#5383: Rename demos/falcon7b to demos/falcon7b_common

Signed-off-by: Salar Hosseini <[email protected]>

CODEOWNERS

@@ -140,7 +140,7 @@ models/demos @eyonland @arakhmati @cfjchu @xanderchin
 models/demos/metal_BERT_large_11 @tt-aho @TT-BrianLiu
 models/demos/wormhole @uaydonat @eyonland @AleksKnezevic @nsmithtt
 models/demos/t3000 @uaydonat @AleksKnezevic @nsmithtt
-models/demos/falcon7b @skhorasganiTT @djordje-tt @uaydonat @pavlejosipovic @pavlepopovic @s-jovic
+models/demos/falcon7b_common @skhorasganiTT @djordje-tt @uaydonat @pavlejosipovic @pavlepopovic @s-jovic
 models/demos/mamba @esmalTT @uaydonat @kpaigwar
 models/demos/wormhole/falcon7b @skhorasganiTT @djordje-tt @uaydonat @pavlejosipovic @pavlepopovic @s-jovic
 models/demos/wormhole/mistral7b @yieldthought @uaydonat @mtairum

docs/source/ttnn/ttnn/get_started.rst

@@ -22,7 +22,7 @@ Install and build the project by following the instructions in the
 ^^^^^^^^^^^^^^^^^^^^^^^^^^
 
 Get started with the Falcon 7B demo. Navigate to the `Falcon 7B demo folder
-<https://github.com/tenstorrent/tt-metal/tree/main/models/demos/falcon7b>`_
+<https://github.com/tenstorrent/tt-metal/tree/main/models/demos/wormhole/falcon7b>`_
 for details.
 
 You can also check our demos for

models/demos/falcon7b/demo/demo.py → models/demos/falcon7b_common/demo/demo.py

@@ -12,10 +12,10 @@
 import ttnn
 import tt_lib
 from loguru import logger
-from models.demos.falcon7b.reference.hf_modeling_falcon import FalconConfig
-from models.demos.falcon7b.tt.falcon_causallm import TtFalconCausalLM
-from models.demos.falcon7b.tt.model_config import get_model_config, model_config_entries
-from models.demos.falcon7b.tests.test_utils import initialize_kv_cache, load_hf_model
+from models.demos.falcon7b_common.reference.hf_modeling_falcon import FalconConfig
+from models.demos.falcon7b_common.tt.falcon_causallm import TtFalconCausalLM
+from models.demos.falcon7b_common.tt.model_config import get_model_config, model_config_entries
+from models.demos.falcon7b_common.tests.test_utils import initialize_kv_cache, load_hf_model
 from models.demos.utils.llm_demo_utils import create_benchmark_data, verify_perf, check_tokens_match
 from models.utility_functions import (
     disable_compilation_reports,

models/demos/falcon7b/reference/cpu_demo.py → models/demos/falcon7b_common/reference/cpu_demo.py

@@ -9,7 +9,7 @@
 
 from transformers import AutoTokenizer
 
-from models.falcon7b.reference.hf_modeling_falcon import FalconForCausalLM
+from models.demos.falcon7b_common.reference.hf_modeling_falcon import FalconForCausalLM
 import time
 
 falcon1b = "tiiuae/falcon-rw-1b"

models/demos/falcon7b/tests/perplexity/run_perplexity_falcon.py → models/demos/falcon7b_common/tests/perplexity/run_perplexity_falcon.py

@@ -9,9 +9,9 @@
 import time
 import numpy as np
 import ttnn
-from models.demos.falcon7b.tt.falcon_causallm import TtFalconCausalLM
-from models.demos.falcon7b.tt.model_config import get_model_config
-from models.demos.falcon7b.tests.test_utils import initialize_kv_cache, load_hf_model
+from models.demos.falcon7b_common.tt.falcon_causallm import TtFalconCausalLM
+from models.demos.falcon7b_common.tt.model_config import get_model_config
+from models.demos.falcon7b_common.tests.test_utils import initialize_kv_cache, load_hf_model
 from models.datasets.llm_dataset_utils import (
     prepare_textgen_dataset,
     prepare_textgen_dataloader,

models/demos/falcon7b/tests/perplexity/test_perplexity_falcon.py → models/demos/falcon7b_common/tests/perplexity/test_perplexity_falcon.py

@@ -3,7 +3,7 @@
 # SPDX-License-Identifier: Apache-2.0
 
 import pytest
-from models.demos.falcon7b.tests.perplexity.run_perplexity_falcon import run_test_perplexity
+from models.demos.falcon7b_common.tests.perplexity.run_perplexity_falcon import run_test_perplexity
 from models.utility_functions import is_wormhole_b0
 
 

models/demos/falcon7b/tests/perplexity/test_perplexity_falcon_ref.py → models/demos/falcon7b_common/tests/perplexity/test_perplexity_falcon_ref.py

@@ -3,7 +3,7 @@
 # SPDX-License-Identifier: Apache-2.0
 
 import pytest
-from models.demos.falcon7b.tests.perplexity.run_perplexity_falcon import run_test_perplexity
+from models.demos.falcon7b_common.tests.perplexity.run_perplexity_falcon import run_test_perplexity
 
 
 @pytest.mark.parametrize(

models/demos/falcon7b/tests/run_falcon_end_to_end.py → models/demos/falcon7b_common/tests/run_falcon_end_to_end.py

@@ -9,13 +9,13 @@
 from sklearn.metrics import top_k_accuracy_score
 
 import tt_lib
-from models.demos.falcon7b.tt.falcon_causallm import TtFalconCausalLM
+from models.demos.falcon7b_common.tt.falcon_causallm import TtFalconCausalLM
 
-from models.demos.falcon7b.tt.falcon_common import (
+from models.demos.falcon7b_common.tt.falcon_common import (
     PytorchFalconCausalLM,
 )
 
-from models.demos.falcon7b.tests.test_utils import (
+from models.demos.falcon7b_common.tests.test_utils import (
     get_rand_falcon_inputs,
     concat_device_out_layer_present,
     load_hf_model,

models/demos/falcon7b/tests/test_falcon_attention.py → models/demos/falcon7b_common/tests/test_falcon_attention.py

@@ -6,9 +6,9 @@
 import pytest
 from loguru import logger
 
-from models.demos.falcon7b.tt.falcon_attention import TtFalconAttentionDecode, TtFalconAttentionPrefill
-from models.demos.falcon7b.tt.model_config import get_model_config
-from models.demos.falcon7b.tests.test_utils import get_rand_falcon_inputs, concat_device_outputs, load_hf_model
+from models.demos.falcon7b_common.tt.falcon_attention import TtFalconAttentionDecode, TtFalconAttentionPrefill
+from models.demos.falcon7b_common.tt.model_config import get_model_config
+from models.demos.falcon7b_common.tests.test_utils import get_rand_falcon_inputs, concat_device_outputs, load_hf_model
 from tests.tt_eager.python_api_testing.sweep_tests.comparison_funcs import (
     comp_pcc,
 )

models/demos/falcon7b/tests/test_falcon_causallm.py → models/demos/falcon7b_common/tests/test_falcon_causallm.py

@@ -6,12 +6,12 @@
 import pytest
 from loguru import logger
 
-from models.demos.falcon7b.tt.falcon_causallm import TtFalconCausalLM
+from models.demos.falcon7b_common.tt.falcon_causallm import TtFalconCausalLM
 
-from models.demos.falcon7b.tt.model_config import (
+from models.demos.falcon7b_common.tt.model_config import (
     get_model_config,
 )
-from models.demos.falcon7b.tests.test_utils import (
+from models.demos.falcon7b_common.tests.test_utils import (
     get_rand_falcon_inputs,
     concat_device_out_layer_present,
     load_hf_model,

models/demos/falcon7b/tests/test_falcon_decoder.py → models/demos/falcon7b_common/tests/test_falcon_decoder.py

@@ -6,9 +6,9 @@
 import pytest
 from loguru import logger
 
-from models.demos.falcon7b.tt.falcon_decoder import TtFalconDecoderLayer
-from models.demos.falcon7b.tt.model_config import get_model_config
-from models.demos.falcon7b.tests.test_utils import get_rand_falcon_inputs, concat_device_outputs, load_hf_model
+from models.demos.falcon7b_common.tt.falcon_decoder import TtFalconDecoderLayer
+from models.demos.falcon7b_common.tt.model_config import get_model_config
+from models.demos.falcon7b_common.tests.test_utils import get_rand_falcon_inputs, concat_device_outputs, load_hf_model
 from tests.tt_eager.python_api_testing.sweep_tests.comparison_funcs import (
     comp_pcc,
 )

models/demos/falcon7b/tests/test_falcon_device_perf.py → models/demos/falcon7b_common/tests/test_falcon_device_perf.py

@@ -4,13 +4,13 @@
 
 import pytest
 
-from models.demos.falcon7b.tests.run_falcon_end_to_end import (
+from models.demos.falcon7b_common.tests.run_falcon_end_to_end import (
     DECODE_CONFIG_TO_PCC,
     PREFILL_CONFIG_TO_PCC,
     DeviceSetup,
     run_test_FalconCausalLM_end_to_end,
 )
-from models.demos.falcon7b.tt.model_config import get_model_config
+from models.demos.falcon7b_common.tt.model_config import get_model_config
 from models.perf.device_perf_utils import check_device_perf, prep_device_perf_report, run_device_perf
 from models.utility_functions import disable_compilation_reports, disable_persistent_kernel_cache
 
@@ -85,7 +85,7 @@ def test_device_perf(seq_len, samples):
     margin = 0.03
     num_iterations = 1
     model_config = "BFLOAT16-DRAM".lower()
-    command = f"pytest models/demos/falcon7b/tests/test_falcon_device_perf.py::test_device_perf_wh_bare_metal -k prefill_seq{seq_len}_{model_config}"
+    command = f"pytest models/demos/falcon7b_common/tests/test_falcon_device_perf.py::test_device_perf_wh_bare_metal -k prefill_seq{seq_len}_{model_config}"
     cols = ["DEVICE FW", "DEVICE KERNEL", "DEVICE BRISC KERNEL"]
     subdir = "falcon7b"
 

models/demos/falcon7b/tests/test_falcon_end_to_end.py → models/demos/falcon7b_common/tests/test_falcon_end_to_end.py

@@ -7,16 +7,16 @@
 import torch
 import tt_lib
 from loguru import logger
-from models.demos.falcon7b.tests.test_utils import (
+from models.demos.falcon7b_common.tests.test_utils import (
     concat_device_out_layer_present,
     get_rand_falcon_inputs,
     load_hf_model,
 )
-from models.demos.falcon7b.tt.falcon_causallm import TtFalconCausalLM
+from models.demos.falcon7b_common.tt.falcon_causallm import TtFalconCausalLM
 
 # TODO: Remove this?
-from models.demos.falcon7b.tt.falcon_common import PytorchFalconCausalLM
-from models.demos.falcon7b.tt.model_config import get_model_config
+from models.demos.falcon7b_common.tt.falcon_common import PytorchFalconCausalLM
+from models.demos.falcon7b_common.tt.model_config import get_model_config
 from models.utility_functions import (
     disable_compilation_reports,
     disable_persistent_kernel_cache,

models/demos/falcon7b/tests/test_falcon_mlp.py → models/demos/falcon7b_common/tests/test_falcon_mlp.py

@@ -5,9 +5,9 @@
 import pytest
 import torch
 from loguru import logger
-from models.demos.falcon7b.tt.falcon_mlp import TtFalconMLPDecode, TtFalconMLPPrefill
-from models.demos.falcon7b.tt.model_config import get_model_config
-from models.demos.falcon7b.tests.test_utils import load_hf_model
+from models.demos.falcon7b_common.tt.falcon_mlp import TtFalconMLPDecode, TtFalconMLPPrefill
+from models.demos.falcon7b_common.tt.model_config import get_model_config
+from models.demos.falcon7b_common.tests.test_utils import load_hf_model
 from models.utility_functions import get_devices_for_t3000, torch2tt_tensor, tt2torch_tensor
 from tests.tt_eager.python_api_testing.sweep_tests.comparison_funcs import comp_allclose, comp_pcc
 

models/demos/falcon7b/tests/test_falcon_model.py → models/demos/falcon7b_common/tests/test_falcon_model.py

@@ -5,11 +5,11 @@
 import torch
 import pytest
 from loguru import logger
-from models.demos.falcon7b.tt.falcon_model import TtFalconModel
-from models.demos.falcon7b.tt.model_config import (
+from models.demos.falcon7b_common.tt.falcon_model import TtFalconModel
+from models.demos.falcon7b_common.tt.model_config import (
     get_model_config,
 )
-from models.demos.falcon7b.tests.test_utils import (
+from models.demos.falcon7b_common.tests.test_utils import (
     get_rand_falcon_inputs,
     concat_device_out_layer_present,
     load_hf_model,

models/demos/falcon7b/tests/test_falcon_prefill_decode.py → models/demos/falcon7b_common/tests/test_falcon_prefill_decode.py

@@ -6,12 +6,12 @@
 import pytest
 from loguru import logger
 
-from models.demos.falcon7b.tt.falcon_causallm import TtFalconCausalLM
+from models.demos.falcon7b_common.tt.falcon_causallm import TtFalconCausalLM
 
-from models.demos.falcon7b.tt.model_config import (
+from models.demos.falcon7b_common.tt.model_config import (
     get_model_config,
 )
-from models.demos.falcon7b.tests.test_utils import load_hf_model
+from models.demos.falcon7b_common.tests.test_utils import load_hf_model
 
 from tests.tt_eager.python_api_testing.sweep_tests.comparison_funcs import (
     comp_allclose,

models/demos/falcon7b/tests/test_perf_falcon.py → models/demos/falcon7b_common/tests/test_perf_falcon.py

@@ -4,13 +4,13 @@
 
 import pytest
 
-from models.demos.falcon7b.tests.run_falcon_end_to_end import (
+from models.demos.falcon7b_common.tests.run_falcon_end_to_end import (
     DECODE_CONFIG_TO_PCC,
     PREFILL_CONFIG_TO_PCC,
     DeviceSetup,
     run_test_FalconCausalLM_end_to_end,
 )
-from models.demos.falcon7b.tt.model_config import (
+from models.demos.falcon7b_common.tt.model_config import (
     get_model_config,
 )
 

models/demos/falcon7b/tests/test_utils.py → models/demos/falcon7b_common/tests/test_utils.py

@@ -4,7 +4,7 @@
 
 import torch
 import ttnn
-from models.demos.falcon7b.reference.hf_modeling_falcon import FalconForCausalLM
+from models.demos.falcon7b_common.reference.hf_modeling_falcon import FalconForCausalLM
 from models.utility_functions import torch2tt_tensor, tt2torch_tensor
 
 

models/demos/falcon7b/tests/unit_tests/test_falcon_lm_head_matmul_2d.py → models/demos/falcon7b_common/tests/unit_tests/test_falcon_lm_head_matmul_2d.py

@@ -8,7 +8,7 @@
 
 import ttnn
 
-from models.demos.falcon7b.tt.falcon_lm_head import falcon_lm_head_matmul_2d
+from models.demos.falcon7b_common.tt.falcon_lm_head import falcon_lm_head_matmul_2d
 from models.utility_functions import comp_pcc, tt2torch_tensor, torch2tt_tensor
 
 

models/demos/falcon7b/tests/unit_tests/test_falcon_matmuls_and_bmms_with_mixed_precision.py → models/demos/falcon7b_common/tests/unit_tests/test_falcon_matmuls_and_bmms_with_mixed_precision.py

@@ -7,9 +7,9 @@
 from loguru import logger
 
 import ttnn
-from models.demos.falcon7b.tt.falcon_causallm import falcon_lm_head_matmul
-from models.demos.falcon7b.tt.falcon_mlp import falcon_dense_4h_to_h_matmul, falcon_dense_h_to_4h_matmul
-from models.demos.falcon7b.tt.model_utils import get_falcon_default_core_grid
+from models.demos.falcon7b_common.tt.falcon_causallm import falcon_lm_head_matmul
+from models.demos.falcon7b_common.tt.falcon_mlp import falcon_dense_4h_to_h_matmul, falcon_dense_h_to_4h_matmul
+from models.demos.falcon7b_common.tt.model_utils import get_falcon_default_core_grid
 from models.utility_functions import comp_pcc, tt2torch_tensor, torch2tt_tensor, skip_for_wormhole_b0
 import torch
 import math

models/demos/falcon7b/tt/falcon_attention.py → models/demos/falcon7b_common/tt/falcon_attention.py

@@ -7,7 +7,7 @@
 from torch import nn
 from typing import List, Optional, Tuple
 
-from models.demos.falcon7b.tt.model_utils import get_falcon_default_core_grid
+from models.demos.falcon7b_common.tt.model_utils import get_falcon_default_core_grid
 import ttnn
 
 from models.utility_functions import (
@@ -17,7 +17,7 @@
     is_wormhole_b0,
 )
 
-from models.demos.falcon7b.tt.model_utils import get_weights_cached
+from models.demos.falcon7b_common.tt.model_utils import get_weights_cached
 from models.utility_functions import torch_tensors_to_tt_tensors
 
 

models/demos/falcon7b/tt/falcon_causallm.py → models/demos/falcon7b_common/tt/falcon_causallm.py

@@ -6,9 +6,9 @@
 
 import torch
 import ttnn
-from models.demos.falcon7b.tt.falcon_lm_head import falcon_lm_head_matmul_2d
-from models.demos.falcon7b.tt.falcon_model import TtFalconModelShared
-from models.demos.falcon7b.tt.model_utils import get_falcon_default_core_grid, get_weights_cached
+from models.demos.falcon7b_common.tt.falcon_lm_head import falcon_lm_head_matmul_2d
+from models.demos.falcon7b_common.tt.falcon_model import TtFalconModelShared
+from models.demos.falcon7b_common.tt.model_utils import get_falcon_default_core_grid, get_weights_cached
 from models.utility_functions import (
     is_grayskull,
     is_wormhole_b0,

models/demos/falcon7b/tt/falcon_decoder.py → models/demos/falcon7b_common/tt/falcon_decoder.py

@@ -6,9 +6,9 @@
 
 import torch
 import ttnn
-from models.demos.falcon7b.tt.falcon_attention import TtFalconAttentionDecode, TtFalconAttentionPrefill
-from models.demos.falcon7b.tt.falcon_mlp import TtFalconMLPDecode, TtFalconMLPPrefill
-from models.demos.falcon7b.tt.model_utils import get_weights_cached, layernorm
+from models.demos.falcon7b_common.tt.falcon_attention import TtFalconAttentionDecode, TtFalconAttentionPrefill
+from models.demos.falcon7b_common.tt.falcon_mlp import TtFalconMLPDecode, TtFalconMLPPrefill
+from models.demos.falcon7b_common.tt.model_utils import get_weights_cached, layernorm
 from torch import nn
 
 

models/demos/falcon7b/tt/falcon_mlp.py → models/demos/falcon7b_common/tt/falcon_mlp.py

@@ -4,7 +4,7 @@
 
 import torch
 import ttnn
-from models.demos.falcon7b.tt.model_utils import get_falcon_default_core_grid, get_weights_cached
+from models.demos.falcon7b_common.tt.model_utils import get_falcon_default_core_grid, get_weights_cached
 from torch import nn
 from models.utility_functions import (
     is_grayskull,

models/demos/falcon7b/tt/falcon_model.py → models/demos/falcon7b_common/tt/falcon_model.py

@@ -10,10 +10,10 @@
 
 import tt_lib
 
-from models.demos.falcon7b.tt.falcon_decoder import TtFalconDecoderLayer
-from models.demos.falcon7b.tt.model_utils import get_weights_cached, layernorm
+from models.demos.falcon7b_common.tt.falcon_decoder import TtFalconDecoderLayer
+from models.demos.falcon7b_common.tt.model_utils import get_weights_cached, layernorm
 from models.utility_functions import nearest_32, torch_tensors_to_tt_tensors
-from models.demos.falcon7b.tests.test_utils import create_prefill_attn_mask_for_sharded_softmax
+from models.demos.falcon7b_common.tests.test_utils import create_prefill_attn_mask_for_sharded_softmax
 from tqdm import tqdm
 
 

models/demos/falcon7b/tt/model_config.py → models/demos/falcon7b_common/tt/model_config.py

@@ -195,7 +195,7 @@ def get_model_config(model_config_str, prefill_seq_len=0, decode_batch_size=32):
         "DEFAULT_DTYPE": dtype,
         "DEFAULT_MEMCFG": mem_config,
         "MOVE_DECODER_OUTPUT_BOOL": False,
-        "DEFAULT_CACHE_PATH": Path(f"models/demos/falcon7b/datasets/"),
+        "DEFAULT_CACHE_PATH": Path(f"models/demos/falcon7b_common/datasets/"),
     }  # DEFAULT_MEMCFG also used to determine banking for ttl.device.InitializeDevice
     model_config.update({f"{key}_MEMCFG": mem_config for key in OP_KEYS if key not in NO_MEMCFG})
     model_config.update({f"{key}_DTYPE": dtype for key in OP_KEYS if key not in NO_DTYPE})

models/demos/grayskull/falcon7b/README.md

@@ -8,7 +8,7 @@ To run the model for a single user you can use the command line input:
 
 To run the demo using prewritten prompts for a batch of 32 users run (currently only supports same token-length inputs):
 
-`pytest --disable-warnings -q -s --input-method=json --input-path='models/demos/falcon7b/demo/input_data.json' models/demos/grayskull/falcon7b/demo_grayskull.py::test_demo[user_input0-default_mode_stochastic]`
+`pytest --disable-warnings -q -s --input-method=json --input-path='models/demos/falcon7b_common/demo/input_data.json' models/demos/grayskull/falcon7b/demo_grayskull.py::test_demo[user_input0-default_mode_stochastic]`
 
 ## Inputs
 

models/demos/grayskull/falcon7b/demo_grayskull.py

@@ -3,7 +3,7 @@
 # SPDX-License-Identifier: Apache-2.0
 
 import pytest
-from models.demos.falcon7b.demo.demo import run_falcon_demo_kv
+from models.demos.falcon7b_common.demo.demo import run_falcon_demo_kv
 
 
 @pytest.mark.parametrize(

models/demos/t3000/falcon7b/demo_t3000.py

@@ -3,7 +3,7 @@
 # SPDX-License-Identifier: Apache-2.0
 
 import pytest
-from models.demos.falcon7b.demo.demo import run_falcon_demo_kv
+from models.demos.falcon7b_common.demo.demo import run_falcon_demo_kv
 from models.utility_functions import is_wormhole_b0, get_devices_for_t3000
 
 

models/demos/wormhole/falcon7b/README.md

@@ -14,15 +14,15 @@ To run the model for a single user you can use the command line input:
 
 To run the demo using prewritten prompts for a batch of 32 users run (currently only supports same token-length inputs):
 
-`pytest --disable-warnings -q -s --input-method=json --input-path='models/demos/falcon7b/demo/input_data.json' models/demos/wormhole/falcon7b/demo_wormhole.py::test_demo[user_input0-default_mode_1024_stochastic]`
+`pytest --disable-warnings -q -s --input-method=json --input-path='models/demos/falcon7b_common/demo/input_data.json' models/demos/wormhole/falcon7b/demo_wormhole.py::test_demo[user_input0-default_mode_1024_stochastic]`
 
 - **Decoding method**: The default decoding method is top-k/top-p (stochastic) sampling, however greedy decoding can also be used by replacing `stochastic` with `greedy` in the command above.
 
 #### Performance Measurement Mode
 
 To measure the performance of generating the `i`'th token while the KV cache is filled with `i-1` rows (where `i` is 128 in the command below):
 
-`pytest --disable-warnings -q -s --input-method=json --input-path='models/demos/falcon7b/demo/input_data.json' models/demos/wormhole/falcon7b/demo_wormhole.py::test_demo[user_input0-perf_mode_128_stochastic]`
+`pytest --disable-warnings -q -s --input-method=json --input-path='models/demos/falcon7b_common/demo/input_data.json' models/demos/wormhole/falcon7b/demo_wormhole.py::test_demo[user_input0-perf_mode_128_stochastic]`
 
 - **Supported sequence lengths**: Currently `i` can only be set to 128, 1024, or 2048 for performance measurement mode.
 

models/demos/wormhole/falcon7b/demo_wormhole.py

@@ -3,7 +3,7 @@
 # SPDX-License-Identifier: Apache-2.0
 
 import pytest
-from models.demos.falcon7b.demo.demo import run_falcon_demo_kv
+from models.demos.falcon7b_common.demo.demo import run_falcon_demo_kv
 from models.utility_functions import is_wormhole_b0