |  |jdddS )u   Decodes a list of tokens into a string.

        Decoded bytes are not guaranteed to be valid UTF-8. In that case, we replace
        the invalid bytes with the replacement character "�".

        >>> enc.decode([388, 372])
        'hello world'
        r   replaceerrors)r(   decoder'   r   r   r   r,   -   s   	zSimpleBytePairEncoding.decodec                    s    fdd|D S )zDecodes a list of tokens into a list of bytes.

        Useful for visualising how a string is tokenised.

        >>> enc.decode_tokens_bytes([388, 372])
        [b'hello', b' world']
        c                    s   g | ]} j | qS r   r"   r#   r$   r   r   
<listcomp>@   r   z>SimpleBytePairEncoding.decode_tokens_bytes.<locals>.<listcomp>r   r'   r   r$   r   decode_tokens_bytes8   s   z*SimpleBytePairEncoding.decode_tokens_bytestraining_data
vocab_sizec                 C   s   t | ||d}t||dS )z#Train a BPE tokeniser on some data!)datar0   r   r   r   )	bpe_trainr   )r/   r0   r   r   r   r   r   trainB   s   zSimpleBytePairEncoding.trainc                 C   s$   t | tr
t| } t| j| jdS )Nr2   )
isinstancestrtiktokenZget_encodingr   Z_pat_strZ_mergeable_ranks)encodingr   r   r   from_tiktokenH   s
   

z$SimpleBytePairEncoding.from_tiktokenr   )__name__
__module____qualname__r6   dictbytesintr   r   listr   r(   r,   r.   staticmethodr4   r9   r   r   r   r   r   
   s     

r   r   r   inputr   r   c           
         s  dd |D }	 |r|dv rt | n|dkrt| d }d }tt|d d |dd  D ]\}} |d |d  }|d urM|d u sI||k rM|}|}q.|d u rSn |d usYJ |d | || ||d   g ||d	 d   }q|rxt   fd
d|D }	|	S )Nc                 S      g | ]}t |gqS r   r?   r   br   r   r   r-   T   r   zbpe_encode.<locals>.<listcomp>Tr   colorsimple   r      c                    s   g | ]} | qS r   r   )r   partr   r   r   r-   q   s    )visualise_tokensprint	enumeratezipget)
r   rC   r   partsmin_idxZmin_rankipairZrankr   r   rO   r   r   Q   s0   
&2r   r1   r0   r   c                    s  |dk rt di }tdD ]	}||t|g< qdd t|| D }t||k rt  |D ]}t|d d |dd  D ]
} |  d7  < q?q0t	  fddd	}	|	d
 |	d  }
t|}|||
< g }|D ]K}g }d
}|t|d k r|| ||d  f|	kr|
|
 |d7 }n|
||  |d7 }|t|d k sw|t|d kr|
||  |
| qi|}|rtd|	d
  d|	d   td|
 dt| d |dv rtd tdd |d d D  n|dkrtd |d d D ]}t| qtd t||k s*|S )N   z;vocab_size must be at least 256, so we can encode all bytesc                 S   s    g | ]}d d | dD qS )c                 S   rD   r   rE   rF   r   r   r   r-      r   z(bpe_train.<locals>.<listcomp>.<listcomp>r   )r   )r   r   r   r   r   r-      s    zbpe_train.<locals>.<listcomp>rK   rL   c                    s    |  S r!   r   )xstatsr   r   <lambda>   s    zbpe_train.<locals>.<lambda>)keyr   rM   z The current most common pair is z + zSo we made z our zth tokenrH   z9Now the first fifty words in our training data look like:c                 S   s   g | ]	}|D ]}|qqS r   r   )r   r   r
   r   r   r   r-          2   rJ   z:Now the first twenty words in our training data look like:   
)
ValueErrorranger?   r   r   lencollectionsCounterrS   maxappendrQ   rP   )r1   r0   r   r   ZranksrW   r   ZpiecerX   Zmost_common_pairr	   r
   Z	new_wordsr   Znew_wordr   r[   r   r3   u   s\   




-r3   token_valuesc                 C   s   dd dD }dd | D }d}d }|D ].}||t |  }||kr2||d t |  }||ks2J |}|t |7 }t|| dd qtd	 d S )
Nc                 S   s   g | ]}d | dqS )z[48;5;mr   )r   rW   r   r   r   r-      s    z$visualise_tokens.<locals>.<listcomp>)         M   P   D      c                 S   s   g | ]	}|j d ddqS )r   r)   r*   )r,   )r   rZ   r   r   r   r-      r_   r   rL    )endz[0m)re   rQ   )rj   
backgroundZunicode_token_valuesZrunning_lengthZ
last_colorr
   rI   r   r   r   rP      s   rP   c                  C   s   d} t td}| }W d    n1 sw   Y  tj|d| d}td |d}||dks5J ||dks>J |	|dd	gksIJ |S )
NzN's|'t|'re|'ve|'m|'ll|'d| ?[\p{L}]+| ?[\p{N}]+| ?[^\s\p{L}\p{N}]+|\s+(?!\S)|\s+riX  )r0   r   zJThis is the sequence of merges performed in order to encode 'hello world':zhello worlds   hello worlds   hellos    world)
open__file__readr   r4   rQ   r   r,   r(   r.   )Zgpt2_patternfr1   encr   r   r   r   train_simple_encoding   s   

r|   r:   )__doc__rf   typingr   r   r7   r   r>   r?   r@   r6   rA   r   r3   rP   r|   r   r   r   r   <module>   s<    H

%

E