Skip to content

Commit 3cfe44d

Browse files
sophies927sophies927
authored
Create new local model versions for models with perf issues (#4162)
* Update paths and versions for models with perf improvements * Update Phi-4-reasoning parent assetIds * Fix Qwen2.5-coder-1.5b generic GPU variant name * Update file size and vram info for new versions * Update model.yaml to fix container_path * Update spec.yaml to fix directoryPath * Update model.yaml to remove changes * Update model.yaml to remove changes * Update model.yaml to revert changes * Update model.yaml to remove new line * Update model.yaml to match existing file * Update model.yaml to remove line 9 * Update model.yaml to revert changes
1 parent 9ff47c4 commit 3cfe44d

35 files changed

Lines changed: 80 additions & 80 deletions

File tree

‎assets/models/system/DeepSeek-R1-Distill-Qwen-1.5B-cuda-gpu/model.yaml‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1,6 +1,6 @@
11
path:
22
container_name: models
3-
container_path: foundrylocal/foundry-local/deepseek-r1-distill-qwen-1.5b/onnx/cuda/cuda-int4-rtn-block-32
3+
container_path: foundrylocal/fl-perf-improvements/deepseek-r1-distill-qwen-1.5b/onnx/cuda/cuda-int4-kquant-block-128-mixed/v2
44
storage_name: automlcesdkdataresources
55
type: azureblob
66
publish:

‎assets/models/system/DeepSeek-R1-Distill-Qwen-1.5B-cuda-gpu/spec.yaml‎

Lines changed: 4 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -1,6 +1,6 @@
11
$schema: https://azuremlschemas.azureedge.net/latest/model.schema.json
22
name: deepseek-r1-distill-qwen-1.5b-cuda-gpu
3-
version: 1
3+
version: 2
44
path: ./
55
tags:
66
foundryLocal: ""
@@ -12,7 +12,7 @@ tags:
1212
task: chat-completion
1313
maxOutputTokens: 2048
1414
alias: deepseek-r1-1.5b
15-
directoryPath: cuda-int4-rtn-block-32
15+
directoryPath: v2
1616
promptTemplate: "{\"assistant\": \"{Content}\", \"prompt\": \"\\\\u003C\\\\uFF5CUser\\\\uFF5C\\\\u003E{Content}\\\\u003C\\\\uFF5CAssistant\\\\uFF5C\\\\u003E\"}"
1717
type: custom_model
1818
variantInfo:
@@ -23,5 +23,5 @@ variantInfo:
2323
quantization: ['RTN']
2424
device: 'gpu'
2525
executionProvider: 'CUDAExecutionProvider'
26-
fileSizeBytes: 1073741824
27-
vRamFootprintBytes: 1362861314
26+
fileSizeBytes: 1535450808
27+
vRamFootprintBytes: 1535670046

‎assets/models/system/DeepSeek-R1-Distill-Qwen-1.5B-generic-cpu/model.yaml‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1,6 +1,6 @@
11
path:
22
container_name: models
3-
container_path: foundrylocal/foundry-local/deepseek-r1-distill-qwen-1.5b/onnx/cpu_and_mobile/cpu-int4-rtn-block-32-acc-level-4
3+
container_path: foundrylocal/fl-perf-improvements/deepseek-r1-distill-qwen-1.5b/onnx/cpu_and_mobile/cpu-int4-kquant-block-128-mixed-acc-level-4/v2
44
storage_name: automlcesdkdataresources
55
type: azureblob
66
publish:

‎assets/models/system/DeepSeek-R1-Distill-Qwen-1.5B-generic-cpu/spec.yaml‎

Lines changed: 4 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -1,6 +1,6 @@
11
$schema: https://azuremlschemas.azureedge.net/latest/model.schema.json
22
name: deepseek-r1-distill-qwen-1.5b-generic-cpu
3-
version: 1
3+
version: 2
44
path: ./
55
tags:
66
foundryLocal: ""
@@ -12,7 +12,7 @@ tags:
1212
task: chat-completion
1313
maxOutputTokens: 2048
1414
alias: deepseek-r1-1.5b
15-
directoryPath: cpu-int4-rtn-block-32-acc-level-4
15+
directoryPath: v2
1616
promptTemplate: "{\"assistant\": \"{Content}\", \"prompt\": \"\\\\u003C\\\\uFF5CUser\\\\uFF5C\\\\u003E{Content}\\\\u003C\\\\uFF5CAssistant\\\\uFF5C\\\\u003E\"}"
1717
type: custom_model
1818
variantInfo:
@@ -23,5 +23,5 @@ variantInfo:
2323
quantization: ['RTN']
2424
device: 'cpu'
2525
executionProvider: 'CPUExecutionProvider'
26-
fileSizeBytes: 1964944541
27-
vRamFootprintBytes: 1965162961
26+
fileSizeBytes: 2061584302
27+
vRamFootprintBytes: 2061807165

‎assets/models/system/DeepSeek-R1-Distill-Qwen-1.5B-generic-gpu/model.yaml‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1,6 +1,6 @@
11
path:
22
container_name: models
3-
container_path: foundrylocal/foundry-local/deepseek-r1-distill-qwen-1.5b/onnx/directml/directml-int4-rtn-block-32-acc-level-4
3+
container_path: foundrylocal/fl-perf-improvements/deepseek-r1-distill-qwen-1.5b/onnx/webgpu/webgpu-int4-kquant-block-32-mixed-acc-level-4/v1
44
storage_name: automlcesdkdataresources
55
type: azureblob
66
publish:

‎assets/models/system/DeepSeek-R1-Distill-Qwen-1.5B-generic-gpu/spec.yaml‎

Lines changed: 4 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -1,6 +1,6 @@
11
$schema: https://azuremlschemas.azureedge.net/latest/model.schema.json
22
name: deepseek-r1-distill-qwen-1.5b-generic-gpu
3-
version: 1
3+
version: 2
44
path: ./
55
tags:
66
foundryLocal: ""
@@ -12,7 +12,7 @@ tags:
1212
task: chat-completion
1313
maxOutputTokens: 2048
1414
alias: deepseek-r1-1.5b
15-
directoryPath: directml-int4-rtn-block-32-acc-level-4
15+
directoryPath: v1
1616
promptTemplate: "{\"assistant\": \"{Content}\", \"prompt\": \"\\\\u003C\\\\uFF5CUser\\\\uFF5C\\\\u003E{Content}\\\\u003C\\\\uFF5CAssistant\\\\uFF5C\\\\u003E\"}"
1717
type: custom_model
1818
variantInfo:
@@ -23,5 +23,5 @@ variantInfo:
2323
quantization: ['RTN']
2424
device: 'gpu'
2525
executionProvider: 'WebGPUExecutionProvider'
26-
fileSizeBytes: 1362648117
27-
vRamFootprintBytes: 1362929128
26+
fileSizeBytes: 1610612736
27+
vRamFootprintBytes: 1610930124

‎assets/models/system/DeepSeek-R1-Distill-Qwen-7B-cuda-gpu/model.yaml‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1,6 +1,6 @@
11
path:
22
container_name: models
3-
container_path: foundrylocal/foundry-local/deepseek-r1-distill-qwen-7b/onnx/cuda/cuda-int4-rtn-block-32
3+
container_path: foundrylocal/fl-perf-improvements/deepseek-r1-distill-qwen-7b/onnx/cuda/cuda-int4-kquant-block-128-mixed/v2
44
storage_name: automlcesdkdataresources
55
type: azureblob
66
publish:

‎assets/models/system/DeepSeek-R1-Distill-Qwen-7B-cuda-gpu/spec.yaml‎

Lines changed: 4 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -1,6 +1,6 @@
11
$schema: https://azuremlschemas.azureedge.net/latest/model.schema.json
22
name: deepseek-r1-distill-qwen-7b-cuda-gpu
3-
version: 1
3+
version: 2
44
path: ./
55
tags:
66
foundryLocal: ""
@@ -12,7 +12,7 @@ tags:
1212
task: chat-completion
1313
maxOutputTokens: 2048
1414
alias: deepseek-r1-7b
15-
directoryPath: cuda-int4-rtn-block-32
15+
directoryPath: v2
1616
promptTemplate: "{\"assistant\": \"{Content}\", \"prompt\": \"\\\\u003C\\\\uFF5CUser\\\\uFF5C\\\\u003E{Content}\\\\u003C\\\\uFF5CAssistant\\\\uFF5C\\\\u003E\"}"
1717
type: custom_model
1818
variantInfo:
@@ -23,5 +23,5 @@ variantInfo:
2323
quantization: ['RTN']
2424
device: 'gpu'
2525
executionProvider: 'CUDAExecutionProvider'
26-
fileSizeBytes: 5096273664
27-
vRamFootprintBytes: 5096487731
26+
fileSizeBytes: 5669356830
27+
vRamFootprintBytes: 5669577287

‎assets/models/system/DeepSeek-R1-Distill-Qwen-7B-generic-cpu/model.yaml‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1,6 +1,6 @@
11
path:
22
container_name: models
3-
container_path: foundrylocal/foundry-local/deepseek-r1-distill-qwen-7b/onnx/cpu_and_mobile/cpu-int4-rtn-block-32-acc-level-4
3+
container_path: foundrylocal/fl-perf-improvements/deepseek-r1-distill-qwen-7b/onnx/cpu_and_mobile/cpu-int4-kquant-block-128-mixed-acc-level-4/v2
44
storage_name: automlcesdkdataresources
55
type: azureblob
66
publish:

‎assets/models/system/DeepSeek-R1-Distill-Qwen-7B-generic-cpu/spec.yaml‎

Lines changed: 4 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -1,6 +1,6 @@
11
$schema: https://azuremlschemas.azureedge.net/latest/model.schema.json
22
name: deepseek-r1-distill-qwen-7b-generic-cpu
3-
version: 1
3+
version: 2
44
path: ./
55
tags:
66
foundryLocal: ""
@@ -12,7 +12,7 @@ tags:
1212
task: chat-completion
1313
maxOutputTokens: 2048
1414
alias: deepseek-r1-7b
15-
directoryPath: cpu-int4-rtn-block-32-acc-level-4
15+
directoryPath: v2
1616
promptTemplate: "{\"assistant\": \"{Content}\", \"prompt\": \"\\\\u003C\\\\uFF5CUser\\\\uFF5C\\\\u003E{Content}\\\\u003C\\\\uFF5CAssistant\\\\uFF5C\\\\u003E\"}"
1717
type: custom_model
1818
variantInfo:
@@ -23,5 +23,5 @@ variantInfo:
2323
quantization: ['RTN']
2424
device: 'cpu'
2525
executionProvider: 'CPUExecutionProvider'
26-
fileSizeBytes: 6667944735
27-
vRamFootprintBytes: 6668163013
26+
fileSizeBytes: 6904159928
27+
vRamFootprintBytes: 6904383723

0 commit comments

Comments
 (0)