From e5423bf0e905266874871e3ca1ea06b13476b254 Mon Sep 17 00:00:00 2001 From: JushBJJ Date: Sun, 31 Mar 2024 04:05:27 +1000 Subject: [PATCH 01/14] Qwen1.5 0.5B pybuda implementation --- .../model_qwen1.5/pytorch_qwen1.5.py | 71 +++++++++++++++++++ model_demos/requirements.txt | 4 ++ 2 files changed, 75 insertions(+) create mode 100644 model_demos/nlp_demos/model_qwen1.5/pytorch_qwen1.5.py diff --git a/model_demos/nlp_demos/model_qwen1.5/pytorch_qwen1.5.py b/model_demos/nlp_demos/model_qwen1.5/pytorch_qwen1.5.py new file mode 100644 index 00000000..a3e46031 --- /dev/null +++ b/model_demos/nlp_demos/model_qwen1.5/pytorch_qwen1.5.py @@ -0,0 +1,71 @@ +import pybuda + +from transformers import Qwen2ForCausalLM, Qwen2Tokenizer, Qwen2Config +from pybuda.transformers.pipeline import pipeline as pybuda_pipeline + +""" +=== Models === + +Qwen/Qwen1.5-0.5B +Qwen/Qwen1.5-0.5B-Chat +Qwen/Qwen1.5-0.5B-Chat-GPTQ-Int4 +Qwen/Qwen1.5-0.5B-Chat-GPTQ-Int8 +""" + +model_name = "Qwen/Qwen1.5-0.5B" + + +def run_qwen_causal_lm(max_length=1024, top_p=0.9, top_k=50, temperature=0.7): + # Set PyBuda configurations + compiler_cfg = pybuda.config._get_global_compiler_config() + compiler_cfg.default_df_override = pybuda._C.DataFormat.Float16_b + + # Config + config = Qwen2Config.from_pretrained(model_name) + config_dict = config.to_dict() + config_dict["return_dict"] = False + config_dict["use_cache"] = False + + config = Qwen2Config(**config_dict) + + # Load the model and tokenizer + model = Qwen2ForCausalLM.from_pretrained( + model_name, config=config, device_map=device) + tokenizer = Qwen2Tokenizer.from_pretrained(model_name, device_map=device) + + # Set pad token + tokenizer.pad_token = tokenizer.eos_token + + # Example usage + prompt = "What is a neural network?" + + # Initialize pipeline + text_generator = pybuda_pipeline( + "text-generation", + model=model, + tokenizer=tokenizer, + config=config, + ) + + # Inference + output = text_generator( + prompt, + do_sample=True, + pad_token_id=tokenizer.pad_token_id, + max_new_tokens=max_length, + temperature=temperature, + top_k=top_k, + top_p=top_p, + ) + + # Display output + print("OUTPUT:\n", output[0]["generated_text"]) + + +if __name__ == "__main__": + run_qwen_causal_lm( + max_length=1024, + top_p=0.9, + top_k=50, + temperature=0.7 + ) \ No newline at end of file diff --git a/model_demos/requirements.txt b/model_demos/requirements.txt index 4b8b2f52..43929eee 100644 --- a/model_demos/requirements.txt +++ b/model_demos/requirements.txt @@ -9,4 +9,8 @@ soundfile==0.12.1 # For Whisper librosa==0.10.0 # For Whisper numba==0.53.1 # For Whisper segmentation-models-pytorch==0.3.3 # For U-Net +pylocron==0.2.1 # For U-Net diffusers==0.14.0 # For Stable Diffusion +transformers==4.37.0 # For Qwen2 +acceleration==0.28.0 # For Qwen2 +auto-gptq==0.7.1 # For Qwen2-GPTQ \ No newline at end of file From 88641a373588ac91fa724c45078c48435a93e41b Mon Sep 17 00:00:00 2001 From: JushBJJ Date: Sun, 31 Mar 2024 04:07:10 +1000 Subject: [PATCH 02/14] remove unneeded requirement --- model_demos/requirements.txt | 1 - 1 file changed, 1 deletion(-) diff --git a/model_demos/requirements.txt b/model_demos/requirements.txt index 43929eee..118c65b9 100644 --- a/model_demos/requirements.txt +++ b/model_demos/requirements.txt @@ -9,7 +9,6 @@ soundfile==0.12.1 # For Whisper librosa==0.10.0 # For Whisper numba==0.53.1 # For Whisper segmentation-models-pytorch==0.3.3 # For U-Net -pylocron==0.2.1 # For U-Net diffusers==0.14.0 # For Stable Diffusion transformers==4.37.0 # For Qwen2 acceleration==0.28.0 # For Qwen2 From e442e192c82b4d3d03b1b08fd1917ca6faf97928 Mon Sep 17 00:00:00 2001 From: JushBJJ Date: Sun, 31 Mar 2024 04:25:12 +1000 Subject: [PATCH 03/14] rename "acceleration" to "accelerate" --- model_demos/requirements.txt | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/model_demos/requirements.txt b/model_demos/requirements.txt index 118c65b9..637db37e 100644 --- a/model_demos/requirements.txt +++ b/model_demos/requirements.txt @@ -11,5 +11,5 @@ numba==0.53.1 # For Whisper segmentation-models-pytorch==0.3.3 # For U-Net diffusers==0.14.0 # For Stable Diffusion transformers==4.37.0 # For Qwen2 -acceleration==0.28.0 # For Qwen2 +accelerate==0.28.0 # For Qwen2 auto-gptq==0.7.1 # For Qwen2-GPTQ \ No newline at end of file From 597a8b95ec85d196b5bd8888c89d15ded86c0ac4 Mon Sep 17 00:00:00 2001 From: JushBJJ Date: Sun, 14 Apr 2024 15:35:50 +1000 Subject: [PATCH 04/14] Update env vars and compiler configs --- model_demos/nlp_demos/model_qwen1.5/pytorch_qwen1.5.py | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-) diff --git a/model_demos/nlp_demos/model_qwen1.5/pytorch_qwen1.5.py b/model_demos/nlp_demos/model_qwen1.5/pytorch_qwen1.5.py index a3e46031..9d450174 100644 --- a/model_demos/nlp_demos/model_qwen1.5/pytorch_qwen1.5.py +++ b/model_demos/nlp_demos/model_qwen1.5/pytorch_qwen1.5.py @@ -1,4 +1,5 @@ import pybuda +import os from transformers import Qwen2ForCausalLM, Qwen2Tokenizer, Qwen2Config from pybuda.transformers.pipeline import pipeline as pybuda_pipeline @@ -16,9 +17,15 @@ def run_qwen_causal_lm(max_length=1024, top_p=0.9, top_k=50, temperature=0.7): + # Set environment variables + os.environ['PYBUDA_DEVMODE'] = '0' + os.environ['TT_BACKEND_TIMEOUT'] = '0' + os.environ["PYBUDA_FORK_JOIN_EXPAND_FORK_OUTPUT_BUF"] = "0" + os.environ["TT_BACKEND_OVERLAY_MAX_EXTRA_BLOB_SIZE"] = "65536" + # Set PyBuda configurations compiler_cfg = pybuda.config._get_global_compiler_config() - compiler_cfg.default_df_override = pybuda._C.DataFormat.Float16_b + compiler_cfg.amp_level = 0 # Config config = Qwen2Config.from_pretrained(model_name) From f136d08b621c245d7ab9f6eb4cf532309bf54123 Mon Sep 17 00:00:00 2001 From: "Jush (yupiop12)" <36951064+JushBJJ@users.noreply.github.com> Date: Mon, 15 Apr 2024 22:11:40 +1000 Subject: [PATCH 05/14] remove undefined device_map --- model_demos/nlp_demos/model_qwen1.5/pytorch_qwen1.5.py | 7 +++---- 1 file changed, 3 insertions(+), 4 deletions(-) diff --git a/model_demos/nlp_demos/model_qwen1.5/pytorch_qwen1.5.py b/model_demos/nlp_demos/model_qwen1.5/pytorch_qwen1.5.py index 9d450174..4e689e75 100644 --- a/model_demos/nlp_demos/model_qwen1.5/pytorch_qwen1.5.py +++ b/model_demos/nlp_demos/model_qwen1.5/pytorch_qwen1.5.py @@ -36,9 +36,8 @@ def run_qwen_causal_lm(max_length=1024, top_p=0.9, top_k=50, temperature=0.7): config = Qwen2Config(**config_dict) # Load the model and tokenizer - model = Qwen2ForCausalLM.from_pretrained( - model_name, config=config, device_map=device) - tokenizer = Qwen2Tokenizer.from_pretrained(model_name, device_map=device) + model = Qwen2ForCausalLM.from_pretrained(model_name, config=config) + tokenizer = Qwen2Tokenizer.from_pretrained(model_name) # Set pad token tokenizer.pad_token = tokenizer.eos_token @@ -75,4 +74,4 @@ def run_qwen_causal_lm(max_length=1024, top_p=0.9, top_k=50, temperature=0.7): top_p=0.9, top_k=50, temperature=0.7 - ) \ No newline at end of file + ) From 800737ad491f02157175b0947868a1b1ed7571da Mon Sep 17 00:00:00 2001 From: "Jush (yupiop12)" <36951064+JushBJJ@users.noreply.github.com> Date: Fri, 26 Apr 2024 21:23:46 +1000 Subject: [PATCH 06/14] Remove misleading and unnecessary environment variables --- model_demos/nlp_demos/model_qwen1.5/pytorch_qwen1.5.py | 3 --- 1 file changed, 3 deletions(-) diff --git a/model_demos/nlp_demos/model_qwen1.5/pytorch_qwen1.5.py b/model_demos/nlp_demos/model_qwen1.5/pytorch_qwen1.5.py index 4e689e75..cd1261a5 100644 --- a/model_demos/nlp_demos/model_qwen1.5/pytorch_qwen1.5.py +++ b/model_demos/nlp_demos/model_qwen1.5/pytorch_qwen1.5.py @@ -18,10 +18,7 @@ def run_qwen_causal_lm(max_length=1024, top_p=0.9, top_k=50, temperature=0.7): # Set environment variables - os.environ['PYBUDA_DEVMODE'] = '0' os.environ['TT_BACKEND_TIMEOUT'] = '0' - os.environ["PYBUDA_FORK_JOIN_EXPAND_FORK_OUTPUT_BUF"] = "0" - os.environ["TT_BACKEND_OVERLAY_MAX_EXTRA_BLOB_SIZE"] = "65536" # Set PyBuda configurations compiler_cfg = pybuda.config._get_global_compiler_config() From 471acec62d521292d09851097b139aa3b793e8dc Mon Sep 17 00:00:00 2001 From: JushBJJ Date: Fri, 26 Jul 2024 10:50:06 +1000 Subject: [PATCH 07/14] remove qwen from phi branch --- .../model_qwen1.5/pytorch_qwen1.5.py | 74 ------------------- 1 file changed, 74 deletions(-) delete mode 100644 model_demos/nlp_demos/model_qwen1.5/pytorch_qwen1.5.py diff --git a/model_demos/nlp_demos/model_qwen1.5/pytorch_qwen1.5.py b/model_demos/nlp_demos/model_qwen1.5/pytorch_qwen1.5.py deleted file mode 100644 index cd1261a5..00000000 --- a/model_demos/nlp_demos/model_qwen1.5/pytorch_qwen1.5.py +++ /dev/null @@ -1,74 +0,0 @@ -import pybuda -import os - -from transformers import Qwen2ForCausalLM, Qwen2Tokenizer, Qwen2Config -from pybuda.transformers.pipeline import pipeline as pybuda_pipeline - -""" -=== Models === - -Qwen/Qwen1.5-0.5B -Qwen/Qwen1.5-0.5B-Chat -Qwen/Qwen1.5-0.5B-Chat-GPTQ-Int4 -Qwen/Qwen1.5-0.5B-Chat-GPTQ-Int8 -""" - -model_name = "Qwen/Qwen1.5-0.5B" - - -def run_qwen_causal_lm(max_length=1024, top_p=0.9, top_k=50, temperature=0.7): - # Set environment variables - os.environ['TT_BACKEND_TIMEOUT'] = '0' - - # Set PyBuda configurations - compiler_cfg = pybuda.config._get_global_compiler_config() - compiler_cfg.amp_level = 0 - - # Config - config = Qwen2Config.from_pretrained(model_name) - config_dict = config.to_dict() - config_dict["return_dict"] = False - config_dict["use_cache"] = False - - config = Qwen2Config(**config_dict) - - # Load the model and tokenizer - model = Qwen2ForCausalLM.from_pretrained(model_name, config=config) - tokenizer = Qwen2Tokenizer.from_pretrained(model_name) - - # Set pad token - tokenizer.pad_token = tokenizer.eos_token - - # Example usage - prompt = "What is a neural network?" - - # Initialize pipeline - text_generator = pybuda_pipeline( - "text-generation", - model=model, - tokenizer=tokenizer, - config=config, - ) - - # Inference - output = text_generator( - prompt, - do_sample=True, - pad_token_id=tokenizer.pad_token_id, - max_new_tokens=max_length, - temperature=temperature, - top_k=top_k, - top_p=top_p, - ) - - # Display output - print("OUTPUT:\n", output[0]["generated_text"]) - - -if __name__ == "__main__": - run_qwen_causal_lm( - max_length=1024, - top_p=0.9, - top_k=50, - temperature=0.7 - ) From dcff324731905542f3979661f5ab5e77c2862f91 Mon Sep 17 00:00:00 2001 From: JushBJJ Date: Fri, 26 Jul 2024 10:50:54 +1000 Subject: [PATCH 08/14] Add Phi 2 --- .../model_phi2/phi2_text_generation.py | 60 +++++++++++++++++++ 1 file changed, 60 insertions(+) create mode 100644 model_demos/nlp_demos/model_phi2/phi2_text_generation.py diff --git a/model_demos/nlp_demos/model_phi2/phi2_text_generation.py b/model_demos/nlp_demos/model_phi2/phi2_text_generation.py new file mode 100644 index 00000000..cd7bd9a6 --- /dev/null +++ b/model_demos/nlp_demos/model_phi2/phi2_text_generation.py @@ -0,0 +1,60 @@ +import os +import pybuda + +from transformers import PhiForCausalLM, AutoTokenizer, PhiConfig +from pybuda.transformers.pipeline import pipeline as pybuda_pipeline + +model_name = "microsoft/phi-2" + +def run_qwen_causal_lm(max_length=1024, top_p=0.9, top_k=50, temperature=0.7): + # Set PyBuda configurations + compiler_cfg = pybuda.config._get_global_compiler_config() + compiler_cfg.default_df_override = pybuda.DataFormat.Float16_b + compiler_cfg.enable_auto_fusing = False + compiler_cfg.balancer_policy = "Ribbon" + + # Load the model configuration + config = PhiConfig.from_pretrained(model_name) + config.use_cache = True + + # Load the model and tokenizer with the updated config + model = PhiForCausalLM.from_pretrained(model_name, config=config) + tokenizer = AutoTokenizer.from_pretrained(model_name) + + # Set pad token + tokenizer.pad_token = tokenizer.eos_token + + # Example usage + prompt = ["My name is Jimmy and"] + + # Initialize pipeline + text_generator = pybuda_pipeline( + "text-generation", + model=model, + tokenizer=tokenizer + ) + + # Inference + output = text_generator( + prompt, + do_sample=True, + num_beams=1, + no_repeat_ngram_size=2, + pad_token_id=tokenizer.pad_token_id, + max_new_tokens=max_length, + temperature=temperature, + top_k=top_k, + top_p=top_p, + ) + + # Display output + print("OUTPUT:\n", output[0][0]["generated_text"]) + + +if __name__ == "__main__": + run_qwen_causal_lm( + max_length=100, + top_p=0.7, + top_k=50, + temperature=0.7 + ) \ No newline at end of file From 25958b868b4aced48029c1b1164bebce344e3596 Mon Sep 17 00:00:00 2001 From: "Jush (yupiop12)" <36951064+JushBJJ@users.noreply.github.com> Date: Fri, 26 Jul 2024 11:20:22 +1000 Subject: [PATCH 09/14] Update requirements.txt --- model_demos/requirements.txt | 3 --- 1 file changed, 3 deletions(-) diff --git a/model_demos/requirements.txt b/model_demos/requirements.txt index 637db37e..4b8b2f52 100644 --- a/model_demos/requirements.txt +++ b/model_demos/requirements.txt @@ -10,6 +10,3 @@ librosa==0.10.0 # For Whisper numba==0.53.1 # For Whisper segmentation-models-pytorch==0.3.3 # For U-Net diffusers==0.14.0 # For Stable Diffusion -transformers==4.37.0 # For Qwen2 -accelerate==0.28.0 # For Qwen2 -auto-gptq==0.7.1 # For Qwen2-GPTQ \ No newline at end of file From 871f749deb74acdaf148204f21d83cf7f8f61171 Mon Sep 17 00:00:00 2001 From: JushBJJ Date: Thu, 1 Aug 2024 00:23:23 +0000 Subject: [PATCH 10/14] Standardize Phi2 demo and added tests --- .../phi2/pytorch_phi2_text_generation.py | 68 +++++++++++++++++++ model_demos/pyproject.toml | 1 + model_demos/tests/test_pytorch_phi2.py | 10 +++ 3 files changed, 79 insertions(+) create mode 100644 model_demos/nlp_demos/phi2/pytorch_phi2_text_generation.py create mode 100644 model_demos/tests/test_pytorch_phi2.py diff --git a/model_demos/nlp_demos/phi2/pytorch_phi2_text_generation.py b/model_demos/nlp_demos/phi2/pytorch_phi2_text_generation.py new file mode 100644 index 00000000..a0f172f3 --- /dev/null +++ b/model_demos/nlp_demos/phi2/pytorch_phi2_text_generation.py @@ -0,0 +1,68 @@ +# SPDX-FileCopyrightText: © 2024 Tenstorrent AI ULC +# SPDX-License-Identifier: Apache-2.0 + +# Phi2 Demo - Text Generation + +import os +import pybuda + +from transformers import PhiForCausalLM, AutoTokenizer, PhiConfig +from pybuda.transformers.pipeline import pipeline as pybuda_pipeline + +def run_phi2_causal_lm(batch_size=1): + os.environ["TT_BACKEND_TIMEOUT"] = '0' + + # Set PyBuda configurations + compiler_cfg = pybuda.config._get_global_compiler_config() + compiler_cfg.default_df_override = pybuda.DataFormat.Float16_b + compiler_cfg.enable_auto_fusing = True + compiler_cfg.balancer_policy = "Ribbon" + + # Setup model configuration + config = PhiConfig.from_pretrained("microsoft/phi-2") + config.use_cache = False + config.return_dict = False + + # Load model and tokenizer with config + model = PhiForCausalLM.from_pretrained("microsoft/phi-2", config=config) + tokenizer = AutoTokenizer.from_pretrained("microsoft/phi-2") + tokenizer.pad_token, tokenizer.pad_token_id = (tokenizer.eos_token, tokenizer.eos_token_id) + + # Disable DynamicCache + # See: https://github.com/tenstorrent/tt-buda/issues/42 + model._supports_cache_class = False + + # Example usage + prompt = ["My name is Jim Keller and"] * batch_size + + # Initialize pipeline + text_generator = pybuda_pipeline( + "text-generation", + model=model, + tokenizer=tokenizer + ) + + # Inference on TT device + response = text_generator( + prompt, + temperature=0.7, + top_k=50, + top_p=0.9, + max_new_tokens=512, + num_beams=1, + do_sample=True, + no_repeat_ngram_size=5, + pad_token_id=tokenizer.pad_token_id, + eos_token_id=tokenizer.eos_token_id, + early_stopping=True + ) + + # Display Responses + for batch_id in range(batch_size): + print(f"Batch: {batch_id}") + print(f"Response: {response[batch_id][0]['generated_text']}") + print() + + +if __name__ == "__main__": + run_phi2_causal_lm() \ No newline at end of file diff --git a/model_demos/pyproject.toml b/model_demos/pyproject.toml index 334d1b69..e691f978 100644 --- a/model_demos/pyproject.toml +++ b/model_demos/pyproject.toml @@ -87,4 +87,5 @@ markers = [ "yolov6: tests that involve yolov6", "segformer: tests that involve SegFormer", "monodle: tests that involve Monodle", + "phi2": tests that involve Phi2", ] diff --git a/model_demos/tests/test_pytorch_phi2.py b/model_demos/tests/test_pytorch_phi2.py new file mode 100644 index 00000000..ba27124e --- /dev/null +++ b/model_demos/tests/test_pytorch_phi2.py @@ -0,0 +1,10 @@ +# SPDX-FileCopyrightText: © 2024 Tenstorrent AI ULC +# SPDX-License-Identifier: Apache-2.0 + +import pytest + +from nlp_demos.phi2.pytorch_phi2_text_generation import run_phi2_causal_lm + +@pytest.mark.qwen1_5 +def test_qwen1_5_causal_lm_pytorch(clear_pybuda, test_device, batch_size): + run_phi2_causal_lm(batch_size=batch_size) \ No newline at end of file From fd76bb5c8e5a11594f12c4de3c1157a9598ecf2e Mon Sep 17 00:00:00 2001 From: JushBJJ Date: Thu, 1 Aug 2024 00:26:19 +0000 Subject: [PATCH 11/14] Remove old phi2 demo --- .../model_phi2/phi2_text_generation.py | 60 ------------------- 1 file changed, 60 deletions(-) delete mode 100644 model_demos/nlp_demos/model_phi2/phi2_text_generation.py diff --git a/model_demos/nlp_demos/model_phi2/phi2_text_generation.py b/model_demos/nlp_demos/model_phi2/phi2_text_generation.py deleted file mode 100644 index cd7bd9a6..00000000 --- a/model_demos/nlp_demos/model_phi2/phi2_text_generation.py +++ /dev/null @@ -1,60 +0,0 @@ -import os -import pybuda - -from transformers import PhiForCausalLM, AutoTokenizer, PhiConfig -from pybuda.transformers.pipeline import pipeline as pybuda_pipeline - -model_name = "microsoft/phi-2" - -def run_qwen_causal_lm(max_length=1024, top_p=0.9, top_k=50, temperature=0.7): - # Set PyBuda configurations - compiler_cfg = pybuda.config._get_global_compiler_config() - compiler_cfg.default_df_override = pybuda.DataFormat.Float16_b - compiler_cfg.enable_auto_fusing = False - compiler_cfg.balancer_policy = "Ribbon" - - # Load the model configuration - config = PhiConfig.from_pretrained(model_name) - config.use_cache = True - - # Load the model and tokenizer with the updated config - model = PhiForCausalLM.from_pretrained(model_name, config=config) - tokenizer = AutoTokenizer.from_pretrained(model_name) - - # Set pad token - tokenizer.pad_token = tokenizer.eos_token - - # Example usage - prompt = ["My name is Jimmy and"] - - # Initialize pipeline - text_generator = pybuda_pipeline( - "text-generation", - model=model, - tokenizer=tokenizer - ) - - # Inference - output = text_generator( - prompt, - do_sample=True, - num_beams=1, - no_repeat_ngram_size=2, - pad_token_id=tokenizer.pad_token_id, - max_new_tokens=max_length, - temperature=temperature, - top_k=top_k, - top_p=top_p, - ) - - # Display output - print("OUTPUT:\n", output[0][0]["generated_text"]) - - -if __name__ == "__main__": - run_qwen_causal_lm( - max_length=100, - top_p=0.7, - top_k=50, - temperature=0.7 - ) \ No newline at end of file From 0fa971c97d9ce3fd6710f1009935fa1f318c5b90 Mon Sep 17 00:00:00 2001 From: JushBJJ Date: Thu, 1 Aug 2024 00:27:11 +0000 Subject: [PATCH 12/14] fix missing quote in pyproject.toml --- model_demos/pyproject.toml | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/model_demos/pyproject.toml b/model_demos/pyproject.toml index e691f978..bb1a4628 100644 --- a/model_demos/pyproject.toml +++ b/model_demos/pyproject.toml @@ -87,5 +87,5 @@ markers = [ "yolov6: tests that involve yolov6", "segformer: tests that involve SegFormer", "monodle: tests that involve Monodle", - "phi2": tests that involve Phi2", + "phi2": "tests that involve Phi2", ] From d40ca54e0848e87d3cd8752ce624294a67b53ea3 Mon Sep 17 00:00:00 2001 From: JushBJJ Date: Thu, 1 Aug 2024 00:28:07 +0000 Subject: [PATCH 13/14] fix --- model_demos/pyproject.toml | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/model_demos/pyproject.toml b/model_demos/pyproject.toml index bb1a4628..ea8239d7 100644 --- a/model_demos/pyproject.toml +++ b/model_demos/pyproject.toml @@ -87,5 +87,5 @@ markers = [ "yolov6: tests that involve yolov6", "segformer: tests that involve SegFormer", "monodle: tests that involve Monodle", - "phi2": "tests that involve Phi2", + "phi2: tests that involve Phi2", ] From d55116db4b54443421180c2959f653baaa2e866e Mon Sep 17 00:00:00 2001 From: JushBJJ Date: Sat, 31 Aug 2024 01:00:26 +0000 Subject: [PATCH 14/14] Fix test saying qwen1_5 instead of phi2 --- model_demos/tests/test_pytorch_phi2.py | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/model_demos/tests/test_pytorch_phi2.py b/model_demos/tests/test_pytorch_phi2.py index ba27124e..394fb41d 100644 --- a/model_demos/tests/test_pytorch_phi2.py +++ b/model_demos/tests/test_pytorch_phi2.py @@ -5,6 +5,6 @@ from nlp_demos.phi2.pytorch_phi2_text_generation import run_phi2_causal_lm -@pytest.mark.qwen1_5 -def test_qwen1_5_causal_lm_pytorch(clear_pybuda, test_device, batch_size): +@pytest.mark.phi2 +def test_phi2_causal_lm_pytorch(clear_pybuda, test_device, batch_size): run_phi2_causal_lm(batch_size=batch_size) \ No newline at end of file