Add model support Multilingual-Multimodal-NLP/IndustrialCoder-32B in huggingface (#455)

* support model Multilingual-Multimodal-NLP/IndustrialCoder-32B with vllm

* Add IndustrialCoder configs and run script

* all

* all

* quantize Industral-Coder-Thinging

* Delete thinkingmodel

* Update .gitignore

* style: satisfy pre-commit after main merge

---------

Co-authored-by: Yang Yong (雍洋) <yongyang1030@163.com>

Author: Charles2530

configs/quantization/video_gen/wan_i2v/awq_w_a.yaml

@@ -46,4 +46,4 @@ quant:
             clip_sym: True
 save:
     save_lightx2v: True
-    save_path: /path/to/x2v/
+    save_path: /path/to/x2v/

configs/quantization/video_gen/wan_t2v/awq_w_a.yaml

@@ -46,4 +46,4 @@ quant:
             clip_sym: True
 save:
     save_lightx2v: True
-    save_path: /path/to/x2v/
+    save_path: /path/to/x2v/

configs/quantization/video_gen/wan_t2v/rtn_w_a.yaml

@@ -29,4 +29,4 @@ quant:
             granularity: per_token
 save:
     save_lightx2v: True
-    save_path: /path/to/x2v/
+    save_path: /path/to/x2v/

configs/quantization/video_gen/wan_t2v/smoothquant_w_a.yaml

@@ -42,4 +42,4 @@ quant:
             alpha: 0.7
 save:
     save_lightx2v: True
-    save_path: /path/to/x2v/
+    save_path: /path/to/x2v/

llmc/compression/quantization/base_blockwise_quantization.py

@@ -35,11 +35,7 @@
                            _TRANSFORMERS_LN_TYPES_, EffcientFakeQuantLinear,
                            FakeQuantLinear, LlmcActFn, OriginFloatLinear,
                            RotateLinear)
-from .quant import (
-    FloatQuantizer,
-    IntegerQuantizer,
-    Weight48IntegerQuantizer,
-)
+from .quant import FloatQuantizer, IntegerQuantizer, Weight48IntegerQuantizer
 
 
 class BaseBlockwiseQuantization(BlockwiseOpt):

llmc/compression/quantization/module_utils.py

@@ -446,6 +446,14 @@ def __repr__(self):
         return 'LlmcQwen2RMSNorm()'
 
 
+class LlmcIndustrialCoderRMSNorm(LlmcLlamaRMSNorm):
+    def __init__(self, weight, eps=1e-6):
+        super().__init__(weight, eps)
+
+    def __repr__(self):
+        return 'LlmcIndustrialCoderRMSNorm()'
+
+
 class LlmcMixtralRMSNorm(LlmcLlamaRMSNorm):
     def __init__(self, weight, eps=1e-6):
         super().__init__(weight, eps)
@@ -1187,6 +1195,7 @@ def __repr__(self):
     'Mixtral': LlmcMixtralRMSNorm,
     'Interlm2': LlmcInternLM2RMSNorm,
     'Qwen2': LlmcQwen2RMSNorm,
+    'IndustrialCoder': LlmcIndustrialCoderRMSNorm,
     'Gemma2': LlmcGemma2RMSNorm,
     'MiniCPM': LlmcMiniCPMRMSNorm,
     'Starcoder': LlmcLayerNorm,

llmc/data/dataset/base_dataset.py

@@ -3,11 +3,12 @@
 from abc import ABCMeta
 
 import torch
-from datasets import load_dataset, load_from_disk
 from loguru import logger
 from PIL import Image
 from torch.nn import functional as F
 
+from datasets import load_dataset, load_from_disk
+
 from .specified_preproc import PREPROC_REGISTRY
 
 
@@ -172,9 +173,10 @@ def get_batch_process(self, samples):
         return calib_model_inputs
 
     def get_calib_dataset(self):
-        samples = self.calib_dataset[
-            int(os.environ['RANK'])::int(os.environ['WORLD_SIZE'])
-        ]
+        samples = self.calib_dataset.shard(
+            num_shards=int(os.environ['WORLD_SIZE']),
+            index=int(os.environ['RANK'])
+        )
         logger.info(f'len(samples) rank : {len(samples)}')
 
         calib_model_inputs = self.get_calib_model_inputs(samples)

llmc/eval/eval_base.py

@@ -5,10 +5,11 @@
 
 import torch
 import torch.nn as nn
-from datasets import load_dataset, load_from_disk
 from human_eval.data import read_problems
 from loguru import logger
 
+from datasets import load_dataset, load_from_disk
+
 
 class BaseEval:
     def __init__(self, model, config):

llmc/eval/eval_ppl.py

@@ -3,10 +3,11 @@
 
 import torch
 import torch.nn as nn
-from datasets import load_dataset, load_from_disk
 from loguru import logger
 from tqdm import tqdm
 
+from datasets import load_dataset, load_from_disk
+
 from .eval_base import BaseEval
 
 

llmc/models/__init__.py

@@ -5,6 +5,7 @@
 from .falcon import Falcon
 from .gemma2 import Gemma2
 from .glm4v import GLM4V
+from .industrialcoder import IndustrialCoder
 from .internlm2 import InternLM2
 from .internomni import InternOmni
 from .internvl2 import InternVL2
@@ -35,6 +36,6 @@
 from .videollava import VideoLLaVA
 from .vila import Vila
 from .vit import Vit
+from .wan2_2_t2v import Wan2T2V
 from .wan_i2v import WanI2V
 from .wan_t2v import WanT2V
-from .wan2_2_t2v import Wan2T2V