psaiops 0.4.7__py3-none-any.whl

psaiops/combine/app.py

@@ -0,0 +1,366 @@
+import functools
+import itertools
+
+import gradio
+import pandas
+import torch
+import torch.cuda
+
+import psaiops.common.tokenizer
+import psaiops.common.data
+
+# META #########################################################################
+
+MODEL = 'openai/gpt-oss-20b'
+
+STYLE = '''.giga-text input { font-size: 32px; }'''
+TITLE = '''Combine Datasets'''
+INTRO = '''Combine and wrap prompts to form new datasets.'''
+
+COUNT = 8
+
+# TEMPLATE #####################################################################
+
+ROLES = ['system', 'developer', 'user', 'assistant', 'tool']
+CHANNELS = ['analysis', 'commentary', 'final']
+
+# INTRO ########################################################################
+
+def create_intro_block(intro: str) -> dict:
+    __intro = gradio.Markdown(intro, line_breaks=True)
+    return {'intro_block': __intro}
+
+# MODEL ########################################################################
+
+def create_template_block() -> dict:
+    __template = gradio.Dropdown(label='Template', value='openai/gpt-oss-20b', choices=['openai/gpt-oss-20b', ''], scale=1, allow_custom_value=False, multiselect=False, interactive=True)
+    return {
+        'template_block': __template,}
+
+# SAMPLING #####################################################################
+
+def create_huggingface_block() -> dict:
+    __token = gradio.Textbox(label='Token', value='', placeholder='Hugging Face authentication token.', lines=1, max_lines=1, scale=4, show_label=True, show_copy_button=False, interactive=True)
+    return {
+        'token_block': __token,}
+
+# SAMPLING #####################################################################
+
+def create_source_block() -> dict:
+    __search = gradio.Dropdown(label='Search', value='', choices=[''], scale=7, allow_custom_value=True, multiselect=False, interactive=True)
+    __append = gradio.Button(value='>', variant='primary', size='lg', scale=1, interactive=True)
+    __dataset = gradio.Dropdown(label='Datasets', value='', choices=[''], scale=8, allow_custom_value=False, multiselect=True, interactive=True)
+    return {
+        'search_block': __search,
+        'append_block': __append,
+        'sources_block': __dataset,}
+
+# SAMPLING #####################################################################
+
+def create_download_block() -> dict:
+    __download = gradio.Button(value='Download', variant='primary', size='lg', scale=1, interactive=True)
+    return {
+        'download_block': __download,}
+
+# ACTIONS ######################################################################
+
+def create_meta_block() -> dict:
+    __name = gradio.Textbox(label='Path', value='', placeholder='Dataset ID: user/name.', lines=1, max_lines=1, scale=1, show_label=True, show_copy_button=False, interactive=True)
+    __col0 = gradio.Textbox(label='Column 0', value='', placeholder='Name of the column 0.', lines=1, max_lines=1, scale=1, show_label=True, show_copy_button=False, interactive=True)
+    __col1 = gradio.Textbox(label='Column 1', value='', placeholder='Name of the column 1.', lines=1, max_lines=1, scale=1, show_label=True, show_copy_button=False, interactive=True)
+    return {
+        'name_block': __name,
+        'column_0_block': __col0,
+        'column_1_block': __col1,}
+
+# INPUTS #######################################################################
+
+def create_inputs_row(index: int=0) -> dict:
+    with gradio.Row(equal_height=True, visible=(index == 0)) as __row:
+        __role = gradio.Dropdown(
+            type='value',
+            label=f'Role',
+            value='user',
+            choices=[__r for __r in ROLES],
+            # elem_classes='giga-text',
+            scale=1,
+            show_label=(index == 0),
+            allow_custom_value=False,
+            multiselect=False,
+            interactive=True,
+            visible=(index == 0))
+        __channel = gradio.Dropdown(
+            type='value',
+            label=f'Channel',
+            value='final',
+            choices=[__c for __c in CHANNELS],
+            # elem_classes='giga-text',
+            scale=1,
+            show_label=(index == 0),
+            allow_custom_value=False,
+            multiselect=False,
+            interactive=True,
+            visible=(index == 0))
+        __source = gradio.Dropdown(
+            type='value',
+            label=f'Source',
+            value='mnaual',
+            choices=['manual'],
+            # elem_classes='giga-text',
+            scale=4,
+            show_label=(index == 0),
+            allow_custom_value=False,
+            multiselect=False,
+            interactive=True,
+            visible=(index == 0))
+        __content = gradio.Textbox(
+            label=f'Prompt',
+            value='',
+            placeholder='Some text.',
+            lines=1,
+            max_lines=1,
+            scale=9,
+            show_label=(index == 0),
+            show_copy_button=True,
+            interactive=True,
+            visible=(index == 0))
+        __hide = gradio.Button(
+            value='X',
+            variant='secondary',
+            size='lg',
+            scale=1,
+            interactive=True,
+            visible=(index == 0))
+    return {
+        f'row_{index}_block': __row,
+        f'role_{index}_block': __role,
+        f'channel_{index}_block': __channel,
+        f'source_{index}_block': __source,
+        f'content_{index}_block': __content,
+        f'button_{index}_block': __hide,}
+
+# OUTPUTS ######################################################################
+
+def create_outputs_block() -> dict:
+    __output = gradio.Textbox(label='Sample', value='', placeholder='Resulting combination of the prompts.', lines=2, max_lines=8, scale=1, show_label=True, show_copy_button=True, interactive=False)
+    return {'output_block': __output,}
+
+# ACTIONS ######################################################################
+
+def create_action_block() -> dict:
+    __show = gradio.Button(value='Add', variant='primary', size='lg', scale=1, interactive=True)
+    __upload = gradio.Button(value='Upload', variant='primary', size='lg', scale=1, interactive=True)
+    return {
+        'show_block': __show,
+        'upload_block': __upload,}
+
+# TABLE ########################################################################
+
+def create_table_block() -> dict:
+    __table = gradio.DataFrame(label='Table', type='numpy', headers=None,  row_count=4, col_count=256, scale=1, interactive=False)
+    return {'table_block': __table,}
+
+# STATE ########################################################################
+
+def default_state(visible: bool=False) -> dict:
+    return {'visible': visible, 'role': 'user', 'channel': 'final', 'source': 'manual', 'content': ''}
+
+def create_state(limit: int=COUNT) -> dict:
+    return {
+        'cache_block': gradio.State(
+            [default_state(True)] + [default_state(False) for _ in range(limit - 1)])}
+
+# LAYOUT #######################################################################
+
+def create_layout(intro: str=INTRO, limit: int=COUNT) -> dict:
+    __fields = {}
+    __fields.update(create_intro_block(intro=intro))
+    with gradio.Row(equal_height=True):
+        __fields.update(create_meta_block())
+    with gradio.Tabs():
+        with gradio.Tab('Column 0') as __col0_tab:
+            __fields.update({'column_0_tab': __col0_tab})
+            for __i in range(limit):
+                __fields.update(create_inputs_row(index=__i))
+            with gradio.Row(equal_height=True):
+                __fields.update(create_outputs_block())
+            with gradio.Row(equal_height=True):
+                __fields.update(create_action_block())
+        with gradio.Tab('Details') as __details_tab:
+            __fields.update({'details_tab': __details_tab})
+            with gradio.Row(equal_height=True):
+                __fields.update(create_table_block())
+        with gradio.Tab('Settings') as __settings_tab:
+            __fields.update({'settings_tab': __settings_tab})
+            with gradio.Row(equal_height=True):
+                __fields.update(create_template_block())
+            with gradio.Row(equal_height=True):
+                __fields.update(create_huggingface_block())
+            with gradio.Row(equal_height=True):
+                __fields.update(create_source_block())
+            with gradio.Row(equal_height=True):
+                __fields.update(create_download_block())
+    return __fields
+
+# DYNAMIC ######################################################################
+
+def get_input_rows(inputs: dict, limit: int=COUNT) -> list:
+    return list(itertools.chain.from_iterable([
+        [
+            inputs.get(f'row_{__i}_block', None),
+            inputs.get(f'role_{__i}_block', None),
+            inputs.get(f'channel_{__i}_block', None),
+            inputs.get(f'source_{__i}_block', None),
+            inputs.get(f'content_{__i}_block', None),
+            inputs.get(f'button_{__i}_block', None),]
+        for __i in range(limit)]))
+
+def render_input_rows(rows: list) -> list:
+    return list(itertools.chain.from_iterable([
+        [
+            gradio.update(visible=__r.get('visible', False)),
+            gradio.update(visible=__r.get('visible', False), value=__r.get('role', 'user')),
+            gradio.update(visible=__r.get('visible', False), value=__r.get('channel', 'final')),
+            gradio.update(visible=__r.get('visible', False), value=__r.get('source', 'manual')),
+            gradio.update(visible=__r.get('visible', False), value=__r.get('content', '')),
+            gradio.update(visible=__r.get('visible', False))]
+        for __r in rows]))
+
+def show_input_row(rows: list) -> tuple:
+    __count = 0
+    __rows = list(rows)
+    for __i in range(len(__rows)):
+        # count the number of hidden rows (before changing their state)
+        __count = __count + int(not __rows[__i]['visible'])
+        # all the visible rows stay the same and the first hidden row is toggled
+        __rows[__i]['visible'] = __rows[__i]['visible'] or (__count < 2)
+    # update state and components
+    return __rows, *render_input_rows(__rows)
+
+def hide_input_row(rows: list, index: int) -> tuple:
+    __rows = list(rows)
+    # always show the first row
+    if 0 < index < len(__rows):
+        # remove the target row
+        __rows.pop(index)
+        # keep the number of rows constant
+        __rows.append(default_state(False))
+    # update state and components
+    return __rows, *render_input_rows(__rows)
+
+# EVENTS #######################################################################
+
+def update_input_cache(cache: list, index: int, value: any, field: str) -> list:
+    __cache = list(cache)
+    __cache[index][field] = value
+    return __cache
+
+def update_role_cache(cache: list, index: int, value: any) -> list:
+    return update_input_cache(cache=cache, index=int(index), value=str(value), field='role')
+
+def update_channel_cache(cache: list, index: int, value: any) -> list:
+    return update_input_cache(cache=cache, index=int(index), value=str(value), field='channel')
+
+def update_source_cache(cache: list, index: int, value: any) -> list:
+    return update_input_cache(cache=cache, index=int(index), value=str(value), field='source')
+
+def update_content_cache(cache: list, index: int, value: any) -> list:
+    return update_input_cache(cache=cache, index=int(index), value=str(value), field='content')
+
+def update_table_data(tokenizer: object) -> callable:
+    # called with unpacked arguments
+    def __update_table_data(*prompts: list) -> list:
+        # array of token IDs
+        __outputs = tokenizer(prompts, return_tensors='pt', padding=True)
+        # array of token strings
+        __tokens = [tokenizer.convert_ids_to_tokens(__s) for __s in __outputs['input_ids']]
+        # shift the special characters
+        return [[__t.replace(chr(0x0120), ' ').replace(chr(0x010a), '\\n') for __t in __s] for __s in __tokens]
+    # fixed to a given tokenizer
+    return __update_table_data
+
+def update_dataset_list(data: str) -> dict:
+    __datasets = []
+    if len(data) > 3:
+        __datasets = psaiops.common.data.query_huggingface(target=data, label='dataset', limit=8)
+    return gradio.update(choices=__datasets, visible=True)
+
+# APP ##########################################################################
+
+def create_app(title: str=TITLE, intro: str=INTRO, style: str=STYLE, limit: int=COUNT, model: str=MODEL) -> gradio.Blocks:
+    __inputs = {}
+    with gradio.Blocks(theme=gradio.themes.Soft(), title=title, css=style) as __app:
+        # load the tokenizer
+        __tokenizer = psaiops.common.tokenizer.get_tokenizer(name=model, device='cpu')
+        # create the UI
+        __inputs.update(create_layout(intro=intro, limit=limit))
+        # init the state
+        __inputs.update(create_state(limit=limit))
+        # apply the configuration
+        __format = update_table_data(tokenizer=__tokenizer)
+        # show hidden row
+        __inputs['show_block'].click(
+            fn=show_input_row,
+            inputs=[__inputs['cache_block']],
+            outputs=[__inputs['cache_block']] + get_input_rows(inputs=__inputs, limit=limit),
+            queue=False,
+            show_progress='hidden')
+        # update the table TODO
+        __inputs['details_tab'].select(
+            fn=__format,
+            inputs=[__inputs[f'content_{__i}_block'] for __i in range(limit)] + [__inputs['output_block']],
+            outputs=__inputs['table_block'],
+            queue=False,
+            show_progress='hidden')
+        # fetch the list of matching datasets
+        __inputs['search_block'].change(
+            fn=update_dataset_list,
+            inputs=__inputs['search_block'],
+            outputs=__inputs['search_block'],
+            queue=False,
+            show_progress='hidden')
+        # link each row of inputs to the cache
+        for __i in range(limit):
+            # update the target role in the cache
+            __inputs[f'role_{__i}_block'].change(
+                fn=update_role_cache,
+                inputs=[__inputs['cache_block'], gradio.State(__i), __inputs[f'role_{__i}_block']],
+                outputs=__inputs['cache_block'],
+                queue=False,
+                show_progress='hidden')
+            # update the target channel in the cache
+            __inputs[f'channel_{__i}_block'].change(
+                fn=update_channel_cache,
+                inputs=[__inputs['cache_block'], gradio.State(__i), __inputs[f'channel_{__i}_block']],
+                outputs=__inputs['cache_block'],
+                queue=False,
+                show_progress='hidden')
+            # update the target column in the cache
+            __inputs[f'source_{__i}_block'].change(
+                fn=update_source_cache,
+                inputs=[__inputs['cache_block'], gradio.State(__i), __inputs[f'source_{__i}_block']],
+                outputs=__inputs['cache_block'],
+                queue=False,
+                show_progress='hidden')
+            # update the target content in the cache
+            __inputs[f'content_{__i}_block'].change(
+                fn=update_content_cache,
+                inputs=[__inputs['cache_block'], gradio.State(__i), __inputs[f'content_{__i}_block']],
+                outputs=__inputs['cache_block'],
+                queue=False,
+                show_progress='hidden')
+            # hide the target row
+            __inputs[f'button_{__i}_block'].click(
+                fn=hide_input_row,
+                inputs=[__inputs['cache_block'], gradio.State(__i)],
+                outputs=[__inputs['cache_block']] + get_input_rows(inputs=__inputs, limit=limit),
+                queue=False,
+                show_progress='hidden')
+        # gradio application
+        return __app
+
+# MAIN #########################################################################
+
+if __name__ == '__main__':
+    __app = create_app()
+    __app.launch(share=True, debug=True)

psaiops/common/data.py

@@ -0,0 +1,31 @@
+import requests
+
+# CONSTANTS ####################################################################
+
+HF_URL = 'https://huggingface.co/api/quicksearch?q={target}&type={label}&limit={limit}'
+
+# HUGGING FACE #################################################################
+
+def query_huggingface(target: str, label: str='model', limit: int=16, endpoint: str=HF_URL) -> list:
+    __results = []
+    # make sure the label has no trailing "s"
+    __label = label.rstrip('s').strip(' ')
+    # the HTTP request or the parsing may fail
+    try:
+        # query HF
+        __response = requests.get(endpoint.format(target=target, label=__label, limit=limit))
+        # filter by type ('models' / 'datasets' / 'spaces')
+        __results = [__d.get('id', '') for __d in __response.json().get(f'{__label}s', [])]
+    # ignore all the errors
+    except:
+        __results = []
+    # list of strings
+    return __results
+
+# EXAMPLES #####################################################################
+
+def update_dropdown(target: str, label: str) -> dict:
+    # query huggingface
+    __data = psaiops.elements.data.query_huggingface(target=target, label=label, limit=16)
+    # list choices in the dropdown
+    return gradio.update(choices=__data, visible=True)

psaiops/common/model.py

@@ -0,0 +1,73 @@
+import functools
+
+import torch
+import torch.nn
+
+import deformers.models.openai.gptoss
+
+# LOAD #########################################################################
+
+@functools.lru_cache(maxsize=1)
+def get_model(name: str, device: str='cpu'):
+    __model = deformers.models.openai.gptoss.GptOssForCausalInference.from_pretrained(
+        name,
+        dtype='auto',
+        device_map=device)
+    # toggle the inference mode (not training)
+    __model.eval()
+    # transformers model
+    return __model
+
+# PREFIX #######################################################################
+
+def get_prefix(
+    parent_obj: object,
+    layer_num: int
+) -> object:
+    # init from the config
+    __child = parent_obj.__class__(parent_obj.config)
+    # share decoder core
+    __child.model.embed_tokens = parent_obj.model.embed_tokens
+    __child.model.norm = parent_obj.model.norm
+    # share prefix layers (same objects)
+    __child.model.layers = torch.nn.ModuleList(parent_obj.model.layers[:layer_num])
+    # keep LM head
+    __child.lm_head = parent_obj.lm_head
+    # config hygiene
+    __child.model.config.num_hiddelayer_num = layer_num
+    __child.config.num_hiddelayer_num = layer_num
+    # layer types
+    if getattr(__child.config, "layer_types", None) is not None:
+        __child.config.layer_types = __child.config.layer_types[:layer_num]
+        __child.model.config.layer_types = __child.config.layer_types
+    # wrapper with the first N hidden layers, pointing at the parent weights
+    return __child
+
+# GENERATE #####################################################################
+
+@functools.lru_cache(maxsize=32)
+def generate_token_ids(
+    model_obj: object,
+    input_ids: torch.Tensor,
+    token_num: int,
+    topk_num: int = 4,
+    topp_num: float = 0.9,
+    attention_mask: torch.Tensor=None,
+) -> torch.Tensor:
+    # generate completion
+    with torch.no_grad():
+        __outputs = model_obj.generate(
+            input_ids=input_ids,
+            attention_mask=attention_mask,
+            max_new_tokens=token_num,
+            do_sample=(0.0 < topp_num < 1.0) or (topk_num > 0),
+            top_k=topk_num if (topk_num > 0) else None,
+            top_p=topp_num if (0.0 < topp_num < 1.0) else None,
+            return_dict_in_generate=True,
+            output_hidden_states=False,
+            output_attentions=False,
+            output_scores=False,
+            # early_stopping=True,
+            use_cache=True)
+    # full sequence
+    return __outputs.sequences # (1, T)

psaiops/common/tokenizer.py

@@ -0,0 +1,41 @@
+import functools
+
+import torch
+import transformers
+
+# LOAD #########################################################################
+
+@functools.lru_cache(maxsize=4)
+def get_tokenizer(name: str, device: str='cpu'):
+    return transformers.AutoTokenizer.from_pretrained(
+        name,
+        use_fast=True,
+        dtype='auto',
+        device_map=device)
+
+# PREPROCESS #####################################################################
+
+@functools.lru_cache(maxsize=32)
+def preprocess_token_ids(
+    tokenizer_obj: object,
+    prompt_str: str,
+    device_str: str='cpu'
+) -> dict:
+    # tokenize
+    __inputs = tokenizer_obj(prompt_str, return_tensors='pt')
+    # move to the main device
+    return {__k: __v.to(device_str) for __k, __v in __inputs.items()}
+
+# POSTPROCESS ####################################################################
+
+@functools.lru_cache(maxsize=32)
+def postprocess_token_ids(
+    tokenizer_obj: object,
+    token_data: torch.Tensor,
+) -> list:
+    # remove the batch axis
+    __indices = token_data.squeeze().tolist()
+    # back to token strings
+    __tokens = tokenizer_obj.convert_ids_to_tokens(__indices)
+    # normalize the tokens
+    return [__t.replace(chr(0x0120), ' ').replace(chr(0x010a), '\n') for __t in __tokens]